Weird impedance balancing behavior

[EmilFrid]

I once decided to experiment with impedance balancing. The only unbalanced gear I had was a little tape machine. According to the service manual it had an output impedance of at least 10k. So I followed the routine, disconnecting the ground wire of the unbalanced source from ground, and connecting it to cold of the balanced destination through a resistor of 10k.

So, I did notice an improvement in terms of noise performance, but it was a very smal improvement. I then tried hooking up the ground wire from source directly to the cold of the destination, and that made a huge improvement.

Anyone know why? I found it counterintuitive, since (in my world) matching the source impedance would make it behave more like a balanced system. Or am I missing something? Was i also supposed to take the destination impedance into account?

This was just a small experiment, but left me with some interesting thoughts...

Emil

 

[JohnRoberts]

Just to be clear "impedance balanced" generally describes the practice of using a passive resistor for the negative leg of a balanced audio output. To maintain balance this resistance should equal the source impedance of the positive leg.

10k ohm sounds awfully high for output source impedance, perhaps they are talking about drive capability or minimum load impedance. Professional gear typically use 50-200 ohm source impedance, consumer or semi-pro gear can use source impedances as high as 2k ohm.

It is important for impedance balance configuration to deliver expected performance that the negative and positive source impedances be the same resistance.

JR

 

[EmilFrid]

Just to be clear "impedance balanced" generally describes the practice of using a passive resistor for the negative leg of a balanced audio output. To maintain balance this resistance should equal the source impedance of the positive leg.

10k ohm sounds awfully high for output source impedance, perhaps they are talking about drive capability or minimum load impedance. Professional gear typically use 50-200 ohm source impedance, consumer or semi-pro gear can use source impedances as high as 2k ohm.

It is important for impedance balance configuration to deliver expected performance that the negative and positive source impedances be the same resistance.

JR

Thanks John. I too thought that sounded like a crazy high output impedance, so (forgot to mention this) I put a 500 ohm trim pot in place of the resistor after the initial disappointment, but the performance was still the best with the pot set to zero. This confused me, and still does.

 

[ruffrecords]

Which tape machine is it?

Cheers

Ian

 

[EmilFrid]

Which tape machine is it?

Cheers

Ian

It was my old fostex g16 (R.I.P.).
I went to the studio storage earlier to find the manual for it, and it turns out I've been a complete idiot. It says LOAD IMPEDANCE 10k or higher. I must've missed that, and not double checked even though it came across as a ridiculously high source impedance.

 

[EmilFrid]

Still interesting that I got the best results by not using a resistor at all. I can't understand why.

 

[Bo Deadly]

Still interesting that I got the best results by not using a resistor at all. I can't understand why.

Because if the destination is properly balanced, then you're connecting the unbalanced ground of the source to the cold leg of the balanced destination. This leverages the balanced input and disconnects the grounds from each unit. This is common method for interfacing unbalanced gear with balanced gear.

But it cannot be done at the source without considering the destination because it doesn't work if you plug into an unbalanced destination. In some cases you'll leave the ground of the source gear floating which will result in massive noise. So it has to be done on a case-by-case basis. I have patchbay cables that do this but I must admit sometimes I forget how the plugs are wired and I have to unscrew the shell to look and see if the cold is connected to ground. So you could make a cable and put a big label on the cable that says "0V to cold" or something like that.

 

[EmilFrid]

Thanks for the response @Bo Deadly , I really appreciate it. So since my destination is properly balanced, this makes the directly-to-cold configuration work the best for me. Is that correct? I always thought you had to match the impedance of the hot leg.

 

[ruffrecords]

It was my old fostex g16 (R.I.P.).
I went to the studio storage earlier to find the manual for it, and it turns out I've been a complete idiot. It says LOAD IMPEDANCE 10k or higher. I must've missed that, and not double checked even though it came across as a ridiculously high source impedance.

I still have a Fostex R8. It uses a standard Dolby chip which feeds the outputs directly at consumer voltage levels (-10dBV) The signal comes staright from an op amp so the output im[pedance could be very very low which explains your results.

Cheers

Ian

 

[Bo Deadly]

Thanks for the response @Bo Deadly , I really appreciate it. So since my destination is properly balanced, this makes the directly-to-cold configuration work the best for me. Is that correct?

Yes. It's a significantly superior arragement but only if you know what's at each end and know it will work.

I always thought you had to match the impedance of the hot leg.

No. Impedance matching is for high frequency stuff like radio where the wavelength of the AC signal is short relative to the length of traces. When that happens the geometry of traces causes reflections and voltage doubling and all sorts of stuff that is fixed if the source impedance is increased to match the characteristic impedance of the line. For anything below 100kHz, that is not applicable.

Almost all audio signal lines are now "bridged" instead of matched where the source impedance is low and the destination impedance is high. For line-level signals input impedacnes of 10K is very common. Typical source impedances for line signals is 30-100 ohms and that "buildout" resistance is only to isolate the driver from the line which might be stressed if loaded too much or oscillate and be generally unstable when loaded with a lot of capacitance. Guitars are 30k-ish so "bridged" inputs for that is 1M. Speakers drivers are a fraction of an ohm into 4-16 ohms so that's "bridged" too. Only really old stuff like like Pultecs, early versions of the 1176 and tube gear used 600 ohm characteristic impedance lines but that was largely because transformers were all designed around 600 ohms. Probably so that when stepping down from a high Z tube stage it wouldn't loose too much signal level. Or maybe it had something to do with the telephone tech at the time. Not sure why exactly actually. Before my time.

 

[EmilFrid]

@Bo Deadly appreciate your thorough response. Thanks. But I think I was unclear when I said "matching". I meant matching the source impedance of the cold leg with the source impedance of the hot leg, so that the balanced destination sees the same impedance to ground on both of them, thereby making for a better CMRR.

In my case it might be like @ruffrecords said; since the g16 has an opamp at its output (like the r8), the impedance is so low that just tying the tape machine's ground to cold was optimal (well, almost)

 

[Bo Deadly]

@Bo Deadly appreciate your thorough response. Thanks. But I think I was unclear when I said "matching". I meant matching the source impedance of the cold leg with the source impedance of the hot leg, so that the balanced destination sees the same impedance to ground on both of them, thereby making for a better CMRR.

Ah, ok. Sure. You could do that. You would have to figure out what the resistor value is though. If you open it up and look at the output you might find a resistor and measure it with a meter. Then insert a resistor of the same size between ground and cold. Or remove the old one and use two that are closely matched. That will give you the best CMRR. Although you probably wouldn't notice a difference. CMRR is overrated compared to separating grounds by either connecting ground to cold when connecting to a balanced in or by said resistor being between ground of the source and ground of the dest when connecting to an unbalanced in.

 

[JohnRoberts]

Thanks for the response @Bo Deadly , I really appreciate it. So since my destination is properly balanced, this makes the directly-to-cold configuration work the best for me. Is that correct? I always thought you had to match the impedance of the hot leg.

What kind of noise are you trying to eliminate?

The balanced output (same) impedance criteria for feeding a balanced input stage is important for reducing noise pickup in the wiring between stages. In simple terms the same impedance + and - leg, both pick up environmental noise the same amount. The balanced input is also differential so any noise voltage picked up by the negative leg will subtract from noise picked up by the positive leg.

There are other sources of interface noise that could dominate common mode (same in both legs) noise.

JR

 

[EmilFrid]

I had a bit of a cable run, so I wanted to keep some environmental noise out of the preamp of my mixer. As I said, it worked like a charm if I kept the resistance at 0 ohms or very close to 0.

For the record, I didn't have a huge noise problem to begin with. I could hear it with the preamp set very hot, and I just wanted to do some experimenting and see what it could result in. But the difference (when everything maxed out) was actually impressing. Couldn't hear much difference between my truly balanced setup and this.

 

[abbey road d enfer]

The actual output impedance of an opamp is very low at low frequencies (a few microohms), and increases with frequency to reach maybe a fraction of ohm at 20kHz. There is usually a capacitor, which impedance decreases with frequency (10-100 ohms at 50 Hz, 1-10 ohms at 20kHz) so the capacitor's impedance is dominant. This is what you need to match first. Matching the opamp's output impedance could involve a small inductor.
There is no way you can properly increase CMRR by trying to match a capacitor with a resistor

 

[MisterCMRR]

I think the confusing issue in this discussion is that "grounding" is what keeps the chassis voltage between the source (unbalanced in this example) equipment and the destination equipment low (zero if zero-resistance wire was a real thing). A lot of equipment with unbalanced outputs is consumer stuff and uses a 2-prong AC power connection. An unbalanced to unbalanced connection allows the chassis to get "grounded" to the destination chassis via the "shield" of the unbalanced cable. But, if you connect that shield to the "low" balanced input and the "signal" of the unbalanced cable to the "high" balanced input, the chassis of the source will "float" at a significant fraction of the AC line voltage. This will challenge even the best balanced input to reject 60 volts of AC! The proper solution (it always works) is to use a standard balanced (low, high, and shield) cable and plug one end into the balanced input. But, at the source end, tie cable shield and signal low to the sleeve of the unbalanced output - and, of course, signal high to signal high. This connection allows the small (generally only a mA or so) chassis-to-chassis current to flow in the cable shield, where it doesn't cause noise. The balanced input is now sensing the voltage directly at the unbalanced output via the cable's twisted pair. The inter-chassis current will not flow in the same conductor as signal low - which is why unbalanced interconnections have noise problems in the first place. Now, to improve the CMRR of this setup, you should find out what components exit in the unbalanced output - between the op-amp output pin and the unbalanced signal connector pin. Making a duplicate of that network (often two resistors and a coupling capacitor, too) and connecting it between the cable signal low and the unbalanced output low side (shield) will make it a truly impedance-balanced output. This often improves CMRR by at least 30 dB). See the schematic and explanations under section 2.4 of Jensen Application Note 3 (attached).

 

[abbey road d enfer]

The balanced input is now sensing the voltage directly at the unbalanced output via the cable's twisted pair.

It works also the other way round, with a balanced output feeding an unbalanced input, on the condition that the balanced output is floating; the unfortunately too common arrangement of two opamps delivering opposite phases without cross-feeding, such as the one in AN003, does not conform to this need.
Fortunately, many manufacturers have abandoned this arrangement in favour of a cross-coupled arrangement, particularly since monolithic versions are commonly available, or for an inexpensive, but quite efficient if well implemented, impedance balanced arrangement.

 

[Newmarket]

The actual output impedance of an opamp is very low at low frequencies (a few microohms), and increases with frequency to reach maybe a fraction of ohm at 20kHz. There is usually a capacitor, which impedance decreases with frequency (10-100 ohms at 50 Hz, 1-10 ohms at 20kHz) so the capacitor's impedance is dominant. This is what you need to match first. Matching the opamp's output impedance could involve a small inductor.
There is no way you can properly increase CMRR by trying to match a capacitor with a resistor

So I take it that you advocate always including the capacitor in the "cold impedance" when implementing a "passive" impedance balancing solution ?
Interesting because IIRC many commercial implementations (and tbh some of my DIY interface buffers etc) don't include this - just a resistance.
I've been wary of including the capacitor given the tolerances of a typical electrolytic (say +/-20%) wrt matching comapared to the +/-1% (or better) resistors that I've used. But plugging in the numbers might make me reconsider that.
Thanks.

 

[abbey road d enfer]

So I take it that you advocate always including the capacitor in the "cold impedance" when implementing a "passive" impedance balancing solution ?

I do.

Interesting because IIRC many commercial implementations (and tbh some of my DIY interface buffers etc) don't include this - just a resistance.

IMO it's wrong.

I've been wary of including the capacitor given the tolerances of a typical electrolytic (say +/-20%) wrt matching comapared to the +/-1% (or better) resistors that I've used.

That is obviously a concern, however, capacitors from a same batch usually have values that are close enough. You can't "match" an additional capacitor with one that already exists. You must replace the existing one.
Remember that it's important for LF CMRR, so it will probably improve hum rejection, but be unsignificant for harsh buzz such as that from dimmers

 

[abbey road d enfer]

See attachment.
For a typical circuit with 100 ohms output Z and 20k input Z.
Green is CMRR with 10 ohms unbalance, purely resistive.
Brown is with a 100uF cap on one leg only.