Other Parts Discussed in Thread: ADS8866, REF5050, TIPD211, ADS8900B, OPA836
I'm hoping to use two daisy-chains of ADS8866 ICs in a multichannel simultaneous sampling system running at 50kHz. I currently have over twenty ADCs in the longer of the two chains, and I'm worried about the aggregate capacitve load on the REF pins, which are driven by a single REF6050 (i.e. one REF6050 per chain).
Both ADS8866 and REF6050 data sheets recommend a 22uF to 47uF capacitor on the ADC REF pin, and the REF6050 data sheet sepcifies a maximum of 47uF for stability. Obviously with >20 ADCs with 47uF each I have over an order of magnitude more capacitance than this.
My questions are:
Has anyone ever tried using a single reference IC with this many ADCs? (I note the TIPD211 reference design has a single REF5050 driving four ADS8910s with 10uF each, but I wouldn't want to extraploate very far from that).
If not, is there something I've missed that means it will work fine and I'm worrying about nothing?
If not, is there a cheap and easy way to fix it?
There is currently nothing but ~100mm of PCB track between the reference output and the ADCs' REF input. I considered adding a series resistor at each ADC, with the ADC REF pin connected on the capacitor side, but even one ohm creates a low-pass corner well below the sampling rate.
My calculations suggest that 47uF is quite generous. Assuming the typical REF current of 35uA is drawn for 10us from a 47uF capacitor:
Q=CV, Q=It -> CV=It -> V=It/C -> V=(35u*10u)/47u=7.447uV
For a 5V reference that's a relative error of 1.489u, or 2**-19.3. To my eye this is better than needed for a 16b ADC, and the REF6050 data sheet states it is good enough for 18b. I could probably live with 13b, so I could reduce every capacitor value to keep the total under 47uF, perhaps putting larger capacitors on the ADCs handling critical signals and smaller ones on those just monitoring quasi-DC levels.
All suggestions welcome.
