OSHW - 24bit ADC measurement system for voltage references

[branadic]

Almost every CAD for printed circuit boards is able to import Eagle Files.

What can you do with the schematics in PDF format? Almost nothing, just viewing it. Please don't repeat this discussion we already had before. If you don't have Eagle or similar CAD software with import function it's not my problem.

[rs20]

Sorry, but think: phone, tablet, my locked down work computer, internet cafe, borrowed devices. Every one of my devices except my home PC, the only one which I have complete dominion over. All places which have no CAD software, and all support PDFs near-natively. And viewing a schematic is all that 90% of people well ever want to do. You're free to keep your design behind a proprietary wall, just don't pretend that it's sensible or in the spirit of a forum.

To be clear, if you had just said "no", I'd have not said anything. But you said "no, it's easy for you to use what I've provided". And I have to point out that this is massively, massively untrue for many people.

[branadic]

And viewing a schematic is all that 90% of people well ever want to do.

And I'm not adressing exactly those 90% of people

Some design parameters of the pcb to be ordered:

- size: 83 x 50 mm²
- thickness: 1.55mm
- smallest width: 0.2mm (due to guard at reference noise reduction pin)
- smallest via: 0.4mm
- finish: Ni/Au
- no silk screen
- stop mask: green on both sides

[timb]

As an FYI, I would be in the target audience for this device, however I browse the forum (and do research, etc.) almost exclusively on my tablet. Further, I don't use Eagle for design work. If you can't be bothered to provide a PDF, or at the *very least* a PNG of the schematic, then I'm not going to bother with your device.

PDF is the de facto standard for distributing documents on the Internet. If I have to install Eagle on my main computer, just to view your schematic, then I'm not going to even consider your device. Period. What it tells me is you're either lazy or don't have any attention for detail. All you have to do is download Ghostscript... (Or one of the other numerous PDF printers freely available, assuming you're on Windows; if you're on OS X you can natively print to a PDF, so there's no excuse there.)

Sorry for being blunt, but that's just the way things are. (This would be like TI distributing datasheets exclusively in Word or PowerPoint formats. Can you imagine?)

[branadic]

The former version with LT1236LS8 has already been shown. Is it really that complicate for you to imagine that LT1236LS8 was replaced by LT1027LS8 and therefore minor changes on the layout had to be performed? The complete rest is identical to the former version.
If your answer is yes, than you really don't need the pcb, trust me and you are one of the 90% of people out there that don't need a 24bit ADC measurement system for voltage references either.

[timb]

I actually would be in the market for something like this, but since you can't be bothered to keep your documents current, I won't bother either.

[TiN]

It takes 5 seconds to export PDF and attach to files, and saves all that time wasted with valid and invalid complains. If every engineer will post their hobby stuff only in software they used, discussion would be less open and wide. Viewing schematics is all that most of interested folks need, as those who will bother building something likely to use their own chassis, own PCB layout, own components set they want anyway, and PDF-schematics is more than enough to test your idea.

Almost every CAD for printed circuit boards is able to import Eagle Files.

3000 EUR OrCAD can do nothing with Eagle files. Yes, it's not your problem.

I was remotely interested as just got few LTC6655's and ADCs not long time ago and going to bodge little PCB for few tests into my order, so hence came to this thread to check out on similar design ideas.. Baa , can't see files.

[AlfBaz]

Hi Branadic, first let me thank you for posting this and Andreas for allowing it.
I'm in the process of duplicating the circuit using altium, roughly copying your PCB layout and routing.

I'm sure this circuit has been tweaked for minimum noise and as I work through the design I see some of the choices you've made but I still have a few questions

The relatively high series resistance to the optos seems to be to round off the edges on the pulses and the even higher resistances (22k and 47k) to allow the pic to be programmed
The one part I'm stuck on is why the parallel 2.2nF cap across the 3k3 resistor from the PIC to the opto. This seems to consume a large spike of current from VDD on transitions given the typical 10ns rise and fall times, am I missing something here?

B1 and B2, are they ferrite beads? If so i'm curious which frequency you chose 70, 100MHz or something else

The choice of micro, was it based on its low current consumption and adc for the thermistor? Did you create a software UART  and SPI for the interface to the ftdi usb and LTC2400 or is it just some rudimentary bit banging? I ask this because the PIC you have chosen is quite low current and don't know if I'll find anything lower

When you run this in conjunction with the 2-1 divider where do you power the whole lot from, DB9 plug or from the JP1 header on the LTC2400 board?

And finally after you tested did you end up cutting out the LTC2400's mounting square and go dead bug like Andreas did. In other words, did you end up noticing the effects of humidity
 

[Andreas]

The one part I'm stuck on is why the parallel 2.2nF cap across the 3k3 resistor from the PIC to the opto.

And finally after you tested did you end up cutting out the LTC2400's mounting square and go dead bug like Andreas did. In other words, did you end up noticing the effects of humidity

Hello,

intention of the 2.2nF is to speed up the rise time at relatively low power (~1mA).
Otherwise 9600 Baud are marginal with PC817 and 1 mA.

The LTC2400 does not need decoupling from the PCB. (Self adjustment during each measurement).
Only the voltage reference shows a dependancy from PCB stress and should be decoupled.

with best regards

Andreas

[branadic]

Hi,

B1 and B2 were choosen to be 600ohms at 100MHz, if I remember correct. Micro was choosen by Andreas because of battery operation and thus low power consumption required.

Power is always supplied via JP1, DB9 powers additional modules connected to it.

I didn't test the ADC to excess as Andreas did, however I have the reference still mounted directly onto pcb but NTC was changed to 0402 type and this way is much closer to the reference compared to through hole NTCs.
What I wanted to try is to place reference on a flexboard, which is then soldered to the board giving some strain-relief. Since real semi rigid pcb's are barely expensive and dead bug style is not a good production process I thought this tradeoff should give best results to keep board stress due to humidity away from the reference. But haven't tried that yet as some other projects came into focus.

-branadic-

[AlfBaz]

The LTC2400 does not need decoupling from the PCB. (Self adjustment during each measurement).
Only the voltage reference shows a dependancy from PCB stress and should be decoupled.

My mistake, LTC2400 on the brain, meant LT1236

Thank you gentlemen, about to order parts and PCB, will post pics when complete. Lots of measurement fun ahead

Just one more thing, is there any reason to go without solder mask? I seem to recall seeing you guys post pictures of your various pcb's without it and always assumed it was an artifact of an in house manufacturing process. I see Dr Frank's LTZ1000 boards I got from you Branadic have it though

Regards, Alf

[branadic]

Andreas made is first boards at home and I once had a cheap and fast german pcb manufaturer with the only drawback of having no solder stop mask nor silk screen. Unfortunately he's no longer in business.

-branadic-

[AlfBaz]

Ok thanks

[Andreas]

Hello,

Thanks for the summary.
But noise is not all.

For a absolute measurement the offset and gain drift and INL is also important.

That is why I love the LTC2400:
Offset + Gain drift is self adjusted.
INL can be easily modeled and canceled out to below 1 ppm.

with best regards

Andreas

[chickenHeadKnob]

2 of the interesting ADC in the list is ADS125H and ADS8699. it has built in DMM gain scale.
i have also included some relative levels of 6.5D MM noise.

there is a chance i made some fundamental errors ... i have not seen any ADC being able to oversample to hell at nV level of noise ... pls feel free to highlight any corrections.
to note : it would be good to view the chart with lesser importance on rank but to analyze which sampling speed to use to maximise brute force lowering of noise

Thank you for the work you put into this, My first impression is that you have been overly optimistic in two dimensions. Both in using the data-sheet specmanship Enob of the ADC's as well as relying on  getting more than 2 extra bits from averaging. Two extra bits with simple averaging in the presence of gaussian noise, easy. The next bits progressively more difficult, then you will hit some system noise floor that becomes an insurmountable limit. 

The ADS8699/8689 is indeed interesting, I have been planning a design for a year now and  I need to get off my ass. One complicating factor is the bandwidth of the analogue front end. From the data sheet (page 25):
 In order to mitigate the noise of the front-end amplifier and gain resistors of the PGA,the AFE circuit of the device features a second-order,anti-aliasing LPF at the output of the PGA.The magnitude and phase response of the analog anti-aliasing filter are shown in Figure51 and Figure52, respectively.For maximum performance,the –3-dBcutoff frequency for the anti-aliasing filter is typically set to 15 kHz.The performance of the filter is consistent across all input ranges supported by the ADC.

 I read this as you won't get near nyquist performance suggested by the sample rate if you are simply brute force adding these puppies in parallel.

[jeffheath]

I found this thread, files were eagle cad... so I installed eagle cad.... now I can open eagle cad files! It's not that hard...