Balancing an unbalanced circuit

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Matt, you're absolutely right. Looking at the block dgm of the 1646, it shows that the + input receives negative FB from the + output, and the + input receives NFB from the - output. If a parasitic resistance is introduced in one of the inputs, balance is compromised.
Whoa! That sounds like a serious design compromise to me.

Cheers

Ian
 
The CMRR of these circuits is seriously impaired when driven by a non-zero impedance.
With a nominal 1k impedance, only 10 ohms limit CMRR at less than 60dB. 100 ohms=>40dB

What fascinates me about the THAT1646 datasheet is that there are no any diagrams, as if it was difficult for them to turn on the AP. Also, only one sentence mentions that the impedance of the source should be small. Why didn't they show the CMR vs. Rsource diagram?
 
Its interesting that now that is considered an old part.... :unsure:

JR
Depends on what you call old. The datasheet has a copyright from 2015, but I believe its older. I would say that an IC that is at least 10 years old qualifies as an old part...
 
It is interesting how the faults of a very old part suddenly start to come out once people really start paying attention to it.

It's not really a "fault" though is it ? Simply a property / requirement of the part of the part that it typically needs to be driven from a very low impedance to ideally match the impedance of the 0V input reference.
 
It's not really a "fault" though is it ? Simply a property / requirement of the part of the part that it typically needs to be driven from a very low impedance to ideally match the impedance of the 0V input reference.
Perhaps "fault" was a poor choice of words, I guess that a better one is "details"
 
Last edited:
What fascinates me about the THAT1646 datasheet is that there are no any diagrams, as if it was difficult for them to turn on the AP. Also, only one sentence mentions that the impedance of the source should be small. Why didn't they show the CMR vs. Rsource diagram?
One learns soon enough... :D
BTW, the Spice model includes this effect rather adequately.
 
In THAT1646 datasheet they say
“The 1646 is pin-compatible with the TI DRV134
and DRV135, as well as the Analog Devices
SSM2142. "
The question is whether they are interchangeable because for SSM2142 I did not see the need to have Rs=0.
Anyway, all that hype just to emulate the behavior of a good transformer doesn’t seem important to me for the DIY community. A real DIYer will always use a good transformer.:cool:
 
Anyway, all that hype just to emulate the behavior of a good transformer doesn’t seem important to me for the DIY community. A real DIYer will always use a good transformer.:cool:
not me, but I guess I am not a real DIYer whatever that means.

I'm too cheap to use iron when I can do it with silicon, my favorite is cheap silicon (op amps).

JR
 
...my favorite is cheap silicon (op amps).

I agree somewhat, I still like to make designs with ordinary opamps, even if it means using them more in quantity than specialized chips. Whenever you need a specialized chip, you can't get them quickly or they've changed the case or the Chinese are faking them.

But I remember being very happy when I got two thick books of datasheets from Analog Devices about 30 years ago, (SSM 2016, 2017, 2018, 2402 etc ..)
 
The question is whether they are interchangeable because for SSM2142 I did not see the need to have Rs=0.
SSM publishes a rather detailed diagram of the 2142, that I have simulated.
Introducing a source resistance impairs CMRR. It can be seen that one of the outputs has a pretty constant output impedance of about 4.15k, but the other varies significantly.
So yes, I would say the 2142 also has a Zs=0 constraint.
 
DIYer will always use a good transformer.:cool:

No. They may or may not depending on priorities. A simple transformer solution may be subjectively excellent. But it probably won't exceed a 'silicon solution' in terms of audio fidelity wrt input = output. Although there are special configurations - "Zero Field" and others - that increase performance.
CMRR is, of course, a major spec' here. Imo the current "THAT" solutions are pretty much state of the art.
Atm Supply Chain is a general problem but it can affect all devices to some extent.
 
What's the problem with a simple impedence-balanced output?

The only thing it won't give you is the option of an inverted, unbalanced out.
 
What's the problem with a simple impedence-balanced output?

The only thing it won't give you is the option of an inverted, unbalanced out.

Covered in post #6:
"For the output you can either use impedance balanced outputs as Ian mentioned, or use one of the That Corp drivers, or just buffer with a dual op-amp, one configured as a unity gain buffer and the other configured as an inverting unity gain buffer.
"The dual op-amp approach really only has an advantage of increased output level swing from the symmetrical output signals. So again, back to understanding the maximum output level you need. If you only need up to +20 dBu then probably impedance balanced output is the best choice."
 
I'd be very curious if SSL had published anything about their common-mode bootstrap "in the mid-'80s."

Indeed, I have never found schematics which verify when that was used. The 4K series schematics on the technical documents forum of this site (ca. 1993) do not use that technique.

But this circuit then, through a coupling capacitor, directly drives input bias resistors CM3 and CM4

Yes, strange that it is drawn as unbuffered. I have never seen the full service documents for that device, so possibly that figure was an over-simplified drawing to fit within the size constraints. Also possible that it just accepted as good enough the increase in CM impedance, even though it would be much lower than could be achieved with a buffered CM signal.

Fig 12 from the same preprint (attached), where "TYPICAL IMPERFECT RECEIVER WITH UNEQUAL INPUT Z" is used to describe the garden-variety op-amp and 4 equal resistor "diff-amp." In the context of common-mode impedances, which are what's wholly responsible for the noise rejection in a balanced input stage, this statement is wrong!

If you look at the accompanying text in section 3.2.1 that is not about CM noise rejection, but about the behavior of the original cross-coupled output topology (i.e. pre-OutSmarts style design). The contention was that the unequal receiver resistances for each signal leg caused unequal current through the output sense resistors, which generates a common mode signal that the output circuit attempts to compensate, but due to limited loop gain in the common mode feedback loop increases the distortion of the output stage with increasing frequency.

The Sony Oxford design used an isolated DC-DC converter to float the entire output circuit so that it was truly floating, not trying to emulate a floating output like the cross coupled output circuit.

PrismSound do (or at least did in the past) do something similar. See for example the description of the output circuit in sections 6.1 and 6.2 of the DA2 converter from the late '90's (from this link, quoted below, emphasis mine):
Prism Sound DA2 manual

6.1. Analogue output connectors
Conventional connections to a balanced analogue input should use a screened twisted
pair lead. The DA-2 output pins 2 and 3 should be wired to the two conductors of the pair
and pin 1 to the cable screen. The other end of the cable should be connected in a similar
manner with the screen connecting to the chassis of the analogue input.
The DA-2 should be connected to an unbalanced input using only pins 1 and 2 of the output
XLR. Pin 2 should connect to the signal input and pin 1 should connect to the input ground.
Using this method the signal level is half the nominal value.
NOTE: It is possible to drive unbalanced inputs from pins 2 and 3 of the DA-2 analogue output but, because the
connection has been unbalanced, this will give slightly worse performance. However, if the output is connected
in that way, it is important that pin 1 is not connected at the unbalanced input end as this will short circuit one leg
of the floating output driver.

6.2. Interconnect screens - the pin 1 conditions
'Screen' connects directly to the chassis for the digital XLR and BNC connectors. The
RCA digital coaxial connections are transformer isolated with the screen coupled to the
chassis using a capacitor at the connector. The analogue output screens are the mid
points of the floating differential outputs. They are coupled to the chassis using a capacitor
from pin 1 to the chassis.


That output shield connection seems strange to me, but was probably to ensure no CM voltage between the output signal wires and shield.
 
SSM publishes a rather detailed diagram of the 2142, that I have simulated.
Introducing a source resistance impairs CMRR. It can be seen that one of the outputs has a pretty constant output impedance of about 4.15k, but the other varies significantly.
So yes, I would say the 2142 also has a Zs=0 constraint.

Agree. From memory I think it was basically assumed to be driven from a low- Z source even if not explicitly stated (or I missed it but that's what we had where I was using in any case).
SSM2142 specifically is in the past now.
But it had problems of consistency. I had a batch that c 75dB THD+N as opposed to typical 90dB.
Response from AD was simply that they met the published spec' (true).
Additionally more sensitive to grounding at the remote end if feeding an unbalanced input.
 
SSM publishes a rather detailed diagram of the 2142, that I have simulated.
Introducing a source resistance impairs CMRR. It can be seen that one of the outputs has a pretty constant output impedance of about 4.15k, but the other varies significantly.
So yes, I would say the 2142 also has a Zs=0 constraint.

Thanks. The datasheet for DRV134 which is an "improved version of SSM2412" states "The DRV134 input should be driven by a low impedance source such as an op amp or buffer."
It would be nice if someone did a simulation of both DRV134 and THAT1646 with Rs = 2k5 (standard 10k potentiometer on input), so to see how much the specifications of both drivers degrade theoretically.
 
No. They may or may not depending on priorities. A simple transformer solution may be subjectively excellent. But it probably won't exceed a 'silicon solution' in terms of audio fidelity wrt input = output. ...

I have always been of the opinion that these ICs (SSM2412, DRV134, THAT1646) were invented not to get a solution that works superior to transformers, but to get a simple solution that would be good enough to replace a transformer and much cheaper and much less weight and much less space to take up. Similar to when inductors are replaced by gyrators. Some characteristics of ICs that are better than transformers seem to me quite secondary.
 
Back
Top