Author Topic: Deviation MCP3914 ADC  (Read 1893 times)

0 Members and 1 Guest are viewing this topic.

Offline FreddyTopic starter

  • Newbie
  • Posts: 4
  • Country: nl
Deviation MCP3914 ADC
« on: November 25, 2019, 09:23:51 pm »
I am developing a measuring instrument that contains 5 MCP3914s. This is an 8 channel 24 bit ADC. So I have in total 40 AD converters. Now all 40 ADCs measure approximatelly 1.3 times less on all gain settings. I have used the smaller brother the MCP3911 (2 channel 24 bit ADC) more often without problems.

The analog and digital supply voltages are OK: 3.3 volts and no significant noise (<1 mV). The reference voltage is also in order 1.2 V without significant noise.

With a differential voltage of 30 mV on an input (neatly symmetrical around zero, gain on 1x) I would expect the code (30mv / 600mV) * 2 ^ 23 = 419431, but it measures around 322240. This is confirmed by looking at the SPI stream on a oscilloscope.

Another strange thing is that het maximum code 8388607 is reached at an differential input voltage of 770 mV, but i expect that the maximum would be 600 mV corresponding to te reference voltage of 1.2 V.

After initiaslising the configuration registers read as follows:

0x09 : 0x000000
0x0A : 0x000000
0x0B : 0x000000
0x0C : 0xB90000
0x0D : 0xFB6050
0x0E : 0x000040
0x0F : 0xFFE93A
0x10 : 0x000000
0x11 : 0xFFE825
0x12 : 0x000000
0x13 : 0xFFE5AE
0x14 : 0x000000
0x15 : 0xFFE762
0x16 : 0x000000
0x17 : 0xFFEA61
0x18 : 0x000000
0x19 : 0xFFEA1B
0x1A : 0x000000
0x1B : 0xFFECDA
0x1C : 0x000000
0x1D : 0xFFEFE1
0x1E : 0x000000
0x1F : 0xA50000

As you can see the offset calibration is set and used, and the digital gain calibration is set at x1 for now. (Registers 0x0F and higher.)

Does anyone know what can cause the times 1.3 deviation?
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 15154
  • Country: de
Re: Deviation MCP3914 ADC
« Reply #1 on: November 25, 2019, 09:33:59 pm »
Those ADCs need a relatively not impedance source, as the sampling is short. With too much impedance the result can be too low, though a factor of 1.3 is a lot.
 

Offline uer166

  • Super Contributor
  • ***
  • Posts: 1026
  • Country: us
Re: Deviation MCP3914 ADC
« Reply #2 on: November 25, 2019, 09:40:52 pm »
Can you post front-end/filtering circuit? I am using those too for some metrology and interested in ADC's pitfalls etc..
 

Offline FreddyTopic starter

  • Newbie
  • Posts: 4
  • Country: nl
Re: Deviation MCP3914 ADC
« Reply #3 on: November 25, 2019, 10:13:23 pm »
Here is a part of the analog front-end. Yes, the 10 k impedance is a little high, but with a ADC differential impedance of 232 k at a gain setting of x1 is doesn't explain this large deviation. The input circuit is not much different when I used the MCP3911, and that one worked fine.
« Last Edit: November 25, 2019, 10:22:15 pm by Freddy »
 

Online iMo

  • Super Contributor
  • ***
  • Posts: 5570
  • Country: va
Re: Deviation MCP3914 ADC
« Reply #4 on: November 25, 2019, 10:23:26 pm »
How do you work with those 30mV in the above schematics then? Do you measure 30mV at the 10k? With all that 5Meg resistors wired in?
« Last Edit: November 25, 2019, 10:26:27 pm by imo »
Readers discretion is advised..
 

Offline FreddyTopic starter

  • Newbie
  • Posts: 4
  • Country: nl
Re: Deviation MCP3914 ADC
« Reply #5 on: November 25, 2019, 10:34:32 pm »
The 30 mV is measured with a multimeter directly across a differential input pair eg. between CH5- and CH5+. An insulated from the circuit 30 V power supply was connected between inputs V13 and V14.
 

Offline uer166

  • Super Contributor
  • ***
  • Posts: 1026
  • Country: us
Re: Deviation MCP3914 ADC
« Reply #6 on: November 25, 2019, 11:53:23 pm »
Do you measure both with DMM and ADC at same time and compare? The DMM will load the front-end with its' own impedance  (10meg or something?), reducing the voltage seen by ADC a bunch.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Deviation MCP3914 ADC
« Reply #7 on: November 26, 2019, 02:24:54 am »
The 30 mV is measured with a multimeter directly across a differential input pair eg. between CH5- and CH5+. An insulated from the circuit 30 V power supply was connected between inputs V13 and V14.

Hmm, do you know if the power supply may be introducing AC noise (most likely, common mode RF) this way?  Shouldn't be, since it's behind the filters anyway...

Huh, also, why double filter the signals, why not share each consecutive +/-?
...Does the converter load the inputs differentially at all?
(Does help for probing individual channels, though.)


Do you measure both with DMM and ADC at same time and compare? The DMM will load the front-end with its' own impedance  (10meg or something?), reducing the voltage seen by ADC a bunch.

Nahh, out of a ~20k Thevenin source, that can only account for a ~0.2% change.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Online iMo

  • Super Contributor
  • ***
  • Posts: 5570
  • Country: va
Re: Deviation MCP3914 ADC
« Reply #8 on: November 26, 2019, 09:39:52 am »
Doublecheck the ADC settings. It could be the 3914 has got some different config bits there against 3911.
Also they use "1.5" coefficient in the math at several places in the DS.
Readers discretion is advised..
 
The following users thanked this post: Freddy

Offline FreddyTopic starter

  • Newbie
  • Posts: 4
  • Country: nl
Re: Deviation MCP3914 ADC
« Reply #9 on: November 26, 2019, 11:14:37 am »
First I have done some double check measurements, now with four 9 volt battery (36.44 V) in serie to avoid common mode noise. The ADC outputs 391690 without multimeter load. With MM load on the differential input it outputs 390970 (little difference) and the MM reads 36.0 mV.

I'm not sure if I read the question about the filter configuration right, but if I connected the CH- and CH+ of neighboring channels to each other (and use a single filter for the common node) then neighboring channels will influence each other because there is a gain depending input impedance of the ADC.

imo, yes, that must be it, I totaly forgot the 1.5 factor in my calculations. With this in mind the ADC result lays within the gain error margin listed in the datasheet. And yes, i'm aware of the differences between the registers of the 3911 and 3914.

Thank you all for looking with me at this problem.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf