Designing a modern mixing console - Part 1 (and introduction): Channel Input

[abbey road d enfer]

I'm not sure I agree with rolling off -2dB at 20Hz. This is the domain of tone control. I would put large lytics there (100-220uF) and deal with subsonics in the filters department.

I thought -2 there seemed somewhat sensible.

That's why I said "Im not sure..." It is indeed sensible, but I wouldn't do that on a commercial unit; that would not make nice specs.

Do I really want subsonics eating up headroom throughout? Or does it even matter?

Subsonics...do they exist? I mean to a point where they actually eat headroom. If yes, you probably need a dedicated HPF anyway. The input stage should be capable of handling the subsonics; that's why headroom is there. I would have a different answer if there were input xfmrs involved.

The gain trim uses a tad too many components for my taste.

  It is a little parts-verbose, true. What I do like about it is that the gain is not tied in any way to the pot value. 20% pot tolerance here still means a 0% variance in total gain range.

The new attached schemo answers that too.

I also like the lack of high value resistors this soon in the audio path. I will probably sim both to see where the noise performance ends up.

  If I understand well, you are worried about the noise contribution of the 100k DC FB resistor. You don't have to. This resistor is always shunted by a much smaller one; that's the equivalent resistance that counts.

I have not been able to find much on the noise contribution of feedback resistors aside from the general advise to use smaller values when possible, up to a point, of course.

Regarding noise, FB resistors are input resistors, going to the negative input. The equivalent value for noise contribution is the equivalent value resulting in paralleling the FB res and the foot res (if any).

As for TIM, I never understood how it could possibly exist in the form often presented. If the output acted in the way described by its proponents, a measurable static phase shift would be present from DC to infinity in addition to the gradual phase shift with slew limit, which we know is not the case.

Absolutely.

 

[Nishmaster]

That's why I said "Im not sure..." It is indeed sensible, but I wouldn't do that on a commercial unit; that would not make nice specs.

Subsonics...do they exist? I mean to a point where they actually eat headroom. If yes, you probably need a dedicated HPF anyway. The input stage should be capable of handling the subsonics; that's why headroom is there. I would have a different answer if there were input xfmrs involved.

Looking at quality transformers, the -3dB point is indeed lower, around .2Hz for a Jensen, for example. Perhaps I am being overzealous here. I'll move the filtering out of this stage.

The new attached schemo answers that too.

Very interesting. Basically the Baxandall configuration eschewing the buffer. At first glance, it looks like the gain law would be a bit odd, though, as the wiper is now loaded by the 4K7 instead of the buffer input. My math may be off here, though. Total gain is obviously no longer affected by pot value, but volume at center of the control is, albeit by only around 3-4% for a +/-20% pot value change. Unity gain also doesn't lie at control center (although that's not really a big deal, and neither does it in my configuration).

EDIT: Math was wrong. Yours has identical pot law except for a slight bend near full up. Noise performance looks to be slightly better for the Baxandall config, at least in the sim.

Regarding noise, FB resistors are input resistors, going to the negative input. The equivalent value for noise contribution is the equivalent value resulting in paralleling the FB res and the foot res (if any).

Gotcha. I didn't know if there was any noise cancellation or other factors involved, but thinking about it more, I can see why that wouldn't be the case.

-Matt

 

[ej_whyte]

I also like the lack of high value resistors this soon in the audio path. I will probably sim both to see where the noise performance ends up.

  If I understand well, you are worried about the noise contribution of the 100k DC FB resistor. You don't have to. This resistor is always shunted by a much smaller one; that's the equivalent resistance that counts.

I have not been able to find much on the noise contribution of feedback resistors aside from the general advise to use smaller values when possible, up to a point, of course.

Regarding noise, FB resistors are input resistors, going to the negative input. The equivalent value for noise contribution is the equivalent value resulting in paralleling the FB res and the foot res (if any).

Are there any publications/books detailing this? Not that i'm questioning your statement, its just not good practice to reference forums for essays and i'm keen to research this a bit more. Also where should I look for more reading on CMOS switching? I plan to get some Douglas Self soon but some online info would be good as well.

Nishmaster, what sim are you using? I haven't found one I like so far.

Cheers,
Elliott

 

[Nishmaster]

Are there any publications/books detailing this? Not that i'm questioning your statement, its just not good practice to reference forums for essays and i'm keen to research this a bit more. Also where should I look for more reading on CMOS switching? I plan to get some Douglas Self soon but some online info would be good as well.

Nishmaster, what sim are you using? I haven't found one I like so far.

I haven't found much quality material online for either of those subjects. The manufacturer application notes aren't terribly helpful for CMOS switches, either. A quantitative noise analysis of more complex opamp circuits can get a bit mathy, and I haven't really gotten cozy with the equations yet. I think TI has a handbook of opamp noise equations out on the net some place. That aside, Doug Self's Small Signal Audio is a treasure trove of goodies, and while I wish that he would go a bit more in depth with most subjects, it offers a wealth of practical knowledge and ideas. He has a whole chapter dedicated to switching.

Handbook for Audio Engineers is the oft repeated recommendation, which is also very good, if a bit older now. Radiotron Designer's Handbook won't do much for your opamp theory, but thar be genius in that tome.

Simming-wise, I've been toggling back and forth between LTSpice and OrCAD. I prefer the simplicity and speed of LTSpice, but OrCAD's libraries are quite extensive. The interfaces of both are rubbish, moreso in the case of LTSpice.

As an aside (and as an interface designer as a small part of my IT work), is there some rule that states engineering (all fields, not just electrical) software must have complete rubbish interfaces? I realize it's all niche industry stuff build on foundations set in the pre-windows days, but jeeze. You'd think somebody would start fresh and give the whole thing a logical intuitive interface. Ahem, sorry, carry on!

-Matt

 

[baadc0de]

As an aside (and as an interface designer as a small part of my IT work), is there some rule that states engineering (all fields, not just electrical) software must have complete rubbish interfaces? I realize it's all niche industry stuff build on foundations set in the pre-windows days, but jeeze. You'd think somebody would start fresh and give the whole thing a logical intuitive interface. Ahem, sorry, carry on!

-Matt

I hear steve jobs is free now

 

[Brian Roth]

In addition, Amek then "branched out" the audio +/- rails into "circuit chunks" of maybe a half dozen or more opamp chips through flameproof 5R6 resistors in series with each rail into the "chunk".  Thus, a failure/short in a "chunk" would "blow" the 5R6 feeding that section, while the rest of the module continued to function.

That's a smart feature. What I need to do is branch my power pins for the opamps out into their own sheets but I haven't figured out how to do that yet in Eagle. Clearly smarter folks than I designed that board, but 5R6 seems high at first glance. Doesn't sag become an issue?

Interesting point, since some of the chunks could possibly draw 50 or 100 mA.  At 100 mA, the drop across a 5R6 resistor would be 0.56 Volts.  But, the chunks with the greater quantity of opamps hanging from the rails are driving internal loads that are comparatively high Z, so there should be very little shift in current consumption between idle and at full signal levels.  OTOH, stages such as drivers for line outputs into The Real World with 600 Ohm drive capability were in much smaller PSU chunks.

Best,

Bri

 

[Brian Roth]

While looking through the various Amek desk schemos, it dawned on me that the input section for this project could be somewhat simplified.  I've attached a sketch inspired by their design which has several advantages compared to the Rev4 version Nish posted.  It eliminates the need for the second THAT input buffer, the MAX FET switch and its follower amp, which is a shorter/cleaner (in theory) signal path.

S1 selects one of two balanced sources and can be a regular switch, or a Panasonic T2Q relay (as discussed previously) if that function needs to be controlled by a "master mode" or similar function.  Since the chosen signal source is still balanced, there is a good place for a polarity ("phase") reverse switch S2.  Additional contacts can be added if a LED indicator is required.

The 11K resistors across each balanced input ensures a relatively constant load for each input, selected or not.

ALSO....

I suggest a THAT1203 for the buffer for two reasons.  It can handle 3 dB more level from the "outside world" while also dropping the incoming signal by 3 dB for greater internal headroom within the module.  Many/most desks drop external levels for better internal headroom, then boost the levels back up at the various output points.

Too bad the THAT1206 doesn't offer the expected 6 dB higher input signal handling capability; the spec sheet shows the same maximum level as the 1203.

Best,

Bri

 

[Nishmaster]

Interesting point, since some of the chunks could possibly draw 50 or 100 mA.  At 100 mA, the drop across a 5R6 resistor would be 0.56 Volts.  But, the chunks with the greater quantity of opamps hanging from the rails are driving internal loads that are comparatively high Z, so there should be very little shift in current consumption between idle and at full signal levels.  OTOH, stages such as drivers for line outputs into The Real World with 600 Ohm drive capability were in much smaller PSU chunks.

That makes sense. Even still, a decent amount of amps on the chunk drops a non-trivial amount of voltage. Something to think about as I move forward, I suppose.

While looking through the various Amek desk schemos, it dawned on me that the input section for this project could be somewhat simplified.

I agree, and actually that was the first configuration that came to mind. The biggest issue with it, though, is that it will require a rather exotic switch configuration to make work with sufficient offness. Even plain switch contacts have some capacitance that will cause some crosstalk issues in that configuration. Shunting to ground would require double the amount of switch contacts, and since I will later need to defeat the bus assignment at the fader when the input "B" is selected, I'll need another two contacts on top of that for the LED and for the feed to the bus assign defeat relay/CMOS/whatever. I see a minimum of 2-3 relays or switches needed, which doesn't really reduce the cost of the input much. It may also mildly upset common mode rejection, although with sufficiently low (and consistent) R switching that would be a very minor problem. If it was a transformer based input, I think I would be more inclined to pursue that configuration, but heck, chips are cheap, exotic switches not so much.

The 1203 did come to mind. I know headroom is always something to keep in mind, but I have no frame of reference for how real desks handle this. It feels to me like dropping 3dB so soon also gives me a 3dB worse noise performance right off the bat. Since the primary use of this desk is coming straight from D/A converters, I can't really see how overload will be a problem even with severe output levels. Even at +20dBu that stage isn't hitting clipping levels yet. Then again, I probably don't know what the heck I'm talking about here. Like I said, I don't know what the usual convention is.

Work is busy, so not much progress over the last few days. Abbey, do you concur with my noise findings regarding the two gain trim configurations or am I off there?

-Matt

 

[Brian Roth]

The Amek desks use the exact relay configuration that I drew in my sketch, and crosstalk hasn't been a problem that I am aware of.  Each input module has some combination of mic pre out/line in/tape in/bus out through the Panasonic relays which then feed into an electronically  balanced input buffer.

As far as internal levels, many/most pro desks intentionally run at lowered signal levels to allow for some "oops" room.  The Amek desks drop the source levels by 4.5 dB right at the balanced input buffer.  Other desks (MCIs come to mind) run even lower at a -6 dBu internal level which is 10 dB below a nominal +4.

While this project may be intended to operate with D/A converter sources, what happens if you patch in something like a LA-2 between the converter outputs and desk inputs?

Best,

Bri

 

[abbey road d enfer]

The 1203 did come to mind. I know headroom is always something to keep in mind, but I have no frame of reference for how real desks handle this. It feels to me like dropping 3dB so soon also gives me a 3dB worse noise performance right off the bat. Since the primary use of this desk is coming straight from D/A converters, I can't really see how overload will be a problem even with severe output levels. Even at +20dBu that stage isn't hitting clipping levels yet. Then again, I probably don't know what the heck I'm talking about here. Like I said, I don't know what the usual convention is.

It was quite usual, even when recording was all analog, to trim the input signal either right after the line receiver or within it. Now, with the general tendency of recording with elevated dBfs levels (-6dBfs peaks  +18dBu peak/+6rms, are typical today), and even in some cases receiving files normalized at 0dBfs, gain trimming is more and more necessary, not for the input stages sake, but in order to leave some space for processing. Dropping level by 6dB, making the internal operating level -2dBu is a convenient answer to the compromise between noise and headroom that any mixer designer has to do. But it's not the only answer. My personal choice was operating at +4. My reasoning is that if there is noise, it is the equipment's fault, if there is clipping, it's the operator's fault. I make quiet mixers... Thousands of Soundcraft and Amek mixers have tried to prove me wrong... What I can say is that when Soundcraft made the 500/600's operating level as low as -12, with justified claims of outstanding headroom, many complained about noise.

Abbey, do you concur with my noise findings regarding the two gain trim configurations or am I off there?

I haven't really simmed it but with the resistor values you're using, the noise contribution is essentially the voltage noise of the opamp.

 

[JohnRoberts]

I don't know if we are talking about the same thing, but I standardized my internal nominal 0VU at -6dB (-2dBu) from external 0VU so that electronically balanced outputs would clip simultaneously with single ended insert points. If I had the luxury of balanced insert points I would probably still run at -6dB internally, but you would have no way to tell from the outside world.

Noise floor is generally not an issue with well executed modern electronics. While there is always noise, a simple mic and room noise is louder.

JR

 

[abbey road d enfer]

I don't know if we are talking about the same thing, but I standardized my internal nominal 0VU at -6dB (-2dBu) from external 0VU so that electronically balanced outputs would clip simultaneously with single ended insert points.

This would be true only if the EBOS are connected to balanced lines. In the case they are grounded on one side, clipping appears 6dB sooner.

Noise floor is generally not an issue with well executed modern electronics. While there is always noise, a simple mic and room noise is louder.

I know it's a pet subject of yours, but the sonic character of summing amps noise is very different than that of the room.; summing amps noise is constant, and for me it doesn't have the soothing drone benefit some want to attribute. It's a constant nuisance, against which the user can only blame the designer. OTOH, mic and room room is there only when you raise the fader; one knows that he is just being challenged by the laws of physics. It's not fair, but it's the way customers perceive it.

 

[JohnRoberts]

I don't know if we are talking about the same thing, but I standardized my internal nominal 0VU at -6dB (-2dBu) from external 0VU so that electronically balanced outputs would clip simultaneously with single ended insert points.

This would be true only if the EBOS are connected to balanced lines. In the case they are grounded on one side, clipping appears 6dB sooner.

Perhaps, but if you run inserts at +4dBu SE, with board set for unity gain, you will clip internally 6 dB before reaching full output on active balanced outputs.

Pick your poison, I choose to run cooler inside, and opamps are even quieter these days than back when I was actively designing such things.

Noise floor is generally not an issue with well executed modern electronics. While there is always noise, a simple mic and room noise is louder.

I know it's a pet subject of yours, but the sonic character of summing amps noise is very different than that of the room.; summing amps noise is constant, and for me it doesn't have the soothing drone benefit some want to attribute. It's a constant nuisance, against which the user can only blame the designer. OTOH, mic and room room is there only when you raise the fader; one knows that he is just being challenged by the laws of physics. It's not fair, but it's the way customers perceive it.

Trust me I know, been there done that...  but that customer will never be satisfied even if you cooled your electronics with liquid nitrogen, because WFO they will still find noise. In practice as long as bus and other other dominant noise sources are  adequately below signal noise floor this is a cosmetic issue.

As I have said before, I know how to make a sum bus with very low noise gain (sum synthesized current sources), but yawn, IMO it isn't worth the trouble (cost) because in use you won't hear the (noise) improvement. Of course you may still hear distortion and phase shift from a poorly executed sum amp, but there are cheaper ways to skin those cats, not to mention digital summing that doesn't have those problems.

So I will take every opportunity I can to try to educate end users and fledgling designers to the practical design considerations, even though I am no longer in the trenches listening to whiney customers, who just bought a new esoteric CR monitor power amp with higher than typical voltage gain, and can't understand the now higher CR noise floor (real story). :-(

Another +1 to Bob Katz's system to normalize SPL in CR monitors, for consistent mixes.

JR   

PS: one feature I put in my last big one (>100 feeds to the two mix), I had an adjustable 0 to -10dB gain trim on master sum, again mostly cosmetic for 10dB lower noise floor when turned down, and for convenience with all those inputs.

 

[Nishmaster]

Thanks to everyone for their input on the internal line level. I'll have to mull it over a bit more, work some numbers, and see where I end up.

In the interim, I have attached the block diagram of the channels. Hopefully that makes the feature set and routing capacity a bit clearer. Please let me know if something looks wonky or untenable; I've been working this sheet for the better part of 3 hours straight and to be honest it looks pretty much like spaghetti to me right now. 

Right now I'm not entirely sold on my Insert In/Out switching schema, as even when the inserts are switched out the signal has to travel through the insert pre eq/post eq switching. I'm sure there's a better way that I'm missing.

-Matt

 

[Brian Roth]

The insert pre/post won't work as drawn.  When that switch is in "post" there's no input into the EQ section.  I'll have to ponder that entire switch section to come up with a better flow.  Also, are you planning to use decent quality mechanical switches/relays  in those paths, or throw a bunch of FETs at it?  If the latter...for what purpose if you are not needing "total recall" of the settings?

I am not seeing a polarity/"phase" reverse switch.  Very important IMHO, unless you plan to do that "inside the DAW".  But, that makes the unit less useful for anyone who is using some sort of multitrack machine.

High cut filter is definitely an optional item IMHO.

The solo system is a bit mystifying to me, since it looks like some sort of combination of "solo in place" which works by muting channels, but at the same time has a stereo solo summing bus pair.

I vote for more than six auxes, but that is just me.  The Ameks I keep mentioning have 16 (!!) which is maybe too far the other direction.  Regardless, each aux send needs an individual pot, and then you get into how to deal with pre/post for each aux.  Pre/post for each aux?  "Group" the auxes so that, say, every two auxes has its own pre/post?  Also, many desks group the auxes into pairs with a pan pot for that aux pair.  One cool thing along those lines the 9098 desks offered on several aux pairs was the ability for the aux pair to be post-main panpot.

As drawn, the feed to the main L/R buses is "either/or" with the 16 groups.  Just have the L/R as its own bus assign, along with the 16 groups.

I do not comprehend the 1-16 grouping as drawn.  The summer for those is on the input modules??  If so, then you can't use that particular channel as a line input.  Perhaps a better idea would be to have 16 group master modules.

BUT, things are moving along a bit....keep up the good work!

Best,

Bri

 

[Brian Roth]

OK...best I can come up with is the attachment...as always,  drawn with RothCrapCAD <g>.

I cannot concoct a way to use less than four switch poles to select "insert pre/post".  But, "insert in/out" only requires a single pole switch.  Same total number of poles that you drew, Nish...but moved around.

Maybe someone more clever than me can delete a pole out of the mish-mash!

The insert pre/post switch MUST be a "break before make", but that is pretty standard with mechanical switches or relays.  Dunno about using a "mess of FETS" for that function....would depend upon the timing of the control voltages to the FETs.

Also.....I would tend to think that the low/high pass filters should be "part" of the "EQ block"...but again, that is just my 2 cents.

Best,

Bri

 

[Henke]

Here is my take on the pre/post eq switching, using 3 poles...

 

[Brian Roth]

egads  hmmm    looks ok

Bri

 

[Nishmaster]

The insert pre/post won't work as drawn.  When that switch is in "post" there's no input into the EQ section.  I'll have to ponder that entire switch section to come up with a better flow.  Also, are you planning to use decent quality mechanical switches/relays  in those paths, or throw a bunch of FETs at it?  If the latter...for what purpose if you are not needing "total recall" of the settings?

Erm, no, it certainly won't work. Henke's configuration is what I was getting at but managed to mis-draw rather badly.

My intention so far is to use the CMOS switching option. Relays are the most simple and pure option, but I'm not sold on the either the expense or power supply tradeoffs. I don't think I'll mess with FETs.

I am not seeing a polarity/"phase" reverse switch.  Very important IMHO, unless you plan to do that "inside the DAW".  But, that makes the unit less useful for anyone who is using some sort of multitrack machine.

I suppose it is simple enough to add one. I figured since this unit is meant as a mixdown-from-pro-tools desk, polarity can be done in the box, but heck, it's only one switch.

High cut filter is definitely an optional item IMHO.

I find the high cut to be quite useful on electric guitars, because I can reduce "fizz" while still having a band of eq to suck out some of the 4-8k resonance that always creeps up. With modern condensers and digital recording, I find vocals can often use some high cut as well. Probably only 6dB/oct slope here.

The solo system is a bit mystifying to me, since it looks like some sort of combination of "solo in place" which works by muting channels, but at the same time has a stereo solo summing bus pair.

Yes, that is mystifying indeed! :-[ Clearly the solo bus is totally unnecessary. I started by drawing a PFL style setup, but then decided that I really actually hate PFL and much prefer solo-in-place. I then didn't get rid of the solo bus.

I vote for more than six auxes, but that is just me.  The Ameks I keep mentioning have 16 (!!) which is maybe too far the other direction.  Regardless, each aux send needs an individual pot, and then you get into how to deal with pre/post for each aux.  Pre/post for each aux?  "Group" the auxes so that, say, every two auxes has its own pre/post?  Also, many desks group the auxes into pairs with a pan pot for that aux pair.  One cool thing along those lines the 9098 desks offered on several aux pairs was the ability for the aux pair to be post-main panpot.

Yeah, I hear you. The aux sends are going to be a pain. I know it's been mentioned that they usually run unbalanced in the desk, and I know they're usually just for effects which don't need as pristine a signal path, but I'd hate for noise/buzz to creep in to the system from that path. My initial plan was to run the auxes balanced. It would be easy enough if I had a balancing amp feed all the aux sends, but then I need balanced fader and pan pots for each send. I could avoid the balanced pots, but then I'd need a balancing amp for each send. There's also the issue of panel real-estate, which is going to get rather tight pretty quickly. Hence 6 auxes was my initial gut feeling. 4-6 stereo auxes may be a more flexible configuration at the cost of increased complexity. I rarely run more than 4 stereo effects myself.

As drawn, the feed to the main L/R buses is "either/or" with the 16 groups.  Just have the L/R as its own bus assign, along with the 16 groups.

I'm not sure I follow. The L/R is it's own bus assign. This may be related to the next bit:

I do not comprehend the 1-16 grouping as drawn.  The summer for those is on the input modules??  If so, then you can't use that particular channel as a line input.  Perhaps a better idea would be to have 16 group master modules.

The desk is a bit of a unique bird; at least, I haven't seen this configuration. It's a bit of a hybrid between a split design and an in-line design. Basically, any input module can be a line input module or a group input module. Here's a sample signal flow:

Let's say I have drums on track 1-12. The track output on those is assigned to group 1-2. The signal goes from those tracks to the master module, which will have the summing amps for each group. Group 1 & 2 are summed at the master module, then the summed output is sent back out a second set of buses, the group input buses.

In "Group Mode", the channel can select to use the output of the group summers as it's input instead of a line input. Therefore, I can take channel 13, put it in group mode, select group 1 as the input, and pan it to the left. I then take channel 14, put it in group mode, select group 2 as the input, and pan it to the right. Selecting group mode automatically selects the L/R bus as output to avoid feedback loops.

Therefore, I have drums on channels 1-12, and the drum bus on 13-14. Yes, this configuration eats line channels as you increase group usage, but it avoids having dedicated group modules as wasted space on the board as well as keeps the design totally shared between both. For those that do a metric ton of grouping, this is probably not optimal, but for me, I use probably 3-4 stereo groups per mix, which means on a 48 input desk I lose 8 inputs. The Trident 80B I was looking to get is 32/8 anyhow. The other great thing is that I don't need to design any stereo eq for the group modules, the eq is on the channel strip already. Finally, this also makes the workflow across the desk nice, as you can put the groups right after the channels that feed that group for more logical placement, or, you can stick them at the end (or middle, or wherever) of the desk like a split console if that feels better to you.

The block labeled "Sum" at the group inputs of the channel is indeed a summing amp. The reason I put it there is so that the user could, if they so wished, select multiple group inputs and have them all appear combined in the channel strip. I personally can't see a good reason to do things this way, but it more importantly avoids having to have some kind of mechanical switchbank or complex logic system that only allows one group input to be selected. If the user (read: me) is smart (debatable), he only selects a single group and the summing amp there is then effectively just an untaxed inverting buffer.

If anyone has some idea on how to switch 16 balanced inputs while allowing only one input to be active at any given time without digital logic, I'm certainly open to ideas. Dual concentric 16 position rotary switch is about all I can think of. I haven't been able to put a finger on anything else yet.

-Matt

 

[keefaz]

Why not use a small microcontroller to manage all the CMOS switches ? It is not really complex logic system. Maybe require some care to the pcb layout though.