Line Mixing Console design with THAT ICs

Looking at THAT Corp’s datasheets, they seem to have designed a set of ICs specifically suited for construction of mixing consoles. I’ve been toying with an idea to design and build a 16 channel line level mixing console with a fully balanced summing buss (since I have several mic preamps I like to use, including my hot-rodded G9), similar to the Speck Lilo or the Neve 5088 (but without all those $$$ transformers… Yikes!). I want to use it primarily for mixing down the outputs of my DAW, but I want some flexible routing options as well.

So, here’s the main question: Is the advantage gained by using a fully balanced summing buss (if any – this is a topic of much debate) offset by the extra silicon the signal has to pass through to achieve it?

Using the THAT chips would require the following extra chips when using a balanced summing buss: a balanced line driver (THAT 1646) to convert the panned and level-controlled signal to balanced to feed the buss, and a balanced line receiver (THAT 1200) to convert the buss signal to unbalanced for master level control (since stereo balanced level control is so expensive). This signal, of course, has to be converted again to balanced for output, but it would in a normal console anyway.

Here’s the signal path: Input goes to a full differential buffer using four NE5524s. From there, it goes to Insert Send/Return (or is hard bypassed by a relay). Next is the THAT 1200 line receiver, which debalances the signal. The channel fader is next, then on to another NE5524 for a 16 dB gain boost prior to the Pan pot. Output of the Pan circuit then goes to a unity gain buffer built on an OP4132 (used here as an impedance converter, since the 1646 requires a low source impedance) and then into the THAT 1646 balanced line driver. The 1646 feeds the balanced summing buss through 10k resistors.

The Summing buss feeds the input of a THAT 1512 gain module, which adds the necessary 20dB of makeup gain, and converts the balanced buss signal to an unbalanced signal for level control. The unbalanced summed signal then goes to a relay for selecting a buss insert (which uses the THAT 1646 line driver and 1200 line receiver for balancing/unbalancing the balanced insert connections) or bypassing the insert. The signal level for the summed buss is set with a fader, then goes to an OP4132 unity gain buffer for impedance conversion prior to a THAT 1646 for conversion back to a balanced signal for a subgroup output. The unbalanced signal after the OP4132 buffer also connects through a 10k resistor to a Master Summing buss, which uses once again the 1512 to provide makeup gain. Then on to a fader for Master level setting, then, finally through another impedance buffer (OP4132) prior to final conversion to a balanced signal for the master output. Whew!

I was thinking of breadboarding a channel and seeing if there was a noticeable sound quality difference in the two summing approaches, but it occurred to me that I would need all sixteen channels to really assess how well it works. This is no cheapie weekend project – just the faceplates from Front Panel Express are $1600! And those THAT chips run about $5 a piece. But, quality line mixing consoles are not cheap either. The Speck Lilo is around $9k and the Neve is (gulp!) around $30k. So, flexibility AND quality exacts a steep price.

Thats a lot of written information.

Ya know, schematics would make this a lot easier.

Does balanced buss mean you kiss her on both cheeks? :roll:

Sounds like a lot of parts, complexity, expense to do pretty basic stuff..

IMO bus should be properly differential, but the primary benefit of balanced is CMR of interference which shouldn't be a huge issue "inside" a mixer chassis.

The few dB of S/N improvement from 2x signal vs. 1.4x noise is likewise IMO not worth the trouble as bus noise floor should not be the limiting noise floor for real world signal sources when using a modern premium opamp for sum amp.

JR

Thanks for the input John. I was hoping you'd respond.

I think I see the difference - differential might give me the signal quality and integrity I want without the complexity of a balanced buss. It also saves a lot of pricey chips. The schematics I came up with for this should be easy to convert to a differential buss. I'll give that a try.

Kit - the schematics are really big, so I didn't bother posting them for fear they would be too hard to read. If you have Eagle layout software, I can send you the files to look at.

You should listen to Bill Whitlock's discussion of the In-Genius balanced input chip here:

http://www.aes.org/sections/uk/meetings/a0705.html

You probably would want to dump the 5534 buffer idea and use this chip instead for your inputs. Your CMRR performance will be top-notch.

That also makes the Out-Smart chip to handle your balanced output needs but I prefer using a lineout transformer with a DOA driver in my current builds.

As far as a balanced mix buss, this increases the complexity with very little increase in performance unless you are planning on building this into a desk. For rack mounted mixers, keep it simple.

If you're dead set on keeping it all balanced and/or differential you may want to look at fully-differential op-amps such as the TI OPA1632. The THAT chips are nice, but a diff-in/diff-out amp makes some of the bits you describe much easier.

JDB.

I agree - schemos would help!

Do you even need the 4amp input buffer of 5534s? What topology needs 4amps anyway?

You could just have the insert point passively switched and maintain balanced path. Let your DAC drive the insert device...unless you have lots of 600ohm inserts should be OK. SPL Mixdream does this. Or just leave out the inserts and put it on a patchbay.

Also why 4132 after pan circuit? thats a quad opamp no? Surely you'd need one amp buffer each pan leg, L and R? with a quad you could also use that as your fader booster amp and have one spare for a meter driver or something?? Save some chips...

Also would you not need two THAT 1646s after the pan pot, one each for left and right..as each buss is balanced.

Forgive me if I'm not following. That was a lot to read - reminds me of some of my threads! LOL

-Tom

I think the main reason for a balanced mix bus are the difficulties to keep a reallly clean ground in large consoles (~48 channels onwards). 16 channels shouldn't be a problem with an unbalanced mix bus and would make the signal path a lot simpler.

Michael

[quote author="Michael Tibes"]I think the main reason for a balanced mix bus are the difficulties to keep a reallly clean ground in large consoles (~48 channels onwards). 16 channels shouldn't be a problem with an unbalanced mix bus and would make the signal path a lot simpler.

Michael[/quote]

Not to harp on this, or repeat myself, yes grounds are a place to dump garbage and local grounds are best to reference to or from, not use as a leg of any high integrity audio path, especially if several feet long. The differential bus I mentioned in passing, indeed carries the bus (-) as a true bus, but it A) is not balanced impedance (the lower the better for noise) and B) it doesn't carry an equal but opposite polarity version of the audio waveform (just a proper summation of the individual channel send ground references).

IMO this should equal the ground noise rejection of a full blown "balanced" bus, while the lower impedance (-) bus will contribute less noise at the sum amp, for a fraction of the cost/complexity.

In a balanced bus any ground difference between the channel and sum amp is common mode and will cancel out. In a differential bus the channel ground term is fed to the (-) bus and properly subtracts out.

If one was so inclined they could make a "balanced impedance" summing bus where the ground combining resistors were the same value as the signal summing resistors but in my judgement you would give up more in wide band noise than you might gain from interference rejection.

A very important strategy that you learn early on when designing consoles is that signals must be properly referenced between where they come from and where they're going, and ground is concept not a voltage.

JR

Hi,

i'm a bit late in this thread, but i will try...

So, noise wise...to have an 8 balanced input summing amp followed by a substractor ¿is it noisier or quieter than 8 Ic diferential line receivers into an unablanced summing amp or differentail ? (like the superbal..)

JAY X

JohnRoberts said:

The differential bus I mentioned in passing, indeed carries the bus (-) as a true bus, but it A) is not balanced impedance (the lower the better for noise) and B) it doesn't carry an equal but opposite polarity version of the audio waveform (just a proper summation of the individual channel send ground references).

Click to expand...

John, don't you think it would be convenient to devise a proper name for that arrangement (that I second wholeheartedly). IMO calling that a balanced bus is a misnomer since the impedance of the positive and negative bus don't need (and in fact benefit) to be identical and the noble signal is applied only to one bus.
Over the years I have come to calling that arrangement "ground-sensing bus".
Then we would have to agree on a consensus regarding which bus is negative or positive. 

abbey road d enfer said:

JohnRoberts said:

The differential bus I mentioned in passing, indeed carries the bus (-) as a true bus, but it A) is not balanced impedance (the lower the better for noise) and B) it doesn't carry an equal but opposite polarity version of the audio waveform (just a proper summation of the individual channel send ground references).

Click to expand...

John, don't you think it would be convenient to devise a proper name for that arrangement (that I second wholeheartedly). IMO calling that a balanced bus is a misnomer since the impedance of the positive and negative bus don't need (and in fact benefit) to be identical and the noble signal is applied only to one bus.
Over the years I have come to calling that arrangement "ground-sensing bus".
Then we would have to agree on a consensus regarding which bus is negative or positive. 

Click to expand...

It kind of already has a name. IMO it can be accurately called a "differential bus". The audio is treated differentially feeding both a - bus and + bus with the audio hot and audio 0V reference signals. The differential amp subtracts the two and forward references the sum to the control room 0v reference.  This is similar to the distinction between a differential input and balanced input

Using low value resistors for the  0V side is just a little extra optimization.

It seems pretty late in the game to quibble about mature circuit topologies. Calling it "ground sensing" is descriptive and suggests the benefit (if you don't mind the confusion surrounding "ground" terminology.)  ;D

Note: to get the full benefit from a "ground sensing" "differential" bus, both the audio hot and audio 0V resistors need to be selected and deselected by each channel send to keep the differential math accurate. In some value designs with modest number of channels I did not deselect the 0V resistor to use a cheaper switch. In fact in a small mixer one 0V ref bus can support several different mixes, while for a larger load 2 buses and two resistors per sub mix send will give the full benefit. 

JR

crazytooguy said:

So, here’s the main question: Is the advantage gained by using a fully balanced summing buss (if any – this is a topic of much debate) offset by the extra silicon the signal has to pass through to achieve it?

Click to expand...

I won't answer this question by yes or no; I'd rather give you some elements of reflection. The issue about "extra silicon" is an anthropomorphic notion. Mixers are machiines, they like whatever quantity of silicon makes them work better. In particular, balanced circuitry, which uses more components than unbalanced, provides actually better performance, on account some of the possible imperfections cancel out by symetry.

Using the THAT chips would require the following extra chips when using a balanced summing buss: a balanced line driver (THAT 1646) to convert the panned and level-controlled signal to balanced to feed the buss

Click to expand...

The 1646's forte is its floating outputs, which are of no use in that position. You'd be better off with a good quality dual opamp.

, and a balanced line receiver (THAT 1200) to convert the buss signal to unbalanced for master level control

Click to expand...

Aren't you forgetting about summing amps? The 1200 is a high-impedance line level device, not suited for summing, which is a low-level low-Z job.

Here’s the signal path:  Input goes to a full differential buffer using four NE5524s.

Click to expand...

Do I read 5534? Why four? Does it include a level control? The standard 3-opamp instrumentation amp can be easily modified for variable gain.

From there, it goes to Insert Send/Return

Click to expand...

Would you use a 1646 for balancing the Send?

The channel fader is next, then on to another NE5524 for a 16 dB gain boost prior to the Pan pot.

Click to expand...

Do you need 16dB gain? Most users are content with 10 or 12.

Output of the Pan circuit then goes to a unity gain buffer built on an OP4132 (used here as an impedance converter, since the 1646 requires a low source impedance)

Click to expand...

You could use with profit the active pan-pot Doug Self patented some years ago.

and then into the THAT 1646 balanced line driver. The 1646 feeds the balanced summing buss through 10k resistors.

Click to expand...

Here we are touching the heart of the discussion!

The Summing buss feeds the input of a THAT 1512 gain module, which adds the necessary 20dB of makeup gain, and converts the balanced buss signal to an unbalanced signal for level control.

Click to expand...

Ah, OK, I see you haven't forgotten the summing amps, although I see you envision passive mixing, which is a pity, considering the resources you intend to put into this endeavour.
It is possible to use the 1512 as a balanced virtual earth mixer - using the Rg1 & Rg2 connections as input and shorting out the regular inputs, but I think using VLN opamps such as the AD797 or LME49990 is a better bet.

The signal level for the summed buss is set with a fader, then goes to an OP4132 unity gain buffer for impedance conversion

Click to expand...

Why a 4132 there instead of a 5534 or even half a 5532? The 4132 is a very competent device, but I see only one significant advantage. Its FET input makes the fader noise insignificant. But then, why not chose it also for the channel fader?
Fader noise is a phenomenon that happens with less than perfect faders, I mean it doesn't happen with a P&G, but it happens with all types of Alps, even the most expensive. The tiny variations of contact resistance modulate the input noise current of the opamp they are connected to. We found that FET input opamps were the only proper answer to that problem. At the time, the only FET input opamps economically viable were the TL0's. We set up a panel of users and the results were mitigated; some prefer the sonic signature of the 5532 and some prefered the silent operation of the TL0. Today the 2134 offers both FET inputs, low noise, superb audio performance and low cost.

I was thinking of breadboarding a channel and seeing if there was a noticeable sound quality difference in the two summing approaches, but it occurred to me that I would need all sixteen channels to really assess how well it works.

Click to expand...

You may evaluate some aspects of the performance by loading the bus with resistors of value equivalent to all the channels applied. That would give a good estimate of the performance that would be achieved if the world was perfect. In short, the ultimate goal that you would never achieve.
The main problem with mixers is all the things that cannot be calculated, like how much PSU ripple actually appears on the channel ground, how much vlongitudinal noise voltage is generated by the presence of all these channels, each having its own polluted reference.
Some kind of "balanced" bus arrangement is the single most important improvement to the real-world noise performance of a mixer. But you don't have to go all the way, as JR mentioned.

JohnRoberts said:

It kind of already has a name. IMO it can be accurately called a "differential bus". The audio is treated differentially feeding both a - bus and + bus with the audio hot and audio 0V reference signals. The differential amp subtracts the two and forward references the sum to the control room 0v reference.  This is similar to the distinction between a differential input and balanced input

Click to expand...

OK, so, as of this day, "differential bus" is the sanctified term. 

What is the difference between "balanced" and "differential" when we are talking about conventional audio inputs and outputs (and internal bus design, for that matter)?
In my perception it is a "distinction without a difference" in philosophical terms.
http://en.wikipedia.org/wiki/Distinction_without_a_difference

Yes, I understand the growing popularity of "balanced impedance" outputs which are not really "differential".
However to the average audio-gear user, the distinction seems immaterial.

That original discussion (from 7 years ago!) was rather "over the top" and typical of the kind of over-engineering we see in zealots who don't really understand real-world circuit design and performance/cost/complexity tradeoffs.

Richard Crowley said:

What is the difference between "balanced" and "differential" when we are talking about conventional audio inputs and outputs (and internal bus design, for that matter)?
In my perception it is a "distinction without a difference" in philosophical terms.
http://en.wikipedia.org/wiki/Distinction_without_a_difference

Click to expand...

IMO it's a distinction with a difference. While a balanced bus (or input) will be differential a differential bus (or input)  is not likely to be balanced or it would be called balanced since balanced infers higher value.

While there is some confusion surrounding "balanced", like balanced impedance or balanced termination (equal impedance), vs. balanced signal (equal but opposite polarity) signal.

In the context of console design this has been pretty well explored, but not concisely named (thus Abbey's inquiry). A balanced bus most likely** refers to summing positive and negative polarity stems into two VE summers. There is very little benefit to balanced impedance inside a console, and some actual noise penalty if not summing two active signals from using higher impedance resistors to sum 0v references. 

If there was no difference between differential and balanced inputs I could have used a lot fewer opamps over the years. I am not inclined to design unnecessary cost inside sharp pencil value products.

I guess we must agree to disagree about this. 

JR

** I have also seen sum bus topology using the two VE sum amps, but only feeding one bus hot audio signal and the other the 0V reference. I guess this could be called "balanced" because it is, while I see little benefit from the extra parts.     

Richard Crowley said:

What is the difference between "balanced" and "differential" when we are talking about conventional audio inputs and outputs (and internal bus design, for that matter)?
In my perception it is a "distinction without a difference" in philosophical terms.

Click to expand...

There are indeed several differences, and I believe the lack of proper terminology for describing the various arrangements is the basis of the confusion. There is a lot of difference between a floating output, a non-floating balanced output and a balanced-impedance output. Describing them as simply "balanced" may be acceptable or not. For example, a non-floating balanced output cannot be grounded without serious deterioration of performance. For me, it's a good reason to have a somewhat more descriptive term than "balanced".
I must add that I would not use the term differential for an output stage.
Just the same, calling an input balanced is very restrictive. It means that the input is not floating and both legs present the same impedance to ground. In particular, a floating transformer input is not "balanced", although it's actually very differential.
However, it seems that usage has connected the term "differential" with transformerless active.

Yes, I understand the growing popularity of "balanced impedance" outputs which are not really "differential".

Click to expand...

In what respect are they not? They are certainly not signal-balanced, but the signal is distributed between two legs that are not the ground, which, IMO, makes it differential.

However to the average audio-gear user, the distinction seems immaterial.

Click to expand...

Not always. Many Tascam users learnt at their expense that non-floating balanced outputs are very difficult to integrate in an environment.
I think many members of this group are not "average audio-gear users" and as such deserve to be a little more enlightened.

That original discussion (from 7 years ago!) was rather "over the top" and typical of the kind of over-engineering we see in zealots who don't really understand real-world circuit design and performance/cost/complexity tradeoffs.

Click to expand...

And 7 years later the same confusion reigns. I don't think Bill Whitlock and the folks at THAT are over-engineering zealots disconnected from the realities of manufacturing.

JohnRoberts said:

I guess we must agree to disagree about this. 

Click to expand...

I wasn't seeking to disagree. I was trying to understand the (apparently subtle and/or not clearly defined) difference between usage.

abbey road d enfer said:

There are indeed several differences, and I believe the lack of proper terminology for describing the various arrangements is the basis of the confusion. There is a lot of difference between a floating output, a non-floating balanced output and a balanced-impedance output. Describing them as simply "balanced" may be acceptable or not. For example, a non-floating balanced output cannot be grounded without serious deterioration of performance. For me, it's a good reason to have a somewhat more descriptive term than "balanced".

Click to expand...

Yes, I agree, in several forums I frequent there appears to be chronic confusion about how to handle interconnection between balanced and unbalanced equipment.  In particular, confusion about what the source output topology is, and how that affects whether (or not) to ground the "cold" side.

I must add that I would not use the term differential for an output stage.

Click to expand...

But in the general electronic sense (not specifically AUDIO) two output nodes with opposite polarity signals are "differential" are they not?  Or are we using the term "differential" strictly to explain how the signal is extracted at the destination?

Just the same, calling an input balanced is very restrictive. It means that the input is not floating and both legs present the same impedance to ground. In particular, a floating transformer input is not "balanced", although it's actually very differential.

Click to expand...

Ah!  I see what you are getting at. "Differential" implies ground-referenced signals, while "balanced" is a sub-set and/or has no such implications.

However, it seems that usage has connected the term "differential" with transformerless active.

Click to expand...

For both outputs an for inputs.

Yes, I understand the growing popularity of "balanced impedance" outputs which are not really "differential".

Click to expand...

In what respect are they not? They are certainly not signal-balanced, but the signal is distributed between two legs that are not the ground, which, IMO, makes it differential.

Click to expand...

I guess I was sub-consciously thinking of "differential" as two active, opposite-polarity, ground referenced nodes (as typical "transformerless balanced".)

Many Tascam users learnt at their expense that non-floating balanced outputs are very difficult to integrate in an environment.
I think many members of this group are not "average audio-gear users" and as such deserve to be a little more enlightened.

Click to expand...

Yes, it would be excellent if gear were explicitly labeled with the output topology (or at least mention in the manual how to handle unbalanced destinations).

That original discussion (from 7 years ago!) was rather "over the top" and typical of the kind of over-engineering we see in zealots who don't really understand real-world circuit design and performance/cost/complexity tradeoffs.

Click to expand...

And 7 years later the same confusion reigns. I don't think Bill Whitlock and the folks at THAT are over-engineering zealots disconnected from the realities of manufacturing.

Click to expand...

I wasn't talking about Whitlock, et.al. I was referring to crazytooguy's seeming desire for a balanced summing bus simply because "balanced must be better".  I don't buy that dogma.

> "Differential" implies ground-referenced signals

No. It really means NO concept of ground. A voltmeter in a (bone-dry) ivory tower, just two leads.

With Differential, one lead "could" be grounded. Or I could be probing the difference between two wires on a street-line where all the ground-wires have been stolen.

> "balanced" is a sub-set and/or has no such implications.

Generally that Voltage AND Impedance are the same on both side. Both sides of what? Possibly ground, though uncertain.

The side-implications are different for Outputs and Inputs.

In the specific case above, of a so-called "balanced" MIX-bus, we could use 10K mix-resistors on both sides. However since one side comes from (channel-strip) "ground", we can often use 10K for the signal mix and 100r for the ground-mix, with some advantage in hiss. 10K/100r signal/ground is far from "balanced" in impedance or voltage. However the mix-amp senses the *difference* (differential) of the two buses.

> there appears to be chronic confusion

People do not bother to learn Basic Electricity and how it applies to the circuits they work with every day.

It is universal. As recently as a century ago there was a strong feeling that street power wires should be UN-grounded for safety. People who should have figured it out just didn't.

> I don't buy that dogma.

Me neither. I got a long way with mostly un-balanced interconnects. (However either very benign rooms or rarely-patched, and I knew tricks.) When that didn't cut the crap, I often went Floating on one end and TASCAM-style the other end... good Floating is near as good as full telephonic balancing.