With the impending takeover and no recent news, does anyone have any insight into the future of the XMEGA range? New models, long term development?
Personally I'd love to see a USB enabled E5.
With the impending takeover and no recent news, does anyone have any insight into the future of the XMEGA range? New models, long term development?
Personally I'd love to see a USB enabled E5.
The Xmega is an excellent chip. It may not have caught on with the hobbyist market.
I would guess that the production cost is purely dependent on the size of the die. In other words, if a complex Xmega or ARM can fit in the same area of Silicon Real Estate as a simple AVR, you have the same cost.
Of course the profitability depends on how much you can sell the chips for. The new owners will make their own decisions.
David.
New models, ...
The December 2015 product selection guide does not have a new product column.
http://www.atmel.com/images/atmel-45154-product-selection-guide_brochure.pdf (page 5)
Personally I'd love to see a USB enabled E5.
All :
Everspin Technologies
Everspin Embedded MRAM
Edit : MRAM URL.
Hey mojo, if they drop the E5 range then we will need to start a law suit. I would have wasted around AU$100,00.00 + on a product which has been in development for the past 3 years and almost getting all the relevant approvals to go on the market.
Hey mojo, if they drop the E5 range then we will need to start a law suit. I would have wasted around AU$100,00.00 + on a product which has been in development for the past 3 years and almost getting all the relevant approvals to go on the market.
I really doubt they would drop a finished die, unless
a) the volumes were very small, or yields were terrible. (Disti stock levels suggest this has good design wins)
or
b) there were some FAB line issues ( that one seems unlikely on a new part )
What can happen with mature parts, is the order codes can rationalize.
In an E5 series that would mean the 16k version may go, leaving the 8k (cheapest) & 32k (largest).
The price of mature devices also seldom decrease, so over time they become less of a 'go to' choice.
I doubt they would drop the E5, or even some of the flash memory sizes. Even if the yields don't supply a lot of 16k or 8k parts through binning they will just cripple some to make up the numbers for people who need them. Also, if anything Microchip are well known for never discontinuing anything (I think there have been about 3-4 PICs discontinued over the years) so it seems unlikely they should start pushing Atmel to do that.
I'm just hoping we see continued development.
I'm just hoping we see continued development.
Personally I wouldn't hold my breath on that one.
Shame Atmel's published figures don't show the split of revenue between the processor families. All the processors together created a revenue of around $1bn in both 2013 and 2014 but we don't know how that splits across brain-dead/tiny/8051/mega/xmega/UC3/SAM. I would guess that Xmega was not a large contributor but I guess it only takes Apple or Sony to have used one in a 100 million selling product to bump the numbers? ;-)
As others have said they probably won't EOL existing designs as long as they have the masks and they can see SOME demand but whether they pour revenue into on-going design is another matter. Remember that now they won't just be weighed against: brain-dead/tiny/8051/mega/xmega/UC3/SAM but also PIC16/PIC18/dsPIC/PIC32 etc etc.
(one attractive thing PICs have are USB devices in DIP packages - easy for hobbyists armed with nothing more than a breadboard! I actually looked at using one to make a very cheap AVR ISP programmer (ironic eh?) just before the USBAsp caught hold in a big way!)
I don't think any of the PIC range are as nice to work with as XMEGA. MPLABX is pretty awful too.
are as nice to work with as XMEGA
If you program in C how can you tell? Then it just becomes a question of the peripherals available and (possibly) how easy to use the manufacturer supplied support code is. Personally I've looked at support from Atmel (ASF) and the Microchip equivalent and I prefer the latter. Your mileage may vary.
I'm not throwing up my hands in despair yet.
I agree Tom.... none of us are any better at reading tea leaves than the next guy. Eventually we will be told.
+1 Tom and Ross.
Jim
mojo-chan wrote:
are as nice to work with as XMEGAIf you program in C how can you tell? Then it just becomes a question of the peripherals available and (possibly) how easy to use the manufacturer supplied support code is. Personally I've looked at support from Atmel (ASF) and the Microchip equivalent and I prefer the latter. Your mileage may vary.
Well, they don't have bits for controlling each peripheral spread randomly over a bunch of registers, for a start. I find things like the register naming and datasheets superior too.
none of us are any better at reading tea leaves
...or coffee residue at the bottom of the cup....
I like the xmega devices like the 128/256A3U, slightly lower cost than the 1284p and of course a lot more features.
There are certain times that 3.3v is a pain but still not a major problem.
I agree the documentation could use a re-work, at times I have to look in the definition file to find what I need.
Just received a newsletter from Atmel. Nothing about Microchip! Strange! I use xmega in my product, so it would be nice to know about their plans.
Just received a newsletter from Atmel. Nothing about Microchip! Strange! I use xmega in my product, so it would be nice to know about their plans.
Why do you think that they owe you a personalised report of their intentions? They probably don't know that level of detail themselves yet.
Sheesh... some people want the impossible! I am getting tired of all the "they ought to tell me everything" gripes and am tempted to put on my moderator's hat and just delete every one of them the moment they get posted.
There I have said it!
Ross
Apart from anything else I think it's the case that the legalities of the transfer mean that Microchip won't actually "OWN" it until May - so it may be a bit presumptuous to start advising customers about the way things might be headed just yet.
Of course this entire thread (like so much else about this takeover) is simply speculation of one amongst a million things that might happen )or might not!).
Who knows, perhaps Microchip bought Atmel for the very reason they want to get hold of the Xmega design and mainstream it?
jmaja1 wrote:Just received a newsletter from Atmel. Nothing about Microchip! Strange! I use xmega in my product, so it would be nice to know about their plans.
Why do you think that they owe you a personalised report of their intentions? They probably don't know that level of detail themselves yet.
I didn't say they should report to me. I just thought it is strange that their newsletter that I got today by e-mail doesn't mention anything about the deal. I wouldn't even know about it, if I hadn't read this thread. Of course it is way too early to tell anything specific, but usually customers are told about this kind of things and most often some frase about that nothing changes now and further information will be given later.
I hope some indication does come soon. I can't imagine XMEGA will go away, but it would be nice to know that development is going to continue.
I hope some indication does come soon. I can't imagine XMEGA will go away, but it would be nice to know that development is going to continue.
It would be nice if someone dropped $1m in small unmarked bills in my letterbox "soon", but it isn't going to happen any sooner than your wish, so stop beating us all up about it.... please. We here have no secret pipeline into the Microchip/Atmel board rooms. So please continue the discussion in your local pubs. The clientele there are likely to be just as well informed as us.
I hope some indication does come soon.
One more time...the deal will not be competed until mid-May. Until that time they are two separate companies; Microchip is Microchip, Atmel is Atmel. They have two separate boards who each have a legal responsibility to maximise shareholder value to two separate sets of shareholders. They have two sets of management responsible to those boards for implementing two separate corporate strategies.
At this stage, whilst the lawyers are going through due-diligence, it's highly likely that the management teams don't even have access to the other companies internal information. It's highly probable that Microchip don't know which Atmel product lines are profitable and which aren't.
And why would they. That's a level of detail several steps below what is looked at when considering corporate mergers.
If guaranteed product availability is vital to you then it's time to stock up. But you should be doing that anyway.
Okay, okay... Just to be clear, I'm not blaming anyone or demanding answers. It's just that some people have had insight in the past, so I thought I'd ask.
... whilst the lawyers are ...
http://www.leginfo.ca.gov/.const/.article_6 ... CALIFORNIA CONSTITUTION ARTICLE 6 JUDICIAL SEC. 9. The State Bar of California is a public corporation. Every person admitted and licensed to practice law in this State is and shall be a member of the State Bar except while holding office as a judge of a court of record....
Edit : attach
Pardon the pedantic,...
I like pedantic. Or is that 'pedantry'? (Or should that be 'Or is that 'pedantry?')
gchapman wrote:Pardon the pedantic,...
I like pedantic. Or is that 'pedantry'? (Or should that be 'Or is that 'pedantry?')
I think the question mark should be outside of the quotation. Indeed, there is no need for the quotation marks at all as you are not quoting the prior usage of the word in this thread.
Oh shoot me!
I like the xmega devices like the 128/256A3U, ...
Gabotronics - Development Boards and Electronic Kits
I'd like to see some more SRAM. We have a project coming up where we need to buffer some data before writing it to SD card, and it's a choice between XMEGA and PIC24F. The PIC range has the advantage of having much larger RAM, but the disadvantage of having a much worse IDE and the only fully free compiler being ancient. Atmel are in a good position if they can keep improving the range.
That's a place where ARMs tend to score higher than AVR/Xmega. In AVR the ratio of RAm to flash flash to RAM tends to be about 16:1. So a 16K flash AVR might have 1K RAM. In ARM it is often more like 4:1. So a 16K flash device might have 4K of RAM.
*edit: corrected the ratio definition. Phantom pedant *
Nice .sig, Cliff ;-)
I just thought it is strange that their newsletter that I got today by e-mail doesn't mention anything about the deal.
As long as it is not legally bound, I would think that most of the information on the deal will come out of Microchip (which have, AFAIK, bassoon'ed this out on their web site). For anyone at Atmel, right now it's "sit down in the boat, and don't rock it" - they are somewhat in a limbo where they can no longer act as Atmel-with-a-future-on-it's-own and still can not acct as Atmel-as-part-of-Microchip.
Look mojo-chan; you just have to wait for any news, and your projects will always be at risk.
Anyone using any product is in as somewhat more risky terrain now than they where half a year ago. But even back then you risked Atmel saying "device(faqmily) X is EOL as of now".
Did any of you guys receive a letter from Atmel with a guarantee of supply lifetime for any device?
No?
No. Basta.
'It would be nice if someone dropped $1m in small unmarked bills in my letterbox "soon"'
P.M. me your address, Ross.
That's a place where ARMs tend to score higher than AVR/Xmega. In AVR the ratio of RAm to flash tends to be about 16:1. So a 16K flash AVR might have 1K RAM. In ARM it is often more like 4:1. So a 16K flash device might have 4K of RAM.
Indeed, we have looked at ARM. Our products run for many years on AA batteries though, and ARM just can't compete for ultra low power processing. Sleep modes are somewhat comparable (but check the electrical requirements), but not active modes.
and ARM just can't compete for ultra low power processing
Not even the M0+? The "+" is in part about power saving while running. Rather curiously that involves dropping from a 3 stage to a 2 stage pipeline among other things.
Look mojo-chan; you just have to wait for any news, and your projects will always be at risk.
Anyone using any product is in as somewhat more risky terrain now than they where half a year ago. But even back then you risked Atmel saying "device(faqmily) X is EOL as of now".
Did any of you guys receive a letter from Atmel with a guarantee of supply lifetime for any device?
No?
No. Basta.
Sure, I appreciate that. I'm not worried about XMEGA going EOL. Microchip famously rarely EOL anything, even ancient OTP PICs in DIP packages that are only used in some obscure Bulgarian toaster. I'd be surprised if they started by obsoleting a bunch of Atmel parts, especially relatively modern and well liked ones like XMEGA. Aside from anything else it would devalue their nice new acquisition massively, as everyone would assume all Atmel parts are at risk, and probably a bunch of Microchip ones too.
No, I'm just interested to know if there are any NEW parts on the horizon, or if anything has been said regarding future developments. I know some people here have their ears to the ground.
mojo-chan wrote:and ARM just can't compete for ultra low power processingNot even the M0+? The "+" is in part about power saving while running. Rather curiously that involves dropping from a 3 stage to a 2 stage pipeline among other things.
On paper they look pretty good, but when you try to implement the same algorithms they end up worse and 8 bit MCUs. I suppose it really depends on what kind of processing you are doing and how you are interacting with the peripherals.
You think the 'people who have their ears to the ground' would ever reveal anything on a public forum?
The only way to guarantee availability is to do like the Soviet Union did in the sixties during the race to the moon: Build a giant facility which pulled in raw materials at one end, and pumped out rocket ships at the other.
These doomsday Micromel/AtChip threads are awfully entertaining, though. I'll have to go stock up on popcorn again soon ;-)
Of course, if anyone wants to know the official Atmel position it's right there on the front page of their website and has been for over a week.
Of course, if anyone wants to know the official Atmel position it's right there on the front page of their website and has been for over a week.
January 19, 2016
Dear Valued Atmel Customers and Channel Partners,
Today Microchip Technology announced the signing of a definitive agreement to acquire Atmel. The transaction is subject to the approval of shareholders of Atmel, regulatory approvals and customary closing conditions. We expect the transaction to close in the second calendar quarter of 2016.
Microchip is a leading provider of embedded control solutions through our Microcontroller, Mixed-Signal, Analog, Flash IP solutions and Memory product lines. This acquisition adds Atmel’s strong portfolio of Microcontrollers, Wireless, Touch, Security, Memory and Automotive products to Microchip. The combined product lines of Microchip and Atmel will offer you a broader range of innovative solutions to serve your needs.
For now, and until the transaction closes, it will be business as usual for customers and channel partners. We will communicate any changes to you in a timely fashion after the transaction has closed. But for the foreseeable future, we request that you continue to do business using your current Atmel contacts and processes.
Together, we remain committed to providing you with outstanding customer service, advanced technology solutions, industry-leading supply chain management, and world-class quality and manufacturing. We look forward to a long and mutually beneficial relationship.
If you have any questions or concerns, please contact your local sales contact.
Best Regards,
Steve Sanghi
President and CEO
Microchip Technology Incorporated
Steven Laub
President and CEO
Atmel Corporation
Cautionary Statement:
Statements about the expected timing, completion and effects of the proposed transaction, and other statements in this letter that are not historical facts, are forward-looking statements made pursuant to the safe harbor provisions of the Private Securities Litigation Reform Act of 1995. These statements involve risks and uncertainties that could cause actual results to differ materially, including, but not limited to the actual timing of the closing of the acquisition, the satisfaction of the conditions to closing in the acquisition agreement, any termination of the acquisition agreement, the effect of the acquisition on Microchip’s and Atmel’s existing relationships with customers and vendors and their operating results and businesses; the costs and outcome of any litigation involving Microchip, Atmel or the acquisition transaction; general economic, industry or political conditions in the U.S. or internationally; and the risks described from time to time in SEC reports including filings on Forms 10-K, 10-Q and 8-K. You can obtain copies of such Forms 10-K, 10-Q and 8-K and other relevant documents for free, as applicable, at Microchip’s website (www.microchip.com), at Atmel’s website (www.atmel.com), the SEC's website (www.sec.gov) or from commercial document retrieval services. You are cautioned not to place undue reliance on our forward- looking statements, which speak only as of the date such statements are made. We do not undertake any obligation to publicly update any forward-looking statements to reflect events, circumstances or new information after the date hereof.
Letter to Customers and Channel Partners – Page 2
Additional Information and Where to Find It
This document does not constitute an offer to sell or the solicitation of an offer to buy any securities or a solicitation of any vote or approval nor shall there be any sale of securities in any jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such jurisdiction. The proposed transaction will be submitted to the stockholders of Atmel for their consideration. Microchip will file a Registration Statement on Form S-4 that will include a proxy statement of Atmel that will also constitute a prospectus of Microchip in connection with the acquisition transaction. Investors and security holders are urged to read this document when it becomes available because it will contain important information about the transaction. Investors and security holders may obtain free copies of this document (when it is available) and other documents filed with the SEC at the SEC's web site at www.sec.gov. Microchip, Atmel and their directors and executive officers may be deemed to be participants in the solicitation of proxies from the stockholders of Atmel in connection with the acquisition transaction. Information regarding the special interests of these directors and executive officers in the transaction will be included in the proxy statement/prospectus described above. Additional information regarding the directors and executive officers of Microchip is also included in Microchip's proxy statement for its 2015 Annual Meeting of Stockholders, which was filed with the SEC on July 10, 2015, and Microchip’s amendment to its Annual Report on Form 10-K for the fiscal year ended March 31, 2015, filed with the SEC on June 8, 2015. Additional information regarding the directors and executive officers of Atmel is also included in Atmel’s proxy statement for its 2015 Annual Meeting of Stockholders, which was filed with the SEC on April 3, 2015. These documents are available free of charge at the SEC's web site at www.sec.gov and as described above.
I'd like to see some more SRAM. We have a project coming up where we need to buffer some data before writing it to SD card, and it's a choice between XMEGA ...
Digital Synthesizer with ATxmega128
by Rolf
https://www.avrfreaks.net/comment/1551226#comment-1551226 (scroll down a bit for the schematic with 1MB SRAM by 2.5-port EBI and the SD)
... and PIC24F.
... and the only fully free compiler being ancient.
Atmel are in a good position if they can keep improving the range.
wiki.kewl.org
tools:xc16
Microchip XC16 compiler
http://wiki.kewl.org/dokuwiki/tools:xc16
...
Free edition
...
The free edition has no limitations and does not require a license for optimisations.
...
Edit : typo
Our products run for many years on AA batteries though, and ARM just can't compete for ultra low power processing.
Power-save current (typical, room temperature) with RTC :
Operating at well above room temperature would make these differences even more apparent.
Some very low power RTC (Ambiq Micro) are on MLCC.
Subthreshold transistors and MCUs
by Jack Ganssle
December 21, 2015
http://www.embedded.com/electronics-blogs/break-points/4441091/Subthreshold-transistors-and-MCUs
Abracon
ABX8XX Application Note
Using Low Cost Ceramic Capacitors for RTC Backup Power
On paper they look pretty good, but when you try to implement the same algorithms they end up worse and 8 bit MCUs.
I suppose it really depends on what kind of processing you are doing and how you are interacting with the peripherals.
ARM usually fits well for a compute-bound problem.
Depends for an I/O-bound problem.
Power-save is usually better for some 8-bit and some 16-bit MCUs.
© 2014 Copyright Atmel Corporation
Atmel Technology Live
8-bit AVR
http://www.atmel.com/images/8-bit%20AVR.pdf (go to page 4 for the start of "Why ...")
Thanks for the link to the Ganssle article. Lead me to Ambiqu:
Features
- Ultra-low active mode power consumption: 30µA/MHz (executing from Flash)
- Ultra-low sleep mode power consumption: 100nA (with RTC on)
- High-performance, 32-bit ARM Cortex-M4F processor
- Up to 24MHz clock frequency
- Floating point unit
- Wake-up interrupt controller with 12 interrupts
- Ultra-low power memory
- Up to 512kB Flash
- Up to 64kB low-leakage RAM
- Ultra-low power interface for off-chip sensors
- 10-bit, 13-channel, 1MS/s ADC
- Temperature sensor with ±2°C accuracy
- Rich set of timing peripherals
- Flexible serial peripherals
- I2C/SPI master for communication with external peripherals
- I2C/SPI slave for optional host communications
- UART for communication with peripherals and legacy devices
- Wide operating range: 1.8 to 3.8V
- Compact package options
- 64-pin BGA with 50 GPIO
- 42-pin CSP with 28 GPIO
30µA/MHz. Holy smokes. Even assuming that figure applies only to the bottom end of VCC (1.8V) and therefore at the bottom end of the speed grade, an AVR like the m328p needs 250µA/MHz at that same VCC, core only (i.e. not including oscillator). And only 8-bits, vs. the Apollo's 32-bits.
No datasheet yet, and no product either AFAICS, just an announcement (their only real product ATM seems to be a family of ULP RTC, although a brief search turns up no stock). Still, exciting.
EDIT: Ah. Datasheet:
http://ambiqmicro.com/sites/default/files/apollo-product-sheet-lowest-power-mcu.pdf
Integrated buck converter. Typical current goes >>down<< with increasing VCC.
Just found a case FOR XMega vs ARM M-series. EEPROM. I am told that, at least for Atmel's ARM chips, some include "EEPROM emulation" which appears to treat a block of flash as eeprom, but that not all do.
Jim
That's always been a limitation of most embedded ARMs that are available (not just Atmel). Either you connect up an external I2C/SPI storage device or you have to emulate it by writing flash. ASF provides this kind of thing:
Just found a case FOR XMega vs ARM M-series. EEPROM. I am told that, at least for Atmel's ARM chips, some include "EEPROM emulation" which appears to treat a block of flash as eeprom, but that not all do.
...
Yes and no! and that's for low-cost purposes and also there are some interesting points in comparing the flash of an ARM MCU like STM32 against EEPROM. look:
I said no because some ARM-based MCUs have EEPROM. e.g.
http://www.st.com/web/en/catalog/mmc/FM141/SC1169/SS1817
http://www.st.com/web/en/catalog/mmc/FM141/SC1169/SS1295
Added points in favor of devices with built-in EEPROM (real or emulated) -
1. No added component cost
2. No added assembly cost
3. No added board area
Jim
30µA/MHz. Holy smokes. Even assuming that figure applies only to the bottom end of VCC (1.8V) and therefore at the bottom end of the speed grade, an AVR like the m328p needs 250µA/MHz at that same VCC, core only (i.e. not including oscillator). And only 8-bits, vs. the Apollo's 32-bits.
Integrated buck converter. Typical current goes >>down<< with increasing VCC.
AVR's have no internal regulator, so their Icc varies greatly with Vcc.
Most ARMs are on finer process and better ones have internal regulators, which keeps Icc and MHz largely fixed
This example has more than one regulator, and the lower Icc when Buck-on, suggests a core Vcc of maybe 1.2V
There is also a growing number of ARMs now offering Wide Vcc, allowing direct 5V operation.
Nuvoton, Spansion, ABOV, Cypress, Infineon ....
This will have quite an effect on sales of generic non-5V capable parts.
They will be squeezed between super-low Icc parts like Apollo, and wide Vcc parts.
AVR's have no internal regulator, so their Icc varies greatly with Vcc.
That was neither a surprise, nor a mystery to me. However it means the claim of 30uA/MHz needs to be evaluated with the higher-end of the VCC range in mind, rather than the lower-end of VCC as you would for devices without an internal buck regulator, where ICC drops as VCC drops.
A better metric would be power/MHz rather than current/MHz.
It makes sense that a subthreshold device would require a carefully regulated supply.
suggests a core Vcc of maybe 1.2V
Possibly much lower. That is implied by the subthreshold design.
30µA/MHz. Holy smokes. Even assuming that figure applies only to the bottom end of VCC (1.8V) and therefore at the bottom end of the speed grade, an AVR like the m328p needs 250µA/MHz at that same VCC, core only (i.e. not including oscillator).
No datasheet yet, and no product either AFAICS, just an announcement ...
Atmel does similar by use of engineering sample MCU onto an Atmel board.
Good way to try something new in a proof-of-concept way (on the bench), work-around the known defects, and identify more defects.
An example of win-win for all.
(their only real product ATM seems to be a family of ULP RTC, although a brief search turns up no stock)
If yes, Ambiq Micro sells the die to Abracon for packaging and test; IIRC Infineon does similar with its RTC.
http://www.abracon.com/products.php?search=rtc&type=RTC%20IC%20-%20Ultra%20Low%20Power
http://www.abracon.com/news2013.php (search for Ambiq Micro)
EDIT: Ah. Datasheet:
Highlights the more than significant compute capability of ARM Cortex-M along with Ambiq Micro's ultra-low current feature.
Radios need a "lot" of current; therefore, compress and encrypt the data, turn on the radio, transmit, turn everything off, sleep.
Edit : 2nd URL
And the same points for the MCU with built-in FRAM.
Would like to see MCU with eMRAM (Embedded MRAM); has some significant advantages over FRAM though reduction of current is not one.
AVR's have no internal regulator, ...
An alternative are some of the Texas Instruments MSP430 at double the XMEGA384C3 32KB SRAM.
I did look at those, but we had some bad experiences with another company's modules that used them. Kept corrupting their own memories. Texas acknowledge it on their forums but don't offer any real solution, other than use another part. Plus, their IDE sucks.
[quote]There are recent free third-party builds of XC16.
I'll look into that, thanks. Seems a bit rubbish to use GCC, but then be dicks by charging for the optimization. It's allowed by the GPL, just not a very nice thing to do. Doesn't exactly encourage people to use their parts either, since if you want any commercial support I bet they will tell you to try the commercial compiler. You could argue it's a small cost for a company, but then why bother collecting small amounts of money in the first place?
Possibly much lower. That is implied by the subthreshold design.
30µA/MHz. Holy smokes.
Yeah, but there are caveats. For example, in one product I worked on we need to log data for an hour a data. Logging involves taking 100 samples per second and doing a bit of averaging. We need a ~16 bit ADC to get the dynamic range we require.
The solution was to use an XMEGA. It runs at 250kHz, waking up 100 times a second from a low power mode. When awake it turns on the ADC, takes 8 readings (idling while the ADC works), averages them and stores the result as 1 sample. After one second it does some rectification/averaging over the 100 samples. Although the XMEGA ADC is only 12 bit, we use the PGA to increase the dynamic range it can measure over.
What this boils down to use <2mA average current over the measurement period. We tried something similar with ARM but it just isn't as efficient when it comes to peripherals and idling, so you can't get near that performance. ARM is great if you need to do more processing of the data, but to get really low power you have to find ways of minimizing that.
We need a ~16 bit ADC to get the dynamic range we require....
Although the XMEGA ADC is only 12 bit, we use the PGA to increase the dynamic range it can measure over.
Some of the better characteristics of analog by CMOS are more easily obtained by larger process geometries (the "old" fab) and more easily operated by use of a battery (coin cell, super cap, MLCC, etc.).
Indeed, if the E5 had been out at the time we would have considered it. Using the automatic 16 bit mode and idling we could probably push the average current down even lower. I should mention that is current from a 3.6V lithium cell. We run the MCU at 2.7V via a Texas LDO with 0.5uA quiescent current. 2.7 is as low as we could go due to the requirements of other parts, and also because with the 0.6V overhead of the ADC it makes a 2.048V reference possible.
That is the sort of thing I'm hoping to see in future. 16 bit mode on the larger devices, USB on the E series, expansion of the custom logic. The enhanced DMA on the E5 is nice too, ideal for handling input from devices like GPS receivers.
What was(and also still is) always question to me is this, Why ARM-based MCUs and XMEGA series are not available in DIP package?(Just saw that NXP has two ARM-based MCUs which is available in DIP package) while you can find some DsPICs which are available in DIP package.
Some dsPICs, PIC24s and PIC32s are available in DIP.
Why ARM-based MCUs and XMEGA series are not available in DIP package?
NXP do an 8pin LPC800 in DIP. It's M0+.
If you only need the dynamic range, why not use a LOG ADC or opamp?
Yes, and the LPC800 is pretty useless too.
If you want to do some prototyping, an ARM evaluation board is practical. They even have onboard JTAG/SWD.
The Xmega Xplained boards are quite good but cannot compete with a $10 NUCLEO.
It is a pity that we do not have an Xmega Arduino. Then we would have Chinese clone boards. And the Xmega would become a hobbyist target.
David.
Maybe you should whisper in Massimo's ear David...
...Why ARM-based MCUs and XMEGA series are not available in DIP package?...
The target market does not need them in DIP so there is no point on doing it.
If you only need the dynamic range, why not use a LOG ADC or opamp?
Cost, board space. The old product used two log amps and a load of other analogue stuff, we eliminated it all by going to fully digital processing and massively reduced power consumption at the same time.
Maybe you should whisper in Massimo's ear David...
I would not dare. He might throw his rattle out of the pram.
I have a 32A4U module from Batsocks.
Looking on Ebay, there are similar DIL modules from Barion and MattairTech.
Nothing like the sheer range of Mega and ARM dev boards that are available.
From a GPIO point of view, the Xmega can compete for peripherals and speed with all the ARM Cortex. Obviously, if there is computation to be done, the 32-bit ARM wins.
David.
What this boils down to use <2mA average current over the measurement period. We tried something similar with ARM but it just isn't as efficient when it comes to peripherals and idling, so you can't get near that performance.
The draw from the ADC, on the other hand:
Yes, and the LPC800 is pretty useless too.
More or less useless than an 8pin DIP packaged AVR Tiny?
Obviously there is only so much you can do with 8 pins but one of the nice things about the LPC800 is that you can choose amongst it's range of peripherals within the chip and decide what ones you want to use and how you want them wired to the pins (with certain limits)
"Average run current", LOL. That's datasheet weasel words for "absolute lowest possible value you will ever see under very specific test conditions."
Sadly we can't compare it to XMEGA, because the datasheet only gives us a value of 1.6mA at 3.0V and 250ksps. At least it is specific about the test conditions.
"Average run current", LOL. That's datasheet weasel words for "absolute lowest possible value you will ever see under very specific test conditions."
More like "These are engineering samples". Like all parts, they will be properly characterised over time. Have a look at any of the first-run XMEGA datasheets and you'll the same kind of thing.
Yep, which is why you can't rely on those numbers at all. It just says "1", without even a decimal point, so I wouldn't rush to design it in to a product on that basis just yet.
Anyway, it also depends what the overhead is like. What sleep modes you can use, how long it takes to wake up, how long it takes to process, how fast you need to run the core to hit your target sample rate etc. That's where being 32 bit tends to be more of a hindrance than a help.
NXP do an 8pin LPC800 in DIP. It's M0+.
Yes Cliff and I said it.
And also it has two as I said.
An 8pin
And a 28(32?)pin
And bear in mind that there are JUst two DIP packaged ARM MCUs.
Yes, and the LPC800 is pretty useless too.
...
Agree
...
If you want to do some prototyping, an ARM evaluation board is practical. They even have onboard JTAG/SWD.
The Xmega Xplained boards are quite good but cannot compete with a $10 NUCLEO.
...
Yes but when you want to use an MCU with DIP package, you don't need any board. just put it on breadboard and program/debug it.
...
It is a pity that we do not have an Xmega Arduino. Then we would have Chinese clone boards. And the Xmega would become a hobbyist target.
Good point. really why?
The target market does not need them in DIP so there is no point on doing it.
What target Brian?
About 1/5 years ago I was chatting with a friend from Italy(in StackExechange). he was a student(Master degree) who was studying in Pisa university. he and his friends were working on a project to make a famous Robot. I asked him"Do you use Microchip MCUs?"
"Yes" he answered.
Again I asked "What sort of packages do you use?"
"As possible as DIP" he answerd.
I asked "Why?"
He said "because the MCUs in this package are easy to use"
This is his profile: http://electronics.stackexchange.com/users/16993/vladimir-cravero
I guess a lot of hobbyists,amateurs and other non-professionals(or maybe even professionals) prefer to use DIP package.
While SMD packages are cheaper but still some people or some companies prefer to use DIP. we have a SuperGeneral split air conditioner at home. this summer when we opened it to service(wash) it, I found out it has an NEC MCU in DIP package.(I don't know why everything I open, it has an NEC MCU. it looks like this company has been successful to pre-empt the market of MCUs)
What target Brian?
Low-technology consumer products, like air conditioners, still use DIP because it is cheaper than SMT, especially on low component count boards. High-volume high-technology consumer products, like set-top boxes uses SMT because it is cheaper than DIP in high-volume assembly (and often the chips used just won't fit in DIP).
The xMegas, even in volume, aren't a cheap chip and so will never be used in those low-volume low-technology products, therefore a DIP package is not required.
For companies developing products which will end up needing SMT assembly the additional cost of prototyping in SMT is not an issue.
What was(and also still is) always question to me is this, Why ARM-based MCUs and XMEGA series are not available in DIP package?(Just saw that NXP has two ARM-based MCUs which is available in DIP package) while you can find some DsPICs which are available in DIP package.
If you believe there is a market, buy Known Good Die, and package some yourself.
There is a reason the big players side-step DIP: the numbers simply fail to stack up, and volume users do not need them.
What I have noticed, is a move driven from Asia, to coarser pitch SMD, and that is driven by Wave Soldering.
QFN and 0.5mm parts cannot wave solder, but 1.27mm parts easily can, and it seems 0.8mm also can, with experience.
That is the 'new dip' -> Wave Solder-able Gull wing.
Fortunately for hobbyists, a part that is Wave Solder-able, is also not too hard to manually solder or design to.
Some vendors offer 64 pin MCUs in 0.5mm and 0.8mm pitch, and I see recently a SOPN20 (from Asia) that is SO16N plastic mold,(150 mil wide) with 0.8mm leads - a nice looking package, I hope catches on.
While SMD packages are cheaper but still some people or some companies prefer to use DIP. we have a SuperGeneral split air conditioner at home. this summer when we opened it to service(wash) it, I found out it has an NEC MCU in DIP package.
See above - was this wave soldered PCB ? They did not 'prefer' DIP.
The xMegas, even in volume, aren't a cheap chip and so will never be used in those low-volume low-technology products, therefore a DIP package is not required.
Why would xmega be too expensive for a low-volume product? A low-volume product will be expensive and paying a dollar more for each MCU is not going to change anything.
I used to hate when DIP was not available, but now I hate to use any through hole components. I guess I was just worried about my soldering skills, which aren't good, but I can manage to solder SMT, even quite small. For the first SMT parts I used I soldered wires to each leg/pad and used breadboard. Now I always make a PCB, which is cheap. Still worried about the possible next step to BGA etc. I have already used DFN.
Does any new design use DIP?
For companies developing products which will end up needing SMT assembly the additional cost of prototyping in SMT is not an issue.
I'm not sure there is even really any "additional cost of prototyping in SMT"
First evaluation can be done 95% of the time using a low-cost development board(s), and SMT pilot runs will always be needed.
Does any new design use DIP?
Not that I have seen, where a wave solder-able alternative exists.
Brian Fairchild wrote:Does any new design use DIP?
No I didn't
Why would xmega be too expensive for a low-volume product? A low-volume product will be expensive...
I said 'low-volume consumer product, which will be a higher volume compared to low-volume professional products. So I class things like air-conditioners as low-volume because compared to things like mobile phones, or laptops, they are.
No I didn't
Oops - now fixed.
More or less useless than an 8pin DIP packaged AVR Tiny?
More "useless" when 2 pins are taken out for SWD instead of 1 pin for DW.
jmaja1 wrote:Why would xmega be too expensive for a low-volume product? A low-volume product will be expensive...
I said 'low-volume consumer product, which will be a higher volume compared to low-volume professional products. So I class things like air-conditioners as low-volume because compared to things like mobile phones, or laptops, they are.
So an air condiotioner costs several hundreds, it's big and has quite expensive mechanical components. Would spending one dollar for MCU really be too much? I can understan that for many cheap things like remote controllers, bicycle computers etc. I would guess finding DIP parts from an air condiotioner just means it is an old desing they didn't want to update.
I would guess finding DIP parts from an air condiotioner just means it is an old desing they didn't want to update.
Possibly that, or, it could mean it is wave soldered, with many other thru hole parts, and that suits their production flow.
White Goods still use thru-hole, but they use SMD wherever they can, to keep size and costs down.
Interesting discussion.
I have a new clothes washer. It costs several hundred dollars!
It has lots of LEDs, and dials, and switches.
It looks like the cockpit for an airplane.
And the primary display for the user is still not directly human readable!
Just a little while ago I walked past the machine and it looked like a "u" but the vertical parts were only 1/2 of a segment high, (i.e. 1/4 of the height of the side of an "8"), and it had an "L" next to it.
Perhaps it means "UnLoad", i.e. its done washing?
I don't know.
I just think its ridiculous to spend hundreds of dollars and have a worthless display for a consumer item.
A simple LED flashing next a label that says DONE would be easier to understand, and cost less, and be more meaningful.
Either add the "extra" segments to the display, or remove the display.
Argh...
Where is that Thread on frustration, maybe I should go co-post!
JC
Although this is all rampant speculation, I personally would/will not mourn the passing of the XMeagre(tm). I have used the Mega series, and the XMeagre chips. For me, the XMeagre is a falling between two stools. It is slightly better than the Mega series(although the counter timer functionallity is crippled), but it's hard to understand why anyone wouldn't jump straight into the arms of an ARM. Having said that, you have my sympathy if you are currently using or specifying an XMeagre, and you are worried about continuity.
All a matter of perspective, I guess.
The vast majority of my work is hobbyist projects, and even the occasional for-hire project is usually one-off, or just several of a board.
But, the XmegaE series is my goto chip these days, instead of a Mega.
Oh well, time will tell.
JC
Why ARM-based MCUs and XMEGA series are not available in DIP package?
The XMEGA A4U default I/O rise/fall-time is 4ns and the AVR32 UC3 reduced that (to about 2ns); that is significant and an EMI/EMC concern.
A follow-on to XMEGA is SAM L, but SAM L does not have the drive that an XMEGA does (iow SAM L has an increased rise/fall-time versus XMEGA).
The customer likes your proof-of-concept or first prototype but then says it's too large.
I have a 32A4U module from Batsocks.Looking on Ebay, there are similar DIL modules from Barion and MattairTech.
Nothing like the sheer range of Mega and ARM dev boards that are available.
Shop Cheap AVR Development Kit from China AVR Development Kit Suppliers at Hangzhou ARMe Electronic Co., Ltd. on Aliexpress.com
AVR Development Kit
http://www.aliexpress.com/store/group/AVR-Development-Kit/800571_209680125.html
(I don't know why everything I open, it has an NEC MCU. it looks like this company has been successful to pre-empt the market of MCUs)
Renesas Fact Sheet
http://am.renesas.com/media/comp/RenesasFactSheet.pdf
...
... as part of an agreement to integrate the business operations of NEC Electronics Corporation and Renesas Technology Corp,
...
So an air condiotioner costs several hundreds, it's big and has quite expensive mechanical components. Would spending one dollar for MCU really be too much?
What on earth are you talking about? It doesn't matter who you are and what you are making a dollar is a dollar. It's either yours or you give it away. You don't get rich giving them away.
I've worked for a company where a typical production run was 1..5 million units. Even our chairman used to have arguments with us about whether we really needed some $0.10 capacitor or would a $0.05 one be OK. (5 million times $0.05 is $250,000). For the same reason the front panel PCBs of our products were single sided PCB and used DIP packaged components to reduce the cost per square inch by a few cents per unit.
+1
I have been in a situation where it was OK to spend 1-2 weeks on a software solution that would save about 0.02$ in components. (and it ended up with some inline 8051 ASM, this was about 20 years ago).
jmaja1 wrote:So an air condiotioner costs several hundreds, it's big and has quite expensive mechanical components. Would spending one dollar for MCU really be too much?What on earth are you talking about? It doesn't matter who you are and what you are making a dollar is a dollar. It's either yours or you give it away. You don't get rich giving them away.
I've worked for a company where a typical production run was 1..5 million units. Even our chairman used to have arguments with us about whether we really needed some $0.10 capacitor or would a $0.05 one be OK. (5 million times $0.05 is $250,000). For the same reason the front panel PCBs of our products were single sided PCB and used DIP packaged components to reduce the cost per square inch by a few cents per unit.
xmega costs less than a dollar for that big series. How much is a cheaper MCU? Does it use the same external components? Couldn't xmega offer something that is worth the extra cost like USB or maybe less man hours for design. I'm not saying that they don't care about a dollar cost in an more expensive product, I'm just saying it can well be used, if it has something more to offer. If you are targetting at well below 10 dollar total production cost for a unit, you simply can't use any 1 dollar components.
Is it cheaper to produce single sided DIP than single sided SMD?
Is it cheaper to produce single sided DIP than single sided SMD?
Depends if all components are through hole, or smd or mixed and where you have the assembly done, components sourced and more importantly the quantities involved. It is not any easy question to answer without a lot of details.
And often for high volume products size and shape matter. (sometimes shape matter the most "so we can hide it in this hole")
And often for high volume products size and shape matter. (sometimes shape matter the most "so we can hide it in this hole")
This part of the discussion started from low volume consumer products. I don't know if 1-5 million can be seen as a low volume. Are air condiotioners typically made in that range?
So what is considered high volume? Mobile phones are made in the order of 10-100 million in the lifespan of a model. So 10 million must be high volume? What about 1 million?
When size matters, DIP is out of the question.
That is why I say that shape matter, for an air conditioner you need push buttons and a display (at least 7 segments) and for that there are needed a big PCB for design reasons, so you just put the cheapest components on.
so here size matter, DIP are fine :)
jmaja1 wrote:
So an air condiotioner costs several hundreds, it's big and has quite expensive mechanical components. Would spending one dollar for MCU really be too much?
Adding $1 to the manufacturing cost of a consumer item can add anywhere between $10 and $50 to the consumer price tag. In a market where there are multiple competing products vying for a piece of the consumer pie, a $10 difference on a ~$400 air-conditioner can make the difference between selling 1000 units per season and selling 50,000 units per season, or far more. In extreme cases, it can make the difference between selling enough to make a profit, or not selling enough to recoup your investment. Failed product. End of life.
There are many factors which come into play to determine what will be a profitable consumer product. Almost all of them have to do with the consumer. Almost none of them have to do with satisfying an engineer's sense of balance or beauty. Save a dollar on the mcu, that dollar can be used to position the product at a more attractive price point, or it can be spent on a sleeker-looking front panel, or a better-quality paint finish, to attract a different 'class' of consumer. It's all a bean-counting balancing act which, frankly, I thank my cotton socks every day I have no involvement in.
Adding $1 to the manufacturing cost of a consumer item can add anywhere between $10 and $50 to the consumer price tag.
I don't think that applies to things like air conditioners. More like $2 or $3 to the consumer price tag, which is nothing, if it adds a feature that helps selling it.
Actually why would any product consumer price tag need to added 10-50 times the added manufacturing cost? I think Nokia used to be able to make mobile phones at $10 cost, but they were sold quite cheaply as well, maybe $100-200. Iphone 6S plus seems to cost $236 to make http://www.cnbc.com/2015/09/30/apple-iphone-6s-plus-costs-236-to-make-sells-for-749.html and sells at $750. Nokia used to make much better percentual margin, but in dollars margin is much better for the high price IPhones. In this case adding $226 to manufacturing cost added consumer price by $600.
I don't think that applies to things like air conditioners. More like $2 or $3 to the consumer price tag, which is nothing, if it adds a feature that helps selling it.
You appear to live in a world of rose colored spectacles! I've worked in consumer electronics for decades. We've always been keen to hit key price points like £49, £99, £199, £249 and so on. That gives you your target for the retail price to the consumer. Those prices including UK sales tax (currently 20%) so for a £99 item (say) you have £80 to the retailer. He typically wants to take about 30% profit so we need to sell to him at £80-30% (£24) so we sell to him at £56. We might want to make about 20% profit ourselves so say we take £11. That means we need the item "landed" at £45. There's likely something like 4% duty on overseas manufactured goods and there are shipping cost which might account for about 10% of the £45. So that means the item has to be Full On Board from the factory at £41.50. The factory who make it for us need to take their own LOP (Labour, Overhead and Profit) and let's call that 15% So they take £6.25 off the £41.50 which means the BOM cost is £35.275 or about $50. That is for an item that retails at £100 in the UK. Now I haven't included things like royalties or marketing costs (though they may be part of the £11 we got out of this.
Anyway there are $50 to make the thing. We then have to weigh up whether we can economically do that. If that means using a single sided PCB and DIP components to save $0.80 or using five $0.05 capacitors instead of five $0.10 ones to hit the $50 BOM cost we will do that.
Any $ we can get below the BOM knocks through as straight profit.
We apply this technique for items from $50 retail to $500+ retail.
At one stage our company was worth more than $1.5bn. The above is why. We counted every last cent we spent.
david.prentice wrote:Another source is Mcuzone on Aliexpress; almost totally XMEGA.I have a 32A4U module from Batsocks.Looking on Ebay, there are similar DIL modules from Barion and MattairTech.
Nothing like the sheer range of Mega and ARM dev boards that are available.
Shop Cheap AVR Development Kit from China AVR Development Kit Suppliers at Hangzhou ARMe Electronic Co., Ltd. on Aliexpress.com
AVR Development Kit
http://www.aliexpress.com/store/group/AVR-Development-Kit/800571_209680125.html
I was not aware of the McuZone modules. The $12 price looks ok until they want $23 for postage.
Somehow, a module from Poland, US or UK looks a lot more attractive.
David.
I haven't owned an air conditioner, but the ones I know cost from $500 to $2000, not £99. And we seem to agree that $1 added to production costs adds $2 to $3 to consumer price, not $10 to $50.
What I have looked at recently are washing machines. Simple feature like timer or a few extra programs may add $100 or even $200 to the consumer price. If you need a different MCU to nicely add such a feature, wouldn't it be good idea to pay $1 more?
Apple is also a good example of getting much more profit, with a cheap add like more memory or mobile data to IPad.
You really need to come down from cloud cuckoo land and join the real world. There is one thing that drives a capitalist economy and that is the pursuit of profit. No manufacturer (and that include the likes of Sony and even Apple) pay any more for any component in any product than they have to.
Apple have sold something like 100m iPhones. Suppose they could save just $0.01 per unit that would have been $1m!! If they paid $1 more per unit than they really needed to that would have cost $100,000,000 - that is 1/10th of 1 billion dollars!!
I've worked in consumer electronics for decades. We've always been keen to hit key price points like £49, £99, £199, £249 and so on. That gives you your target for the retail price to the consumer.
Do you know what lead pitch can currently be Wave soldered ?
I've seen QFP packages spun 45' on consumer products to wave solder, with longer pads.
You really need to come down from cloud cuckoo land and join the real world. There is one thing that drives a capitalist economy and that is the pursuit of profit. No manufacturer (and that include the likes of Sony and even Apple) pay any more for any component in any product than they have to.
Of course they don't pay more than they have to, but they don't choose the cheapest components. Certainly Apple could cut down the production cost from that $236 to say $100, but then it would be a much less desirable product that doesn't sell at the same margin. They don't use a cheap display nor a cheap processor. You can't just calculate that they would now make $136 * 100M more profit, since they wouldn't get the same price or sell the same numbers. It's a different thing, if you can save without any difference to the product (that a customer would notice).
An air conditioner or a waching machine is of course a totally different product, but they are getting more "inteligent" and thus may need a better MCU than the absolutely cheapest one. I really can't see xmega as a terrible expensive one that could not be used in them.
When I lived in US one summer I had to buy a air conditioner (about 8 years ago), and they had one for $69 and a quick test Wallmart has one for $112.
You don't know perhaps you get the extra feature by changing a switch on the mother board. (They make you get into the store with the low price but happy leave the store paying $100 for some added features), but other times there actually are some of the features that cost more to produce.
30µA/MHz. Holy smokes. ......
Integrated buck converter. ...
SAM L MCUs
http://www.atmel.com/products/microcontrollers/arm/sam-l.aspx
...
The Atmel® | SMART SAM L family of microcontrollers (MCUs) delivers power consumption down to 35 µA/MHz in active mode and 200nA in Sleep mode.
...
Atmel SAM L is close and has a buck converter but its BACKUP mode with RTC is about one microamp (14uA max at 85C); ATxmega256A3U power-save with RTC is one microamp typical to 3uA max. (85C or 105C).
I'd like to see more XMEGAs with battery backed RTCs. The current selection is quite small, but it's an important feature.
Someone is bound to say that there is the 16 bit RTC peripheral, so allow me to explain why a more independent one is desirable. The problem with the XMEGA 16 bit RTC is that you the MCU resets it normally loses its count. You can kinda get around it by marking the in-memory variables as "no init", but basically a reset is going to screw your RTC up to a lesser to greater extent.
Lead me to Ambiqu:
I'd like to see more XMEGAs with battery backed RTCs. The current selection is quite small, but it's an important feature.
Energy harvesting demands full-spectrum microcontroller efficiency
Mike Salas (Vice President, Marketing and Strategy, Ambiq Micro)
February 05, 2016
...
Development over many years at the University of Michigan and Ambiq Micro has led to many sub-threshold innovations ...
...
There are ultra-low energy real-time clock designs that can check for external issues, such as issues raised by hardware interrupts or changes in input voltage sensed by a comparator.
...
Ambiq
I know you love these guys, but their website doesn't offer up datasheets for their MCUs without registering. All the claims seem to be speculative rather than based on real world measurements.
Their RTCs look okay but not that exciting. Another item for my BOM, a single interrupt input line... It's not a great substitute for a MCU's RTC, and I can't seem to buy them around here anyway.
mojo-chan wrote:
but their website doesn't offer up datasheets for their MCUs without registering.
Ahem:
joeymorin wrote:
EDIT: Ah. Datasheet:
http://ambiqmicro.com/sites/default/files/apollo-product-sheet-lowest-power-mcu.pdf
FYI I wasn't required to (nor did I in fact) register.
mojo-chan wrote:
All the claims seem to be speculative rather than based on real world measurements.
gchapman wrote:
There is an evaluation board.
Is it a real part? You betcha. I have an eval board in my lab. More details will follow when I have a chance to play with it.
mojo-chan wrote:
Their RTCs look okay but not that exciting. Another item for my BOM, a single interrupt input line... It's not a great substitute for a MCU's RTC, and I can't seem to buy them around here anyway.
gchapman wrote:
Ambiq Micro sells the die to Abracon for packaging and test
DigiKey shows thousands in stock:
Where are you that you can't get parts from Digi-Key, Mouser, Newark, Farnell, etc...?
EDIT: Ah. Datasheet:
http://ambiqmicro.com/sites/default/files/apollo-product-sheet-lowest-power-mcu.pdf
FYI I wasn't required to (nor did I in fact) register.
That's not a datasheet, it's a single page summary.
The fact that this micro isn't available anywhere and that there is no final public datasheet rather suggests that it's pre-production, which means all bets on performance are off.
gchapman wrote:
Ambiq Micro sells the die to Abracon for packaging and test
The datasheet lists the part numbers as AM1805 and AM1815. Farnell UK and Mouser UK don't list them under those part numbers. How am I supposed to know that I need to search for Abracon parts? I found it on the Ambiq web site, I tried searching suppliers for Ambiq and came up with nothing.
These guys are clowns.
mojo-chan wrote:
That's not a datasheet, it's a single page summary.
That's my error. Full datasheet (368 pages):
http://ambiqmicro.com/system/files/Apollo_MCU_Data_SheetDS0010V0p45.pdf
mojo-chan wrote:
The fact that this micro isn't available anywhere and that there is no final public datasheet rather suggests that it's pre-production
Looks like for < $4 you can have one of your own. Similar quantities for the WLCSP package type. Seems they're behind schedule on the 256KB, 128KB, and 64KB versions (meant to be available 4Q15, but parts searches turn up nothing).
Future Electronics also shows up in the search results for that part number from ECIA and Octopart.
Future Electronics also carries the rest of Ambiq's silicon, and a couple of eval kits for the micros:
http://www.futureelectronics.com/en/Search.aspx?dsNav=Ny:True,Nea:True,N:4294852175
mojo-chan wrote:
How am I supposed to know that I need to search for Abracon parts?
By reading :)
Our Sales Representatives and Distributors
Forte Technical Sales
2033 Gateway Place
5th Floor
San Jose, CA 95110Phone: 408.573.6131
Contact form
Future Electronics
237 Hymus Blvd.
Pointe Claire, QC H9R 5C7
Tel. (514) 694-7710
Abracon Corporation
Rancho Santa Margarita, California
+1 (949) 546-8000Abracon branded products are in stock and immediately available from:
More Distributors under: Europe, China, Japan.
mojo-chan wrote:
These guys are clowns.
I see no reason to bring the performance arts into this discussion. Besides, some of my best friends are clowns.
Again, the datasheet is marked "preliminary" and listing Abracon as a distributor doesn't tell me I need to search for a different Abracon part number.
mojo-chan wrote:
listing Abracon as a distributor doesn't tell me I need to search for a different Abracon part number.
Why would you expect an Ambiq part no. on an Abracon branded part?
Abracon aren't the only distribution channel. While their Abracon-sourced parts are rebranded, other channels carry the parts with their original Ambiq part no:
joeymorin wrote:
Future Electronics also carries the rest of Ambiq's silicon, and a couple of eval kits for the micros:
http://www.futureelectronics.com/en/Search.aspx?dsNav=Ny:True,Nea:True,N:4294852175
Again, the datasheet is marked "preliminary" ...
'Wow, really ? Some datasheets are tagged "preliminary" for years...
OK guys... I think "both sides" have had their say... time to return to arguing against any common enemy instead of each other. Besides Valentine's Day is just around the corner.
The datasheet lists the part numbers as AM1805 and AM1815. Farnell UK and Mouser UK don't list them under those part numbers. How am I supposed to know that I need to search for Abracon parts?
http://dilp.netcomponents.com/cgi-bin/abracon.asp?partnumber1=AB1805-T3
We need a ~16 bit ADC to get the dynamic range we require.
http://www.microchip.com/wwwproducts/en/PIC18F87J72
http://www.microchip.com/wwwproducts/en/PIC24FJ128GC010
mojo-chan wrote:There are a relative few PIC18 and PIC24 with a 16-bit delta-sigma ADC.We need a ~16 bit ADC to get the dynamic range we require.http://www.microchip.com/wwwproducts/en/PIC18F87J72
http://www.microchip.com/wwwproducts/en/PIC24FJ128GC010
We did look at those. The 18F has an old, terrible architecture that would make our application unnecessarily complex, doesn't have enough RAM and is rather expensive. The 24F can only do 1ksps on the 16 bit channels, and we need at least 5ksps. Believe me, we looked long and hard before selecting XMEGA.
In general you can't put make ADC with real 16bit, on a micro, if that is needed then get an external ADC.
Why is that?
You do other things on a micro at the same time, if you can sleep it could work, but in general you could have IO's go high and low around the time of sampling, and that could easy make 1mV noise internal in the chip. (just a simple thing like receiving on the UART could come at anytime)
12 bit resolution with a 2.048V reference is 500uV/bit already. So yeah, 16 bit is quite a challenge. Fortunately we have quite limited bandwidth.
Analog Devices does restrict the clock speed of their ARM Cortex-M3 that have the high resolution ADC.
Constraining the Analog Devices product selector for ARM Cortex-M3 to greater than 5ksps for 16b and 20b resulted in 3 possibles.
One that restricts the clock to 16MHz is the ADUCM350 with a 16bit SAR (?) ADC; the other two are higher speed core but with 20b delta-sigma ADC (an inherently noise tolerant ADC).
Analog Devices has had high resolution ADC on an 8051 core for quite sometime but these are relatively "slow" ADC; an example is ADuC848 1.
Analog Devices
ADUCM350 Datasheet and Product Info
Analog Devices
ADUCM331 Datasheet and Product Info
1 The Art of Electronics, third edition, Paul Horowitz, Winfield Hill, 15.6 Design example 4: thermal controller (RTD in bridge; then, differential amp, PGA, chopper 16b ADC, all in an ADuc848), pages 1069-1077.
Analog Devices
ADUC848 Datasheet and Product Info
http://www.analog.com/en/products/processors-dsp/microcontrollers/8052-core-products/aduc848.html
Edit : last URL
... and we need at least 5ksps.
A 14-bit 900Ksps SAR ADC on an improved 8051 :
Silicon Labs
EFM8 Laser Bee 8-bit Microcontrollers (MCUs)
http://www.silabs.com/products/mcu/8-bit/efm8-laser-bee/Pages/efm8-laser-bee.aspx
One of the packages is SOP (pitch 0.635mm, 25mil); that's handy.
Just to clarify, I'm not looking for alternatives. There are other options, but we selected XMEGA because it's got the right features (not just ADC) and at the right price. The IDE is excellent, light years ahead of most of the competition (MPLAB X is crap, our PIC guy has an elaborate system to compile the code in Visual Studio 2008 because the code browsing is so much better) and the architecture is one of the best I've ever seen for 8 bit. We are now making thousands of these things every month and are very happy with the MCU, and have used other XMEGA parts in other products.
As such, I'd like to see Atmel continue to develop the line.
Has anyone got an update on this thread? Still hoping for an ATXMEGA with OTG USB!
It's pretty clear from recent releases that Atmel are moving in a more "downgraded" than upgraded direction for Xmega. They are pouring their effort into SAM chips for the more advanced end of things.
Something very bad is happening at Microchip. Prices on a lot of stuff doubled since the take-over, and it seems like they going to ditch a lot of product lines. Their main interest in Atmel seems to have been ARM/wireless stuff, and the new ASF 4 only has very basic support for XMEGA.
Follow the money...
At least they are upgrading the Tiny to be more like the xmega.
Maybe there will be some new mega upgrades in a similar fashion.
As Cliff suggests, there seems little likelihood in upgrading either the Mega or XMega - they have SAM for that.
Hence effort will be expended at extending the low end downwards - where SAM does not fit.
Has anyone got an update on this thread? Still hoping for an ATXMEGA with OTG USB!
That may be optimistic.
A good quick reality check update for USB trends, is to search Digikey for Microcontroller, then sub-search USB.
Cheapest is a recent 8 bit SiLabs part, followed by new STM32F070F6P6, & new ATSAMD11C14A, then PIC16F1454, and rather down the list PIC18F13K50
If you then switch search to USB OTG, the cheapest is RX100 from Renesas, followed by PIC32MX210F016, but Digikey's OTG filter is not great.
Checking on Atmel's filter, the SAMD21 looks to have Device, Host support.
The other fine-print gotchas around USB, are things like no 5V regulator, 5V tolerant, & any Crystal dictate
eg STM32F070 is 5V tolerant, but says
It requires a precise 48 MHz clock which can be generated from the internal main PLL (the clock source must use an HSE crystal oscillator).
SAMD21 is not 5V tolerant, but says "If crystal-less operation is used in USB device mode, refer to USB Clock Recovery Module."
If you want all of 5V regulator, 5V tolerant / 5V IO and No Crystal, in 32b USB device, then maybe a NUC125 is closest.
As long as I can buy reasonably priced Xmegas32E5 for the next 50 years I'll be happy...in more ways that one...
So if new mid range 8 bit PICs are in development, why not new mid range 8 bit AVRs?
Google--> lucio di jasio Microchip avr
Google--> lucio di jasio Microchip avr
And, surely, it was the new Tinies that he was referring to?
Again suggesting that the focus for 8-bit is low end - leaving the high end for SAM.
Which certainly makes sense to me!
That blog post was from a year ago, and all that has appeared are the new xTiny parts... Which are very nice, can't wait to use them, but I still love XMEGA.
I'm going to have to make a real effort to get going with SAM. I just hate having to use libraries for everything, they are never as flexible or debuggable as just hitting the registers directly. I don't know why ARM has to be such a pain in the arse in that respect... Other 32 bit parts aren't that bad.
Bring back 68000 I say
"And he says interest in the 8bit MCU remains strong. “Volume and revenue from the 8bit MCU sector continue to increase,” he said, “and we’re talking about billions. If you, as a manufacturer, can show you’re continuing to invest in a sector, then you will get a return on that investment – and Microchip is investing more today in the AVR portfolio than Atmel did over the last three years,” he concluded."
I was reading somewhere and can't find it right now, that 32bit was on the increase in sales but seems to now be slowing compared to 8bit.
Who knows what what will really happen in the next few years?
I just hate having to use libraries for everything
You don't have to use libraries for anything!
The registers are still there - the processor neither knows nor cares how you get to them.
You don't have to use libraries for anything!
True, but here is a "minimal" example for ARM: http://pandafruits.com/stm32_pri...
That blog post was from a year ago, and all that has appeared are the new xTiny parts... Which are very nice, can't wait to use them, but I still love XMEGA.
Right, that shows how long it takes to bring them to market, some of those xTiny parts are still impossible to find.
Who knows what's in the works right now?
I was reading somewhere and can't find it right now, that 32bit was on the increase in sales but seems to now be slowing compared to 8bit.
I would imagine that's because "32-bit" (specifically, ARM) is probably reaching (or has already reached?) saturation by now?
And there's probably some who were simply following the fad but have now realised that there are still places where 8-bit fits better - so there's a bit of adjustment going on there. We saw that from NXP recently ...
I was reading somewhere and can't find it right now, that 32bit was on the increase in sales but seems to now be slowing compared to 8bit.
At the low end 8 bit parts are cheap enough now that you can just throw them into a design where previously you would have had some glue logic or a special purpose IC. I think there are more people who can do both electronic design and 8 bit code as well now. Being able to buy pre-programmed parts cheaply helps as well.
The proliferation of battery powered devices is also creating demand for 8 bit. IoT and wearables.
Right, you may not really need an ARM in a toaster oven unless it's an IOT or has a video camera and sound.