If you short pins 5 & 6 on the LTC2440, what does the digital output look like?

If you apply 1/2 vref to in+ (pin 5), then what does it look like?

Oh, and what is your conversion rate?

Joey, you own this thread. Rock on.

 

I just used a ADS1220 24-bit sigma-delta converter in a project to deliver 20-bits of needed resolution to measure solar panel power in the microwatt range during moonlight with great stability from a very noisy signal source so it should be possible to use one for your application. Use the information the vendor provides to calculate your design requirements with their device.
TI has very good information on how to configure its products for proper operation that can usually be used in general with other devices.
https://github.com/nsaspook/nidaq700/blob/master/supermoon/ads1220-556792.pdf

http://www.edn.com/design/analog/44...igma-delta-ADCs---A-non-mathematical-approach

 

The quantity of disinformation one can glean from this site is incredible!

Not to mention outright horsepucky-- like the notion that sigma-delta ADCs somehow aren't "appropriate" for this kind of application.

Huh? Look at the datasheet for any commercially available sigma-delta ADC-- whether from Analog Devices, Texas Instruments, Burr-Brown, Linear Technology, or Microchip-- or any of the relevant application notes, and what sorts of application do you see given as examples? You see exactly the applications these converters were designed for: load cell signal processing. It's what they're designed for, and what they've been used for, thousands and thousands of times, and they've been around for years. As converters go, nothing else comes close.

The LTC2440 is a superb ADC. There are better converters-- joeyd999 knows all the very best ones-- but the LTC2440 is damn good and if you're not getting the performance from it that's specified on the datasheet, you're doing something wrong.

Check power supply noise. Reference voltage noise. Input amplifier noise. And digital-analog signal crosstalk. It's gotta be something. Whatever the problem is, I guarantee you it is absolutely, positively NOT the LTC2440.

 

Thank you for the information on sigma-delta. It's been years since I looked at them.

I stand corrected.

 

Thank you for the information on sigma-delta. It's been years since I looked at them.

The early ones, say around 20 years ago, weren't anywhere near as good as the ones that are available today, which are definitely worth looking at for anything other than high-speed applications.

 

Some 19.5 SR FFT data from the moonlight 24-bit sigma-delta detector with the diff inputs shorted and with a 4.7k resistor on the inputs. Gain 4X with 50/60Hz internal filter, internal gain block off. For the moon data collection I only generated samples every 3 seconds so I used a software averaging filter on the raw ADC output to smooth data for that. Not bad for something built on a vector board.

I modified the logger program to generate SIGVIEW ascii import files of 3000 samples per file with no processing. X time in seconds, Y value in microvolts
http://www.sigview.com/ 21 day eval.

Input shorted for noise floor for ~10Hz . Top two RAW signal, bottom two smoothed, signal levels in micro-volts.

The 4.7K load resistor generates a small DC offset.

Re-sampled noise signal spectrum:

 

Joey, you own this thread. Rock on.

I'm looking forward to what he has to add to what he's already said... sometimes I wonder why Joey's not charging this site for posting his opinions...

 

Here are pictures of the new adc, it uses one amp to get the reference to 5V and another to scale the sensed voltage into the 0-5V range for the ADC. I tried getting more decoupling on the power rails and better low pass filtering on the reference and input, but to no avail. Plus the datasheet from the original adc. http://forum.allaboutcircuits.com/attachments/db5601_e-pdf.56381/

Here's a couple of cents coming from a rather sub-amateur-level member of this forum, so anyone please feel free to correct me if my opinion is off the mark:
PCB layout and tracing is critical when working with high resolution ADC. Just taking a glimpse at your layout told me this: never run traces under an ADC chip if you can avoid it, especially if those traces belong to the digital part of the ADC. Clock, sync and digital I/O signals can induce a significant amount of noise in the chip and they should be traced in opposite directions from the analog part of the circuit.

 

I'm looking forward to what he has to add to what he's already said... sometimes I wonder why Joey's not charging this site for posting his opinions...

I pay for it -- every time one of my profoundly insightful political posts gets deleted or a thread that just hit the proper tempo gets closed...

 

I pay for it -- every time one of my profoundly insightful political posts gets deleted or a thread that just hit the proper tempo gets closed...

Maybe one of the few members subscribed to an additional assured deletion service.

 

Here's a couple of cents coming from a rather sub-amateur-level member of this forum, so anyone please feel free to correct me if my opinion is off the mark:
PCB layout and tracing is critical when working with high resolution ADC. Just by taking a glimpse at your layout told me this: never run traces under an ADC chip if you can avoid it, especially if those traces belong to the digital part of the ADC. Clock, sync and digital I/O signals can induce a significant amount of noise in the chip and they should be traced in opposite directions from the analog part of the circuit.

I think that is what I was trying to do, but it is a bit hard on a two layer pcb.
Still dint´t have the time to do any measurements, but I think I am going to split this into two boards, one with the opamps and one with the ADC and CPU which will likely be a four layer board.

 

...I think I am going to split this into two boards, one with the opamps and one with the ADC and CPU which will likely be a four layer board.

I strongly suggest doing it the other way around: keep the ADC and opamps together on one board with a good clean layout (i.e., no traces under the ADC chip except a solid ground pour, and get the ADC's SCK, SDI, SDO and CS- signals as far away from the analog stuff as possible) and put the CPU and other noisy stuff on another board. By the way, you're NOT using a switching regulator to power this stuff, are you? I hope not...

 

No, there are original linear regs that I am using, both for the +/-15V and the +5V.
Thanks, that seems like a logical advice to keep it that way. But first I will do some more testing on the board I have and decide from that what to do better on the next revision.

 

One change to consider might be replacing that LM399 voltage reference with something a bit quieter. Although the LM399 is very temperature stable due to its on-chip temperature regulator, it is very noisy, with up to 50 μV rms noise, which translates to 300 μV peak-to-peak. A much better replacement would be an ADR445 (http://www.analog.com/media/en/technical-documentation/data-sheets/ADR440_441_443_444_445.pdf), which has about 1/10th the noise.

 

One change to consider might be replacing that LM399 voltage reference with something a bit quieter. Although the LM399 is very temperature stable due to its on-chip temperature regulator, it is very noisy, with up to 50 μV rms noise, which translates to 300 μV peak-to-peak. A much better replacement would be an ADR445 (http://www.analog.com/media/en/technical-documentation/data-sheets/ADR440_441_443_444_445.pdf), which has about 1/10th the noise.

Yes, but you just traded temperature stability for noise performance. For a precision balance, DC accuracy is preferable to low AC noise. SNR can always be increased by taking more samples, but DC errors are unrecoverable.

 

Yes, but you just traded temperature stability for noise performance. For a precision balance, DC accuracy is preferable to low AC noise. SNR can always be increased by taking more samples, but DC errors are unrecoverable.

A good point, but I'm not seeing a big difference in temperature stability between the two: 2 ppm/°C max. for the LM399H, vs. 3 ppm/°C max for the ADR445B. They're both pretty darned good.

 

A good point, but I'm not seeing a big difference in temperature stability between the two: 2 ppm/°C max. for the LM399H, vs. 3 ppm/°C max for the ADR445B. They're both pretty darned good.

There is a difference, but not as much as I thought. I didn't dig deep. I was looking at 0.5 ppm vs. 3 ppm at the top of the datasheets.

 

I pay for it -- every time one of my profoundly insightful political posts gets deleted or a thread that just hit the proper tempo gets closed...

And... that's the price of playing with fire, my friend.