DDC11XEVM-PDK: DDC112 EVM application software v4.0.16 data questions

I'd also like to know the minimum value for the external integration capacitors on the DDC112.

Hi Philip,

• Each Input channel has 2 integrating sides (i.e. 4 integrators in total). While one side (two integrators, one on channel 1 and one on channel 2) are integrating 1A and 2A, the other pair is providing 1B and 2B data to the ADC. Basically A and B are alternating samples of the same channel.
• As for displaying more points, I believe this is an EVM buffer limitation
• As for the min value of the external integration cap, the "typical" value in Table 1 is 12.5 pC. If you're going to use 12.5 then you should just use Range 1. When an external cap is use it is to increase the range, not the sensitivity. Since there is no DS minimum, the realistic minimum you could use would have to be able to live with the parasitics of the device + board. 
• Why do you want to use a small external cap? If you could explain or describe your setup, I may be able to provide further help.

Matt,

Please see my responses in red  below.

• Each Input channel has 2 integrating sides (i.e. 4 integrators in total). While one side (two integrators, one on channel 1 and one on channel 2) are integrating 1A and 2A, the other pair is providing 1B and 2B data to the ADC. Basically A and B are alternating samples of the same channel.

From the DDC112 data sheet, that is what I expect.  However looking at the data from the EVM data file that doesn't appear to be what is happening.  I have configured the EVM with 1000 pF capacitors connected to both the IN1 and IN2 inputs, with the other terminal of the capacitors grounded (1000 pF cap across P3 and P4  and another 1000 pF cap across P11 and P12 on the EVM).  I then configure the DDC112 for 1 mS sample rate and external integration capacitor (RANGE2 = RANGE1 = RANGE0 = 0).  I acquire the data and save to a .csv file.  I've attached graphs of the data as it's saved to the data file, in other words the first 512 samples are channel 2A, the next 512 samples are channel 1A, then 2B, and finally 1B.  As you can see channels 2A and 1A have similar average offset values as do channels 2B and 2A.  I'm consistently seeing this behavior.  From this data it looks to me like channels 2A and 1A are a pair, and channels 2B and 1B are a pair.  Can you please confirm on your EVM if you are seeing similar behavior?

chart1.pdf

chart2.pdf

• As for displaying more points, I believe this is an EVM buffer limitation

I expect that the EVM is using the onboard SRAM chip for the buffer.  The SRAM device is a 16 Mbit device (2MB).  Four channels of 20-bit data requires 49152 bits (assuming 24-bit storage type for each sample).  There is enough room the SRAM to store ~175000 24-bit samples @ four channels.  It seems to me that the buffer space on the EVM is much larger than 512 samples.  Can you please check with the EVM developers to find out if it is possible to increase the buffer size?  I'd like to get 8K points/channel or greater.

• As for the min value of the external integration cap, the "typical" value in Table 1 is 12.5 pC. If you're going to use 12.5 then you should just use Range 1. When an external cap is use it is to increase the range, not the sensitivity. Since there is no DS minimum, the realistic minimum you could use would have to be able to live with the parasitics of the device + board. 
• Why do you want to use a small external cap? If you could explain or describe your setup, I may be able to provide further help.

I am investigating connecting a sensor to the DDC112 which sources ~1 nA of current.  I need to amplify the signal by ~4.6e+9 gain to attain a full-scale signal to a 5Vpp D/A.  To achieve this level of amplification I've calculated that I need to use a 270 fF integration cap on the DDC112.  Since caps of such low value are not readily available I'm testing with 1 pF caps.

With traditional op amps I am seeing a lot of 1/f noise in low frequencies.  I expect that the switched capacitor integration design of the DDC112 will have less 1/f noise.  I am looking at the DDC112 EVM to test that theory.

Matt,

Please ignore my first response above about channel designations.  I've done some more testing, this time with an active signal and I see that 1A and 1B are from the same input, as are 2A and 2B.

This testing has brought to light more questions on the DDC112 behavior.  I'll update this case when I've had a chance to look at the data more closely.

-phil

I've successfully captured data from our target sensor using the DDC112 EVM, so I think I understand how to do that now.  The biggest issue I have now is the limited sample size.  Can you please help me increase the sample size captured and stored to the .csv file?  I'm looking for something around 32k samples, much larger than the current 512 sample limit.

Thanks,

-phil

Matt,

Thank you for figuring this out, it works perfectly.  It seems to me that there is a 16K sample limit which while not exactly what I need I should be able to work with that.

In my previous life as a TI apps engineer I used to hate the RTFM questions from customers so thank you for putting in the effort to answer this one.  In my defense I found the documentation a bit misleading.  The document states "nDVALID Read: This is the number of nDVALID pulses after which to capture data."  I understood the word after to mean that this was a delay AFTER which the sampling would begin.

If SLAU234A is revised in the future I'd suggest that this statement be reworded to something like "nDVALID Read: This is the number of data samples to capture on each channel."  It would also be good to include valid limits (min and max values) for this and the other register settings.

Thanks,

-phil