Author Topic: 32-bit ADC playground for precision measurement tasks.  (Read 7481 times)

maat and 1 Guest are viewing this topic.

Offline TiN

  • Super Contributor
  • ***
  • Posts: 2948
  • Country: tw
  • Volt-bolt
    • xDevs.com
32-bit ADC playground for precision measurement tasks.
« on: June 20, 2017, 03:02:23 AM »
 :popcorn:

There are plenty new toys for a modern voltnut available now, such as Delta-Sigma TI ADS1262/1263's, ADI AD7177 and now fast SAR with DF - Linear 2500-32. It might be interesting to play with these converters, instead of always relying on bulky 3458A...

Long time we didn't play "guess the thing" game, eh? I got some parts today for new old idea, so here are some hint photos...



Some extra close-ups.





Also open for suggestions :)
Need to make HSMC to 0.1" header adapter though, as my DE1-SoC does not have HSMC.

xDevs.com | Have test gear documentation to share? Upload here! No size limits, firmware dumps and teardown photos welcome.
 
The following users thanked this post: Muxr

Online Echo88

  • Regular Contributor
  • *
  • Posts: 228
  • Country: de
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #1 on: June 20, 2017, 03:38:28 AM »
Seems the LTC2500-32 ist the new brother of the LTC2508-32 with a few better specs it seems. Will be interesting to see it compared to your 3458A.  :popcorn:  :-+

Also: LTZ1000 instead of the (assumed) LTC6655.
« Last Edit: June 20, 2017, 03:40:16 AM by Echo88 »
 

Offline SilverSolder

  • Contributor
  • Posts: 40
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #2 on: June 20, 2017, 04:19:49 AM »
Interesting idea!

Which reference is actually better for this kind of thing - LTC6655 or LTZ1000?
 

Offline Vgkid

  • Super Contributor
  • ***
  • Posts: 2243
  • Country: us
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #3 on: June 20, 2017, 05:11:01 AM »
I look forward to reading this. The LT offering looks especially interesting.
If you own any North Hills Electronics gear, message me.
 

Offline alm

  • Super Contributor
  • ***
  • Posts: 1101
  • Country: 00
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #4 on: June 20, 2017, 05:44:14 AM »
I wonder how it would compare to rolling your own integrating ADC with a CPLD and a few precision op-amps, as in most high-resolution DMMs (except the ones who do the same with an ASIC). Obviously it will be much simpler.
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 3529
  • Country: de
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #5 on: June 20, 2017, 05:54:18 AM »
The LTC6655 could directly fit to the ADC, while the LTZ1000 would need a kind of divider (e.g. divide by two, may include filter) and buffer. So while the LTZ1000 can be more stable (lower drift) it might get difficult to reach the same level after the divider and buffer.

Using something like 4 of the LTC6655 might bring the noise down to a level lower that the LTZ1000. This might still work without an extra buffer, though for more filtering a buffer might be needed here too.

So it kind of depends on the application, which reference would be better. The ltc6655 is definitely easier even if several in parallel are used - but long time drift might be a problem.

The noise level and INL reached by this ADC is quite impressive. Noise wise it could be better than many ADCs used in high end DMMs. Something like 0.1 ppm RMS noise at 60 SPS is impressive. However the input stage might add some extra noise.
The INL of the integrating ADCs might be better, if well build with a lot of experience. It is a little difficult to compare due to the different range (e.g. +- 2,5 V vs something like +-12 V). Ready made DMMs might also use some software corrections, not just the raw ADC readings.
 

Offline branadic

  • Frequent Contributor
  • **
  • Posts: 736
  • Country: de
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #6 on: June 20, 2017, 06:23:49 AM »
±0.5ppm INL (Typ) and ±2ppm INL (Min/Max) is not that impressive. Even the good old LTC2400 delivers 4ppm INL and 0.3ppm RMS noise but with a predictable and therefor adjustable error. Thus the error can be adjusted to <1ppm. LTC2500-32 seem to have not that easy adjustable INL error.
But maybe TiN can prove us that I'm wrong.
Prema 5000 | Prema 5017 SC | Keithley 181 | Tek 2465A | VNWA2.x with TCXO upgrade and access to: Keysight 3458A, Keithley 2002, Prema 5017 SC, 34401A, 34410A, HDO6054, Keysight 53230A and other goodies at work
 

Online NANDBlog

  • Super Contributor
  • ***
  • Posts: 3625
  • Country: be
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #7 on: June 20, 2017, 06:41:56 AM »
From specs, it looks to me, it is using the same core as the LTC2376 family. I did play with those, damn impressive. For sure you need a 3458A to fully appreciate them. When you see your measured value and the 3458A track each other withing PPMs for days.
Good times. I should design analog front ends again, I miss it  :-[
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 1664
  • Country: de
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #8 on: June 20, 2017, 06:56:00 AM »
±0.5ppm INL (Typ) and ±2ppm INL (Min/Max) is not that impressive. Even the good old LTC2400 delivers 4ppm INL and 0.3ppm RMS noise but with a predictable and therefor adjustable error. Thus the error can be adjusted to <1ppm. LTC2500-32 seem to have not that easy adjustable INL error.
But maybe TiN can prove us that I'm wrong.
Hello,

Look at the patends cited in the datasheet.
They have done much to improve DNL by adding artificial noise. (dithering).
(That explains the "noise floor" on the INL-diagram).
So I think there is not much to improve.
If you have luck you can halve the resulting INL.

with best regards

Andreas

 

Offline SilverSolder

  • Contributor
  • Posts: 40
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #9 on: June 20, 2017, 11:21:03 AM »
The LTC6655 could directly fit to the ADC, while the LTZ1000 would need a kind of divider (e.g. divide by two, may include filter) and buffer. So while the LTZ1000 can be more stable (lower drift) it might get difficult to reach the same level after the divider and buffer.

Using something like 4 of the LTC6655 might bring the noise down to a level lower that the LTZ1000. This might still work without an extra buffer, though for more filtering a buffer might be needed here too.

So it kind of depends on the application, which reference would be better. The ltc6655 is definitely easier even if several in parallel are used - but long time drift might be a problem.

Can we do both?  we take 4x LTC6655 as the main reference.  We also have an LTZ1000 on board, and we read its output when the system starts (after warming up of course) to take care of any long term drift?  We could even have multiple LTZ1000 on board and average readings of all of them using the ADC...  and we could arrange that power to the LTZ's is never cut, for long term stability.
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 3529
  • Country: de
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #10 on: June 20, 2017, 08:44:36 PM »
Using a second reference for doing long time stability tests is possibly and may be a good idea in some applications. This is especially true if there are additional amplification stages that are also included in the test / adjustment. Though I would not use a parallel connection of several references for this. If more than one ref is needed one would use separate references, maybe even different ones.

The Datron 1281 uses two LTZ1000 and one LM399 (or maybe LM199): using the average of both LTZ1000 as the working reference but have additional paths to individually measure them separately.

The slight difficulty with this ADC and the LTZ1000 (and other 7 V references) is that one would likely need to divide down to something like 1/3 or 1/2. This divider would also need to be very stable, like a capacitive or maybe a precision transformer circuit.
 

Online Alex Nikitin

  • Frequent Contributor
  • **
  • Posts: 860
  • Country: gb
  • Femtoampnut and Tapehead.
    • A.N.T. Audio
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #11 on: June 20, 2017, 08:56:27 PM »
The slight difficulty with this ADC and the LTZ1000 (and other 7 V references) is that one would likely need to divide down to something like 1/3 or 1/2. This divider would also need to be very stable, like a capacitive or maybe a precision transformer circuit.

In this situation you already have an ADC which is a very accurate ratio measurement device. It shouldn't be difficult to auto calibrate out (and compensate for with a cheap DAC) any drift in the main precision divider by just measuring its output compared with the reference output, using a couple of buffers and another divider with a good short term stability and a low enough noise.

Cheers

Alex
 

Online NANDBlog

  • Super Contributor
  • ***
  • Posts: 3625
  • Country: be
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #12 on: June 20, 2017, 09:26:32 PM »
You pretty much only need to place the following before the ADC:
a buffer (with the capacitive load capability, high bandwidth, 5V)
divider
multiplexer
Nx RC filtering low speed
Nx a buffer (low speed, high accuracy, high voltage +/-15V-ish)

Now, just be careful with the switching, let enough time for the RC and the opamps to settle. It is not trivial, because you need the settling time to 0.5 ENOB. I've used actually a much more complicated AFE to increase the throughput of the system, but that goes into the dark magic category.With a switched front end, and a very accurate reference voltage it is actually possible to get rid of the 0.1Hz-10Hz noise, (though LTC6655 does not have too much) and the drift of the reference (the one which is driving the ADC).
 

Offline SilverSolder

  • Contributor
  • Posts: 40
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #13 on: June 20, 2017, 09:53:43 PM »
With a switched front end, and a very accurate reference voltage it is actually possible to get rid of the 0.1Hz-10Hz noise, (though LTC6655 does not have too much) and the drift of the reference (the one which is driving the ADC).

So it might be enough with a single LTC6655 - as long as we have at least one "serious" reference in the system?  Which we read immediately before the measurement (and maybe immediately after, as well - and take an average)?   That sounds a little Keithley-alike, which is probably not a bad thing!
 

Online NANDBlog

  • Super Contributor
  • ***
  • Posts: 3625
  • Country: be
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #14 on: June 20, 2017, 10:08:13 PM »
That is probably enough for a simple gain adjustment. In a real situation, just like a DMM you want to do also a zero measurement. And the inverted reference, if you measure bipolar signals. Then, you can compensate for common mode voltages, if the system does not handle it well, and your measured values are not zero centered (most likely not). There are a number of parameters which can be compensated, but its up to the engineer to decide which ones are significant and which are not. 100dB CMMR, which is a lot, on an opamp could be enough for a 16 bit system, but not for the ppm level.
Keithley is doing this, because practically we have very accurate 100+ KSPS ADCs, but the passive parts are just not accurate enough. Its an economic system. Maybe there are resistor dividers in heated virgin oil baths, weighting 2kg, that can divide a voltage accurately, but its not economic to use them.
 

Offline MisterDiodes

  • Frequent Contributor
  • **
  • Posts: 316
  • Country: us
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #15 on: June 21, 2017, 12:32:16 AM »
Oh boy...a LTC250x != 3458a....

Those 32-bit ADC's are -NOT-, repeat NOT intended for absolute measure - they are not even close to 32 bits for that application.  These chips are intended for ratio-metric inputs only, i.e. resistance bridge weigh scale, where the same Vref for the ADC is also providing the excitation voltage for the bridge.  For best results at DC absolute use LTZ-based Vrefs and better yet multiple LTZ systems.  6655's are too noisy even in multiples unless your measurement period is very short and you don't need repeat-ability over longer time spans.

Talk to LTC and they are the first to admit that, and really they don't offer any solution for a 32-bit (or even 24 bit, really) direct Vref for DC measures.  This part could really use multiple LTZ's Vref for lower noise. OR build multiple complete VRefs and ADC's for cleaner data.  Quiet data at > 22 bits is not cheap and will require averaging across multiple independent systems.

32 bits is a flippin' joke when you actually run these, more like 20~23 bits or so even on LTZ.  3458a is down into the 28 bit range.  Not even in the same league.  Try it!

Another heads up:  Those great datasheet noise specs are at zero volts input - i.e. inputs shorted.  You won't get that on a non-zero level signal.  Good luck averaging out the crap also on a single chip. 

These will never be a replacement for a 3458a, and frankly - during our tests they aren't that great compared to a regular '2400 at DC.  As others have pointed out, the '2400 series has a very predictable and repeatable correction curve.  You can't do that with the newer fast 32 bits ADC's.  What you DO get is some speed of measurement for AC sigs of course.  And it's nice that some of the filter math is done for you on chip, but nothing you can't do in your own code.

But for fastest speed capabilities these need to be running under FPGA control because it's tough for a lot of CPU's to keep up with a 100MHz max serial datastream, head's up.

These have not proven to have much bang for the buck compared to '2400 / 2408 if you're talking about measuring DC absolute values.  For the cost of one '250x you could have multiple '240x running with excellent quiet data output at 22~23 bits.

Save the 2500 series chips for your ratio-metric application (weigh scales, pressure sensors, etc) is my (and LTC's) suggestion, use the 2400's for DC measures.   

 
The following users thanked this post: Mickle T., hwj-d, zhtoor, thermistor-guy

Online NANDBlog

  • Super Contributor
  • ***
  • Posts: 3625
  • Country: be
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #16 on: June 21, 2017, 01:11:30 AM »
The 3458A produces that 28 bit sample every 20 seconds. This produces that ~20 bit noise free bits every 10 us (if it is anything like the 2376, that I've thoroughly tested). For sure, it is not going to be exacly the same. But this is where it gets interesting: The 3458A is preatty much only limited by the LTZ1000 and the sampling speed. With this, you can build a system, which is limited by the LTZ1000 and its sampling speed will be multiples of the 3458A.
Personally, I think the 32 bit is a bit overkill. I had the 18 bit (and the 20 bit part after it was released) and they behaved very well in the system. There is 0.5ppm INL, that is 21bit. The change seems to be the built in digital filters, for the full utilization of the 23xx series, you had to read out every sample, and do the calculations yourself.
And yes, they are an upgrade to the LTC2400. That has 4ppm INL.
Of course both of them is still far from the 3458A linearity.
 

Online zhtoor

  • Regular Contributor
  • *
  • Posts: 212
  • Country: pk
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #17 on: June 21, 2017, 01:36:05 AM »
Oh boy...a LTC250x != 3458a....

so what would be a 3458a like dc-measurement solution for a poor man? looks like some kind of automated KVD? (if any exists?)
or a cutkowsky-like divider?

regards.
« Last Edit: June 21, 2017, 01:37:38 AM by zhtoor »
 

Online Echo88

  • Regular Contributor
  • *
  • Posts: 228
  • Country: de
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #18 on: June 21, 2017, 02:23:34 AM »
There is no low-cost-solution for substituting a 3458A, no matter how much Manganin-Wire or Discrete Transistors you throw at the problem.

Also: how would you measure your resistors which must be super precise/ultra low TC for Cutkowsky-Dividers/KVDs?
 

Offline SilverSolder

  • Contributor
  • Posts: 40
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #19 on: June 21, 2017, 03:14:20 AM »
Oh boy...a LTC250x != 3458a....

Even making a good 6 1/2 digit measurement is actually a reasonably challenging endeavour - don't breathe too hard...

But there is no harm in trying to make a "poor man's 3458a", it would certainly be very educational even if it "fails" by only working to 6.5 or 7.5 digits!
 
The following users thanked this post: MisterDiodes, zhtoor

Online zhtoor

  • Regular Contributor
  • *
  • Posts: 212
  • Country: pk
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #20 on: June 21, 2017, 03:23:16 AM »
There is no low-cost-solution for substituting a 3458A, no matter how much Manganin-Wire or Discrete Transistors you throw at the problem.

jim williams definitely threw transistors at the problem, especially the 2N2369 one  ;)

regards.
 

Offline MisterDiodes

  • Frequent Contributor
  • **
  • Posts: 316
  • Country: us
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #21 on: June 21, 2017, 03:24:00 AM »
This produces that ~20 bit noise free bits every 10 us (if it is anything like the 2376, that I've thoroughly tested)...

Once you actually use a 250x part, you're not really going to see that, honestly. Maybe 18 or 19 bits.  Probably like 2376.  But your Vref is modulating your input data, and at low ppm's that becomes more and more critical. As your signal increases dynamic range then the Vref noise becomes even more critical.  Then if you're running these 250x fast with an FPGA good luck keeping the that digital noise out of the faster measure (if you're chasing PPMs)...Uggh. It can be done, but you have to be careful.

Speaking of noise: Also notice that the AC specs (SNR, THD) of the '250x parts are pretty lackluster.  That's more like a 16 or 18 bitter.

An '2400 / 2408 has 4ppm INL but it is normally corrected out in software, and has been since the 80's. Linear even gives you the code for it, it's not like you have to re-invent the wheel.  You should get to ppm absolute or sub PPM relative measures on every part in a production run.  Very repeatable and stable correction curve is very similar even across parts - and is stable over time (decades).  We just use a 10 point test cal correction during manufacture and it's quick and easy.  You can get to ppm (or 20 bits) relatively easily, and if you keep your system very very quiet you can get into the 21~22 bit range.  No good way to correct a '2500 / '2508 long term, they seem to bounce around every 1kHrs or so.  We've been looking at them for the past year or so and feeding suggestions back to LT, but they seem a bit confused these days.  This seems to be a part that marketing wanted to have a datasheet for to compete with AD, but now they are the same company.  The will probably drop either the LT or the AD 32-bit part at some point, no need for two parts competeing against each other.

Again:  These '2500 serires are not meant to be a 32 bit, sub-PPM absolute measure device, and LinearT is the first to tell you that.  Use these when you need to measure a high-resolution (not necessarily with high absolute accuracy) ratiometric sensor.  Use a '2400 / '2404 / '2408 with a 6655 or LTZ's for accurate DC measuring at a MUCH better profit margin.

Zhtoor:  Metrology and "poor man's" anything are generally mutually exclusive.  You really need to understand that in the ppm world you get exactly what you pay for - and as you get below 10ppm the cost goes up more than exponentially every ppm below that.  By the time you're taking accurate measures under 8~5ppm absolute requires more than a single 3458a... And by the time you're at real, calibrated 2ppm (the realistic limit if you're not a jjarray-refered cal lab) you've got a several 3458a's & 732bs or equivalent, with some  thousands of $$$ per year in cal services alone.  IF you're doing it right.  And at that point you're talking a solid, noise free 19 bits measure no matter how you slice it.  To get to a real, calibrated 20 bit accurate absolute voltage measure (1 ppm accuracy) takes an extreme level of measurement ability, and really talking about a JJ-Array reference at that point - or a bank of 732b' refe'd to a JJ-array at regular intervals.

The other thing to understand is that very few applications require ppm-level anything unless it's VoltNuts entertainment or usually trying to calibrate / test something else. That means a very small market with no real profit center, and that means very expensive parts.  We use ppm-level test jigs on a laser diode semiconductor process line where each device in production sells for over $500 - but I can't think of a single instance where I'd need that in daily life at home.
« Last Edit: June 21, 2017, 03:36:20 AM by MisterDiodes »
 
The following users thanked this post: Echo88, zhtoor, baldwin1974

Offline MisterDiodes

  • Frequent Contributor
  • **
  • Posts: 316
  • Country: us
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #22 on: June 21, 2017, 03:48:26 AM »
Oh boy...a LTC250x != 3458a....

Even making a good 6 1/2 digit measurement is actually a reasonably challenging endeavour - don't breathe too hard...

But there is no harm in trying to make a "poor man's 3458a", it would certainly be very educational even if it "fails" by only working to 6.5 or 7.5 digits!

Yup.  You can certainly make something -close- to a 6 digit meter with off the shelf parts - but it'll be at one (or very limited) input voltage range, probably won't work for AC, and compared to a real piece of equipment it'll still basically be a piece of crap (sorry).  A 3456a / 3458a takes man-centuries of engineering time to develop, and you're never going to replace that with a home brew PC board and a bag of parts from DigiKey.   

Educational value for making you own?  Priceless. You will find out exactly how hard it is to replicate even a 6 digit meter.  And then you'll realize what a good value there is in real quality, well designed equipment.

For an actual piece of equipment you'd use in the lab:  A older piece of equipment that is calibrated, working, documented and well-aged is even more priceless and useful.
 
The following users thanked this post: zhtoor

Online Cerebus

  • Super Contributor
  • ***
  • Posts: 1916
  • Country: gb
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #23 on: June 21, 2017, 04:42:08 AM »
±0.5ppm INL (Typ) and ±2ppm INL (Min/Max) is not that impressive. Even the good old LTC2400 delivers 4ppm INL and 0.3ppm RMS noise but with a predictable and therefor adjustable error. Thus the error can be adjusted to <1ppm. LTC2500-32 seem to have not that easy adjustable INL error.
But maybe TiN can prove us that I'm wrong.

What's more revealing is if you convert those figures into counts (~4300 and ~17000 respectively) and think about what this really means in the context of a "32-bit" ADC. (Hint: log2(4300) = 12 and log2(17000) = 14).
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline SilverSolder

  • Contributor
  • Posts: 40
Re: 32-bit ADC playground for precision measurement tasks.
« Reply #24 on: June 21, 2017, 06:57:12 AM »
For an actual piece of equipment you'd use in the lab:  A older piece of equipment that is calibrated, working, documented and well-aged is even more priceless and useful.

If you need an extra digit on your DMM...   is it realistic to try to build a very high quality x10 amplifier to put in front of a good 5 or 6 digit meter?
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf