INA1651 input resistors

[sdvd]

Hello,

I would like to design a balanced input line receiver using INA1651.
I have a question about the input resistors of 100k on the XLR inputs.
With my calculations, I found that the CMRR will degrade to aproximatly 80dB with 10 Ohm mismatch on ouput sender impedance and 74dB with 20 Ohm mismatch.
Without those resistors, the CMRR should be aproximatly 90dB with 20 Ohm mismatch.

What could be the issues of replacing those resistors by 500k each and putting the common jonction of those with a 1MOhm to GND. (A T configuration).
The global input impedance will be higher but the CMRR will be close to 90dB. Is this high impedance input a problem? I know there will be a lot of noise when source unconnected but, is there another issue?

Thank you.
David

 

[JohnRoberts]

Johnson (thermal) noise in the resistors, and ein current from active devices working into source and network impedances. Perhaps not a problem for hot line level signals, but not as quiet as it could be with lower value resistors.

JR

 

[ccaudle]

I have a question about the input resistors of 100k on the XLR inputs.

The INA1651 has built in input resistors, you should not use any external resistors at all.
To quote from the datasheet, "these resistors are optional."

 

[sdvd]

Johnson (thermal) noise in the resistors, and ein current from active devices working into source and network impedances. Perhaps not a problem for hot line level signals, but not as quiet as it could be with lower value resistors.

JR

The input will be drived by a low impedance like 50 ohm, so the global input impedance of the system will be close to the source impedance when source connected. The noise should be present only when input disconnected. Am I wrong? Did I understood correctly noise theory?
I didn’t understood the second part of your idea. Do you mean a current noise source comming from the source?

 

[sdvd]

The INA1651 has built in input resistors, you should not use any external resistors at all.
To quote from the datasheet, "these resistors are optional."

I thought those resistors were useful for discharging the input capacitors eventually charged.

 

[ccaudle]

I thought those resistors were useful for discharging the input capacitors eventually charged.

Do they get charged up? Most decent equipment shouldn't put any DC on the input. If the input capacitors do get charged up slightly, do you care how quickly they discharge? You could put 10M resistors on the input side of the capacitors. The current values would have a 1s time constant, do you care if it is 100s instead?

And to your earlier question, yes, you could use a T configuration. The data sheet even suggests that as a good configuration for the device itself, which you show in your schematic.

Is this high impedance input a problem?

In general not a problem. If you do want a lower input impedance for consistency with other equipment, you can put a resistor across the input connections, i.e. between pins 2 and 3. That will set the differential input impedance independently of the common mode impedance.

 

[Bo Deadly]

The input will be drived by a low impedance like 50 ohm, so the global input impedance of the system will be close to the source impedance when source connected. The noise should be present only when input disconnected. Am I wrong? Did I understood correctly noise theory?
I didn’t understood the second part of your idea. Do you mean a current noise source comming from the source?

Sounds right to me. Yes, the resistors are just "drain resistors". Generally it's bad form to leave a capacitor dangling. If you really want to increase CMRR to it's max, you might connect them to a third large common resistor and connect that to COM / pin 1. Although in practice I would think there should be some input protection like low-capacitance TVS diodes or small series resistors and shunt caps in which case the resistors shown might become unnecessary. Also, I'm not sure I would connect them to pin 1 at all. But I'm not familiar with the specifics of that INA1651 part.

 

[KA-Electonics.com]

I think the OP's approach is solid.

Let's take a look at the INA165X block diagram:



Inserting a 1MΩ from the Com pin to ground raises the CM impedance to about 1.25MΩ from 250KΩ.
The bias current flowing into the 1MΩ will develop a small(*) DC CM offset that will be rejected by the following Difference Amp.
*(A quick scan of the data sheet didn't show an actual input bias current spec.)
The noise current that develops in the 1MΩ will also appear in CM at the Diff amp and be rejected.
Assuming reasonable bias currents I don't see a significant down-side to adding it.

As previously suggested you might utilize the same approach on the left-hand side of the input coupling caps using a "T" network for capacitor discharge.
Using 2X 10M from each leg to ground or 2X 100K to a 10M "T" may however be a toss-up. (approx 1M||5M vs 1M||10M)

As you (the OP) have pointed out the driving source impedance will swamp the high differential impedance reducing its' noise contribution.

If both T networks are used the Differential input impedance will be (500K+500K)||(100K+100K).
The common mode impedance will be 1M||10M if you use a larger discharge resistor.

One of the advantages of the input buffers and high value bias resistors, in addition to improved CM rejection with mismatched source impedances, is that you can use lower value film caps for the inputs taking advantage of both the the higher Zcm and Zdiff. A Simple AC-Coupled Balanced Line Reciever - Pro Audio Design Forum

 

[sdvd]

Thank you all for the answers.

 

[thor.zmt]

Hello,

I would like to design a balanced input line receiver using INA1651.
I have a question about the input resistors of 100k on the XLR inputs.
With my calculations, I found that the CMRR will degrade to aproximatly 80dB with 10 Ohm mismatch on ouput sender impedance and 74dB with 20 Ohm mismatch.

Add a common mode bootstrap?

I'd be tempted to use a different receiver though, possibly using a Quad OPA1654 per channel which allows to in effect make the same circuit including a common mode bootstrap buffer and allows lower value resistors for the differential Amplifier, lowering noise.


This is a schematic in TINA-TI for what I suggest. Note that common mode bootstrap is patented, though IIRC the patents have expired. Unless you go into huge mass production I see little issue.

Does it work? For ideal 0% tolerance resistors, here some graphs:


Clockwise from top left:

Common Mode input impedance (Yes, well > 10MegaOhm below 1kHz)
Differential Mode input impedance (20k target, we are a trifle over)
Frequency Response (well, nothing to see really)
Noise, unweighted re 0dBU (with 108dB we are more than 10dB better than INA165X)

CMRR not shown, in reality it's down to resistor match, 0.1% resistors will give -60dB and 0.01% resistors would guarantee -80dB worst case, I would expect > 90dB typical using 0.01% Resistors.

Modern SMD thin film resistors are pretty good and inexpensive. If buying 0.01% resistors in reels of 500 you are looking at 10 USD or less for a reel.

Thor

 

[Newmarket]

CMRR not shown, in reality it's down to resistor match, 0.1% resistors will give -60dB and 0.01% resistors would guarantee -80dB worst case, I would expect > 90dB typical using 0.01% Resistors.

Modern SMD thin film resistors are pretty good and inexpensive. If buying 0.01% resistors in reels of 500 you are looking at 10 USD or less for a reel.

Thor

If being fussy then need to consider tempco too. Ime Nothing beats the ratio matching of laser trimmed resistors on same substrate. Either as part of an ic or stand alone parts eg Vishay. Absolute tempco may be 10ppm but ratiometrically you get 2ppm.

 

[ccaudle]

Add a common mode bootstrap?

Common mode bootstrap is useful if you have a variable gain front end, but with the unity gain buffers shown in the schematic you can just bootstrap each input buffer independently and get an equivalently high common mode impedance, without needing the very closely matched resistor pair to derive the common mode signal. Split the input bias resistor for each buffer, and connect a capacitor from the unity feedback to the junction of the resistor split. Put a lower value resistor input-to-input to set the differential mode impedance if you do not want super high differential impedance as well.

 

[thor.zmt]

If being fussy then need to consider tempco too.

0.01% tolerance resistors all have single digit tempco.

Ime Nothing beats the ratio matching of laser trimmed resistors on same substrate.

Using 0.01% tolerance resistors guarantee 80dB CMRR. there are measures that can be taken to improve. For example, take 3pcs 12k/0.01% resistors in parallel and you get 4k/0.005% and garanteed -86dB CMRR. Typical values actually will be on average 6dB better.

Of course, the last stage of the circuit I presented can be replaced by THAT124X, INA134 et al. but distortion and noise will usually be worse.

By not using laser trimmed on chip polysilicone resistors we get overall better resistors with lower noise and HD. It is all a tradeoff.

Either as part of an ic or stand alone parts eg Vishay. Absolute tempco may be 10ppm but ratiometrically you get 2ppm.

All 0.01% rated SMD resistors I know are << 10ppm singly.

I strongly consider that it is possible to create a better performing balanced line receiver using something like my suggested circuit for less budget than using IC's.

Of course, off the shelf IC's with key parts inside may be better in case of one off DIY due to simplicity.

Thor

 

[abbey road d enfer]

Common mode bootstrap is useful if you have a variable gain front end, but with the unity gain buffers shown in the schematic you can just bootstrap each input buffer independently and get an equivalently high common mode impedance,

That is true but unfortunately, on the specified IC (INA165X), the outputs of the input buffers are not accessible.
This is the case with all monolithic line receivers, except the THAT1200 series, where the CM impedance bootstrap is included (patented by Bill Whitlock (member Mr CMRR here).
These could be the answer the OP is looking for.
Now I have one question: when does anyone need >80dB CMRR?

 

[thor.zmt]

Common mode bootstrap is useful if you have a variable gain front end, but with the unity gain buffers shown in the schematic you can just bootstrap each input buffer independently and get an equivalently high common mode impedance, without needing the very closely matched resistor pair to derive the common mode signal.

Good catch. Yes, I am aware of this as well. And you are correct about benefits. I used up the spare OPA as Servo, could also be made into the Birt circuit to give balanced out from balanced in.





I don't like the impedance peaking so much. Noise improves 1dB by lowering the resistors in the differential amp.

Which one do I prefer?

Neither.

I like to use this:


Objective performance, while respectable enough is merely "ok" with around -100dB SNR @ 0dBU and -100dB H2 / -120dB H3 @ +20dBU Out.

CMRR is 75dB (could be trimmed to be better) with 5k common mode input impedance.

This circuit actually emphasises and optimises the single ended input (sw "BAL" is open if using SE in) which has lower noise and distortion than balanced.

Normally there is also a LDR in parallel with R17 to implement a limiter. Different approach and priorities.

Thor

 

[abbey road d enfer]

I like to use this:

View attachment 102861
Objective performance, while respectable enough is merely "ok" with around -100dB SNR @ 0dBU and -100dB H2 / -120dB H3 @ +20dBU Out.

In the absence of component values, it's hard assessing any kind of performance.

CMRR is 75dB (could be trimmed to be better) with 5k common mode input impedance.

Doesn't it depends on perfect matchiong of a few resistors? And maybe capacitors?

 

[Newmarket]

0.01% tolerance resistors all have single digit tempco.

Fair enough. I hadn't considered 0.01% for cost reasons - having previously dealt with them in previous (non-audio) work.

Using 0.01% tolerance resistors guarantee 80dB CMRR. there are measures that can be taken to improve. For example, take 3pcs 12k/0.01% resistors in parallel and you get 4k/0.005% and garanteed -86dB CMRR. Typical values actually will be on average 6dB better.

Okay. Getting spendy !

Of course, the last stage of the circuit I presented can be replaced by THAT124X, INA134 et al. but distortion and noise will usually be worse.

By not using laser trimmed on chip polysilicone resistors we get overall better resistors with lower noise and HD. It is all a tradeoff.

In my understanding the laser trimmed resistors are "Thin Film" rather than Polysilicone type (that I understand not to be laser trimmed) but this could be a matter of convention / terminology.

All 0.01% rated SMD resistors I know are << 10ppm singly.

I strongly consider that it is possible to create a better performing balanced line receiver using something like my suggested circuit for less budget than using IC's.

Of course, off the shelf IC's with key parts inside may be better in case of one off DIY due to simplicity.

Budget ? As in cost ? (or indeed pcb estate)
With 0.01% resistors ? If you have an inexpensive source then please let me know

Cheers

https://uk.farnell.com/w/c/passive-...sistance-tolerance=posneg-0.01pc&sort=P_PRICE

Thor

 

[ccaudle]

That is true but unfortunately, on the specified IC (INA165X), the outputs of the input buffers are not accessible.

Sure, but I was commenting on the multi-op-amp circuit which Thor added to the thread.

 

[abbey road d enfer]

With 0.01% resistors ? If you have an inexpensive source then please let me know

LT/AD have duals and quads at a cost of about $2 per resistor (1ku list price)
https://www.analog.com/en/parametricsearch/11449#/

 

[KA-Electonics.com]

Abbey: "Now I have one question: when does anyone need >80dB CMRR?"
Probably never in a line input.
But we get huge CMRR numbers for free in mic preamp owing to it being referenced to the input and having gain added to the spec.

Thor: "By not using laser trimmed on chip polysilicone resistors we get overall better resistors with lower noise and HD. It is all a tradeoff."
IIRC THAT use SiCr not polySi.

TI has an app note regarding their on-chip thin film.
On-Chip Thin Film Resistors Enable High-Performance Audio Circuitry https://www.ti.com/lit/an/sboa312/sboa312.pdf

There is one instance in which a super-high common mode impedance can be problematic with an active input.

That's when you have two devices at different chassis potentials: One mains grounded at or near 0V and the other floating at 60V (or 120V in 240V countries) because it has a 2 wire cord.
When the two are connected we certainly hope the leakage current is small, but when that leakage current is multiplied by a high Zcm, the common mode bootstrap amp or common mode rejection stage can overload.
The best example is an unbalanced "2-wire-AC-power-cord" device connected to a balanced input without a shield connection to the grounded device.
Tying the + balanced input to the RCA center pin and - to the RCA ground sends the leakage current through the inputs in common mode.

I'm not sure a 10MΩ Zcm is 10X mo-betta' than 1MΩ.

I like the OPs "T-bias" approach best. It raises Zcm 5X at the cost of a single low-tolerance resistor.
Less is more.