Balancing an unbalanced circuit

[k brown]

All depends on which transformers you compare them to. The chips are certainly superior sounding to most small, inexpensive transformers.

 

[Newmarket]

Back in the German golden age of audio there was a weird conglomerate of Neuman, Telefunken, Siemens and among them, WSW

Thanks. New name to myself unlike the other three there.

 

[Newmarket]

Why didn't they show the CMR vs. Rsource diagram?

It's of technical interest but not particularly useful imo. It has been stated that a low impedance (low as in opamp output Z etc). So why would anyone choose to drive it from a significant source impedance ? Having said that, illustrations of how small source Impedance variations, as might be experienced with opamp or other circuits, affect CMRR would possibly be of use.

 

[moamps]

It's of technical interest but not particularly useful imo. It has been stated that a low impedance (low as in opamp output Z etc). So why would anyone choose to drive it from a significant source impedance?

It is obvious that for the proper functioning of this IC it is vital that the Rs be very small. If Rs is not small, some characteristics degrade considerably. If so, then this should be strongly emphasized in the datasheet and it should be shown what will happen if that Rs is not zero ohm, and what is the maximum value to obtain the average characteristics.
Not that every time someone wants to do something with that chip, more readers jump in with the famous sentence "you know, that IC should be connected directly to the buffer's output etc .." As if it's that secret ingredient that will make the dish famous. That IC should be sold exclusively with one NE5534 .. (NE5534 shoud be for free, of course)
As I said before, the absence of any diagrams in the THAT1646 datasheet doesn’t look professional to me.

 

[user 37518]

As I said before, the absence of any diagrams in the THAT1646 datasheet doesn’t look professional to me.

Unfortunately, THAT is great for making chips, not that great for making data sheets. I've noticed an error in one of their line receivers data sheets, they provide a circuit and state a wrong figure for the common mode input impedance using the resistor values they suggest in the diagram.

 

[Newmarket]

Well I thought it was a point at least fairly clearly documented in the datasheet and associated literature. But that's from memory. I have designed quite a bit with various permutations of diff amps and IAs etc. So maybe it's just sort of 'baked in' for myself.
You're right though. That buffer point does come up regularly

 

[Newmarket]

Unfortunately, THAT is great for making chips, not that great for making data sheets. I've noticed an error in one of their line receivers data sheets, they provide a circuit and state a wrong figure for the common mode input impedance using the resistor values they suggest in the diagram.

tbf all chip makers - analogue / digital / micros etc are subject to errors and corrections
They can throw you though. Esp when it's eg a microcontroller with complex / optional functionality.

 

[abbey road d enfer]

As I said before, the absence of any diagrams in the THAT1646 datasheet doesn’t look professional to me.

A diagram of the internals of the chip would not be terribly helpful.
I once tried to construct an equivalent using standard spice models, to no avail. There is a specialized balanced transconductance amp in the1646 that has no easy equivalent in Spice. Constructing a spice model, either with behavioral sources or to the transistor level would require a detailed silicon map.
The model supplied by THAT combines both levels (only 4 transistors but 11 behavioral sources).

 

[JohnRoberts]

I miss the good old days when chip makers published schematics of their IC internals, but I suspect ICs were simpler back then. IC technology supports devices with multiple junctions that don't have simple discrete device equivalents.

JR

 

[user 37518]

tbf all chip makers - analogue / digital / micros etc are subject to errors and corrections
They can throw you though. Esp when it's eg a microcontroller with complex / optional functionality.

It is true. There is nothing more frustrating than finding a value in a data sheet or book, which doesn't agree with your calculations and you loose a lot of time re-working your calculations and thinking that you must be wrong because the book or the data sheet can't possibly be wrong....

 

[JohnRoberts]

It was pretty common to find errors in old hand drawn schematics. Product data sheets should be reliable but sometime imprecise qualifications (like weighting or bandwidth for noise) make numbers hard to compare.

JR

 

[ruffrecords]

It's of technical interest but not particularly useful imo. It has been stated that a low impedance (low as in opamp output Z etc). So why would anyone choose to drive it from a significant source impedance ? x

because you might want to drive it from a fader for example where the source impedance might vary from zero to about 2K5 for example.

Cheers

Ian

 

[Newmarket]

because you might want to drive it from a fader for example where the source impedance might vary from zero to about 2K5 for example.

Cheers

Ian

Yes. But imo what you need to know is that you shouldn't do that. You need to buffer the wiper signal.
If you're not going to do that then there seems little point in using those parts with their precision trimmed values.

 

[ruffrecords]

Yes. But imo what you need to know is that you shouldn't do that. You need to buffer the wiper signal.
If you're not going to do that then there seems little point in using those parts with their precision trimmed values.

I agree. I have just been working on a design that uses that very device. I read the app notes to configure the chip correctly but nowhere do I recall such a warning. It might be somewhere in the datasheet but something that important really ought to be highlighted.

Cheers

Ian

 

[Newmarket]

I agree. I have just been working on a design that uses that very device. I read the app notes to configure the chip correctly but nowhere do I recall such a warning. It might be somewhere in the datasheet but something that important really ought to be highlighted.

Cheers

Ian

Under the "Applications" heading it does state as the second point.

"2. Both devices must be driven from a low￾impedance source, preferably directly from opamp
outputs, to maintain the specified performance. "

But fair comment if people think that should be highlighted more.

 

[JohnRoberts]

They could have built in active buffers, its only silicon and money . Anticipating fader/volume control use, they would want to have differential inputs to insure good pot kill. Looks like they decided to keep it simple.

I still haven't used one and am now unlikely to, since I only melt solder to repair stuff these days.

JR

 

[Newmarket]

They could have built in active buffers, its only silicon and money . Anticipating fader/volume control use, they would want to have differential inputs to insure good pot kill. Looks like they decided to keep it simple.

I still haven't used one and am now unlikely to, since I only melt solder to repair stuff these days.

JR

The THAT1606 has differential inputs. And a simpler CM implementation using one small value cap'.

 

[JohnRoberts]

The THAT1606 has differential inputs. And a simpler CM implementation using one small value cap'.

I don't want to drive this veer too far off into the weeds but audio paths that already have differential outputs (like many mic preamp topologies). A volume control could bridge between the two hots but that would require exceptional CMRR to deliver very good pot kill.

Sorry this is arguably overthinking this, and I am not a typical customer for these parts.

JR

 

[Newmarket]

I don't want to drive this veer too far off into the weeds but audio paths

I like an interesting veer tbh

 

[Matt Syson]

Lurching back to an 'advantage' of a real transformer, the common mode 'interference' that may be picked up on the interconnection will 'subtract' from the voltage headroom of an 'electronic' balanced output so for example a sub audio oscillation (drift) could push audio peaks into clipping. or if it is supersonic would introduce distortion on asymmetric audio signals as they approach the supply rails.
This then takes the requirements for a 'best' interconnection, to also include sub and supersonic filtering that probably should not include 'silicon' or other devices that may saturate at high voltage levels. Ferrite 'feed through filters' that have an internal zener like function (popular on computer comms ports) have a limit of 30 to 50 Volts for example.