Asking for feedback on simple ADC front-end/preamp idea

[deva]

Hi everyone!

I need to build a front-end to an ADC that can both accept a (balanced) line input, or work as a very simple microphone preamplifier.

I was thinking of using the classical three-opamp instrumentation amplifier approach (pic for reference).

However, the ADC I will be using has differential inputs, whereas the standard INA topology has a single-ended output.
My first thought was to replace the differential amplifier (U3) with a fully differential opamp to drive the ADC; then I thought, can't I just remove U3 and associated resistors at all, and drive the ADC directly from U1-U2?

I think I can see it working on paper, as gain for differential signals is still provided by U1 & U2; but I'm afraid I might be missing out on some real-world limitations or issues that could make this simpler / lower part count approach undesirable.
Any comments on this?

Also, as far as the preamp gain goes, I was thinking of implementing a stepped control with 3 or 4 steps, using relays to swap resistors in place of VR1 (I have hundreds of relays laying around...).
What would you recommend as "reasonable" gain steps? I'm planning to use this thing mostly with dynamic mics (on stage - live sound), but I might also need the occasional line input, which I guess warrants a 0dB step in there.

Thanks!
- Dave

EDIT: and I guess I managed to post this in the wrong board. Sorry guys! I guess it should be moved to the Drawing Board or something...

 

[JohnRoberts]

While it's extra parts I'd be tempted to add two differential amps so common mode will get subtracted out before you send to A/D. Some A/D have a common Vref so perhaps connect to that instead of 0V/ground for the differential amps.

If using 5534 with more than unity (noise) gain you might be able to get away with less than 22pf compensation caps. 

JR

 

[joaquins]

Also you need an attenuator, with a balanced resistive divider, let's say double L pad, to adapt the level to the ADC while using all the dynamic range of the opamps, for best S/N ratio. You have to make some math on that, depending on your rails and other parts of the design. Don't use too big resistors, the thermal noise of 10K resistance seen by the ADC would be affecting the noise performance of a good converter, if everything else suits that noise floor (hard to get there) you'll be limiting the noise performance because of a single resistor, would be a shame. But for 18V rails, let's say you have 16Vp-p in each opamp, so twice that, for both opamps, 32Vpp, less the 2V of common mode for the ADC so 28Vpp and at the ADC you'll have something like 4Vpp each side, so 8Vpp. So you need 10dB attenuation, with three 1K resistors each opamp would see 1.5K load, you have to be able to drive them comfortably. This is just to make some numbers, you could use bigger resistors, 3K3 would be fine and won't load significantly a 5534. Also, for noise performance you should for low noise analog part, maybe scale down some resistors there, I don't know what ADC you'll be using but usually the S/N performance is limited by the analog circuit, so good care there, you need to look at the equivalent input noise of your converter, then calculate all the noise sources on the analog and scale them down 10dB because of the output pad, then you'll know if your circuit is limiting the S/N or not.

JS

 

[deva]

Thanks for the feedback!

JR,

While it's extra parts I'd be tempted to add two differential amps so common mode will get subtracted out before you send to A/D.

are there any advantages on this approach compared to just inverting the signal off U3 with an unity gain inverting stage? Seems like using two differential stages would make it more of a pain as it'd need more matched resistors.

 

[JohnRoberts]

Thanks for the feedback!

JR,

While it's extra parts I'd be tempted to add two differential amps so common mode will get subtracted out before you send to A/D.

are there any advantages on this approach compared to just inverting the signal off U3 with an unity gain inverting stage? Seems like using two differential stages would make it more of a pain as it'd need more matched resistors.

The advantage is very subtle... Adding an opamp inverter in series with first output means that the second path is slightly different. This differences is probably insignificant but....  :'(
=====
In general the attraction of instrumentation amp topology is high input impedance, your circuit is roughly 4.5k ohm + to - input. (marginally high for mic input and marginally low for line inputs.) If you were to instead use two opamps configured as simple differential inputs  (like your third opamp stage). The + side differential would take difference of + minus -, and - side would take difference of - minus +.  One less opamp and completely symmetrical path for both inputs.

Using 5534 as mic preamp will not be very quiet... IIRC I did the math decades ago and the noise figure with low Z mics was around 9dB, as compared to a good transformer or dedicated IC preamp at < 3dB NF.

If you need mic preamp capability perhaps add a preamp chip, then allow the capability to patch line level inputs after the preamp.

JR     

 

[joaquins]

Thanks for the feedback!

JR,

While it's extra parts I'd be tempted to add two differential amps so common mode will get subtracted out before you send to A/D.

are there any advantages on this approach compared to just inverting the signal off U3 with an unity gain inverting stage? Seems like using two differential stages would make it more of a pain as it'd need more matched resistors.

The advantage is very subtle... Adding an opamp inverter in series with first output means that the second path is slightly different. This differences is probably insignificant but....  :'(
=====
In general the attraction of instrumentation amp topology is high input impedance, your circuit is roughly 4.5k ohm + to - input. (marginally high for mic input and marginally low for line inputs.) If you were to instead use two opamps configured as simple differential inputs  (like your third opamp stage). The + side differential would take difference of + minus -, and - side would take difference of - minus +.  One less opamp and completely symmetrical path for both inputs.

Using 5534 as mic preamp will not be very quiet... IIRC I did the math decades ago and the noise figure with low Z mics was around 9dB, as compared to a good transformer or dedicated IC preamp at < 3dB NF.

If you need mic preamp capability perhaps add a preamp chip, then allow the capability to patch line level inputs after the preamp.

JR   

The instrumentation amplifier, as posted on the first post has 4k4 for NM signal, but 101K for CM signal, the advantage of instrumentation of high input impedance is maintained, for the CM which is we are interested, so CMRR will remain high even with unbalanced input. As extreme example, if one of the inputs is left open and the other has a signal from a 0Ω source, still the input will be seen as -33dB of the signal level, is a problem if you have unbalanced source and leave the other terminal open, but why would someone do that? in the case of not so unbalanced sourced, from a mic 150Ω as really high unbalance the CMRR will be over 55dB. As you pointed out instead of those 33dB you have 6dB and the same happens to the unbalanced sources.

Doing what you are saying for U3 would be the better, except for having two more opamps, but CMRR is way better, and the signal path will be now a symmetric path.

Other option as pointed is to take out U3 and feed the ADC from U1 and U2, for unity gain will have as good CMRR as the converter, I don't know how good they are at this point, but as signal gain increases, the CM gain is always one, so the CMRR will get better and better, and if we think about it, higher level signals are more immune to noise than lower level, so is not a bad compromise if can't use 4 opamps for some reason.

FWIW I did build this as shown, never used as a mic pre but measured CMRR (I was experimenting with this as a good topology for the matter, and get more than 70dB of rejection with balanced sources and I don't remember the exact number but I guess about those 33dB for worst case, as pointed up, unbalanced sources. Input noise is a bit high for mic preamp though, I don't remember the numbers now.

JS

 

[JohnRoberts]

Thanks for the feedback!

JR,

While it's extra parts I'd be tempted to add two differential amps so common mode will get subtracted out before you send to A/D.

are there any advantages on this approach compared to just inverting the signal off U3 with an unity gain inverting stage? Seems like using two differential stages would make it more of a pain as it'd need more matched resistors.

The advantage is very subtle... Adding an opamp inverter in series with first output means that the second path is slightly different. This differences is probably insignificant but....  :'(
=====
In general the attraction of instrumentation amp topology is high input impedance, your circuit is roughly 4.5k ohm + to - input. (marginally high for mic input and marginally low for line inputs.) If you were to instead use two opamps configured as simple differential inputs  (like your third opamp stage). The + side differential would take difference of + minus -, and - side would take difference of - minus +.  One less opamp and completely symmetrical path for both inputs.

Using 5534 as mic preamp will not be very quiet... IIRC I did the math decades ago and the noise figure with low Z mics was around 9dB, as compared to a good transformer or dedicated IC preamp at < 3dB NF.

If you need mic preamp capability perhaps add a preamp chip, then allow the capability to patch line level inputs after the preamp.

JR   

The instrumentation amplifier, as posted on the first post has 4k4 for NM signal, but 101K for CM signal, the advantage of instrumentation of high input impedance is maintained, for the CM which is we are interested, so CMRR will remain high even with unbalanced input. As extreme example, if one of the inputs is left open and the other has a signal from a 0Ω source, still the input will be seen as -33dB of the signal level, is a problem if you have unbalanced source and leave the other terminal open, but why would someone do that? in the case of not so unbalanced sourced, from a mic 150Ω as really high unbalance the CMRR will be over 55dB. As you pointed out instead of those 33dB you have 6dB and the same happens to the unbalanced sources.

Doing what you are saying for U3 would be the better, except for having two more opamps, but CMRR is way better, and the signal path will be now a symmetric path.

Other option as pointed is to take out U3 and feed the ADC from U1 and U2, for unity gain will have as good CMRR as the converter, I don't know how good they are at this point, but as signal gain increases, the CM gain is always one, so the CMRR will get better and better, and if we think about it, higher level signals are more immune to noise than lower level, so is not a bad compromise if can't use 4 opamps for some reason.

FWIW I did build this as shown, never used as a mic pre but measured CMRR (I was experimenting with this as a good topology for the matter, and get more than 70dB of rejection with balanced sources and I don't remember the exact number but I guess about those 33dB for worst case, as pointed up, unbalanced sources. Input noise is a bit high for mic preamp though, I don't remember the numbers now.

JS

Yes I understand how that circuit works but to be a front end for an A/D you rarely need gain, generally you need to pad down the signal.

To answer the OP's question about just losing the 3rd opamp differential stage, that can cause headroom issues since the signal will be superimposed on top of any CM signal that that topology presents such a high impedance to.

If you need a mic preamp use a proper mic preamp, if you need a pad and CMRR two simple differentials can do that without consuming A/D headroom.

YMMV

JR

 

[joaquins]

You are right about headroom and CM, I knew I was forgetting something. To use another pre when needed is good, but  what he was looking for is a preamp too.

JS

 

[deva]

Hi everyone,

Apologies for the delay in getting back here - just gotten back to Europe from the States, still jet-lagged.

To answer the OP's question about just losing the 3rd opamp differential stage, that can cause headroom issues since the signal will be superimposed on top of any CM signal that that topology presents such a high impedance to.

Thanks! I did overlook this when first thinking about it.

...to be a front end for an A/D you rarely need gain, generally you need to pad down the signal. [...]
If you need a mic preamp use a proper mic preamp, if you need a pad and CMRR two simple differentials can do that without consuming A/D headroom.

To use another pre when needed is good, but  what he was looking for is a preamp too.

To clarify a bit on this issue: what I'm trying to build is a digital audio snake.
So, for most channels on the stage side I do need gain, except for maybe the odd line input; on the FoH side, I just need line inputs.
Using the same front-end/ADC structure for all inputs would be neat; that's why I was thinking of using that instrumentation amp topology, with some gain control that could cover from "line in" to "mic pre".

- Dave

 

[JohnRoberts]

S/N for microphones will not be very good using schematic as drawn, easily 6+dB worse than dedicated mic preamp.

I would use one of the several proper mic preamp chips for mic inputs. On stage side you can add pads to accommodate the few line level inputs. For the FOH side I would use a dedicated line input topology.

JR