## mega128L clock speed versus vcc

22 posts / 0 new
Author
Message

I notice the newer AVRs have clock speed versus vcc curves allowing them to be run at faster speeds for higher supply voltages. Does anyone know whether the mega128L can be run at higher than 8MHz if above 2.7V? It makes sense that there would be some kind of derating curve for this.

Mark.

Heh.Well, it's only meant for the internal RC oscillator. The frequency of a external crystal oscillator or clock source will not be modified in any way. Although you can't run at 16MHz when @ 2.7V.

There are pointy haired bald people.
Time flies when you have a bad prescaler selected.

I meant how fast can you clock it at, say 3.3V? The mega1281 can be run at around 10MHz or so according to the datasheet. Does the mega128L have this behaviour too but just not formally documented?

Curiously Figure 161 in the mega128L full datasheet shows active current at various voltages and speed, the graph seems to show some kind of relationship between voltage and speed..

On the first page of the Datasheet, it states 2.7V-5.5V for ATmega128L and 4.5V-5.5V for ATmega128. It also states 0-8MHz for ATmega128L and 0-16MHz for ATmega128. Doesn't seem to be any clearer to me! Also note, the only graphs that show voltage vs. speed are for the internal oscillator which stops at 8MHz.

edited spelling

Well if you already know about 161 then you already knwo the answer. From that you can deduce that at 2.7V you can run up to 8MHz, at 3.3V up to 10MHz, at 4V up to 12MHz and at 4.5V or above up to 20MHz. However as page 1 of the datasheet states the 128 has a maximum speed 16MHz (at up to 5.5V) and the 128L has a maximum of 8MHz

So you can't run a 128L in spec beyond 8MHz.

Cliff

rstahlhu wrote:
On the first page of the Datasheet, it states 2.7V-5.5V for ATmega128L and 4.5V-5.5V for ATmega128. It also states 0-8MHz for ATmega128L and 0-16MHz for ATmega128. Doesn't seem to be any clearer to me! Also note, the only graphs that show voltage vs. speed are for the internal oscillator which stops at 8MHz.

edited spelling

Well '0-8MHz for a mega128L' assumes it can be powered anywhere between 2.7V and 5.5V. My question is whether this is like the mega1281, so if you're running at 5.5V what exactly is the maximum speed? The spec says 8MHz I agree, but is that actually a real maximum? Common sense says there's a relationship here even if not documented or guaranteed.

clawson wrote:
Well if you already know about 161 then you already knwo the answer. From that you can deduce that at 2.7V you can run up to 8MHz, at 3.3V up to 10MHz, at 4V up to 12MHz and at 4.5V or above up to 20MHz. However as page 1 of the datasheet states the 128 has a maximum speed 16MHz (at up to 5.5V) and the 128L has a maximum of 8MHz

So you can't run a 128L in spec beyond 8MHz.

Cliff

Mmmm. What do you mean "3.3V up to 10MHz"? Which processor can run at 10MHz? The mega128 is only qualified at between 4.5V and 5.5V, so that only leaves the mega128L, which you say can't run above 8MHz. I think this is my point, that graph shows a relationship that's not documented at all, but *is* documented for the mega1281.

You are getting confused by the newer chips.

There is no clock maximum speed vs Vcc similarity between the Atmega128L and ATmega1281 chips. The ATmega1281 is a newer processor that works differently. The reason the Atmega128L has no graph similar to the ATmega1281 clock speed vs Vcc graph, is because the Atmega128L does not work that way. The Atmega128L is in fact documented and needs no graph because it is specified from 0 to 8 MHz at 2.7 to 5.5 volts Vcc, end of story.

The newer processors have more complex clock speed vs Vcc relationships which the graphs explain much better than 1000 words can explain it. My guess is that is why the graphs appear in newer data sheets.

In the ATmega128 data sheet, figure 161 is simply a current consumption projection for a given Vcc at a given clock speed. If you take this chart literally then the ATmega128 must be a 20 MHz part above 4.5 volts Vcc. In reality it is not specified above 16 MHz for any Vcc. Figure 161 is a projection of a “what if” condition, not a specification stating a real AVR chip can actually go that fast. In fact people do push their AVR chips beyond the specified maximum. This is acceptable for one of or hobby circuits, but should never be done for production units sold to customers or anything you expect support on from ATMEL (it would also be nice to let the people here on AVRfreaks know if you are asking for help on an over clocked processor).

Mike B wrote:

The ATmega1281 is a newer processor that works differently.

Assuming it's still a CMOS process (fairly safe bet), a characteristic of CMOS is that edge rates speed up with lowering temperature and increasing voltage. I don't accept that it works differently - the feature dimensions might be smaller yielding lower power and faster processors but they are no different to other CMOS chips.

Mike B wrote:

The newer processors have more complex clock speed vs Vcc relationships which the graphs explain much better than 1000 words can explain it. My guess is that is why the graphs appear in newer data sheets.

Again I don't accept that. Atmel probably didn't bother to qualify the older chips to those specifications, that doesn't mean to say they don't behave like newer chips.

Mike B wrote:

In the ATmega128 data sheet, figure 161 is simply a current consumption projection for a given Vcc at a given clock speed. If you take this chart literally then the ATmega128 must be a 20 MHz part above 4.5 volts Vcc. In reality it is not specified above 16 MHz for any Vcc. Figure 161 is a projection of a “what if” condition, not a specification stating a real AVR chip can actually go that fast.

I agree, however it's curious the 'projections' stop at different frequencies depending on voltage. Why not draw each one to 20MHz? Could it be that there really is a voltage dependence on max frequency? I think so, it just hasn't been qualified or documented.

Well, you don't know how it will behave if you run these AVRs out of what datasheets say.

It might work, then again, it may just do something really weird which you spend a month debugging, just to find out that one of the several AVRs you have was not able to go that fast with that supply voltage, because it was from a different batch than the rest of the AVRs.

Okay, if this is for your own hobby, fine, go ahead and experiment. If this is for a product which you or your customer are planning to sell by masses, well, I'd just hate it if I got a malfunctioning product because someone didn't bother to believe the specs.

So what are you trying to achieve by overclocking the Mega128L?

Only real difference between Mega128 and the L version is that the Mega128 has only 4 volt brown out detector, while L version has both 4 volt and 2.6 volt detectors.

I think they just test if the chip works down to the lower BOD reset level without going berserk, and some chips just stop operating properly before they get reset by the lower BOD level. So former devices would become L versions and latter would become non-L versions. This lets them specify that L:s work with bigger voltage range, and can be used with low voltage, but might not work with full speed at the lowest voltages. The non-L:s will have at least 4.5 volts, so they know it will work with higher frequency.

- Jani

EDIT: And yes, it is normal that greater frequencies need greater supply voltage. So that is why figure 161 looks like what it looks like. They can only say that with this voltage X, the maximum safe frequency for you to use is this Y.

Last Edited: Fri. Mar 23, 2007 - 11:28 PM

Its my understanding that the 128 and 128L are the same chip, but that the L is specifically tested for low current at low voltage operation.

It is reasonable to expect that supply voltages above 2.7 would allow the 128L clock to run with the same limits as a 128. However, Atmel appears not to include that in the specs, so there is no certainty (as if there ever is).

If this is for a commercial product, I wouldn't; for a hobby project, go for it.

Jim

Until Black Lives Matter, we do not have "All Lives Matter"!

ka7ehk wrote:
Its my understanding that the 128 and 128L are the same chip, but that the L is specifically tested for low current at low voltage operation.

It is reasonable to expect that supply voltages above 2.7 would allow the 128L clock to run with the same limits as a 128. However, Atmel appears not to include that in the specs, so there is no certainty (as if there ever is).

If this is for a commercial product, I wouldn't; for a hobby project, go for it.

Jim

I was curious as to whether anyone knew any differently. This is used in a demonstrator for a set of commercial products to come, I certainly won't be using it out of spec. I'm going to swap for a mega1281 instead, it's pin compatible and has a specified safe region.

mapelec wrote:
Assuming it's still a CMOS process (fairly safe bet), a characteristic of CMOS is that edge rates speed up with lowering temperature and increasing voltage. I don't accept that it works differently - the feature dimensions might be smaller yielding lower power and faster processors but they are no different to other CMOS chips.
I was talking about the entire chip as a total unit working differently. Your reference to CMOS construction is only relevant to people involved in the actual engineering of the silicon dies in both chips. What are feature dimensions? I am aware of silicon process sizes (component density) and smaller dies, which are not solely responsible for faster processors, but do ultimately make it possible. According to you, because they are all CMOS none of them work differently, except that feature dimensions do change power consumption and speed. Changing power consumption and speed is not making it ”work” differently???? I thought clock speed was part of the core of this thread. In my world microprocessor chips are a complex interaction of parts and processes that are impacted by even small innocuous changes, which the people that actually design and build each chip are smart enough to figure out when they characterize the chip. Your assumption that all CMOS is the same is far too simplistic and has led you to make unsupportable assumptions about the characteristics of these two different chips. I think you have confused basic properties of CMOS with characteristics of more complex CMOS circuits.
mapelec wrote:
Again I don't accept that. Atmel probably didn't bother to qualify the older chips to those specifications, that doesn't mean to say they don't behave like newer chips.
It does not mean they do behave like newer chips either. Then why did ATMEL bother to characterize the new chips differently? Your assumption is characterizing a part is limited to a predetermined result that the characterization process is designed to meet then stop at. I though characterizing a part was discovering its limits and then determining the reliable/stable operational limits from complete test data. I think that gathering full test data is a good reason why the information in table 161 exceeds the published specification for clock speed. Just because the chip can over clock beyond specifications does not mean it can still do it under all conditions (i.e. maximum current loads on all the I/O pins, temperature ranges, etc.). The characterization outcome, i.e. specification, is not only about maximum clock speeds, it is about the complete range of combined reliable operating limits for the chip.

You contradict your own argument when you say you are switching to the ATmega1281. Since you know these "CMOS" chips are actually the same, especially the maximum clock speeds, you should use your far superior knowledge to run the ATmega128 under the ATmega1281 clock speed and voltage specifications. In fact, why not use the specifications from one of the 20 MHz ATmega data sheets and make your ATmega128 into a 20 MHz “CMOS” part, or do you know the 20 MHz parts have different feature dimensions? Do you believe the ATmega128L is really a 1.8 volt Vcc part and ATMEL never really had to make any changes at all to get these new ATmega1281V parts down to that voltage?

Since you believe these two parts are really the same and ATMEL just didn't bother to point it out, then they must have the same feature dimensions. Then explain why the same “CMOS” in the EEPROM results in the ATmega128 having typical write speeds of 8.5 ms, while the ATmega1281 typically writes EEPROM in 3.3 ms. If it is really so simple and only depends on it being CMOS, they must be the same, yet they are radically different. If you think a change in feature dimensions is required for higher speeds, then maybe these two chips really do not have identical feature dimensions.

Nice to see you have blind faith about the inner workings of of these chips (I assume you were not part of the ATMEL design teams). I just read the data sheets is all, and noticed “documented” differences which lead me to the obviously false :lol: conclusion these chips work differently. Sorry, I know I shouldn't confuse you with documentation because it appears you have no faith in documentation.

Mike B wrote:
mapelec wrote:
Assuming it's ....

Er, OK. I guess if I qualify with 'fairly safe bet' and state it's an 'assumption' it means I believe it's true without documentation to back it up. Oh well.

Edit: BTW at no point have I ever said I think the mega128 and mega1281 have the same characteristics. Throughout all this I've been trying to find out if anyone knew anything other than what's written in the datasheet for the mega128. To me it's a fair bet that they "behave" the same way, i.e. as the voltage goes up the maximum frequency you can run it also goes up, but I've never assumed actual values. The ATmega128 is *not* qualified for this speed increase, but the ATmega1281 is. No contradiction in that.

Sorry, I'm only arguing this case because you are leaving a record in the forum that kind of looked like ATMEL doesn't bother to tell you what the chip really does, CMOS being the same means the chips work the same, etc. As opinions go everyone will make up their own minds no matter what anyone argues, but I didn't want to leave a one sided account for everyone. I'm concerned about others thinking your views are definitive and using them as permission to get themselves into trouble.

Anyone that gets to the professional level of designing AVR based products for manufacturing production follows the specifications or risks the consequences. The biggest consequence is ATMEL has absolutely no obligation whatsoever to provide any support if you run into problems with AVRs being used outside the their published specifications. Another one is you really have no reason to “expect” the AVR to perform like you want it to. If not following the specifications causes problems, its all your fault (not a good thing for the boss to figure out :wink:, especially if it causes an expensive product recall or field repair). Also, manufacturing microprocessor chips is not a perfect process and the specifications take this into account. ATMEL shouldn't care if production variations occur within published specifications because the chips still work like they said. Anyone not following specifications may find their AVR circuits suddenly break when using chips with a particular date code or revision, while the rest of the world that follows the specifications have no problems with the same date code or revision chips. BTW, it is important to look at the errata along with the specifications.

For all other uses, intensionally exceeding the specifications is your own choice if you are willing to go it alone. However, people with more embedded experience have an easier time doing it because they have learned how to detect, isolate and fix problems with their AVR circuits and software. Someone without much embedded experience might be getting themselves into a nightmare if problems directly related to not following the specification pop up. With little or no experience, how do they even figure out if not following the specifications is contributing to their problems or not?

Personally, I look at the specifications like they are part of a product label. It tells me what I can expect. If that does not cover what I need, then its a wake up call that I'm looking at the wrong product.

Oh look, I know damn well if it's not in the datasheet you can't rely on it. I also know if you design something in that doesn't meet the spec you could be in deep trouble, one batch may work that way and others might not. Next week they could change the process or use another fab, and your undocumented 'feature' might just stop working. I shall *never* use undocumented features in a production board, and *never* use out of specification either because it just cannot be relied on. All I wanted to know is whether anyone knew whether these chips behave like the new ones or not, for my information only. That's it! And yes, Atmel may well not bother to document something that they could have documented, they might not have known for instance that this was a reliable characteristic at the time. It might well be. I think you've jumped to quite a few conclusions here...

Mapelec, you made some sweeping statements that MikeB jumped onto. Realistically, for questions like you have, refer to Atmel, anything else is just conjecture. I recently read a discussion on cpu speeds etc here and there was reference to the speed of the flash memory and how it should be similar to the Atmel Arm devices - obviously this comparison is simplistic as more than the flash speed alone can determine the maximum cpu speed.

Kartman wrote:
Mapelec, you made some sweeping statements that MikeB jumped onto. Realistically, for questions like you have, refer to Atmel, anything else is just conjecture. I recently read a discussion on cpu speeds etc here and there was reference to the speed of the flash memory and how it should be similar to the Atmel Arm devices - obviously this comparison is simplistic as more than the flash speed alone can determine the maximum cpu speed.

I know it's conjecture!! I've never made any 'sweeping statements', MikeB jumped to the wrong conclusions. I've been talking about assumptions, and fair bets, and at no point did I *ever* state that the mega1281 and mega128 are the same chip. (God, why don't you lot read the posts properly?)

I have the same problem but i seem to be more confused after reading these posts:? .

In the datasheet it is mentioned that Atmega128L can work from 0-16MHz ,Atmega128 bw 0-8 MHz and the respective working vcc ranges are 2.7 to and 5.5 and 4.5 to 5.5,so does it mean that i can run the uC at any frequency the given range at any given voltage in the given range...If yes is it safe and reliable and if no how to determine the max safest value of the crystal at a given voltage (say 3.4V for a mega 128L or mega128)?

There are two variables... speed and voltage.... the map of 'safe operating area' might not extand into all four corners of this graph... we know it will go lo v and slow and hi v and fast. What about the other two combinations? Airplanes will fly low and slow and hi and fast, but get into trouble in the other extremes of the flight envelope.

Imagecraft compiler user