Balanced line input output

[Schnitzel]

I'm in the process of modifying and upgrading my Midas XL200 mixing console. The small balanced input output sub PCBs  truly are a bottleneck in this desk (schematic attached). They're used for the channel inserts, direct outs, etc. and can be bypassed with a switch. When the insert is engaged it really makes the audio sound kinda flat.
After researching for a bit it came down to two designs: The so called 'voltage mode' configuration which can be seen on this website in figure 1 http://sound.whsites.net/project87.htm and 'current mode' configuration as can be seen the original circuit. It can also be found in the 1176 and many other pieces of equipment. What are the differences between those two designs? And most important which one will perform and sound better? Same question with the output driver. The cross coupled circuit  is supposed to offer transformer like behaviour (if one side is shorted to ground). It's often described as unstable with real world cables though I couldn't notice any problems concerning stability yet. The alternative would be a simpler circuit with a non-inverting and an inverting op amp which does not offer transformer like behaviour (figure 3A). Since 99% of my studio gear offers balanced inputs this would be worth considering.
I hope someone can share some experiences or thoughts on this issue.

 

[abbey road d enfer]

When the insert is engaged it really makes the audio sound kinda flat.

This shouldn't be; I would investigate on why this happens. Have you done any measurements (frequency response, distortion...)? Or maybe comparisons by recording the same signals with and without teh insert engaged?

After researching for a bit it came down to two designs: The so called 'voltage mode' configuration which can be seen on this website in figure 1 http://sound.whsites.net/project87.htm and 'current mode' configuration as can be seen the original circuit.

I don't really understand what you describe; a schematic would help understanding your distinctions.

The cross coupled circuit  is supposed to offer transformer like behaviour (if one side is shorted to ground). It's often described as unstable with real world cables though I couldn't notice any problems concerning stability yet.

Performance of the CCBD (Cross-Coupled Balanced Driver) is constrained by component tolerances. Very often they need some adjustment to optimize their operation. Stability shouldn't be an issue anymore with "modern" opamps (think 5534!).

Actually, you'd be better off using a THAT 1646, which performance is flawless and provides very good rejection of interference, even when connected to an unbalanced input.
Or the remote ground sensing output used by Soundcraft in some of their mixers. Check in the Documents->Soundcraft 6000
https://groupdiy.com/index.php?topic=44786.msg815215#msg815215
page 3 -> Ground cancel studio output

The alternative would be a simpler circuit with a non-inverting and an inverting op amp which does not offer transformer like behaviour (figure 3A).

  Though this solution has one advantage - simplicity - it has no other, in particular it does not have the capacity of rejecting interference, because its common-mode impedance is zero. Since balanced connections are used almost exclusively because of their interference rejection properties, this is a no-go. A basic unbalanced output will work equally well.

Since 99% of my studio gear offers balanced inputs this would be worth considering.

I wouldn't do that; issues with this arrangement have been familiarly known as the "Tascam problem". That's alway in the heat of the moment that you need to connect your balanced source to an input that you forgot is unbalanced, and you wonder why you have distortion and signal loss (particularly if you figure that out when the musicians are gone and you're mixing).

 

[moamps]

Though this solution has one advantage - simplicity - it has no other, in particular it does not have the capacity of rejecting interference, because its common-mode impedance is zero. Since balanced connections are used almost exclusively because of their interference rejection properties, this is a no-go. A basic unbalanced output will work equally well.

Just to be clear, do you referring to this circuit?

 

[ruffrecords]

Referring to fig 3A, Abbey wrote:

  Though this solution has one advantage - simplicity - it has no other, in particular it does not have the capacity of rejecting interference, because its common-mode impedance is zero.

Can you explain this please? If common mode impedance is the source impedance to common mode signals then isn't that the two 100 ohm output resistors? Also, if the common mode resistance was zero wouldn't that short out any common mode signals?

Cheers

Ian

 

[Schnitzel]

This shouldn't be; I would investigate on why this happens. Have you done any measurements (frequency response, distortion...)? Or maybe comparisons by recording the same signals with and without teh insert engaged?

It's subtle. But it's the same on every channel. I guess this is mainly caused be the sh*tty surface mount electrolytic caps used on those small sub PCBs. Unfortunately I don't have any measuring equipment for frequency response or distortion. But as I said the insert can be bypassed with a switch so you can easily compare. I have swapped almost all the electrolytics in the signal path with panasonic FR caps and it really made the console come to life. I have already owned many consoles and absolutely love the Midas EQ. That's why I want to stick to this desk and try to get it to sound a little bit more like a large format console. I mean it sounds great already but there's always room for improvement.

I don't really understand what you describe; a schematic would help understanding your distinctions.

http://www.jhbrandt.net/wp-content/uploads/2014/11/Design_of_High-Performance_Balanced_Audio_Interfaces.pdf
The two basic designs can be found here on page 5 (voltage mode vs current mode configutarion). The circuit from Elliott Sound Projects is pretty much the voltage mode configuration and the original circuit in my mixing desk (schematic I posted) is designed like the current mode configuration. I'd like to know which one is better or discuss the advantages and disadvantages of both designs.

Actually, you'd be better off using a THAT 1646, which performance is flawless and provides very good rejection of interference, even when connected to an unbalanced input.

I thought about that. The only disadvantage would be slightly higher cost...  What about the DRV134 from TI?

Or the remote ground sensing output used by Soundcraft in some of their mixers. Check in the Documents->Soundcraft 6000
https://groupdiy.com/index.php?topic=44786.msg815215#msg815215
page 3 -> Ground cancel studio output

Thanks, that's an interesting circuit! I'll have a closer look at it.

 

[abbey road d enfer]

Just to be clear, do you referring to this circuit?

Yes.

 

[abbey road d enfer]

Referring to fig 3A, Abbey wrote:

Can you explain this please? If common mode impedance is the source impedance to common mode signals then isn't that the two 100 ohm output resistors?

Yes, the common mode impedance is 50 ohms, which is nothing compared to what is necessary in order to ensure decent CMRR.

Also, if the common mode resistance was zero wouldn't that short out any common mode signals?

That would be correct if it was really zero, but conductors and contact impedance defeat it. In the case of differential ground voltage, we are dealing with very low source impedance of the interfering signal.

 

[abbey road d enfer]

http://www.jhbrandt.net/wp-content/uploads/2014/11/Design_of_High-Performance_Balanced_Audio_Interfaces.pdf
The two basic designs can be found here on page 5 (voltage mode vs current mode configutarion).

OK; since the terminology there is not a commonly-accepted one, it's agood thing you post the link.

The circuit from Elliott Sound Projects is pretty much the voltage mode configuration

Actually it's a hybrid mode that combines both "modes". I'm not sure there are any significant advantages in this arrangement.

and the original circuit in my mixing desk (schematic I posted) is designed like the current mode configuration. I'd like to know which one is better or discuss the advantages and disadvantages of both designs.

They're all sensitive to component tolerance for achieving good CMRR; they all can be trimmed. However HF CMRR may be slightly better behaved with the "current mode" arrangement. Anyway, they all need RFI/EMI protection at the input for optimum performance. There was a time when IMD/THD performance of the "current" mode was measurable, but now, with the modern opamps, it's a moot point.

  What about the DRV134 from TI?

DRV134 (as well as the SSM2142) is a laser-trimmed implementation of the original Pontis (Hewlett-Packard) circuit. The 1646 uses a significantly different arrangement that results in much better performance in the presence of unbalanced or poorly balanced inputs. Particularly, a typical CCBD generates significant distorted short-circuit currents when incorrectly balanced, which the Outsmart technology of the 1646 does not.

 

[ruffrecords]

Yes, the common mode impedance is 50 ohms, which is nothing compared to what is necessary in order to ensure decent CMRR.

Still not totally clear to me why this is a bad thing. CMRR depends on the total impedance of the transmitter plus the receiver in each leg being the same.

I don't think I know of any differential semiconductor driver with a common mode impedance determined by anything other than its output impedances. Does this mean  they are all inherently poor?

Cheers

Ian

 

[PRR]

> mainly caused be the sh*tty surface mount electrolytic caps used on those small sub PCBs.
> I thought about that. The only disadvantage would be slightly higher cost...

So on the one hand you deplore penny-pinching, OTOH you fret about higher cost.

Yes, cleverness "can" re-factor this conflict. But in general, if your work has any value, I can't see that the price of any of the likely clever chips is a real problem.

> Unfortunately I don't have any measuring equipment for frequency response or distortion.

Maybe it's just me. But in my opinion, after you get past plug-and-play, THE first requirement is trustworthy measurement tools.

A carpenter or mason lives by his tape and level.

 

[abbey road d enfer]

Still not totally clear to me why this is a bad thing. CMRR depends on the total impedance of the transmitter plus the receiver in each leg being the same.

Not exactly. It's a balanced bridge, so it's the ratio of impedances that is at work there.
You're right to remember that CMRR is pertinent to a connection, which means that both source and receiver are involved.

I don't think I know of any differential semiconductor driver with a common mode impedance determined by anything other than its output impedances.

The typical Cross Coupled Balanced Driver has a CM impedance that's much higher than its actual output impedance.

Does this mean  they are all inherently poor?

Only the "dual unbalanced" type is inherently poor (=non-existent) in terms of rejection.

 

[ruairioflaherty]

It's subtle. But it's the same on every channel. I guess this is mainly caused be the sh*tty surface mount electrolytic caps used on those small sub PCBs. Unfortunately I don't have any measuring equipment for frequency response or distortion. But as I said the insert can be bypassed with a switch so you can easily compare. I have swapped almost all the electrolytics in the signal path with panasonic FR caps and it really made the console come to life. I have already owned many consoles and absolutely love the Midas EQ. That's why I want to stick to this desk and try to get it to sound a little bit more like a large format console. I mean it sounds great already but there's always room for improvement.

Even a tiny level difference can make something feel less impressive/smaller when doing A/B test.  The very first test would be to make sure that levels are truly matched when you are comparing the two signal paths.

 

[Schnitzel]

So on the one hand you deplore penny-pinching, OTOH you fret about higher cost.

Yes, cleverness "can" re-factor this conflict. But in general, if your work has any value, I can't see that the price of any of the likely clever chips is a real problem.

Yeah, you're right. I'll go with the THAT chips.

Even a tiny level difference can make something feel less impressive/smaller when doing A/B test.  The very first test would be to make sure that levels are truly matched when you are comparing the two signal paths.

Thanks, that's definitely another possibility. I will check!

 

[Monte McGuire]

Not exactly. It's a balanced bridge, so it's the ratio of impedances that is at work there.
You're right to remember that CMRR is pertinent to a connection, which means that both source and receiver are involved.
The typical Cross Coupled Balanced Driver has a CM impedance that's much higher than its actual output impedance.
Only the "dual unbalanced" type is inherently poor (=non-existent) in terms of rejection.

The high common mode impedance of the cross coupled stages makes the connection less affected by impedance imbalances, but it's not reasonable to state that you don't get CMRR because of the low common mode source impedance, or that interference is somehow magnified by this. The penalty is simply that increasing resistor imbalances on each leg reduce CMRR pretty rapidly, especially compared to the cross coupled case, so that the cross coupled stages have better numbers with sloppy gear. But, the CMRR you can get is far from zero.

One downside of the cross coupled stages is that they can be more sensitive to cable capacitance (and thus long lines). But like anything, it's all a matter of degree and the cross coupled stages work pretty well without much effort.

 

[abbey road d enfer]

The high common mode impedance of the cross coupled stages makes the connection less affected by impedance imbalances, but it's not reasonable to state that you don't get CMRR because of the low common mode source impedance, or that interference is somehow magnified by this. The penalty is simply that increasing resistor imbalances on each leg reduce CMRR pretty rapidly,

In one stroke of the pen keyboard, you write something and its contrary!

  especially compared to the cross coupled case, so that the cross coupled stages have better numbers with sloppy gear.

Just the same, a real xfmr would have better figures than a CCBD.
If the question is "is it better?", the answer is undeniably yes. If the question is "is it good enough?", it depends on many factors (level of interference, sensitivity/acceptance, actual S/N), as you know well.
Anyway, my primary concern with the "absolutely non-floating balanced driver" is its propensity to generate distorted currents when met with an unbalanced receiver.

  But, the CMRR you can get is far from zero.

How far? As I wrote earlier, we're dealing with a balanced bridge.
If the impedance of each leg of the source and the CM impedance of each leg of the receiver were perfectly balanced, CMRR would be perfect, irrespective of the CM Z of the source, but it's never the case in practice.
In the real world, , the higher the CM impedance, the higher the actual rejection.

 

[jensenmann]

Schnitzel, what did you end up doing with your XL200?

 

[Schnitzel]

Kind of a late response, sorry. I designed small sub-PCBs with the THAT 1200/1646 integrated circuits to replace the original balanced IO sub-PCBs from Midas. I'm still testing but so far I'm very pleased with the sound improvement! Inserts are now completely transparent, works great.