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

[abbey road d enfer]

This detector generates current spikes that are very likely to propagate in the whole channel, even in the whole mixer. Common practice is to have a detector operating at very low current, followed by a LED driver. Can be two transistors.

I like the simplicity there, quite nice. However, I do like the full-wave capability of the dual comparator setup. Would adding a driver to LED1 be sufficient? I do definitely see how there would be some in-rush current when LED1 is driven in the current setup.

You need to detect and hold at low current and use a LED driver.

I suppose the comparators slamming the rails might be noisy also, but I do have it decoupled. Is it current to ground that is your concern?

Both current to gnd and current to rails.

 

[abbey road d enfer]

There should be a 1Meg to GND on the ground leg of the relay which I forgot to add.

That would be unnecessary. You may use an SPDT relay there, or parallel both sections for improved reliability. BTW, you may consider using CMOS switches there, MAX 312 type. Much more reliable and less consumption.

That brings up another question, though. What do I do with the inputs of U1 and U2 when they aren't connected to anything?

Nothing.

 

[Nishmaster]

That would be unnecessary. You may use an SPDT relay there, or parallel both sections for improved reliability. BTW, you may consider using CMOS switches there, MAX 312 type. Much more reliable and less consumption.

I was under the impression that leaving the outputs floating was bad practice.

CMOS switching is certainly attractive from a power supply load standpoint, and avoids needing a second ground for the relays. However, the better chips seem to be as expensive as small signal relays, and don't offer the same level of isolation. They also add distortion. However, I may be splitting hairs about theoreticals here. .001 THD at 20Khz is pretty good.

Nothing.

Interesting. I am curious about the situation that may arise where U2 may be disconnected from the bus inputs as well as switched out of the circuit by the relay. In this case, + in, - in, and output would all be floating. Isn't the amp going to drift into one of the power rails eventually?

In my slow work towards my mixer, I am simply stealing a pair meter buffers from an old Wheatstone console to drive some old modutecs.
They include peak indicators at the VU. I know going about it this way is- in the spirit of your design choices- a form of cheating, but it saves me a little more time to concentrate on the music-sending parts of the console.

Very cool. I do see the VU as one of the simpler parts of the whole; they're really just a unity gain buffer with some filtering and maybe a couple diodes. I may wholesale lift a circuit if I go the VU route since there's no need to reinvent that wheel (although I am probably reinventing countless others, oh well). Most likely, for myself, I'll opt for some kind of led bargraph, though. VUs are not terribly accurate but are quite pricey.

Don't look at it like an integrated "console". Approach it as if it's a bunch of modules you are wiring together via patchbay...

This is my basic approach. In my first broad stroke design revision, I had envisioned some kind of rackmounted modular mix system. However, after running various configurations by our studio's engineers and a few other folks, the consensus seemed to be that at least 50% of what makes mixing on a console "better" is the user interface. Having to move around to your various racks to patch things in and set up EQ moves would get old pretty fast, and take you out of the mix sweet spot.

That being said, it will certainly be possible to build each circuit block into whatever modular configuration your heart desires. For the purposes of this project, all mix subgroup and master buses will be active and differential. If somebody would like to tack a gain block instead of a summing amplifier at the end of the bus, they're more then welcome to design it and do so.

If you use standard 19" RU dimensions you can make a very nice console frame out of rack rails and wood. You can also part it out or repurpose fairly easily if you change your mind and only get 6 modules built.

My plan is to use 19" RU dimensions for the channel strips, but have the faders on their own. Since the module spacing will be identical, it would be fairly trivial to cut the width between the rails to allow for the shorter fader modules. However, panel design and frame layout is all very premature.

-Matt

 

[abbey road d enfer]

That would be unnecessary. You may use an SPDT relay there, or parallel both sections for improved reliability. BTW, you may consider using CMOS switches there, MAX 312 type. Much more reliable and less consumption.

I was under the impression that leaving the outputs floating was bad practice.

CMOS switching is certainly attractive from a power supply load standpoint, and avoids needing a second ground for the relays. However, the better chips seem to be as expensive as small signal relays,

Once you've dialed in the cost of another PSU and debugging all the switching transients, the choice is obvious.

and don't offer the same level of isolation.

The level of isolation you want to have is not really limited by the components, but rather by the circuitry and layout. Read carefully the specs, in particular the x-talk measurement set-up. It's measured on a disconnected switch. This would never happen in a properly designed piece of gear. 

Nothing.

Interesting. I am curious about the situation that may arise where U2 may be disconnected from the bus inputs as well as switched out of the circuit by the relay. In this case, + in, - in, and output would all be floating. Isn't the amp going to drift into one of the power rails eventually?

Check the internal structure of the 1200; the input are referenced to the (guess what?) Reference point.

VUs are not terribly accurate but are quite pricey.

IMO, LED bargraphs are more readable too (except in high light enviroment), particularly if they are in a more or less vertical arrangement and use different colors.

 

[0dbfs]

If you use standard 19" RU dimensions you can make a very nice console frame out of rack rails and wood. You can also part it out or repurpose fairly easily if you change your mind and only get 6 modules built.

My plan is to use 19" RU dimensions for the channel strips, but have the faders on their own. Since the module spacing will be identical, it would be fairly trivial to cut the width between the rails to allow for the shorter fader modules. However, panel design and frame layout is all very premature.

-Matt

Sweet. Faders can use the same xxRU (1.75") width as well as meters... Yes, they don't need to be 19" tall though.
I also thought about instead of using a square chassis for channelstrips, to slope the rear at an angle so you get that nice console profile. You don't get as much space/depth near the bottom for your PCB's though. I/O connectors then go on the rear/side. I would post a sketch but don't have the means to do that at the moment.

Cheers,
j

 

[boji]

VUs are not terribly accurate but are quite pricey.

IMO, LED bargraphs are more readable too (except in high light enviroment), particularly if they are in a more or less vertical arrangement and use different colors.

Yes yes.  But wait till you see my VU's backlight colors dance in any color you wish! (I replaced the bulbs with RGB led's)  ;D

 

[JohnRoberts]

That would be unnecessary. You may use an SPDT relay there, or parallel both sections for improved reliability. BTW, you may consider using CMOS switches there, MAX 312 type. Much more reliable and less consumption.

I was under the impression that leaving the outputs floating was bad practice.

np problemo

CMOS switching is certainly attractive from a power supply load standpoint, and avoids needing a second ground for the relays. However, the better chips seem to be as expensive as small signal relays, and don't offer the same level of isolation. They also add distortion. However, I may be splitting hairs about theoreticals here. .001 THD at 20Khz is pretty good.

I switched a whole bank of subs between modes using inexpensive CMOS transfer gates, located inside the feedback loop of an inverting opamp. The distortion was unmeasurable with our bench equipment, but this was late '70s (sound tech analyzer).  You need some extra glue components to switch quietly and prevent crosstalk, but this worked very nicely, and is clearly more attractive than relays.

For onesy twosy electronic switching like PFL in master sections i've seen these dedicated switches used, but for larger scale switching needs there are cheaper ways (I was cheap even back then).

Very cool. I do see the VU as one of the simpler parts of the whole; they're really just a unity gain buffer with some filtering and maybe a couple diodes. I may wholesale lift a circuit if I go the VU route since there's no need to reinvent that wheel (although I am probably reinventing countless others, oh well). Most likely, for myself, I'll opt for some kind of led bargraph, though. VUs are not terribly accurate but are quite pricey.

I have long been a proponent of simultaneous peak/VU metering on an LED bar graph. Old school VU meters are eye candy, but not great for seeing what is really going on with a signal, and your financial diet. A simple peak bar graph should be adequate for modern digital media, unless that media has meters that can be monitored instead of reinventing that wheel.

This is my basic approach. In my first broad stroke design revision, I had envisioned some kind of rackmounted modular mix system. However, after running various configurations by our studio's engineers and a few other folks, the consensus seemed to be that at least 50% of what makes mixing on a console "better" is the user interface. Having to move around to your various racks to patch things in and set up EQ moves would get old pretty fast, and take you out of the mix sweet spot.

That being said, it will certainly be possible to build each circuit block into whatever modular configuration your heart desires. For the purposes of this project, all mix subgroup and master buses will be active and differential. If somebody would like to tack a gain block instead of a summing amplifier at the end of the bus, they're more then welcome to design it and do so.

This suggests perhaps some refinement of a standard I/O between all blocks. While simple differentials are adequate within a single chassis to reference between local grounds, if these building blocks need to be free standing, I would use a balanced interface between them which increases the total system cost and power consumption. probably logical to distribute power supply regulation to 3 terminal regulators inside each block. This approach is attractive for DIY but not embraced by major console companies because of the cost. Metal work, connectors, etc it all adds up when multiplied by number of channels. In console design you have to think of all costs in terms of a times N factor. 

If you use standard 19" RU dimensions you can make a very nice console frame out of rack rails and wood. You can also part it out or repurpose fairly easily if you change your mind and only get 6 modules built.

My plan is to use 19" RU dimensions for the channel strips, but have the faders on their own. Since the module spacing will be identical, it would be fairly trivial to cut the width between the rails to allow for the shorter fader modules. However, panel design and frame layout is all very premature.

-Matt

Iu spacing is generous for fader centers but you can put some switches down between them.

JR

 

[haima]

Allen Farmelo has made a console using 1U API channel strips:

http://www.tapeop.com/articles/82/designing-api-analog-console-scratch/

http://www.sonicscoop.com/2011/02/24/the-console-is-king-um-projects-and-allen-farmelo-re-imagine-the-classic-console/

 

[kepeb]

been following, although some is over my head. its easy to envision the final appearance and easy customize-able, flexible options. and its DIY! building a sweet desk up strip by strip has got to be good for you.

I have a vote for the as yet 'premature' physical part if I may be so bold.
if you could fit the top vu/led and bottom fader parts into 3U height layouts like 5oo series
and keep the simple signal/peak light on the main strip even if you do end up with vu's option.

good stuff!

 

[Brian Roth]

http://www.mouser.com/ProductDetail/Panasonic-Electric-Works/TQ2-5V/?qs=sGAEpiMZZMs3UE%252bXNiFaVCLW1ZA8WQBLGF3Vwl4YaEE%3d
\

 

[Nishmaster]

For onesy twosy electronic switching like PFL in master sections i've seen these dedicated switches used, but for larger scale switching needs there are cheaper ways (I was cheap even back then).

Well, if it gets the vote from you guys, I'm assuming the CMOS switching may not present the problems that traditionally I've heard about with them. My knowledge of these devices is coming from the stompbox world, though, and I'm sure you all know how inaccurate some of that information is.

I have long been a proponent of simultaneous peak/VU metering on an LED bar graph. Old school VU meters are eye candy, but not great for seeing what is really going on with a signal, and your financial diet. A simple peak bar graph should be adequate for modern digital media, unless that media has meters that can be monitored instead of reinventing that wheel.

Someplace I remember seeing a bargraph driver chip that had simultaneous peak/VU style metering, which would be an inexpensive compromise. I know VU's are the prettier (i.e. producer drool factor) option, but since this is a desk for me, I only have my own drool to wipe up, and I just don't see the benefit.

This suggests perhaps some refinement of a standard I/O between all blocks. While simple differentials are adequate within a single chassis to reference between local grounds, if these building blocks need to be free standing, I would use a balanced interface between them which increases the total system cost and power consumption. probably logical to distribute power supply regulation to 3 terminal regulators inside each block. This approach is attractive for DIY but not embraced by major console companies because of the cost. Metal work, connectors, etc it all adds up when multiplied by number of channels. In console design you have to think of all costs in terms of a times N factor. 

I think I was perhaps not clear in my description of my first design revision by some people's responses. For exactly the reasons you describe, John, this is all going on some cards in traditional console form. While it would be quite neat to offer some standardized 1U modular interface (think BladeServers for you IT folks) that can be swapped with ease, the metalwork, weight, heat transfer, power supply, and interfacing problems would just be too much engineering for something that will probably never leave the room it's constructed in or be used by anyone other than a few individuals over the course of its life.

However, as I said before, all the schematics will be available for anyone to build into whatever modular format that one desires.

Is there any benefit to having regulators on a per-channel basis? The ones I see are reduction of ripple and rail impedance for the channel, but I also see the drawback as being added heat dissipation within the desk itself instead of at the PS. Does that more or less sum it up?

Iu spacing is generous for fader centers but you can put some switches down between them.

Most big desks I see are right around 1U spacing. The mute and solo switches can perhaps float in there someplace, but I also like the big, wide fader knobs so I don't see spacing as a huge problem.

I also thought about instead of using a square chassis for channelstrips, to slope the rear at an angle so you get that nice console profile. You don't get as much space/depth near the bottom for your PCB's though. I/O connectors then go on the rear/side. I would post a sketch but don't have the means to do that at the moment.

The basic framework as I see it right now has three sections: the fader modules, which will be flat; the 19" channel modules, which will be sloped; and the meters, which will be either vertical or slightly angled back from vertical, depending on how the eye-height works out. There's no reason that the edge connector backplane needs to be parallel to the ground, so the whole backplane can be at an angle to allow for the module height to be consistent along the whole length of the module.

Allen Farmelo has made a console using 1U API channel strips:

That is one of the cooler things I've seen in a long time. I like the design, although if I were to put a look I have in mind down to a few words, I would say less IKEA and more Airbus. I do love modern furniture design, though.

I have a vote for the as yet 'premature' physical part if I may be so bold.
if you could fit the top vu/led and bottom fader parts into 3U height layouts like 5oo series
and keep the simple signal/peak light on the main strip even if you do end up with vu's option.

3U height may be indeed a perfect height. We will see when it comes meter design time.

http://www.mouser.com/ProductDetail/Panasonic-Electric-Works/TQ2-5V/?qs=sGAEpiMZZMs3UE%252bXNiFaVCLW1ZA8WQBLGF3Vwl4YaEE%3d
\

Brian, those were the exact relays I was looking at. Certainly less thirsty than some others, but even so, adding 30ma per channel to the power supply requirements is no small beans. That's at least around 1.5A, just for relays. Perhaps when I add up all the current requirements of each channel that won't be a large percentage of the whole, but moreso than sharp pencil design for budget's sake, I'm trying to do sharp pencil design for power supply requirement's sake. That being said, I have no point of reference when it comes to console supplies. Are we talking 5A? 10A? 20A?

And to the rest:

It's been an extremely taxing weekend (worked 7-3:30 on Friday, drove out to Ohio immediately to attend wedding, then back again yesterday at 7am to play a wedding back in Chicagoland, eesh) so I apologize that I haven't moved this train along for a few days. I'm going to look into the CMOS switching option, which looks quite attractive, and I think the overload LED I may actually move to after the EQ section, where it makes much more sense to have it.

-Matt

 

[Brian Roth]

I mentioned that particular relay because Amek used a zillion of them in the huge 9098 ("Rupert Neve the Designer") desks.  In those machines, the audio buckets were powered from +/- 17.5 VDC for audio and +5 for all the switching logic and relays.  Panasonic also makes those relays with 12 and 24V coils, but oddly are more expensive from Mouser.

Bri

 

[JohnRoberts]

For onesy twosy electronic switching like PFL in master sections i've seen these dedicated switches used, but for larger scale switching needs there are cheaper ways (I was cheap even back then).

Well, if it gets the vote from you guys, I'm assuming the CMOS switching may not present the problems that traditionally I've heard about with them. My knowledge of these devices is coming from the stompbox world, though, and I'm sure you all know how inaccurate some of that information is.

Just for clarity, I didn't use these simple CMOS TG the same way they were used in floor pedals. The basic TG has a resistance change with voltage that causes distortion used open loop. I placed these TGs inside the feedback loop so voltage and current through them was near zero. I also used some extra glue to reduce clicks and crosstalk.

Using them in a virtual earth node reduces the voltage change across them so better, but not the current change too, which is best.

IIRC there were some old threads discussing this approach if the old archives are searchable.

I have long been a proponent of simultaneous peak/VU metering on an LED bar graph. Old school VU meters are eye candy, but not great for seeing what is really going on with a signal, and your financial diet. A simple peak bar graph should be adequate for modern digital media, unless that media has meters that can be monitored instead of reinventing that wheel.

Someplace I remember seeing a bargraph driver chip that had simultaneous peak/VU style metering, which would be an inexpensive compromise. I know VU's are the prettier (i.e. producer drool factor) option, but since this is a desk for me, I only have my own drool to wipe up, and I just don't see the benefit.

I have seen peak with peak hold combo chips, and either /or chips. I am not aware of any chip solutions that copied my old invention directly.

Peak with peak hold may look similar, but it doesn't reveal crest factor or average (loudness) information.

This suggests perhaps some refinement of a standard I/O between all blocks. While simple differentials are adequate within a single chassis to reference between local grounds, if these building blocks need to be free standing, I would use a balanced interface between them which increases the total system cost and power consumption. probably logical to distribute power supply regulation to 3 terminal regulators inside each block. This approach is attractive for DIY but not embraced by major console companies because of the cost. Metal work, connectors, etc it all adds up when multiplied by number of channels. In console design you have to think of all costs in terms of a times N factor. 

I think I was perhaps not clear in my description of my first design revision by some people's responses. For exactly the reasons you describe, John, this is all going on some cards in traditional console form. While it would be quite neat to offer some standardized 1U modular interface (think BladeServers for you IT folks) that can be swapped with ease, the metalwork, weight, heat transfer, power supply, and interfacing problems would just be too much engineering for something that will probably never leave the room it's constructed in or be used by anyone other than a few individuals over the course of its life.

However, as I said before, all the schematics will be available for anyone to build into whatever modular format that one desires.

Is there any benefit to having regulators on a per-channel basis? The ones I see are reduction of ripple and rail impedance for the channel, but I also see the drawback as being added heat dissipation within the desk itself instead of at the PS. Does that more or less sum it up?

The 3 terminal regulator also provides per module short circuit current limiting and over temperature shut down protection.  It also simplifies scalability without having to tweak a large regulated PS for future upgrades. Multiple unregulated supplies can be integrated into a loose system with less brain cells having to die. 

JR

 

[abbey road d enfer]

Is there any benefit to having regulators on a per-channel basis? The ones I see are reduction of ripple and rail impedance for the channel, but I also see the drawback as being added heat dissipation within the desk itself instead of at the PS.

You have to be extremely careful with on-card regulation. Regulators are noisy, and the resulting noise current is dumped into the ground via the stabilization caps. This noise injected in the ground pollutes the mixing amps. I've done comparisons between on-card 7815/7915, on-card M317/337 and plain and simple 22R fuse resistors. The resistors are the winners, then the LM317/337 and the 78/79 make the mixer noise performance unacceptable. I even experimented with an additional dirty ground bus that received only the stabilization caps; it was not as clean as the resistors.

 

[JohnRoberts]

Is there any benefit to having regulators on a per-channel basis? The ones I see are reduction of ripple and rail impedance for the channel, but I also see the drawback as being added heat dissipation within the desk itself instead of at the PS.

You have to be extremely careful with on-card regulation. Regulators are noisy, and the resulting noise current is dumped into the ground via the stabilization caps. This noise injected in the ground pollutes the mixing amps. I've done comparisons between on-card 7815/7915, on-card M317/337 and plain and simple 22R fuse resistors. The resistors are the winners, then the LM317/337 and the 78/79 make the mixer noise performance unacceptable. I even experimented with an additional dirty ground bus that received only the stabilization caps; it was not as clean as the resistors.

+1, I encountered this problem once in a large console and ended up black listing one vendor's 3 terminal regulators that were problematic in this one design, when they refused to even provide a noise specification, but this is not a simple "always happens". It can and should also be addressed by proper ground design and differential treatment of signals between local grounds. Ideally successful differential handling should ignore even noise on local grounds, but ALL system noise has consequences. Noise on regulated supplies will get pumped into sundry grounds through the PS decoupling capacitors, and even opamp PSRR will inject a fraction of this noise into signals. Even when using local on board regulation, I ripple filtered the unregulated supply to keep hum out of the chassis.

While consoles are made up of simple circuits, the scale when numbers of these simple circuits are combined sometimes reveals unexpected problems, like this.  I repeat that consoles are one of the more difficult simple products to design. 

JR

 

[Nishmaster]

I know the mantra is that ground is not "ground" on this scale but simply a local reference voltage potential, which is what I keep repeating to myself. With that in mind, how many differential blocks need be constructed? I was thinking that unbalanced throughout the channel strip and then balanced to anything outside of the strip would be sufficient. Currently, I have as follows (in the case of a signal from input to L/R bus to main out):

Diff Input -> Line Receiver -> Trim -> Filters -> EQ -> Pan/Level -> Diff Line Driver -> L/R Bus -> Diff Mix Amp -> Driver -> Out

Also, here is a dumb question I should know the answer to: Is +4dBu +/-1.25v or 1.25v P-P (ie +/-.625v)?

 

[JohnRoberts]

I know the mantra is that ground is not "ground" on this scale but simply a local reference voltage potential, which is what I keep repeating to myself. With that in mind, how many differential blocks need be constructed? I was thinking that unbalanced throughout the channel strip and then balanced to anything outside of the strip would be sufficient. Currently, I have as follows (in the case of a signal from input to L/R bus to main out):

Diff Input -> Line Receiver -> Trim -> Filters -> EQ -> Pan/Level -> Diff Line Driver -> L/R Bus -> Diff Mix Amp -> Driver -> Out

Also, here is a dumb question I should know the answer to: Is +4dBu +/-1.25v or 1.25v P-P (ie +/-.625v)?

+4dBu is 1.2V ave not p-p.

The answer about grounds os it depends.. on things like how much and what kind of current is flowing in those specific grounds.  For example several send pots could be dumping mA of signal current into a local ground that across some distance could generate a potential voltage or cause crosstalk.

JR

JR

 

[Nishmaster]

Would it make the most sense then to have separate grounds per block (ie, one ground for the input, one for EQ, one for sends, one for pan/level/bus out) and then star them on the backplane to the main ground bus?

Already I am potentially dumping some 2.5mA to ground at the trim pot. Across a maximum of 19" of PCB ground fill before going balanced, that doesn't seem too terrible to me, but that also won't be the only ground dump. There will be at least 4 aux send pots (maybe six), as well as the pan, fader, and whatever the filters and EQ dump to ground.

The way I see it, if I have a separate ground leg for each module, then the module ground potential can be at most the maximum current to ground x ground plane resistance on that module to PS ground. So, let's do the math. Lets assume the send pots are 5k, like the trim pot, with unity trim on the channel. Let's also say ground leg resistance is 10R (not sure if this is too much or too little assuming a big ground bus bar across the console with good ground fill on the backplane and channel modules).

1.25v * 5K = 2.5mA * 6 Send Pots = 15mA to ground
15mA * 10R = 0.015v ground potential

Does that look even vaguely correct? I get .015v being -38.4dB (gain of -83.3) from the original signal. Surely that doesn't mean my crosstalk is only -38.4dB, does it? The ground potential is obviously modulated by the current to ground, but how much does this get back into the other amps?

-Matt

 

[JohnRoberts]

Don't over think it. Ground is like a sewer system so don't worry about it too much about the poop after it is dumped.

Generally when I look at console blocks almost every active stage is an opportunity for differential forward or back, Some like EQs are less accommodating, but fader gain stages and such can serve extra duty as a differential.

once again it depends... but signals are usually coming from somewhere or going somewhere.

JR

 

[alexc]

> Don't over think it. Ground is like a sewer system so don't worry about it too much about the poop after it is dumped.

Classic !  ;D

Now I understand better!