In-vehicle Datalogger part 8

Go To Last Post
70 posts / 0 new

Pages

Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Well, it's been a while since an update :) This is part8 of the In-Vehicle Datalogger project that's been going for some time. So, first a brief recap ...

This originally started as a project to develop a cheap datalogger for automotive use. Myself and a bunch of friends all drive rather modified Supras, pushing some obscene horsepower numbers. So engine and system tuning is pretty important. And that's the problem - you can't really know what's going on inside the engine, even with the addon boost/fuel computers etc.

So after some discussion, I decided to have a stab at building a datalogger. It went through several revisions, from an 8515 with outboard A/D, to an 8535 with onboard A/D, to the current Mega128. Heh, note, I haven't actually *built* one yet.

Basic specs are as follows :

- 16 channels analog input, muxed down to the 8 channels in the Mega128. Fully buffered inputs (high source imp, low A/D imp) with 3 jumper-selectable voltage ranges. 0-5V, 0-14V and 9-14V. Those values were chosen from the very spare specifications in the Toyota service manuals. I had wanted to include CalMicro ESD transient pretoection diode arrays on the inputs, but there just isn't any more dang room on the pcb. I'll still try, but I doubt it.

- 6 channels of opto-isolated pulse inputs. Current limited with 680 ohm resistors.

- 3 onboard remote temp sensors.

- USB and standard serial interfaces. Serial is just a 3-pin header, so will need an adapter cable. Don't know yet if this can all be powered from the USB bus - would be nice :)

- MMC card for data storage. Easy to pop a new one in for unlimited storage.

- Battery backed real-time clock for date stamping of data runs.

- Remote LCD screen interface for in car. Initially will be a simpler 4x40 lcd, but I'm planning on a 256x128 graphics screen later. Maybe.

- Onboard ADXL202 G-sensor. This means that you would need some way of *really* securely mounting the box, and as accurately perpendicular/parallel to the vehicle axis as possible. It is possible to compensate for tilt, but icky.

- Single 44-pin plug for in-car use. See question below ...

- One-button operation :) Start/stop logging, and lap marker.

There will very little actual computation going on in the box, since it's main function is to log data. Then you transfer the datasets to a PC for analysis in Excel or whatever. All out of car control is via the USB or serial connection.

Some of the methods of madness behind the design. The power supply section actually has several stages. There's an LM2676 to knock voltage down to 5.5V initially - that's a switching (and hence noisy) regulator. That feeds into 3 separate LT1121 linear LDO regulators, one 3.3V for the MMC card, one 5V for the analog supply only, and the other 5V for the rest of the circuit. Hopefully the power won't contribute to noise ...

The USB interface is the Ftdi 245BM (the new one) in parallel. Lots faster than the 232 serial emulation version. The Sleep and SND/WakeUp pins are connected, though I haven't fully decided on using that functionality yet.

I must have examined half a dozen different 5V-3.3V converter schemes :) FInally Herb Poppe directed me to the MAX3377. I could swear that all the Maxim chips I looked at earlier were only available in 2500 lots, but that one was $1.86 in singles. Cool.

** One concern I have is using the single 44-pin connector for power as well. Not the current draw, but the possibility of other pins being at a potential before the +12V/GND pins connect. How bad could would that be ? If bad, then I guess you will just have to connect power via a standard 2.5mm barrel connector first. But the single connector would be nice ....

Anyway, take a look at the schematic(s). Comments, critiques etc are all welcome - finding problems first is a good thing. You guys have been a *wealth* of information for the other earlier parts as this design has evolved. Hopefully hardware glitches/gotchas will be a minimum from all that help.

The pcb layout is 95% done, on a 6.3x2.9 board. Single board now - was using a daughterboard earlier. I'll post the layout in a couple of days when it's done, though I think I want to flip the power supply section around. It's all 10mil tracks except the ones between the pins on the 44 connector - those are 8mil.

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I take it that these cars do not have the OBD plug for telematics data?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Some do, some don't. The problem with OBD-II is that it's slow as molasses. Not too bad if you're collecting one parameter, but grab a few, and the rate drops precipitously. Besides which, I hear it's a bear of a protocol to deal with :) John Wren here is working on a USB project - he can say more to it.

We want to be able to pull everything at up to 60Hz. 16 analog channels, 6 pulse and 3 temps.

Oh, forgot to mention - DLP has a new USB module out. Smaller (24 pin down from 32pin) and already has the eeprom/resistors on the module. Moving up in the world ...

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Dean,

I would put ground connections in both ends of the connector.
It would stay grounded even if you make an 'uneven' connection (i.e. right-hand side first). And preferreble the outside (ground) pins would be longer, to connect first/last.

Good luck !

Tobias

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

>FInally Herb Poppe directed me to the MAX3377. I could swear that >all the Maxim chips I looked at earlier were only available in 2500 >lots, but that one was $1.86 in singles. Cool.

Who thru?

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Don:

Heh - directly through Maxim :) Who would'a thunk it ?

Tobias:

I'll check with the manufacturer to see if they can supply longer pins. I don't think so though - that would mean that the female receptacle mounted on the pcb would probably need a slightly different construction to accomodate the longer pins.

In an ideal situation, GND gets connected first, then +12V, then the rest. If GND goes first, but then a few signal pins before the +12V goes, what will that do ?

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

actually you can cut the pins down, for a quick & dirty method of staggering vSup, Gnd and the signals. But for a clean solution yes different pins are required (and sometimes a different recepticle as well) there are many different connectors out there that have staggered pins, or different pin options.

if the 44pin connector is just a standard 0.1" IDC header, then all you need to do is pull back the pins.

good luck

Writing code is like having sex.... make one little mistake, and you're supporting it for life.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Currently planning on using the Norcomp 181 series.

http://www.norcomp.net/PAGES/181...

They do make D-sub connectors with dedicated power pins, but only with 21 data pins :( I poked around a lot on the AMP site, but no luck yet.

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

finding staged pins on a D-Sub will be difficult. you may want to consider alternative connectors. SCA type connectors might be a good choice as they have staged pins. But the price will be much higher. Your best option looks to be the barrel type or even a molex type connector, for the power.

Writing code is like having sex.... make one little mistake, and you're supporting it for life.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hrmm. Probably. I already have a barrel connector footprint for power, and it may be just that you will need to use that for in-car power as well.

I'll leave the power traces connected to the 44pin, but that doesn't mean that they have to be used in the wiring harness. Heh, I wish there was a non-potentially-destructive way to test this :) I suppose the only way to be sure is to try it, multiple times.

Dean.

No schematic comments from anyone ? This thing looks *perfect* ? No "I would really do it THIS way ..." ?

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

In order to avoid the second connector you could always put an inline blade fuse on the lead. But then this relies on people doing the right thing when doing a changeover... Or simply doing the most common thing and gritting the teeth whilst doing a hot swap ;-)

parts from Maxim? shock horror...

Cheers

Don

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Aw man, don't tell me. I've never ordered directly from Maxim before. Problems a-waiting ?

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hey Dean,

Have a look at samtec.com. They have some really cool connectors for mixed-use i.e. Power and I/O, plus they offer an excellent sample request centre.

If I was in your position, I would first look at a dedicated power connector solution.

Secondly, if live-power-insertion is a problem , there are some semiconductor solutions on the market that can take care of your problems. Don't remember offhand those guys, but take a look at Texas Instruments, Micrel and Linear Technology. Maybe in one of the App Notes....You can also try searching for topic "Hot Swapping" under PCI, whichh deals specifically with the power insertion / extraction problem while changing card modules...don't remember though exactly like I said.

P.S. Generally keep in mind that mixed-use connectors are alot more expensive than ordinary ones, and sometimes a lot harder to come by.

Regards,
Arno.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Well, price is definitely a consideration point here. If we can save $30 by using a 2.5mm barrel connector, then so be it.

By the way, what *is* the most popular wall-wart plug size ? 2.5mm ? Hmm, might be good to use a 3-connector plug, since I also use +12V-ignition as an "I'm running in the car and the car is on" indicator. Better to have all the direct power on one plug.

Any recommendations ? Which one(s) to stay away from ?

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If this is for automotive use then I would avoid at > cost using a consumer power connector which has no mechanical restraint.

It is not just the issue of keeping the plug in the hole, it is also about not allowing the pin & socket to move in relation to each other as eventually the vibration will cause a film will build up on the surfaces resulting in poor contact.

A quick browse of the Farkenell Cattledog reveals lots of options in locking power connectors.

None will be as cheap as the 2.5mm barrel but then you won't have the intermittent power connector problem either...

Cheers

Don

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hmmm. I suppose you mean Farnell :) I just did a brief 4 minutes on the DigiKey site, but haven't found anything that looks appropriate. They do have a male locking plug, but no corresponding jack ...

Pcb mount jack preferably ...

Ok, time to hit Arrow/Jameco/All/etc

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi,

I am at home now and cannot give you the make and part number but we (at work) just started using a circular mini DIN series that has a locking feature (plug/socket) and the socket we are using is a right angle PCB mount. We are using a 4 pin version and connecting an external universal switching type of 24V power supply. If this is anything remotely interesting to you I will get the part numbers, etc. for you.

Regards,
Steve

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Absolutely. Any information you can supply would be great. Who is your supplier, and would they consider non-thousands amounts ?

Dean.

Oh, hey people. This is all great, but no-one has found any problems in the schematics ? I can't believe I have it *all* right ...

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

A quick look at the opto-isolators, I notice it looks as if the internal pull ups are being used, if so you may find that you will get pw errors, the rise time will be much slower than the fall times, this may not matter is the times are long , or the resolution not critical.

Suggest you add 1k0 external pull ups anyway ( you don't have to fit them )

Your drivng the input diode with 17ma ( which is quite high ) so a sink current of 5ma only requires a ctr of 30% which is well in spec of the device.

The circuit shows 0.5pf in parallel with the 680R input resistor, the time constant 340 x 10^-12 seems too short to be of any use , is there a misprint on the 0.5pf, maybe I misread it?

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi Dean,

I will get that info tomorrow. Don't worry, we deal in quantities of maybe 20, 30 max. a month.

Regards,
Steve

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Arthur -

I'm not sure I understand why you say using the internal pullups will introduce rise/fall time errors ? You're saying to add additional 1k pullups to the PWM0-5 nets (the inputs to the Mega128) ?

I looked at several optoisolators - these seems to be prevalent and easy to find, and reasonably fast switching times, 3-5us when switching 10V at 2mA and 100ohm load. There were a couple out there that were waaay slower ...

I dithered some about the resistor value choice for the optos. At a nominal 12V, they will fire cleanly with a 1k resistor at 12mA. At 5V though it's only 5mA, about the least they'll fire with. 680 seemed like a compromise to make sure they will fire regardless of the voltage being measured.

Argh. Nice catch with the 0.5pF - that is a typo. They should be 22nF caps there. Not sure how that got changed, I remember entering 22n at one point. Global replace somewhere did a little more that expected I suppose.

That was a suggestion from John Doe :) though there has been mention as to whether they're really needed or not.

https://www.avrfreaks.net/phorum/...

Steve -

Thanks for the help, much appreciated.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The circular locking mini-dins may be Sensor connectors. These come in a range of sizes from a wide range of suppliers. They are generally waterproof & commonly avail with a molded lead fitted. They provide good positive locking but as always there is a catch... You pay for what you get ;-)

Have a look in the Farnell Catalogue under sensor connectors. They are not in the general connector section.

Cheers

Don

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

the Opto-couplers rise time is very fast with 100R load.

Looking at the spec sheet the rise time is approx 200us with 10k load and 500us ( graph a bit blurred here ) with 50k load.

I believe the pull up circuits on the chip are equivalent to resistors in the order of 50 -100k

Whereas the devices will switch on quickly they switch off dependant on load, ie could be asymetric.

I didn't know if delays to switch off of this size were important or not.

I also agree and think the caps will do nothing

Nice project please publish code so we can all crib it.

arthur raine

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi Dean,

The connector manufacturer I spoke of is Kycon, Inc. (www.kycon.com). The part numbers are:

Right angle PCB mount socket = KPJ-4S (shielded version = KPJ-4S-S)
Available from Sager (I am sure others also) for $1.64 in minimum qty. of 1
I didn't check the price of the shielded version.

Locking plug = KPP-4P
Sager - 2.92 in minimum qty. of 1

(Gotta love a company that keeps part numbers so simple also!!!)

You can get .pdf drawings from their site (just click on Products and then on Search)

This part almost seems too good to be true... small, cheap and available (Sager had 1800+ of the KPJ and 2300+ of the KPP). I hope you can use it in your application and I hope it doesn't turn out like so many things that are "too good to be true".

I forgot to see what Sager's minimum order requirements are.

Regards,
Steve

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi,

I added up the current draw for your circuit, and I think that it might go over the 150 mA of the regulator at times. Using data-sheet numbers for typical current draw and calculated current draw thorugh resistors, i got around 150 mA of possible current draw. This is sorta worst-case. This depends on everything that can be on being on at one time, but i think that could happen (very unlikely though). Actually I think its impossible for that current draw to happen, but I still haven't included a few other things in my calculations that will probably put it back over the top.

Also this assumes that there is NO outside shorts occuring, which would add more strain to the system (one of the MAX chips said it allowed up to 60mA of short-circuit current). Also i have NOT included the current draw for the LCD display.

Thats all for now... ;)

-Colin

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quick thing- the LCD current is probably only like 1.5 to 3ma, so its not too big a deal (checked datasheet for generic LCD display).

-Colin

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hello yet again ;)

I was checking a few more things, and noticed another possible problem with the LT1121CZ-5. If you check the data-sheet, you will find hidden in it two important graphs. They show the gaurenteed (spelt wrong i bet) drop out voltage, and drop out voltage VS output current and temperature. It is possible that the drop-out voltage could go to 0.5v or higher i think under the wrong conditions.

Try the LT1129 instead (LT1129IST-5 for example). It has about the same dropout, but at a much higher current draw (700 mA). At around 300mA the dropout is somewhere around 0.35 V (50 celcius operating temperature).

The other two LT1121CZ parts are within spec i think, its only the -5 that might need to be replaced.

-Colin

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi Colin,

>gaurenteed<

You must have been a Justin Wilson fan :-)

In case you (or anyone else) is not familiar with Justin Wilson, he hosted a popular Cajun cooking show on PBS. He had a way of saying " I'll gaurenteeeee" that was funny (among many other sayings).

Sorry for OT comments.

Regards,
Steve

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If anyone is interested, I've included a pdf of the imcomplete layout from some days ago. I'll post it again when it's complete. I still have to route the DG333a muxes with the resistors to the opamps, and the pulse inputs to the optos. Then the 3 GND planes, AGND, DGND and EARTH. I've checked, the blue bottom layer isn't too busy, and the planes cover just about everything.

You may catch that the MAX3235 is on the bottom layer, under the USB module. Only place it would fit :) And the serial connector is a 3-pin SIP right-angle. Not normally accessible, you will have to take off the enclosure end to reach it. Oh, the enclosure is a Hammond 1455k1601.

http://www.hammondmfg.com/1455.htm

Colin -

Hah ! That's exactly the part I have open in another window :) There are actually two 5V regulators and one 3.3V one. One 5V feeds the analog portion (AVCC) and the other the rest of the circuit. I seem to recall coming up somewhat short of the 150mA way back when, but I think I'll switch to the 1129 anyway. More headroom.

Steve -

Those connectors look great. I'm ordering a few samples to check out, but they look ideal. The only hassle is that a generic wall-wart supply would have to be modified, or I would have to include the 2.5mm barrel as well. Might do that.

Arthur -

You're the second person to decry the caps over the opto inputs. I think I'll leave the traces in place, just in case, but not install them unless I see problems. The pulse inputs are there for capturing typical automotive pulse signals, which don't tend to be in the uS range. Heh, hopefully they aren't. Toyota service manuals are distinctly vague about specs, both for pulse as well as level signals. They're there mainly so a dealer can tell if a part is good or bad, but not to diagnose/analyze it.

If anyone has a better way to watch pulse signals, I would love to hear it. Umm, keep in mind the pcb is pretty much packed as is. The only way to squeeze more in would be to go to a 4-layer board (lots of trace area) or to the 6x4 board and the 6x4x2.5 box. The existing 6x3 board in the 6x3x1.6 box is an ideal size, and I would really like to stick to it.

That's why I don't have the ESD diode arrays on the analog inputs. Well, there is a way. Switch to a 42-pin header, and use a panel mounted connector via ribbon cable. Kind of ucky.

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hey.

What about powering it from the USB port when connecting to the PC ?

The analog section wouldn't need power, so that can be left off. 5V can be fed directly to the +5V power net, and via a diode to the input to the 3.3V regulator. Actually, that power would be via a switching transistor so the USB module can power the rest of the circuit up.

Ftdi has a couple nice little appnotes on it.

http://www.ftdichip.com/Files/Hi...
http://www.ftdichip.com/Document...

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi,

Wow, seems like a small board! This is one cool project, and it looks like it's finally going to be finished... i still remember reading this thread...

so you decided to go with my case suggestion I see? Let me know how it turns out, and be sure to post a few pictures of the project when it is all done!!!

I know I've learned a lot by following your design and seeing what people think of it!! anyway getting back on topic more...

You could power the entire thing off the USB port, but you have to make sure that it does have the low-current mode... (I assume thats what your talking about when you say using a switching transistor to turn on the rest of the circuit)

Also you could just power parts of the circuit off the USB port. For example the FTDI chip took 50 mA, which was the highest number i found (AtMega128 at 16 Mhz was second at around 30mA i think). That would reduce the current load by quite a bit, and might keep it within safe operating of the 150mA regulator.

-Colin

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Heh, ahh, the good old days. How things have changed :) I think that this is part of the beauty of a forum like this. It can help take something from a conception through refinement to something that hopefully is really close to "not-many-ECOs" before anything is actually built.

I wanted this to have two applications : those that would mount it permanently under the dash, and use the MMC card to move data to the PC (perhaps serial/USB to a laptop in the car) and those that would be popping it in and out all the time. You choose how you want to use it.

The more I think about it, the more I like the idea of powering it off the USB port. Originally USB was just going to be an option ($27 more for the module). I'll see if I can fit the power socket on the left next to the 44pin socket, and flip the whole power supply room around, so it feeds left to right. The optos will have to shove a bit right as well.

VCC supplied from the USB will go to the switching transistor, and the USB load control will go to the base, as in the Ftdi diagrams. The load output from that will feed right into the +5V power net, and feed via a diode into the input side of the 3.3V regulator.

** Is there any problem not supplying power to AVCC for the Mega128 ? That would be good, since then all the analog circuitry is just inert.

I'll replace those regulators with LT1129 ones, which have shutdown capability, so the USB module can control them directly. That way it can go into suspend mode and take the whole circuit down.

** Would I need a diode in the +5.5V feed into the regulators to make sure it doesn't backfeed ? I'm thinking of someone leaving the box in the car powered up, and connecting via USB to a laptop.

If anyone has done USB powered stuff, feel free to add info here.

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Just had a doh!

Move the input side of the 3.3V regulator from the +5.5V switcher output to the output of the digital +5V. That way I don't need a diode to prevent it from backfeeding the analog portion etc.

So the switcher feeds +5.5V to the LT1129 for digital +5V and to the LT1121 for analog +5V. The digital +5V feeds the LT1121.
for +3.3V for the MMC.

USB question ... Say the datalogger is in the car, powered up and running. You want to copy stuff to the laptop, so you plug in via USB. You do stuff. Now the USB bus wants to put your device into suspend mode - maybe you're using a USB camera or something that takes power from the bus.

Since the device is self powered (while in the car), does it have to go to suspend ? Out of the car taking power from the bus, it definitely would have to.

If so, then I suppose I can run the sleep pin from the USB module to the /SHDN pin on the main LM2676 regulator. Hmmm, have to check if that pin stays low until the module is configured - that wouldn't work since it wouldn't configure in the car until talking to a laptop.

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Correct me if I'm wrong, but according to my datasheet on the DS1305 RTC, the unused VCC2 pin should be connected to GND when a non-rechargeable backup battery is used.

Kind Regards

Steve

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Steve -

Thanks. You're right. VCC2 should be to GND. Weird, I don't remember the connection one way or another, that is I don't remember *deciding* to connect it to VCC or to GND.

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hmmm. Nuts.

The 2.5mm barrel connector will fit next to the Norcomp 44pin connector nicely. Keeps the connections on one side of the box while in the car.

Out of the car, the USB, serial and MMC are all on the other side. This is aesthetically nice :)

I really like the Kycon connectors - should have a few coming as samples next week to take a look at. But dang it, the jack won't fit next to the 44pin :( It will have to go on the other side, where you see the space for it in the pcb layout above.

Ehhh, I don't suppose it make much of a difference having it on the other side. It just grates :(

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The PCB layout is pretty hard to read (Does acrobat have layers?), but it looks like you haven't followed spec on the capacitors for the LT1121 and LT1129's. With LDO v-regs it's incredibly important to do exactly what the datasheet says or you will not be please with the results.

BTW: What layout program are you using? I like the color.

admin's test signature
 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I normally use Protel 99SE. I have the newer Protel DXP trial in place to mess around with, so I've been exporting and importing into that for printing. I print to a Color Laserjet 4500 postscript, but print to file, then convert that postscript file to pdf with the ps2pdf tool from ghostscript on Linux.

Kind of roundabout, but it works :) The older pdf's you saw were all right from 99SE to postscript to pdf, no DXP. The layout only has the LT1121 ldo regulators on it. The +5V digital supply one will(is already here) be replaceced with the LT1129.

What part of the spec am I missing ? As far as I know, the datasheet calls for ceramic solid tantalum caps on the output of the LT1121's. Less that 1uF should have a little bit of series resistance, and 1uF and above should have none. It should be a minimum of 1uF, but given that the LT1121 are being used as a secondary regulation stage, I thought that would be adequate. I suppose they can be bumped up to 10uF or 22uF.

The LT1129 calls for a minimum of 3.3uF, again with a ceramic solid tantalum. Same deal, that can be bumped up too.

Thanks for the comments - this is how we find problems *before* we etch ...

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Just a very brief update.

I got in the Norcomp 44-pin pcb socket samples. Good thing :) The pcb footprint I built in Protel was accurate, but the faceplate wasn't. The whole socket had to move to the left in order to protrude from the end-plate of the enclosure. So that messes up the traces there. No biggie, but nice to catch NOW instead of when putting the first board into the enclosure ...

I'm waiting on samples from Kycon for the power connector sockets and plugs. From those I'll decide on whether to use them, or a more typical 2.5mm barrel connector. Pros and cons both ways.

The 2.5mm connector will fit right next to the 44pin harness connector. That keeps the connections all to one side in the car, which makes for easier mounting and dealing with. It also makes the EARTH ground plane nice and neat and compact, since that would include the opto-isolators. The whole power supply section would just flip around the horizontal axis, input to the left, final regulators to the right.

There is the risk of vibration loosening the 2.5mm plug, and creating a film on the connecting surfaces. Not good, particularly since some people will elect to install these things in the car permanently, or at least not so they pop them in and out all the time.

The Kycon lockable connector won't have those problems. But it won't fit there next to the 44pin. It will have to go on the other side, next to the MMC socket and USB socket. Not too bad a thing, just a little cumbersome in the car. Layout won't change too much, except that there will be a thin (30mil) trace from the right end where the EARTH gnd plane would be over to the left to be under the opto-isolators. That part is a little icky, and doesn't really follow the concept of a *plane* so well.

How bad is it to have two larger blobs of a gnd plane connected via a 30mil trace ?

Little bit of shuffling to get the LT1129 in to replace the LT1121 for the main +5V supply. It's a 5-pin TO-220v case instead of the nice little 3-pin TO-92 case.

Any opinions on how good it would be to include ESD diode array protection on the inputs ? This is where each input has two protection diodes, one from GND to input, and one from input to VCC, with breakdown voltages of say, 18V. Nab those spikes in the act !

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hey Guys -

No one has any comments on my two questions above ?

"How bad is it to have two larger blobs of a gnd plane connected via a 30mil trace ?"

"Any opinions on how good it would be to include ESD diode array protection on the inputs ? This is where each input has two protection diodes, one from GND to input, and one from input to VCC, with breakdown voltages of say, 18V. Nab those spikes in the act !"

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I can see nobody have said anything about the ADXL202...

You see you need a series resistor of 10 - 100 ohm or a ferrit bead in series with its supply voltage...

If not the digital noise on the VDD will inject into its delicate circuit and lot of strange things can happen.. Like popping PWM at a rate of ~10Hz (Thats what i have measured before i changed my circuit...)..
Its best to read the datasheet about the supply decoupling...

The PWM frequency is also a little problem ... It can change a lot from part to part with a "fixed frequency resistor" "rset" .. up to +/-50%

Regards Sonny

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Sonny :

Thanks for the tip. I'll move the ADXL202 section north a little to make room for a ferrite bead and resistor in the VCC line. Given the room, should I use both ? I was going to move it anyway to make room to move the DS1305 up and away from the power supply section.

What do you suggest about the rset resistor ? Hmmm, might be worth looking at using another digipot there for fine-tuning/calibration. That would take about the same amount of room as a multi-turn pot. The only trouble is that digipots only really have a 100k max range.

The parts values for the ADXL202 there are just placeholders - I haven't gone through the process to choose the cycle, duty cycle etc yet. I just know that it will take the output caps and the resistor.

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

> No one has any comments on my two questions above ?

> "How bad is it to have two larger blobs of a
> gndplane connected via a 30mil trace ?"

Its not a great idea in general. Better to connect the two planes at one wide point. That said, there are times when you do want to split a ground plane. This is usually done when you have "two" grounds: analog and digital and don't want them to interfere with one another. You would connect them at only one point.

Having said _that_ I'll go back and say that its a bad idea. A quick check at a high-resolution ADC datasheet shows that they don't even recommend a split ground plane. (High-res ADCs are extremely sensitive to noise.) In fact, the only time I've seen a truly split ground plane is when you also split the power plane and bypass both to make a pretend low-impedence source. This will usually do well to keep local current supply problems isolated.

When you connect the two ground planes by one small trace (30 mil is still fairly small), you risk ground bounce because you've just created a high impedence path to ground. You've also just made sure that _all_ return paths go thru the same point. What does that mean? It means that you are likely to get conductively coupled noise.

I suggest either connecting the ground planes or placing the power supply (and thus the signal return) in the middle of the two planes so that each plane is most likely to return all its signal to the center and nowhere near the other plane. Connect the three planes with fat connections and bypass the circuit a lot at the connection of each plane to the power supply plane. (My description is terrible, I know. But it works.)

-- Patrick

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

for a duty cycle time of 4msec. i use a ~600Kohm resistor.

I have a resolution/noiselevel of a few millig's.... if you could live with 10millig's it would rougthly be ~ 50 - 150k.

You filtercaps will also get smaller... ~10 times

Calculate your needs and use the datasheet

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Patrick :

Heh, actually I'm planning on using 3 ground planes on this board, connected via ferrite beads.

There's the standard analog GND (with analog VCC supply) that's under the connector, muxes, opamps and analog portion of the Mega128.

There's an EARTH GND that's under the directly connected vehicle chassis GND. This is for the input diode, inductor and cap, as well as for the optoisolators for the +12V-ignition line and the pulse inputs.

Then there's the standard digital GND under the rest of the circuit.

The analog GND connects to the digital GND via ferrite bead, and the digital connects to the EARTH GND via bead. This keeps the analog one as far away as possible from the raw noisy EARTH GND. The hassle was that the pulse inputs and their optoisolators are at the left end of the board, while the power input socket is on the right side (if I go with the locking Kycon power socket). Most of the area between the two is digital GND, so the blob-planes at the two ends would have to be connected via trace. There's a ferrite connecting them on the right side of the digital gnd.

/----------\ /--------------------------\ /-----------\
| | | digital | | |
| | | gnd | | earth |
| | \--------------------------/ | gnd |
| |---------------------------------- | |
\----------------------------------------------------------/

I've heard several times that in a situation like this, you *want* the grounds to be connected at only one point, with a ferrite or a zero-ohm resistor. This way the noisy digital signals are kept to their area and have a higher-impedance path to affect the quiet analog area.

You're saying that the high-res ADC datasheet says *not* to use a double-GND plane ? Just have one ? I actually have 3 power supplies on this. One for 3.3V for the MMC card, one 5V for the digital stuff and one 5V for the analog stuff. They're all LDO linears fed from a single 5.5V switcher.

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Sonny :

Thanks for the tips. The ADXL202 is really there for two purposes - to provide in-line G forces to be able to calculate HP, and for lateral G forces, to get an idea of cornering forces and efficiency. I haven't gone through the online calculator yet to figure out just what resolution I'll need, but I don't think it will be that high.

Oh, I got the Kycon connectors yesterday. Looks like the socket will actually fit next to the 44pin socket, so power will be able to live on the left side. Excellent. I should be able to get the layout finished up this weekend, so long as the HoneyDo list isn't too big ...

Are people interested in seeing the layout ?

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Ah Nuts.

The MAX6627 remote temp sensor doesn't seem to be available anymore in anything less than a 2500 count tape reel.

It was on the Maxim site a month ago or so in single quantities :( I've asked for a quote for small numbers - we'll see.

Nearest replacement is the MAX6674 - a cold-junction type instead of a simple 2n3904 transistor. It's also in a SO/.150 case instead of the tiny 8/SOT23. So that will need some rearranging on the pcb.

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hey -

I have an initial stab at the parts list, attached above. It looks like this is about $190-$210 in parts (at single item quantities) fully loaded. Note - the list isn't complete. Still have to do a bunch of caps and a few resistors.

That doesn't include the pcb - that will run around $25, $33 or more depending on the fab house.

The expensive and/or optional parts are :

ADXL202 - $30
DLP USB - $25
MAX3235 serial - $5.56
DS1305 RTC - $3.85
DS2890 LCD digipot - #3.05

Any of these can be done away with without critically impairing functionality.

Just plain expensive, but have to have 'em ...

Mega128 - $16.35
44pin connector - $9.13
OpAmps - $23 for two

Dang. I had been hoping for a $120 parts cost ...

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

A few people have been asking to see the latest layout. Here it is.

MAX6627 temp sensor chips are gone.
CalMicro pacdn017 ESD clamp diode array is in place, but not routed yet. It's really hard to see, but look on the left, under the four resistor arrays just to the right of the 44pin connector. It's the blue 24-pin SOIC package.

The question is whether or not to use that ESD array. Without it, the analog GND plane is nice and fairly unbroken. The only place it will fit is on the bottom layer, connecting directly to the 44pin connector. That chops up the analog GND plane right there something fierce :(

What's more important - a nice clean GND plane, or ESD protection ?

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hey again -

Now that there a few extra pins lying around since the MAX6627's were dropped, why not add a piezo buzzer for beeps and boops ?

http://www.digikey.com/scripts/u...

I figure this can be driven from +5V right from an IO pin through a 270ohm resistor.

Still mulling the question of using an ESD clamp diode array or individual caps (or array) to GND for each analog input. Check the "Grounding for car/vehicle device" thread at

https://www.avrfreaks.net/phorum/...

Dean.

Dean 94TT
"Life is just one damn thing after another" Elbert Hubbard (1856 - 1915)

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Actually inside your car you won't hear it. Outside it gets quite annoying. And the good ones are quite big (approx the size of the avr).

Most of the modules don't need a resistor, they work quite nicely directly off one of the ports. (Check total port limits though)

Alex

admin's test signature
 

Pages