Why balanced XLR needs an inverted signal?

[OldHvyMec]

The AES48 is the standard that causes the cables to be neutral sounding and adds or subtracts nothing from the signal because of materials, length
or a difference in where it was used. As long as the component uses the same standard in their wiring. There is NOT a difference in how the product performs.
Atmas-sphere is one of the few that adhere to that standard. It's pretty easy to identify if it is being used. There is NOT an increase in volume going from a SE
RCA to an XLR. Mac and Krell both used the AES48 standard when they started using it. Atmasphere never changed. Mcintosh did. A 25.00 usd patch cable
using standard copper 100 feet long will sound exactly the same on every piece it's hooked to. It's simply not the case with how many other cables and
products are wired. It's also why RCAs can be influenced by how and what they are made from and what they are hooked to.

A phono stage is a good example of a place where you better pay attention to how the RCA was made and how long it physically is. UNLESS they have a balanced
circuit from the cart to the phono stage and adhere to the AES48 standard in the phono stage.

Ralph from Atmasphere has a few pretty good explanations/posts on various forums, one being AG and the other WBF. He may have posted here I'm not sure.

The main reason for AES48 was going from one recording studio to another and having the same results and no argument WHY they sound different or the same.
1940s maybe? Pro cabling is nothing fancy it is just wired to the same standard, BUT I've seen a LOT in the autophile world that do everything BUT wire to the
AES48 standard.

Ralph explained it rather eloquently to a fella named Ted Denny (SR) on WBF forums. Ted Denny listened or at least he quit posting on that thread. I believe he
might have learned something.

I still make my cable and have for 50 years. I did buy from a Mom and Pop in China, because of my hands. Very nice work, good-looking products, fast delivery
and I've never had a problem with well over 1000 cables being made. the products were made from what I wanted them made from and they had a few
suggestions also, like weaves and PTFE over silver or silver clad/copper.

It's been 3-4 years since I used them, but they sure were easy to work with and I NEVER had any issues with solder or screws coming loose after we had a little
chat and I sent them a bottle of Cat/White thread lock (heat required for removal) and the solder I wanted to use.

Regards

 

[antcollinet]

But why do I really need to invert the signal on pin 3? Cant I just connect pin 3 to GND and still have a balanced connection?

Because it is not about balancing the voltages, it is about balancing the impedances.

The cold signal needs to have the same output and input impedances as the hot signal. Otherwise the noise picked up in one is not equal to the noise picked up on the other, and they can't cancel out.

For more information:

https://www.hypex.nl/media/fa/d8/a3/1682342122/The%20G%20word.pdf

Start at section "Balance"


EDIT - ninjad multiple times above. Still, I'll leave it here for re-enforcement.

 

[sergeauckland]

Well, that's the thing - it actually doesn't. It just needs hot and cold to have equal impedances. There can be nothing at all on the cold pin but water and it works.

Exactly. That's why it's called balanced. The impedances are identical, (or should be) such that noise induces equally in both legs, which is then cancelled by the common-mode rejection of the receiver, which itself is balanced. That most outputs are differential, i.e. equal signals, but opposite polarity on both legs is incidental to the noise reduction from balancing the impedances.

S.

 

[DonH56]

Well, that's the thing - it actually doesn't. It just needs hot and cold to have equal impedances. There can be nothing at all on the cold pin but noise and it works.

That will provide some common-mode noise rejection, though not as good as when both sides are driven, but lower SNR and no distortion improvement compared to a truly differential circuit. That is the main difference between balanced (equal impedances on each leg) and differential (complementary -- in-phase and inverted -- signals).

 

[dfuller]

That will provide some common-mode noise rejection, though not as good as when both sides are driven, but lower SNR and no distortion improvement compared to a truly differential circuit. That is the main difference between balanced (equal impedances on each leg) and differential (complementary -- in-phase and inverted -- signals).

Right, differential is better. But impedance balancing works just fine in 99% of cases.

 

[antcollinet]

That will provide some common-mode noise rejection, though not as good as when both sides are driven, but lower SNR and no distortion improvement compared to a truly differential circuit. That is the main difference between balanced (equal impedances on each leg) and differential (complementary -- in-phase and inverted -- signals).

Impedance balancing is about common mode noise rejection.

Differential basically only gives you a higher (differential) voltage (this is why balanced outputs are typically 4V when unbalanced are 2V). This will only improve SNR by virtue of a larger signal with the same noise. This is why we normally see an approx 6dB better SINAD with balanced connection compared with Unbalanced (when there is no common mode noise and when SINAD is noise dominated)

I don't think either will give distortion benefit.

 

[Scytales]

Hello dear friends,

I understand that the balanced XLR needs an inverted signal wire (pin 3) so at the end, in the mixer, the pin 3 is inverted again and summed with the pin 1. This will make any noise that got picked up in the cable to get subtracted from pin 1 and pin 3. Awesome idea!

As others have already pointed out, a balanced transmission line isn't a matter of transmitting symmetrical signals on two wires, but a matter of balancing impedance. A balanced line works equally well with only one wire live where the design make that possible. Other resources to add to those already provided above :

1. https://thatcorp.com/datashts/A_Matter_of_Balance.pdf
2. https://sound-au.com/balance.htm
3. https://sound-au.com/articles/balanced-2.htm
4. https://sound-au.com/articles/balanced-io.htm
5. https://sound-au.com/articles/balanced-4.htm

Generating and transmitting two symmetrical signals opposite to each other is another thing altogether and is primarily done to increased signal level. For instance, the output of many current DAC chips living on single low voltage supply is differential in order to increase the output level. It's the same reason as to why some power amplifiers are bridged: to increase the output level.

However, with some balanced line designs, this two different aspects (impedance balance and symmetrical signals) are inextricably mixed. Hence the unfortunate generalization that this two aspects are the same thing, I guess.

Neither techniques is without challenges to work properly. As with all things, implementation is key to obtain the desired results.

 

[DonH56]

Impedance balancing is about common mode noise rejection.

Differential basically only gives you a higher (differential) voltage (this is why balanced outputs are typically 4V when unbalanced are 2V). This will only improve SNR by virtue of a larger signal with the same noise. This is why we normally see an approx 6dB better SINAD with balanced connection compared with Unbalanced (when there is no common mode noise and when SINAD is noise dominated)

I don't think either will give distortion benefit.

Differential operation cancels even-order harmonic distortion as well as rejecting common-mode noise. It also doubles the signal level compared to signal-ended, a 6 dB increase, but uncorrelated noise increases by 3 dB, yielding a 3 dB net improvement in SNR (all dB voltage referenced).

 

[bmc0]

Bruno Putzeys' "The G Word" (linked by others above) is an excellent article for those with a little background knowledge in electronics. It addresses the original question, though it's not really the main focus.

There is no requirement for both the "+" (hot) and "-" (cold) pins to be actively driven. Outputs where only the "+" pin is actively driven are commonly called "impedance balanced" to distinguish them from symmetrical balanced outputs. As long as the impedances of the + and - outputs are closely matched, rejection of common-mode interference by the differential receiver (i.e. balanced input) works all the same. A balanced output can be as simple as this:

Where R is some low value—perhaps 10-100 ohms for a line output.

Some equipment (certain power amps, in particular) with "balanced" inputs actually require symmetrical signals to work correctly. In my view, these are not proper balanced inputs.

 

[Scytales]

Some equipment (certain power amps, in particular) with "balanced" inputs actually require symmetrical signals to work correctly. In my view, these are not proper balanced inputs.

Yes. As an example, the Grandinote Supremo integrated amplifier I have very briefly reversed engineered here (in French) : https://www.homecinema-fr.com/forum/post181145688.html#p181145688

 

[DonH56]

Bruno Putzeys' "The G Word" (linked by others above) is an excellent article for those with a little background knowledge in electronics. It addresses the original question, though it's not really the main focus.

There is no requirement for both the "+" (hot) and "-" (cold) pins to be actively driven. Outputs where only the "+" pin is actively driven are commonly called "impedance balanced" to distinguish them from symmetrical balanced outputs. As long as the impedances of the + and - outputs are closely matched, rejection of common-mode interference by the differential receiver (i.e. balanced input) works all the same. A balanced output can be as simple as this:
View attachment 435520
Where R is some low value—perhaps 10-100 ohms for a line output.

A lot of equipment, including a lot of pro equipment, takes this approach to save cost and (sometimes) component count. It works well for low frequency common mode noise, such as that from AC power lines, but becomes less effective at rejecting RFI as the active stage's output impedance rises (due to falling loop gain) and ruins the impedance match. Still good enough for most applications, including long cable runs in concert venues, and RFI is handled by other means (simple filters, often just small capacitors across the inputs).

Some equipment (certain power amps, in particular) with "balanced" inputs actually require symmetrical signals to work correctly. In my view, these are not proper balanced inputs.

Virtually all differential circuits take a slight hit in distortion and noise if the signal is not completely symmetric, but it is usually a small hit. It happens because the input stage gets "unbalanced" by the asymmetry causing shifts in biasing and operating point that raises the distortion a bit (imperfect cancellation of even order products plus a slight rise in odd-order products) and adds a bit more noise. Usually in the mud (insignificant). It takes a bit more circuitry, sometimes an extra stage, to make the impact insignificant, so some designs eschew the extra parts to save money (plus some "audiophile" amps claim benefits from fewer components in the signal path, something I find debatable).

One concern is when there is no AC-coupling in the signal path, thus an output or input biased at other than ground (or the same common-mode voltage) can lead to significantly increased distortion. This is rare in the professional world IME, but I have seem it in cheap pro gear and some home (consumer) gear, usually when a single supply is used to save cost so the driver or receiver is at some fairly large voltage above ground. Most (majority IME) such devices use coupling caps, but a few cheap ones do not and depend upon the other end being AC coupled. One (more) reason to be wary of "DC coupled" components.

FWIWFM - Don

 

[Gorgonzola]

Note that is from a SE driver (output) on the left to a balanced source input on the right. It is not appropriate going the other way.

Thanks! Yes. This from the RANE reference ...

 

[DVDdoug]

@DVDdoug thanks for your reply! I am glad you fixed the noise problem on your "special application" without connecting PIN 3 to GND! But would you know why it worked, because it should have created a ton of noise, right? See: if you leave a wire (the pin 3 wire) connected to nothing, it will become an antenna and will pick a lot of noise, I mean, a lot!

I don't have XLR connectors and no pin-3 and no "ground" on the input. I've got RCA out (of my car stereo) RCA into the special gizmo* I built. It looks like any normal RCA-to-RCA connection. Any noise common to signal and ground is canceled out (because there is no "difference" into the differential amplifier). Once it's plugged-in, one of the differential inputs is grounded but the amplifier doesn't care and it still operates differentially.



*This gizmo is a sound activated lighting effect in my Van. It has automatic sensitivity adjustment in the software to adjust for different volume levels. At low volume levels it was getting activated by noise rather than responding to the music. I couldn't hear the noise because it was only affecting the lighting but I deduced what was happening. I tried several things to fix it, and making the input differential turned-out to be solution.

 

[bmc0]

As an example, the Grandinote Supremo integrated amplifier

For the price, you'd think they'd include a fancy input transformer too. That would double up on the "transformer mojo" some believe in and guarantee symmetrical drive to the power amps.

Virtually all differential circuits take a slight hit in distortion and noise if the signal is not completely symmetric, but it is usually a small hit.

A good point, though I was referring to equipment that exhibits severely degraded performance when driven single ended.

It works well for low frequency common mode noise, such as that from AC power lines, but becomes less effective at rejecting RFI as the active stage's output impedance rises (due to falling loop gain) and ruins the impedance match.

I'm sure you're aware, but a lot of balanced inputs (especially in cheaper gear) have rather poor common mode rejection at high frequencies anyway.

 

[DonH56]

A good point, though I was referring to equipment that exhibits severely degraded performance when driven single ended.

I'd expect around 6 dB or so hit plus a fudge factor when driven single-ended (that is the worst-case asymmetric input, especially if there's a DC offset as well). The saving grace is most components are so far below audible distortion and noise it just doesn't matter.

I'm sure you're aware, but a lot of balanced inputs (especially in cheaper gear) have rather poor common mode rejection at high frequencies anyway.

I'm aware. Even expensive gear, as it is a loop bandwidth issue, and wide common-mode bandwidth is rarely a design priority. RFI is solved by other means IME. What I have personally found is a lack of consideration of the signal return path and proper ground technique in cheaper gear.

 

[Lambda]

But why do I really need to invert the signal on pin 3?

Short answer: You don’t—depending on the input.
Sound on Sound Article

Longer answer:
For the best common-mode rejection ratio (CMRR), you want the impedance on hot and cold to be the same. Since the real impedance of an output buffer is complex, a good way to achieve this is by using two identical output buffers.

Also, the acceptable voltage swing of inputs and outputs is usually limited.


By driving both, you can get more voltage and therefore more signal-to-noise ratio (SNR) (usually).


See:
Audio Science Review Discussion

Why balanced XLR needs an inverted signal?

The AES48 is the standard that causes the cables to be neutral sounding and adds or subtracts nothing from the signal because of materials, length or a difference in where it was used. As long as the component uses the same standard in their wiring. There is NOT a difference in how the product...

www.audiosciencereview.com

It also doubles the signal level compared to signal-ended, a 6 dB increase, but uncorrelated noise increases by 3 dB, yielding a 3 dB net improvement in SNR (all dB voltage referenced).

One second question: some people say that if I cant invert the phase, I should not connect the pin 3 to GND because it makes noise worse. BUT if I dont connect pin 3 to GND then the signal on the pin 3 will "float", it will not be a stable 0 signal. If the signal on pin 3 floats, then when it gets inverted and summed with pin 1, a lot of "noise" will be heard but this noise will not be due to eletromagnetic field, but purely because pin 3 was floating. Right?

Audio Science Review Discussion
Short answer: It depends on the topology.
For example, if it's a transformer without a ceter tap (or equivalent), letting one output float will result in no (proper) output from the other side as well.

Shorting one side of a balanced/differential signal to ground makes it unbalanced or single-ended so you lose the benefits of differential operation

That said, you might be overgeneralizing or oversimplifying. If we look at the input-output topology in the picture above:

• With ideal components: Shorting one side of a balanced/differential signal to ground technically makes it unbalanced or single-ended, assuming the output impedance is not also zero.
• With realistic components: The impedance mismatch that results will reduce CMRR, but you still retain some CMRR—you don’t lose all the benefits of differential operation, just a portion of them.

Shorting the output to ground is undesirable, but again depends upon the internal circuit.

I agree—"depends on the internal circuit" is key.
Audio Science Review Discussion

but you lose common-mode noise rejection, reduce SNR, and increase distortion by operating a balanced circuit single-ended.

If you add "usually" and "(circuit-dependent)" to this statement, I would fully agree.
For example, if we look at a simple transformer-based version, I don’t see why this would necessarily increase distortion.

 

[KSTR]

IME, the by far dominating effect of balanced connection in the real world (of domestic HiFi and smaller Studio etc) is hum/buzz/noise reduction that comes from the differential sensing (at the source end, of course), avoiding mains leakage currents flowing through signal conductors which is the issue with unbalanced connections.

This differential sensing, measuring the difference between pin 2 and 3 at the source and re-referencing this difference to the local ground at the receiver is the main asset, and the only real requirement, of balanced signal transmission.

Impedance matching of the legs, and symmetrical drive are just the icing on the cake and may contribute less than 5% to the improvement, except for the occasional corner case.

 

[Sokel]

Impedance balancing is about common mode noise rejection.

Differential basically only gives you a higher (differential) voltage (this is why balanced outputs are typically 4V when unbalanced are 2V). This will only improve SNR by virtue of a larger signal with the same noise. This is why we normally see an approx 6dB better SINAD with balanced connection compared with Unbalanced (when there is no common mode noise and when SINAD is noise dominated)

I don't think either will give distortion benefit.

Differential operation cancels even-order harmonic distortion as well as rejecting common-mode noise. It also doubles the signal level compared to signal-ended, a 6 dB increase, but uncorrelated noise increases by 3 dB, yielding a 3 dB net improvement in SNR (all dB voltage referenced).

Yes,you can try it with REW or Multitone Analyzer, just apply opposite phase distortion to the existing and see it go vastly lower.
It's also a nice trick some top measuring devices use (at certain freqs) when the chip allows for THD compensation, right from the start.

 

[Sokel]

Example:

That's after adding opposite

..and the difference (normal is about -101.8dB THD+N for E-MU's loop, holden back by its DAC)

 

[Lambda]

Yes,you can try it with REW or Multitone Analyzer, just apply opposite phase distortion to the existing and see it go vastly lower.

I’d like to make a distinction here.
In practice, balanced differential transmission often results in lower distortion compared to single-ended unbalanced signals.

However, the causation is not that straightforward. The reduction in distortion may be a result of the interaction between the specific transmitter and receiver circuitry, rather than being inherently due to the balanced transmission itself.

Consider these simple cases:

• With a transformer-isolated balanced input and output, THD does not change whether the impedance of the hot and cold legs are matched or not.
• It also remains unchanged whether one side is referenced to ground or not.
• The current loop stays the same regardless.

So, to be precise, the improvement in THD may comes from the interaction between the specific input and output stages, rather than from the signal itself being (impedance) balanced and or symmetrical