ADC PCB layout help

Jonathan

Hi All,

I am finally going to join the modern world and stop mucking about making my own PCB's. To this end I am modifying one of the existing Prop boards (on Ray's site) or perhaps making a couple of daughter boards. I think I am going to modify a Prop baord to put an xbee on it and have the ADC on it's own board.

Anyway, since layout is important for good ADC performance, I was hoping that someone could share a layout. I'm happy using just about any 12 bit ADC, as long as there is an object for it. Even if it isn't connected to a Prop a good layout example would be very helpful. I'm not talking about anything for an automotive or other very noisy environment, just a decent place to start.

Of course, I'll send Ray any layouts I come up with for all to use.

Jonathan

Cluso99

Sometimes the datasheet will come with recommendations. Also search the manufacturers website for app notes on the ADC device.
This is because there are often special requirements for ADC chips.
Otherwise there are some good references on the web - google is your friend.
You will find these sources much better than we can describe because they usually include diagrams.

Jonathan

Translating data sheet recommendations to an actual PCB layout is above my pay grade. I have googled around and haven't found an actual CAD layout. But lotsa folks have done this, so I am hoping to find one, done by a monkey much smarter than me.

I'll keep looking and thanks for the help!

Jonathan

Jonathan

Cluso,

I have been looking and I find many app notes but no actual layout examples. Surely there must be some example layouts somewhere! Maybe I am looking for the wrong key words.

Jonathan

Sir Gawain

I would suggest that you take the signal you are measuring and run it through an op-amp at unity (voltage follower), then to your ADC. Depending on the circumstances, it means the measuring won't affect the signal.

LoopyByteloose

You might try referring to the Propeller ASC+ board, which includes an ADC chip.

The reality is that ADCs are not that hard unless you desire extremes of precision and higher frequency capture. At those points, noise and adequate sampling rate can become important.

I am guessing that you don't actually have a specific use in mind, but desire to have a general purpose ADC available. Adding a unity-gain op amp to the front end may not be necessary.... though it certainly won't hurt.

hatallica

http://www.ti.com/lit/ml/slyp167/slyp167.pdf

The link above is a short presentation from Texas Instruments regarding PCB layout for ADC. They focus on a handful of core principles and then show examples. TI has other great whitepapers, but guys tend to be visual learners with attention spans that require bullet-points.

Note that this is not a copy-and-paste layout for a specific ADC chip. It is more of a list of Do/Don'ts that will keep you to avoid adding noise to your signal. Without following good practices, almost any environment will be sufficiently noisy to affect the last few bits of a 12-bit ADC.


Regarding the previous suggestions to add a unity op amp buffer, it can help but it can also certainly hurt. You are still adding stuff in between your signal and the measurement point, so carefully consider the non-idealities of the op amp (offsets, signal coupling, etc.). That said, an op amp may be necessary if source impedance is within a few orders of magnitude of the ADC input impedance (say, less than 1:5000 ratio).

Jonathan

@loopy- Looking at the ASC+ board was helpful. Looks like it is just plopped as close to the header as possible with nothing between. You are correct that I am just looking for a general ADC for easy signals to read. Still, smoother is better.

@hatallica - That PDF was indeed interesting. Still trying to puzzle my way through it.

While both those suggestions were very helpful, I'm still left with a lot of questions, like should the opposite side of the board from the adc be ground plane? Or is it OK to route ADC inputs on the other side? I sure wish I could take a look at an ADC design in a CAD program.

Thanks folks!

Jonathan

hatallica

Sorry, Jonathan ... I was going to attach a snapshot of an old PCB design that I did with BS2 and MCP3301. However, I admit that I have not invested the effort to figure out the quirks that this forum has for attaching images from my computer.

Practically speaking, though, there is no one answer. The specific ADC that you use might have one ground pin, or it might have separate analog and digital ground pins. It might measure referenced to analog ground, or it might have differential inputs.

So, I apologize for yet another generic reference, but it shows that there can be several solutions. The basic gist is to keep digital and analog ground paths separate, but still referenced at one single point.

Regarding ground plane - you only need it on one side and the signals can certainly be routed on the other (I am assuming that you will be using a through hole design, which simplified trace routing). In a really simple example that only has a Propeller and a MCP3301, a separate ground plane would be beneath the analog (MCP3301) portion and the digital (Propeller) portion. The two planes would connect at only one point.

Maybe if you have a specific ADC in mind, then I could whip something up really quick in Diptrace and see if I can figure out how to attach the file.

Jonathan

@Hatallica - That link was great in explaining ground planes! Rather than just saying that A-ground and D-ground should connect at one point it gave great graphic examples. Thanks!

I am planning on using the MCP3208. Spent a bit of time yesterday modifying one of the boards on Ray's site for the Prop. Hope to finish that today. Only changes are that I like using RJ45 and RJ11 connectors, so I am switching the headers over.

Thanks a lot for the help! Really appreciated.

JOnathan

denno

Over the years, and lots of them, I have used the LTC1298 in many devices to measure voltage, and/or amps (with a current transformer). The device is simple to use, and in my devices, I just design the PCB in such a manner, to keep the LTC1298 close to the processor. These devices were all powered by a Basic Stamp..sx. If you are measuring more then 5 volts, use a voltage divider of (2) 5k ohm resistors, and a 10k ohm multi turn pot. Your voltage reading, if you are measuring over 5 volts, will be made across the 10k ohm pot to ground. (Single ended). Measure the voltage in question with a meter, and adjust the trimmer pot to match that voltage...calibration. Again, if the voltage is over 5 volts, and less then 15 volts, in your software, simply multiply the voltage across the trimmer pot by 3, or multiply the ADCindex value by 3. Divide this ADCindex value by .00122...12 bit resolution...to obtain a voltage value between 10 volts and 15 volts. Several ways to do this, for voltages over 5 volts, but I am sure you will get the hang if it.

I'm still trying to teach myself SPIN, after decades with the Stamp...which BTW...all you SPIN users, should not discount as a very good controller, especially the SX!

Again, the LTC1298 is a good device, old, like me, but still available, but getting a little more pricy. The MCP3208 is also good, if you need lots of inputs to do the same thing.

I have been using ExpressPCB for just as long, for making boards. They have there own software, which you can download, using there own server. Easy to use IDE, after a short learning curve. Prices are good, which you can Google there web site to check out.

Hope the above helps...DennO

Jonathan

Ahh, the LTC1298. I think that was my first ADC...

So, most folks I see put the ADC inputs as close to the connectors as they can. You were putting the ADC as close to the uC as possible?

Just sent off my first board to ExpressPCB, a mod of one of the Prop boards on Ray's site. I'll be starting an ADC board this week.

Thanks for all the help. Still surprised one can't just find an existing layout.

Jonathan

frank freedman

@hatallicia,

gender-slap aside (personally I tend towards hyper-focus and learn by many methods), one of the best and ongoing references on this topic come from Bonnie Baker. A lot of good articles etc via TI, Microchip, EDN, and such. Google her and "Planet Analog"

F

hatallica wrote: »

http://www.ti.com/lit/ml/slyp167/slyp167.pdf

The link above is a short presentation from Texas Instruments regarding PCB layout for ADC. They focus on a handful of core principles and then show examples. TI has other great whitepapers, but guys tend to be visual learners with attention spans that require bullet-points.

Note that this is not a copy-and-paste layout for a specific ADC chip. It is more of a list of Do/Don'ts that will keep you to avoid adding noise to your signal. Without following good practices, almost any environment will be sufficiently noisy to affect the last few bits of a 12-bit ADC.


Regarding the previous suggestions to add a unity op amp buffer, it can help but it can also certainly hurt. You are still adding stuff in between your signal and the measurement point, so carefully consider the non-idealities of the op amp (offsets, signal coupling, etc.). That said, an op amp may be necessary if source impedance is within a few orders of magnitude of the ADC input impedance (say, less than 1:5000 ratio).

Domanik

Jonathan wrote: »

I am planning on using the MCP3208.

Sorry, I'm late to the game but the manufacture's layout is at the end of this datasheet.
MCP3208 Evaluation Board
Also, I really like this attached schematic from Analog Devices, Inc.:

... very important idea of star ground. If an opamp is used as a buffer it should have low offset (micro volt), high input impedance, low output impedance (~50 ohms), keep max out more than Vref of ADC, be sure reference for AGND is same star-gnd as ADC and supplied power if from Va not Vd. Some opamps are touted as rail to rail but run out of gas as they get close to the rail. Therefore if you have +5V-A supply and an analog input range of 3-4 volts it should be okay.
Or leave it off completely. Do you really need a buffer?
An opamp could be designed in and left off the board with a strap from input to output instead.


PS: I agree Bonnie Baker has many good write ups in EDN.

Jonathan

Domanik,

Great! Just the sort of thing I was looking for. The ADC layout answered many of my questions.

I'm not worried about the op-amp buffer, the signals I am reading are pretty easy. I may throw the place for it on the board. I am certainly going to leave places for a voltage divider and LC filters.

I see in that layout that they didn't leave room for a cap on the inputs. I have often found that caps are helpful in smoothing readings. Mind you, most have been hand soldered on perfboard and had no ground plane. Still, I think I'll leave places for some caps too.

Thanks for all the help, folks. I'm feeling better about laying out a PCB.

Jonathan