0

I want to use the ADS8317 16-Bit SAR A/D converter and have read through the datasheet. In there, ti recommend using a low ESR 47µF tantalum capacitor to buffer the reference input. Now, my reference is already buffered with a 220µF low ESR aluminum electrolytic capacitor (Panasonic FR 220µF/16V; 130mOhm ESR at 100kHz), and I'm wondering whether that's enough, or if the tantalum has some important advantage.

As far as I can tell, the important factors should be that the capacitor is large enough and "fast" enough to charge the internal capacitors of the ADC to 16 bit accuracy within the required time, similar to the signal inputs. However, I think the reference input might see a charge transfer for every bit of the SAR instead of only once during the acquisition time - if that is the case, the charge transfer would have to be faster. The smallest time I can imagine for this would be half a clock period - 500ns in my case.

Ok, but - is my aluminum electrolytic capacitor up to the task (perhaps assisted by a smaller ceramic right next to the input)? As far as I know they are not good for "high frequency" applications, but I don't know any actual numbers.

Medo42
  • 607
  • 5
  • 17
  • 2
    Why such a large buffer? – Ignacio Vazquez-Abrams May 21 '15 at 16:00
  • 1
    Worth reading: http://electronics.stackexchange.com/questions/99320/are-tantalum-capacitors-safe-for-use-in-new-designs – ilkhd May 21 '15 at 16:22
  • Also aluminum caps "dry out". Need replacement every 10 years. – ilkhd May 21 '15 at 16:25
  • The buffer was chosen so large because I'm sometimes switching a 330pF capacitor back and forth between that reference and another voltage elsewhere, and I wanted the voltage change from that to be negligible. I could use an op amp buffer instead, but a larger cap is cheaper than a precision op amp :) – Medo42 May 21 '15 at 17:01
  • 1
    First of all, ask TI. They have application engineers to help with these things. There was a time around 1999 or 2000 when tantalum became un-obtainable. Personally, I would not ever use tantalum for that reason (in mass production). If you only need 47 uF, I would use ceramic (Stick with X5R or X7R dielectric). You can add series resistance to mimic tantalum, if the series resistance is required for stability. Electrolytic capacitors have a limited lifetime. But if you really need 220 uF, then I guess electrolytic is your only reasonable choice. – user57037 May 21 '15 at 18:30
  • 330p/47u = 7ppm. Isn't that negligible enough? Just use 47uF of X5R or X7R capacitance. You may need to use 2 x 22uF. Add a zero Ohm resistor at the output of the op-amp buffer to help stabilize it if needed. – user57037 May 21 '15 at 18:49
  • @mkeith, you're right, using a ceramic cap is probably a good solution. The only downside I can see is the potential for microphonic effects adding noise, and now I worry about that... argh, I'm too paranoid for this job ;). Couldn't I also just drive the Ref input the same way as a signal input, with a fast op amp and a small RC filter? – Medo42 May 22 '15 at 11:36
  • I don't know anything about microphonics in ceramic capacitors. I think what you are trying to do is create a stable, low noise voltage reference, right? The caps seem like a perfectly good way to do it. A high bandwidth amp could do it in theory, provided you create a stable reference for the amp. But it doesn't seem necessary. Might not even be as good as what you have. I suggest you contact TI and say "can I use ceramic caps instead of tantalum?" And also google microphonics in ceramic caps. – user57037 May 23 '15 at 03:08

1 Answers1

1

They don't appear to be much different electrically.

Tantalums are physically smaller and SMT, which is a big advantage in some applications.

I think you'll be fine from the pov of the ADC if you use that, perhaps with 1uF ceramic X7R in parallel close to the ADC.

Minor nit- I would not call a bypass capacitor a 'buffer'. The OPA350 amplifier is the buffer.

Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842