balanced line input with THAT 1286

[Dimitree]

this is the analog front end that I'm going to use with a PCM4222 ADC (a fun project, not for a commercial product), that I draw with the help here on the forum and reading datasheets.



what I learn is that conventional balances input stage have poor real life CMRR even with slightly unbalanced source impedances,
hence the buffers in front of the 4-resistors diff-amp. I choose THAT 1286 because I had several of those already, I could have choose THAT 1200 (full InGenius solution) and avoid buffers, but those chips are more expensive and then I would have needed another opamp to provide the negative side to the ADC.

I want to expand this circuit with switchable input sensitivity, and also solve some doubts I had:

1) now it can handle +24dBu differential input, and (thanks to the -6dB configuration of THAT 1286 and R9/R10/R11) attenuate it that to 2V RMS (the ADC has 5.6vPP full scale input). I'd like to scale the input with some switchable options (probably with analog switches, or mechanical switches) like +12dBu, +18dBu, and so on.
What's the best approach to do this? One option I see is Figure 11 of THAT 1280 datasheet, just adding three resistors on the buffers to add gain when needed. Another option could be switching R10 with other values?

2) how is C8 calculated? if I switch R10 (question above) then I need to switch C8 too?

3) what is your suggestion for the input buffer impedance? I could be wrong, but reading the THAT InGenius explaination, the higher the value of the input resistors, the higher is the noise (that’s why a full InGenius approach would employ bootstrapping). So, is 1M still a good compromise that doesn't not degrade performances, or should I lower it? If so, how about the coupling caps value?

4) I noticed AD8273 and AD8279 by Analog Devices looks functionally similar to THAT 1286. The AD8279 employs higher value resistors..what's the consequence?


thanks!

 

[Newmarket]

Quick note. re 3) In use the 1M0 resistors are bypassed by the signal source impedance so do not define the associated noise performance.
It will be noisier with the input open (don't know if that matters in your application). You can use a switching jack input to short out or 'ground' the resistors if preferred.

 

[user 41968]

+1 on what Newmarket said.

In response to your OP:

1) Make IC11 an instrumentation amp. A similar circuit to yours was discussed here recently.
help me understand differences between THAT line receivers for ADC
2) The value of C8 is normally defined by the manufacturer.
3) 1M seems reasonable to me. As Newmarket points out Rsource parallels, through Cin, the bias resistor. Neglecting Cin reactance the CM impedance will be 500KΩ.
4) Higher-value resistors will make it noisier though how much is too much depends on your goals.

WRT to 3) You could use a bipolar input op amp with these higher value resistors and let some common mode DC offset develop due to the higher bias current. That DC CM voltage gets "burnt off" (rejected) by the following cross-coupled differential amp. If the reduction in headroom from CM offset is of concern, you can pre-bias the 1MΩ resistors "ground" connection to "offset the offset" using an assumed nominal bias current. AP used this trick in the System One front end with a 5534. IIRC they used a 1 µA Ib assumption. (Beware doing this with an LME49720 as it requires several times the nominal bias current at start-up and may latch up as well as be EMI-prone.)

If you do use a bipolar op amp with back-to-back input diodes and continue to use IC11 as a follower, there needs to be some resistance, about 1K, in the feedback loop to limit input current under faults. Most 5532/4 datasheets show this. If you apply this rule and use input diode current limiting its actually better to split the current limiting resistor into two resistors. One 499Ω between the op amp input pin and Cin and another 499Ω in the follower's feedback loop. Doing so will give you 1K current limiting - safety at the expense of some noise - but will also quasi-balance the source impedance seen by the op amp which will reduce its' CM distortion. (The source impedance of the source itself, which is an unknown but usually <<499Ω makes it quasi-balanced.)

EDIT: If you wanted a 1MΩ CM impedance and could up-size Cin you could use two 4K99 input resistors with a 1MΩ connected from the middle of the two 4K99 to ground or to the bias current assistance scheme suggested above. Zdiff will be 10K, Zcm will be about 1MΩ. With an open input this "T-bias" scheme will be quieter. The noise current in the 1MΩ and bias current generated offset will both be in CM and rejected by the 1286s.

References:

https://www.proaudiodesignforum.com/forum/php/viewtopic.php?t=1119https://www.proaudiodesignforum.com/forum/php/viewtopic.php?t=641https://www.proaudiodesignforum.com/forum/php/viewtopic.php?p=11083https://www.proaudiodesignforum.com/forum/php/viewtopic.php?p=16959https://www.audiosciencereview.com/...pecially-of-the-lm4562-lme497x0-family.10687/

 

[ccaudle]

I want to expand this circuit with switchable input sensitivity

I think the traditional instrumentation amp with gain style you suggested is good. Feedback resistors around the input op-amps, and then either open between those for unity gain at the front-end, or a resistor bridging the feedback resistors to add gain. That is fairly traditional.

Another option could be switching R10 with other values?

That would probably work as well, just have less attenuation at the ADC. Double check that in the worst case of a full level input, and you forgot to change back to the standard attenuation, it will not drive too much current into the ADC input buffers.

how is C8 calculated?

It is mostly just a typical RC low pass filter. You can analyze a differential filter like that as two single ended filters joined in the middle.
Series element is still the 910 Ohm resistor, shunt elements are 180 Ohm resistor to gnd, and 4.4nF capacitor to gnd in parallel with the 180 Ohm resistor.
Play with a calculator and LTSpice to see what effect different shut resistors have on the attenuation and frequency response.
An additional function it provides is supplying current into the ADC input buffers if they have any non-linearity (e.g. impedance drops at the sampling clock edge or something like that).
I didn't have time to look through the datasheet details, but start there, then look for TI app notes describing the RC filter use at the input of delta-sigma ADC.

if I switch R10 (question above) then I need to switch C8 too?

I think so if you want to keep the frequency response the same, but usually that filter is set to a high frequency just to act as an anti-alias filter for the first delta-sigma stage, so only need to have substantial attenuation above maybe 1MHz or so.

what is your suggestion for the input buffer impedance?

1M Ohm is OK with those 2134 FET input amps you have shown. In a commercial design that has to support running at high maximum temperatures you would want to make sure that worst case leakage current at high temperature wasn't a problem, so I guess it may depend on whether you think you will always use it indoors, or whether you may need to use it outside in the summer. If you look at the datasheet you see that the input bias current is about 10pA per amp at 25C, but 250pA at 75C. That still is only around 0.25mV through 1M Ohm, so shouldn't be a problem even then.

how about the coupling caps value?

1uF is a -3dB point around 6Hz, so should be OK. Make it 2uF if you are worried about any low frequency issues. Current noise of those amps isn't too high even at low frequencies, but as the caps increase in impedance it will increase the noise which comes from current noise through the 1M input bias resistors.
That does not have much audible effect, so seems OK to me. I have been using basically that same front end on some active speakers for many years now, works perfectly fine (not the attenuator after, and only one diff amp since it feeds a single ended crossover circuit, but dual 2134 amps with 1M bias resistor feeding into an integrated diff-amp).

by Analog Devices looks functionally similar to THAT 1286

And also the Texas Instruments INA receivers. All essentially the same.

higher value resistors..what's the consequence?

Slightly higher noise. You can compare the noise values of the various part numbers. That 1286, TI INA137, AD8273 all use 12k/6k. AD8279 might be useful if you want very low current from the driving stage, but even in parallel the 12k/6k parts should be fine. Noise level of -104dBu, so -101dBu when using two (uncorrelated noise of same level should be 3dB higher), but with 24dB input that is still over 120dB SNR at that stage.

Make IC11 an instrumentation amp.

Uhm...that is what he has built, but changed the typical 3-op-amp in-amp into a 4-op-amp in-amp so that it has differential output.

You could use a bipolar input op amp with these higher value resistors and let some common mode DC offset develop

The problem is there is no guarantee it will be common mode DC, it could be differential DC. You would probably want coupling caps directly in front of the ADC if you do that.
Although if it worked well enough for the AP1 front end I'm not going to argue with it, probably will work just fine if you use a dual op-amp package so that both parts are kept very close in temperature, and probably came from the same wafer lot (or are even on the same die).

What you have looks fine to me, I say go for it! Looks like a fun project.

 

[user 41968]

What our OP has shown is effectively a buffered cross-coupled differential amplifier front-end.
When I said make IC11 an instrumentation amp I meant add resistors to provide switchable gain and not use followers.

With regard to Ios and differential output offset:

I've found that 5532s and NJM2068s both have very low Ios when both sections are used to make a two op amp INA front end.
Other bipolar duals are likely similar.

I typically see an Ios of the op amp pair - made into an INA - about the same as a single section.
The on-die match is very good.
I have experience with this across maybe 100 boards in an INA-based phono preamp.
With the NJM2068 it's usually in the 10 nA range.

Using "T-bias," with a 10KΩ differential input impedance (2X 4K99) and a 1MΩ CM resistor, minimizes Ios x Rdiff compared to 1MΩ/leg/input.
The input bias currents, Ib, are steered into the single 1MΩ and then appears in CM.
Doing so of course requires larger input capacitors since Zdiff is now 10KΩ.

EDIT: I found an example for a line input using T-bias to raise Zcm that I did back in 2012 prior to learning the ills of the LME49720/49860:
A Simple AC-Coupled Balanced Line Reciever - Pro Audio Design Forum

 

[Dimitree]

thank you for the help!

so if this looks ok, I could start with the calculations for the gain resistors. I added a DG411 analog switch to implement three preset resistors for the gain




other questions that come to mind:

1) Where is the best place to split this circuit in two PCBs? The idea is that the ADC ICs (4x) is on a digital motherboard PCB, and the jack/inputs (8x) on another board, so I need to decide where to split the signal and place a connector for the two boards.
2) Is phase shift a concern in this design, given the fact that the signal passes through buffers before reaching the THAT chip?
3) how critical is passive parts matching? I mean C21/C17 and R12/R13 mostly, but also the matching of the two buffers (even if they are on the same package, I guess they are not 100% matched)

 

[Newmarket]

re 2) phase...phase shift of no concern in this application as far as I see it. Circuit is symmetrical. It would only matter at all if you were recombining the output with the original.

 

[zamproject]

I could start with the calculations for the gain resistors. I added a DG411 analog switch to implement three preset resistors for the gain

You can probably do 8 step gain, or 16 with the fourth DG switch

 

[ccaudle]

Where is the best place to split this circuit in two PCBs?

You can work from each end and make a list of limitations.
For instance, RFI filtering needs short and wide connections for lowest inductance, so C13-C15 must be located directly next to the connectors.
The impedance to the reference pin of That1286 devices should be low and wide bandwidth, which implies that IC9 and IC10 should be near to the ADC.

That leaves IC11 and the gain switches as devices which could potentially go on either side. My thinking is that there will likely be some kind of microcontroller communicating with the ADC, and if you have switchable gain the micro would probably be controlling the gain switches, so put all the active devices on the ADC PCB and just have passive filtering devices located at the connectors.

Depending on how your control design is arranged you could get to a different conclusion, but that is my thinking process about the question.

 

[FIX]

The best place is the output of the 2134s. It's low Z, so minimum interference pickup.

 

[Dimitree]

Thank you everyone, I’ll let you know how it will turn out

 

[Dimitree]

so I came up with those values for the instrumentation amp..how does it look?
I started with 24K since I mistakenly thought it was the value used by THAT1200 internally..then I checked the THAT1286 datasheet and saw that they have an example using 4.99K for those resistors.

falstad simulation

what is the real difference, once you have adjusted RGAIN?

I also had to reduce to 220 the resistor in front of the ADC

 

[ccaudle]

I have heard Falstad is not a good simulator for checking values, I would stick with calculator and LTSpice (or other "real" simulation tool like pspice, Tina-TI, etc.).

what is the real difference, once you have adjusted RGAIN?

In different values of feedback resistors? Primarily thermal noise, and also how much current the front end amplifiers have to provide.

 

[thor.zmt]

this is the analog front end that I'm going to use with a PCM4222 ADC (a fun project, not for a commercial product), that I draw with the help here on the forum and reading datasheets.

View attachment 101129

The Datasheet of the PCM4222 shows an "active attenuator" configuration. Using the passive attenuator as you show will increase HD substantially.

1) now it can handle +24dBu differential input, and (thanks to the -6dB configuration of THAT 1286 and R9/R10/R11) attenuate it that to 2V RMS (the ADC has 5.6vPP full scale input). I'd like to scale the input with some switchable options (probably with analog switches, or mechanical switches) like +12dBu, +18dBu, and so on.

Use the datasheet circuity with a suitable pair of input buffer Op-Amp's with gain. Somthing similar to this:

Another option could be switching R10 with other values?

2) how is C8 calculated? if I switch R10 (question above) then I need to switch C8 too?

I would not suggest to do that. The input deglitching capacitor value usually is given in the datasheet.

3) what is your suggestion for the input buffer impedance? I could be wrong, but reading the THAT InGenius explaination, the higher the value of the input resistors, the higher is the noise (that’s why a full InGenius approach would employ bootstrapping). So, is 1M still a good compromise that doesn't not degrade performances, or should I lower it? If so, how about the coupling caps value?

The input resistors bias the input Op-Amp's. They can be very large in value without negative consequences, if fet/jfet input op-amp's are used. For bipolar op-amp's the input current needs to be accounted for.

4) I noticed AD8273 and AD8279 by Analog Devices looks functionally similar to THAT 1286. The AD8279 employs higher value resistors..what's the consequence?

More noise.

Thor