how to use THAT 1293 with ADC

[Dimitree]

hi everyone,
I'm wondering if I can use a THAT 1293 balanced line receiver [datasheet] as front end for a PCM 4222 ADC [datasheet].
The PCM4222 has differential input with a full scale input range of 5.6Vpp.
This is an example from the 1293 datasheet, configured so that the output would be 2Vrms = 5.65Vpp.
So I'm wondering if I can use this schematic as front end for the ADC.
Is really that simple?
thank you

 

[ccaudle]

You can, but the CMRR of that type of input stage is very sensitive to the output impedance match of the output stage which comes before. A better input circuit is to use a pair of unity gain op-amps in front as buffers for the hot and cold inputs, with as high a value for the input resistors as you can tolerate given the input bias current and input offset voltage of your input buffers. With FET input op-amps you can go very high, between 100K Ohms and 1M Ohms can work.
Don't forget RFI and ESD protection in front, the 1293 style devices have resistors in front of the op-amp inputs, but if you have buffers in front of those that have the non-inverting inputs connected directly to the input connector, you risk blowing up the input pair if there is ESD discharge to the input pins.

Also, you have the option of using the parts with -6dB configuration so that you still end up with 24dBu at the output of that stage. It may not make a difference, but check the noise levels both ways. The -6dB parts have slightly lower noise floor because of lower value resistors, but it may not change SNR much depending on how you configure the attenuation after.

 

[ccaudle]

By the way, if you want to research some terms, that configuration of unity-gain (or with gain) input buffers followed by a difference amp configuration is known as an instrumentation amplifier, and specifically the three op-amp instrumentation amplifier. Usually an instrumentation amplifier is balanced in, unbalanced out, but the configuration you suggest with dual difference amplifiers to generate balanced output is like the output stage of the "double balanced microphone amplifier" published by Graeme Cohen in 1984 as AES preprint 2106. That configuration is the basis of some of the microphone amp circuits published by That Corp as well as the Millenia Media HV3 series mic amps.

In addition to the design notes and tech papers published by That Corp on their Design Support page, a really good discussion of requirements for ADC input stages can be found in AES preprint 4126 by Paul Frindle describing the converter system for the Sony Oxford large format digital console. Highly recommended. Did I mention it is really good and highly recommended? I would definitely get preprint 4126 and read it, it comes highly recommended.

 

[Dimitree]

thank you!
do you think the buffers on the input are still needed given the fact that those input are only used as line inputs (not for microphones) ?

also, there are no ac-couple capacitors shown on the example schematic. I guess I need to add that. If I add the opamp buffer, they need to be before the buffer or between the buffer and the 1293?

 

[ccaudle]

do you think the buffers on the input are still needed given the fact that those input are only used as line inputs (not for microphones) ?

I think so, you can get relatively high levels of power line derived common-mode noise. It comes down to how much CMRR you would like to achieve, and how much you can tolerate changes in common mode rejection with different devices connected (due to imbalance in output impedance). Read some of the That Corp app notes on their InGenius input stages to see some derivation of how much CMRR changes with various levels of imbalance (as well as the AES papers from the 1995 June special issue by Bill Whitlock).

there are no ac-couple capacitors shown on the example schematic. I guess I need to add that.

If you don't expect someone to accidentally connect 48V phantom power to the inputs, you could also use a DC servo stage to servo out the DC. That just moves the capacitors to a different point in the circuit so you don't necessarily have to use such large bulky electrolytic caps.

If I add the opamp buffer, they need to be before the buffer or between the buffer and the 1293?

Based on the signal levels you have noted I guess you are using bipolar 15V (or maybe 18V) supplies for this circuit.
In that case you have the option to put the caps on the input or output. I would not put them between the buffers and the 1293 because then you are back to the concern of impedance imbalance affecting the CMRR, which is the entire reason you put buffers in front of the 1293 to begin with.
With high value input bias resistors on the buffers the required capacitance will be lower, so that can be an advantage to putting them in front of the buffers.
You could put the capacitors between the 1293 and the attenuator, or between the attenuator and the PCM4222, but they would get really large because of the lower impedance at that point.

 

[Dimitree]

thank you again for the help,
I tried to re-read the tips multiple times and tried to sketch a schematic of what I (probably) managed to understand, it may be totally wrong, so I apologize for that

 

[Bo Deadly]

There are single ICs that do most of what you need. Something like INA2134 would work just fine. Clearly this is not an exercise in lowest possible THD / noise since there are now ADC chips that are pointlessly low noise and THD so don't get too carried away. Time would be better invested in input protection / filtering.

 

[user 41968]

The 1293 will be noisy in this application due to the high value of the internal feedback resistors.

I would use discrete op amps with low-value feedback resistors to take full advantage of the PCM4222's dynamic range.

 

[Dimitree]

Now I'm even more confused than before

discrete opamps is overkill for what I want to achieve.
I'm using PCM4222 because I've got many with me already. I don't think I need to achieve the best performance possible with this ADC..

INA2134 looks promising. I'm looking into that.
Also, what about THAT1200 series? I used to think that THAT1290 series was just the dual version of 1200, but now I realize they are different ICs

the schematic proposed on the last page of this Design Note:
dn133.pdf

 

[user 41968]

My bad. When I said discrete op amps I meant separate conventional op amps like a 5532 without internal resistors. That way you can optimize the values to minimize noise.

If you're willing to sacrifice some available dynamic range of the PCM4222 and want something simple then use the THAT1293 or INA2134.

You will need to AC-couple the inputs.

 

[ccaudle]

Yep, that is what I was describing. I built something several years ago that was essentially the same thing using OPA2134 for the input buffers and INA137 for the diff amps.

Something like INA2134 would work just fine

The INA2134 is basically the same thing as the THAT1293. You seem to have skipped over the entire discussion of CMRR sensitivity.

The 1293 will be noisy in this application due to the high value of the internal feedback resistors

Not to pick on you in particular, but I get a little annoyed when someone posts something that should be numerically describable very easily, but just describes it as "too" something, or "not enough" that.
Let's roll up our sleeves and do the 30 seconds of work to see if this is in the ball park or not:

THAT1293 has typical output noise of -104 dBu, or -107 dBu for the -6dB variant.
We'll start with the unity gain version first.
-104dBu, but you have two of them so the noise is doubled. Talking just of thermal noise (i.e. ignoring any power supply noise leak through that might be common to both channels), the noise should be uncorrelated, so double is in the power sense, 3dB increase, not 6dB you would get for correlated signal.
So the noise at the differential output should be around -101 dBu.
But the signal at that point is +30 dBu (given a +24 dBu input), so the signal to noise ratio is 131 dB. Seems OK to me.

Sometimes the -6 dB variants are slightly lower noise, so lets see how that works out:
Typical output noise of -107 dBu, doubled uncorrelated noise, so differential output noise of -104 dBu.
But the signal is still just 24 dBu, rather than +30 dBu in the previous example, so the signal to noise ratio is "only" 128 dB. Still seems OK to me.

The dynamic range of the PCM4222 is 124 dB, so either configuration isn't going to degrade the capabilities of the ADC very much.

I would use discrete op amps with low-value feedback resistors

And what matching will you get with the discrete resistors? The 1293 has +/- 0.5% matching because they are the cheap version (around $1 in large quantity). If you step up to the 1280 series you get 40 dB better common mode rejection from the +/- 0.005% matching resistors. Those are about $2 in large quantities, and there is no way you can get discrete 0.005% resistors for $2, much less resistors and op-amps together for $2.

what about THAT1200 series?

You get really high effective common mode impedance with a 1200 receiver, but the diff amp stage is on die, and you only have one, so you would have to generate the other side of the differential output to the ADC stage. You would need to do something like figure 3 in the "Achieving Optimum CMRR with Differential Input A/D Converters" app note:
ThatCorp Design Note 133

 

[Dimitree]

You get really high effective common mode impedance with a 1200 receiver, but the diff amp stage is on die, and you only have one, so you would have to generate the other side of the differential output to the ADC stage. You would need to do something like figure 3 in the "Achieving Optimum CMRR with Differential Input A/D Converters" app note:
ThatCorp Design Note 133

than I'd say I will use the one I sketched on #6 , based on your suggestions.
Using the 1200 receiver route, will have a larger part count, and more expensive too (with no benefits if I understand correctly).

 

[user 41968]

The matching with 1% resistors will certainly not be as good as those of a 1293. I use the 124X series by the hundreds. You're preaching to the choir.

And if avoiding source impedance imbalance CMRR degradation is your thing then use InGenius or input buffers.

I agree that the 124 dB DR of the converter isn't degraded much.

Ya'll have fun.

 

[Dimitree]

now that I think about the schematic on #6 with the added buffers,
I'm wondering, using +/-15V, how could it handle 24dBU input? I guess it can't.
24dBU should be about 34Vpp, so I'd say the power rails should be at least +/-18V (depending on the opamp used) right?

 

[Bo Deadly]

The INA2134 is basically the same thing as the THAT1293. You seem to have skipped over the entire discussion of CMRR sensitivity.

Yeah, I guess I did. I retract everything I said. Definitely need amps in front.

 

[Bo Deadly]

now that I think about the schematic on #6 with the added buffers,
I'm wondering, using +/-15V, how could it handle 24dBU input? I guess it can't.
24dBU should be about 34Vpp, so I'd say the power rails should be at least +/-18V (depending on the opamp used) right?

It's not terribly common to see +24dBU signals these days. Unless you're deliberately trying to overdrive some outboard gear, you really don't need super high signals these days since everything is converted to / from digital at like 5dBU.

 

[ccaudle]

I'm wondering, using +/-15V, how could it handle 24dBU input?

There is an important caveat that the 24 dBu signal needs to be symmetrically driven balanced output.

24dBU should be about 34Vpp

Yes, but if the input is symmetrically driven then that would be 17Vpp on the hot leg and 17Vpp on the cold leg, or +/- 8.5V peak on each leg, so well within the capabilities of devices running on 15V.

If the output is from a transformer coupled device the balance isn't necessarily well defined. Probably OK unless one side of the transformer gets connected to ground, then you would get the entire 24dBu on one leg, which will definitely clip.

If the output is from a solid state device (err...solid state device without a transformer, so not an old Neve transistor design) it is very likely to have symmetrical output if it can drive levels that high. Some gear has impedance balanced outputs with asymmetrical drive, but I don't think I have ever seen any gear like that with 24 dBu output level capability.

With 15V rails you can handle 20dBu, that is just 11V peak, so if you do have a device with asymmetric output you don't have to drop the levels too much to be able to handle it without clipping.

 

[Dimitree]

so do you think it would be helpfull to implement a "switchable" sensitivity on the inputs? if most sources are lower level than 24 dBu, seems to me that I'm wasting dynamic range using the ADC this way..right?

 

[ccaudle]

if the input is symmetrically driven then that would be...+/- 8.5V peak on each leg

One thing it did not occur to mention last night was that the diff amp stages see the entire input amplitude, so you cannot handle 24 dBu with a 1290, you would need a 1293 or 1296 (or the 128x equivalent, or the TI equivalent, etc.).

do you think it would be helpfull to implement a "switchable" sensitivity on the inputs?

That is up to what kind of workflow you like. If you are just going to connect this to a mixer output, or a mic preamp, then a fixed sensitivity probably makes sense because the mic preamp or mixer would have adjustable gain. If you want this to be general purpose, then you have to figure out what source levels you want to accomodate. You could choose anything from switchable 12/18/20/24 dBu, to full adjustable from 0dBu to 24dBu, to a full range input with mic amps. Since you started with a fixed input, I assume you aren't really interested in an integrated mic amp.

Fortunately since you have the buffers in place now, making an adjustable gain instrumentation amp is just a matter of three more resistors (well, three for fixed gain, then as many others as you want additional gain settings).

 

[Bo Deadly]

Yes. That's why many ADC inputs include a pad. On my MOTU for example it switches between +4 and -10 I think. So a 14 dB pad.

Although that only corresponds to about 2 bits of resolution so a pad isn't really used to better utilize the dynamic range of the ADC. The noise floor of a source device is virtually guaranteed to be more than 14 dB above the noise floor of the ADC. Probably more like 40 dB. The pad is really just supposed to make the pro and consumer source levels more similar so that you're not adjusting levels in extreme ways in the interface software and DAW.