Value/Merits of Different Attenuator Types in Design of Small Mixer

I'm designing a small mixer and I have a question about attenuators:

I'd like to put an attenuator (fader) between each one of the mic amps and the mixing network. The mic amps are all of the 600 ohm balanced variety. My intuition would be to use a ladder-type attenuator and keep my signals balanced throughout, but I realize that a T-attenuator or a bridged-T attenuator would also do the job. Am I correct in believing that my signal will no longer be balanced if it goes trough a T-type attenuator?

What benefit would a ladder-type attenuator have over the T-type attenuators in this situation? Does a completely-balanced mixer topology have any significant advantages over one which unbalances the signal in the attenuator and mixing network?

kswan said:

I'm designing a small mixer and I have a question about attenuators:

I'd like to put an attenuator (fader) between each one of the mic amps and the mixing network. The mic amps are all of the 600 ohm balanced variety. My intuition would be to use a ladder-type attenuator and keep my signals balanced throughout, but I realize that a T-attenuator or a bridged-T attenuator would also do the job. Am I correct in believing that my signal will no longer be balanced if it goes trough a T-type attenuator?

What benefit would a ladder-type attenuator have over the T-type attenuators in this situation? Does a completely-balanced mixer topology have any significant advantages over one which unbalances the signal in the attenuator and mixing network?

Click to expand...

The simple answer would be that a straight or bridged T will unbalance the signal.  We could elaborate on that, but lets not go there now.
I trust you are not confusing ladder attenuators with the typical fader potentiometer.  True ladders have the unique characteristic of contact noise decreasing in proportion to attenuation. 
Somewhere on the web is the 1961 Langevin catalog.  The section on faders and attenuators is a must read.  Tremaine (1st and 2nd edition) also covers the subject in depth including balanced mixing networks.

gridcurrent said:

kswan said:

I'm designing a small mixer and I have a question about attenuators:

I'd like to put an attenuator (fader) between each one of the mic amps and the mixing network. The mic amps are all of the 600 ohm balanced variety. My intuition would be to use a ladder-type attenuator and keep my signals balanced throughout, but I realize that a T-attenuator or a bridged-T attenuator would also do the job. Am I correct in believing that my signal will no longer be balanced if it goes trough a T-type attenuator?

What benefit would a ladder-type attenuator have over the T-type attenuators in this situation? Does a completely-balanced mixer topology have any significant advantages over one which unbalances the signal in the attenuator and mixing network?

Click to expand...

The simple answer would be that a straight or bridged T will unbalance the signal. 

Click to expand...

They will only if their output is connected to a non-floating node; a balanced (H or staight) attenuator would also unbalance the signal if it was connected to unbalanced nodes. A balanced connection has to be balanced from the source to the destination.
In practical terms, you want balanced attenuators AND balanced buss.
In more practical terms, you may not need to stay balanced, if the outputs of your preamps are floating.
I reckon you won't use very long connections, or have much interference, so you don't really need impedance balance. What you need is floating, in order to cancel differential voltage between units.

I had a look at Tremaine's The Audio Cyclopedia (Second Edition), and here's what he had to say about the matter:

9.9 Are balanced configurations used in mixer-pot designs? - Yes. In special cases balanced mixer pots are used. These are generally of the balanced bridged-T type as shown in Fig. 5-38. Balanced-H configurations may also be used; however, because of the cost and the fact a balanced-H configuration requires six rows of contacts compared to four for the balanced bridged-T pot, the latter is preferred. A bridged-T pot also has a lower noise level and requires less maintenance.

Click to expand...

It seems like the balanced bridged-T attenuators into a balanced mixing bus is the way to go, no?
But, almost all the schematics and block diagrams that Tremaine shows in Chapter 9, the chapter about mixers, use an unbalanced bridged-T attenuator. Those that do not, use a regular old T-attenuator. Could someone please explain to me, if you would, how and why a signal can be attenuated by a bridged-T or regular-T attenuator and still be perfectly OK for use in a mixing network? As suggested by user abbey road d enfer, I think that this probably has to do with floating output preamps, that is, preamps that do not reference common ground, only themselves.

Also, I poked around for that Langevin catalog and couldn't find it. Could someone point me to it?

Many thanks to everyone for their help.

kswan said:

Could someone please explain to me, if you would, how and why a signal can be attenuated by a bridged-T or regular-T attenuator and still be perfectly OK for use in a mixing network?

Click to expand...

Conversely, can you explain why it would be a problem to use any standard T or ladder? 

There are virtually no examples of balanced passive attenuators to be found in any production console.  The only places that one ever sees them tends to be at the send or receive end of a very long line, like a telco patch to a transmitter, or similar.  The majority of antique broadcast limiters used unbalanced ladders in floating configuration, and many of them drove telco lines to transmitter sites.  Some companies offered balanced attenuators as an option, and all recommended using another repeat coil before or after the unit if complete balance was required in practice. 

The unbal ladder is most commonly found due to cost, and Daven claims it to have the advantage of switching noise that lowers along with attenuation position.  It's chief disadvantage in my mind is it's minimum loss of 6 dB. 

I don't have any balanced attenuators anywhere amongst my piles of vintage passive gear, nor do I ever find it to be an issue. 

abbey road d enfer said:

What you need is floating, in order to cancel differential voltage between units.

Click to expand...

And that is how all of my circuits are connected. 

Inserting an unbalanced T or ladder between floating sources and loads will certainly function.  The input circuitry on a 1176 is a good example.  I'll argue that placing a unbalanced attenuator before a balanced or floating input will degrade the CMRR.  Might be acceptable, but not elegant.

Back to the original question "Does a completely-balanced mixer topology have any significant advantages over one which unbalances the signal in the attenuator and mixing network?"
Yes, in theory.  But the legendary consoles of yesteryear, to the best of my knowledge, were pretty much unbalanced internally.

If your balanced mixer design would allow rotary faders, simple dual section potentiometers would serve the purpose. 

Balanced circuitry is useful for sending signals over distances, since equal termination impedance and differential processing at the receiver will ignore noise picked up equally in both lines.  Inside a mixer there are far better ways to reduce interference than doubling or more the circuit complexity and cost.

In large (physical) consoles, it can be helpful to use a differential bus structure, where the local signals and their local grounds are both bused forward to the master section for differential processing, which accounts for differences between the master and local channel grounds (ground is a relative concept not a single voltage everywhere). This differential bus is typically not balanced, and doesn't use equal impedances in both buses, mainly because it doesn't need to inside a shielded chassis. Further the - or ground leg of the differential bus is typically summed using very low value resistors, to keep their thermal and opamp input noise current product contributions minimal. In fact using equal impedance buses would make more noise than the unbalanced version, and be a step backwards not an improvement.

The complexity and cost of performing pan and fader operations with balanced circuitry is an expense and complexity that just does not make sense inside a console unless you are looking for an audiophool marketing hook, to merchandise an exotic product with. Then it's more imaginary sizzle than steak. It would explain the higher price, but IMO not justify that expense.

Balanced attenuators primary utility is to be located outside, in between two balanced chassis, to not unbalance those signals, not for general use inside a product.

JR

JohnRoberts said:

This differential bus is typically not balanced, and doesn't use equal impedances in both buses, mainly because it doesn't need to inside a shielded chassis. In fact using equal impedance buses would make more noise than the unbalanced version, and be a step backwards not an improvement.
JR

Click to expand...

Could you verify that?  I was given the impression that SSL's busses were equal impedances.  And why would equal impedance busses make more noise?
Maybe a SSL expert could chime in.

How does this block diagram look? Am I correct in assuming that the third wire from the mic amp output transformers is connected to nothing, resulting in a floating output? Also, I am correct in assuming that a center-tapped output transformer in the make-up gain amplifier can be used to re-balance the mixed signal?



Thanks for the help.

kswan said:

How does this block diagram look? Am I correct in assuming that the third wire from the mic amp output transformers is connected to nothing, resulting in a floating output? Also, I am correct in assuming that a center-tapped output transformer in the make-up gain amplifier can be used to re-balance the mixed signal?

Thanks for the help.

Click to expand...

Looks correct to me.  I would float 2 wire on each.  With a 2 circuit low Z network running very short wire in a small chassis I wouldn't expect much trouble with noise or interference issues.  Relative to everything else(of bigger size)  you're kind of at 'ground zero'.

You don't need a center tap grounded output transformer; no one does that anymore.  Unless, again, telephone lines.  If you have a transformer, it's balanced or unbalanced.

Give me a reason why you need a T or a ladder over a simple 1K pot.  All I definitively see in going with a T or ladder is vast expense. 

Remember also, with a T, it's intended for matched conditions, otherwise it attenuates less than labeled, and with a different curve.  A ladder is preloaded, and doesn't change much. 

The input side of your preamps also suggest CT transformer ground; not wise if you use phantom power!  No one does that anymore either.

gridcurrent said:

JohnRoberts said:

This differential bus is typically not balanced, and doesn't use equal impedances in both buses, mainly because it doesn't need to inside a shielded chassis. In fact using equal impedance buses would make more noise than the unbalanced version, and be a step backwards not an improvement.
JR

Click to expand...

Could you verify that?  I was given the impression that SSL's busses were equal impedances.  And why would equal impedance busses make more noise?
Maybe a SSL expert could chime in.

Click to expand...

First, let's start with the definitions:
Balanced busses means balanced driver (in-phase and out-of) feeding two injection resistors per send and two busses.
What you have in an SSL (and many others, such as Soundcraft 2400) is ground-sense busses, consisting in one injection resistor fed by the active output from the channel, and another injection resistor sensing the channel's "ground" voltage.
In this configuration, since the bus is almost invariably differential, both halves produce the same noise level, resulting in 3dB more noise in the output, compared to a common-ground summing amp.
By reducing the value of the ground-sensing resistors, the lower source impedance applied to one half of the mixing stage helps reducing its noise, not dramatically however. *
But in most cases, the improvement in longitudinal noise reduction is tremendous, and largely compensates for the slight increase in silicon-generated noise.
Real balanced busses bring a definite improvement in performance, not only in terms of noise, but also in reducing the effects of stray currents, which generate x-talk and THD, at the cost of huge complexity.

* Typically, a bus receiving 64 channels will see a source impedance of ca. 100 ohms. The summing amp's input stage has to be optimized for such a low impedance, which will operate at ca.1nV/sqrtHz and 10pA/sqrtHz resulting in 200nV noise on both sides of the diff, for 280nV overall => -130dBu EIN which, once multiplied by the noise gain of 65 (36dB) gives -93dBu at the output.
By using low value res in the ground-sensing bus, the contribution of noise-current in that leg becomes negligible, so its noise is reduced to 140nV. The overall noise becomes sqrt (140exp2 + 200exp2) or 245nV, a reduction of about 1.15dB. One could argue that the ground-sensing side of the mix amp could be optimised for a much lower impedance, but that would mean operating devices at very high current (100+ mA).

I concede that I haven't actually run experiments with large bus structures comparing balanced vs, unbalanced differential buses, but was surprised to actually see apparently electrostatic coupling between virtual earth buses, in one unbalanced part of a larger console I did years ago so perhaps the equal resistor value differential could reduce crosstalk (not an option in that console). Another factor perhaps to consider for bleeding edge design is symmetry in terminating the summing opamp with similar impedance at both input ports. This is a pretty obscure and esoteric consideration so I will just mention it in passing and not attempt to put practical numbers on it, but in theory there could be a slight benefit from equal termination impedance there.  
----
I do not know what SSL actually does inside, a haven't spent a lot of time studying other console designs. There are enough SSL techs around here I expect somebody will share that info. I would be surprised if there aren't several different flavors of bus combining topologies used in their different models (I know I've used several).  
--------
While I don't feel like writing a book on the subject here, I did write an article on the subject 30 years ago and nothing much has changed my thinking on the subject, other than that digital combining gets a bad rap these days. There are practical tradeoffs with even simple differential buses. For smaller mixers it is acceptable to hard wire the ground or negative bus, and background the resistors not feeding signal to the bus to maintain the differential integrity, for larger consoles you may not want to accept effectively running at max bus noise all the time even with no signal being sent to bus. To optimize this requires switching both resistors. Certainly manageable but at a cost.  I will quibble about definitions another time. Sending dual opposite polarity feeds are in theory the better solution since we pick up coherent summing of the two signals vs, incoherent summing of the doubled bus noise, for 3 dB net gain.  While the cost increase is 6dB or more.

-------

While most people only focus on bus noise, the same mechanism that increases the noise floor, also increases phase shift, and distortion. Several decades ago I came up with a different way to reduce the noise gain of the sum bus. My solution was to substitute current sources for the resistors so the bus amps didn't see all that noise gain build up. Of course the current sources need to be adequately quiet.

I actually wish, digital combining really was as flawed as people argue, then there might be justification to dust off my old technology and make a better "analog" bus topology. Unfortunately even my improved analog approach has compromises that digital doesn't, in theory at least, and for competent executions I presume.

Sorry for the veer..

========
In response to Kyles posted schematic.. If your pre outputs are transformer balanced performance should be quite good. It could matter if the transformer - and bottom of pad is connected to the bus amp - through individual wires, or shares a common -bus. While crosstalk in a mono mixer is academic, i guess fader kill will care about that - path being shared or not.

Presumably that - node is connected to ground or has some impedance to ground. I need to be careful about making assumptions without seeing the innards of the make up gain stage.

JR

To emmr: I don't know what I was thinking in regard to those center-tapped transformers I decided to draw. I had a look at my mic amp schematic, and sure enough, no center-tapped transformers on either side of the mic amps. Don't know what I was thinking. See mic amp schematic below:

In this situation, my third pin on of both the input of the mic amp and the output of the buffer amp would just be connected to chassis, no?