1

If I were making an audio ADC based upon the PCM1804 for example , what would be the impact on the output audio from the PCM1804 if I used an audio op amp like the OPA1632?

What exactly does the op amp do to the audio signal ? I realise it will operationally amplify I.e amplify the difference in voltage between 2 points, but what will this mean for the input signal to the PCM1804?

If anyone can actually explain what happens if I connect the OPA1632 to the PCM1804 instead of not bothering with the OPA1632 it would help a lot! I would hope any answers would explain the technical differences, then conclude which option would result in the best audio signal explaining why.

Thank you!

RenegadeAndy
  • 444
  • 6
  • 14

1 Answers1

8

Your Analog Front End is a very important tool in getting quality conversion results from your ADC

There are four basic problems with connecting an ADC input directly to the "outside world":

  1. ADC inputs are relatively fragile (only capable of handling a few volts DC without damage), and ADC chips can be expensive/hard-to-replace (several dollars and in fine pitch SMT or BGA)
  2. ADC inputs have a relatively low (tens of kOhms) and variable (due to the internal sample and hold frontend stage) input impedance, which isn't compatible with the expectations of an outside world input (which needs to be at a high, fixed impedance in order to avoid losing signal voltage-divider-style)
  3. ADCs need an anti-aliasing filter to keep high frequencies from being undersampled and thus aliased into the desired passband, as aliases cannot be rejected in the digital domain
  4. Normal real-world signal levels often don't make good use of the ADC input's dynamic range -- sensor outputs and even line-level signals may only use a fraction of the input voltage range, wasting ADC bits and reference accuracy you paid dearly for.

An analog front end stage (AFE) is needed to solve all of these problems. At its simplest, an AFE can be a simple op amp integrator or low-pass filter (LPF) stage (a single low pass pole for anti-aliasing) with some fixed passband gain or attenuation to match the input signal level to the ADC's. More complex AFEs use multiple op amps to better separate gain and filter functions, reject other unwanted signals (such as DC offsets or infrasound in some audio applications), add extra LPF poles (to allow operation of the ADC closer to its Nyquist bandwith of half the sample rate), or interface to more complex inputs (such as balanced inputs, sensor bridges, etal -- think of a mic preamp or line receiver stage, for the audio world). All of this adds up to better signal quality -- you don't have your local AM talk radio station bleeding into your favorite song, loud popping noises from DC offsets interruping the program, or your volume up to 11 just to hear anything at all.

A simple (perhaps even pre-Muntzed, if you will) AFE for your PCM1804 is depicted below -- note that the OPAx134 is designed for dual supply operation, so an op-amp with better common-mode range and output swing than the OPAx134 should be favored for this job, given your supply voltage constraints. R1 and R2 are 1% resistors, while C1 is a film or C0G type and a tantalum-polymer is the best choice for DC block C3. The gain is about 6 with a single pole rolloff at 20kHz -- this brings the 0.8V p-p consumer line level up to 4.8V p-p in order to fill out the PCM1804's dynamic range fairly neatly. Vcom is the common mode voltage (2.5V) supplied by the PCM1804.

schematic

simulate this circuit – Schematic created using CircuitLab

Also, while all of this input conditioning is happening, the AFE's op amps are easier to protect from outside mayhem (such as having 48V of phantom power discharged into them all at once) than a fragile ADC input and can be made easily replaceable (very good op amps like the OPA2134 suggested by the PCM1804's datasheet are still available in standard SOIC-8/SOIC-14 and even DIP-8/DIP-14, the latter sometimes seen in sockets in industrial work where a quick op amp swap can get an analog card back up and going).

ThreePhaseEel
  • 8,858
  • 4
  • 26
  • 41
  • Really detailed answer - thank you! Much of it is difficult for my to currently understand. If we zoom out of the detail abit - can you answer the question on quality I.e would this improve the signal quality....and if so...how!? Also, if we take the 1st basic problem, how does the op amp guard against this? Does it actually scale down the inputs!? – RenegadeAndy Sep 07 '16 at 04:15
  • @RenegadeAndy -- re: the transient problem -- it's far easier to put a robust input (that can survive being kicked in the seat of the pants so to speak by voltage spikes etal with little external help) on a relatively simple op-amp, than on a complex ADC made in what is often a denser process. In addition, op-amps are cheap compared to ADCs -- if you accidentally fry an input channel, which would *you* rather replace, the input op-amp for $2 in a DIP or the ADC for $5 or more in a TSSOP, TQFP, or worse yet a QFN or BGA? – ThreePhaseEel Sep 07 '16 at 04:22
  • Ok totally get that bit. So the op-amp acts as a cheap guard. I like that. But what does it actually down to guard it? Does it scale down the inputs so protect the input to the ADC? Also will the op-amp do any more than just protect the expensive chip? – RenegadeAndy Sep 07 '16 at 04:26
  • I am wondering if something like the OPA2134 actually solves all 4 of the problems you list here... – RenegadeAndy Sep 07 '16 at 04:40
  • @RenegadeAndy -- the op-amp doesn't "scale the signal down" unless you explicitly wire it to do so; however, it has more robust *protection* on-board than an ADC input will. And yes, an op-amp AFE, implemented correctly, will solve all 4 problems I mentioned. – ThreePhaseEel Sep 07 '16 at 11:42
  • Thank you very much ThreePhaseEel. I guess you can see this question coming, can you show me given a OPA2134 implemented correctly to a PCM1804 to solve each of the 4 common problems? Or is that a separate question (if so I am very happy to ask that in a new thread!) – RenegadeAndy Sep 07 '16 at 11:49
  • I absolutely did read the answer! Please remove the TL;DR statement. – RenegadeAndy Sep 07 '16 at 22:27
  • @RenegadeAndy -- the TL;DR is for some other bloke who isn't in the mood to read :) – ThreePhaseEel Sep 07 '16 at 22:54
  • Ahh okay! Thought it was intended for me!. Would it be possible for you to show an implementation? I was also mayb considering the LM339 - but equally, need an example circuit to work from. – RenegadeAndy Sep 08 '16 at 01:04
  • @RenegadeAndy -- implementation's edited in -- LM339's a comparator though not an OA so it won't work very well at all :P – ThreePhaseEel Sep 08 '16 at 01:29
  • Ah! Sorry! I didn't mean the LM339 - my head's in too many different places at at once! I presume you just ignored that! Also I am not particularly restrained to 5V -- I could use another value if needed. – RenegadeAndy Sep 08 '16 at 02:25
  • @RenegadeAndy -- the OPA2134 would do pretty fantastically well on a split supply (say +/-6V or so minimum), although you'll still want the DC block and Vcm injection single supply style in order to drive the single-supply differential input on the PCM1804 correctly. – ThreePhaseEel Sep 08 '16 at 02:59
  • ThreePhaseEel I cannot seem to find a PCM1804 DIP which is a real pain. Are you saying the OPA2134 won't work on DC? – RenegadeAndy Sep 08 '16 at 12:07
  • @RenegadeAndy -- PCM1804 is only available in SSOP-28, but SSOP to DIP adapters are readily available these days. And the OPA2134 is fine DC-coupled, but you don't want the 2.5V DC CM offset from the PCM1804 propagating back to a line driver that's expecting to drive a ground referenced signal. – ThreePhaseEel Sep 08 '16 at 22:09
  • ThreePhaseEel can't that be fixed with diodes? – RenegadeAndy Sep 09 '16 at 01:44
  • @RenegadeAndy -- no as they'd affect the desired signal too much – ThreePhaseEel Sep 09 '16 at 01:46
  • I don't think I understand the actual behaviours too much to determine if this shuts this combination of chips down or not! – RenegadeAndy Sep 09 '16 at 01:58
  • @RenegadeAndy -- a diode would clip off parts of the signal when it turns off, which would cause massive distortion – ThreePhaseEel Sep 09 '16 at 02:04
  • Isnt the point of a diode to make sure current only flows in one direction? Would it distort in the time it takes to switch? – RenegadeAndy Sep 09 '16 at 02:09
  • @RenegadeAndy -- the problem is that an AC current (like an audio signal) *already flows in both directions* -- it oscillates back-and-forth IOW – ThreePhaseEel Sep 09 '16 at 02:10
  • Does it!?! I had no idea that it did or would....how would I have realised an audio Signal is AC!? – RenegadeAndy Sep 09 '16 at 02:11