20 Channel Line "Summing" Mixer

[ashcatlt]

For about 13 years I've been dragging around one of these:


Mine doesn't have the meter bridge, but it's bigger than it looks - about 4 x 3.5 feet – and weighs a ton! It really is a great board with all the pro features. It's a “inline” console with a separate Tape channel for each of the 32 regular channels, so it can act as a full 64 channel board. All kinds of sends and inserts, 8 sub busses. It reads and writes SMPTE timecode and offers mute automation via either snapshots (synced to the SMPTE) or MIDI note on messages. It's even sends MIDI control change messages from 4 of the bus faders. Doesn't respond to them, but it can be used to control other equipment this way.

Back when I was recording to 16 tracks of ADAT and running like 10 different hardware MIDI synth modules this thing was completely indespensible. Combined with a couple patchbays I had all the routing power necessary to make any horrible noises I happened to be shooting for. I've had it in several different studios from Tampa to New Orleans to Superior.

Currently it takes up about half of Sanidar's basement, but since I'm now recording and mixing ITB exclusively, it's become nothing much more than an oversized monitor controller. That and it's getting a bit old. Most of the time we've had it, it's been protected from cigarette smoke, but if I could get in and scrape all the resin out of it, I could probably stay stoned for a couple years. So it's getting a little flaky. I don't think there's anything wrong with it that a bit of contact cleaner wouldn't fix, but since there's about 700,000 pots, jacks, and switches which need cleaned...

It's just got to go.

So I built this:


The smaller box is a +/- 12VDC supply which I built from the General Guitar Gadgets Bipolar Power Supply schematic. Got lucky to find exactly the regulators I needed in a Proverb which I've had around for over 20 years, but have been afraid of since it runs very hot. I also happened to have a 16VAC power transformer in my box of misfit walwarts. Switched the end to a 1/8” TS so I could interface it with a jack I had. The power for the mixer is connected internally, but eventually I'm going to need at least a couple of mic pres, so I went ahead and provided for power to them via the jacks on top. Rather than invest in a 3 or 4 pin connector for this, I decided to just use 2 TS jacks – one for + and ground, and the other for the -.

Originally I was going to build this thing right into a “modular” patchbay that I had laying around. The idea was that I would drill holes through each of the modules to run the Left, Right, and Ground bus wires through. I'd solder a resistor between the tip of one of the jacks and the appropriate bus, with the sleeve soldered directly to the bus. With 20 inputs in a 24 channel patchbay this would have forced me to get the amplifier module into the space of 2 modules, so I could use the last 2 modules for the outputs. That alone would not have been too difficult, but I still would have needed a separate box for the PSU and another for monitor routing/volume control. When I actually opened up the patchbay I found that it was pretty much set up exactly wrong to follow this plan. I could have probably done it, but it would have been a bh and a half!

So I jumped on another plan. I pulled a 6 ft long 8 channel recording snake out of the system and hacked it in half. That's 16. Then I took a stereo pair and hacked it in half, now I've got my 20. (Why 20? I'll explain a little later) In testing, I found that half of the white cable in the snake was flaky, so I replaced it with half of a single cable which had once been part of a stereo pair. Now I had 20 TS plugs on one end and bare wire on the other end. Stripped and twisted all 40 wires (inner conductors and shields) and soldered them to a bus arrangement across the board. From there, the amplifier section is easy – just two of these (the 2K resistor at the front represents one of ten inputs):


The opamp I used is a TL072 and it sounds fine to me, but it's socketed in case I ever decide to go cork sniffing.

The jacks are all stereo jacks, and the pots are dual ganged. The single pot over by itself is the Master Gain. The jack beside it is a passive output. Comes from the node where all the 2K resistors combine, and is there in case some time I want to use something else (a tube preamp or whatever) as a recovery stage. Not that I have any plans for that sort of thing right now, but I like to keep my options open, and for $0.50 and about 30 seconds to drill the hole...

The other four jacks are the active outputs, and each has it's own independent Volume control. In the studio right now I've got two working pair of monitors and a third awaiting a power amp, plus a headphone distribution amp. So I've got independent control of volume for all of these.

I haven't actually gotten the chance to test it since I got all the inputs connected, but I did some tests beforehand by connecting the headphone output from my iphone to the input resistors and the outputs to a small set of computer speakers on my new workbench. It sounds fine to me. With gain all the way down it's basically a unity gain buffer and is flat for as far as the eye can see – like Nebraska! When the gain is way up, theres a bit of rollof at either end of the audio spectrum: -3db at about 20 Hz and about 18KHz. But there's a lot of gain! Somebody told me that each doubling of the inputs will drop the output by 6db once everything is connected. Of course, if two signals are about the same level, mixing them together causes the output to be about 6db up, so that should offset.

Anyway, I don't mind running it at unity. It's got a whole pile of headroom though. 24V peak to peak is enough to crush just about any line-level device. Even with the iPhone and the Gain cranked, I couldn't hear any distortion once I turned the Volume down to a level the poor little amplifier could handle. Not exactly scientific, I guess, but good enough for me.

So how's it going to be used? Well nowadays I'm using a Fostex D2424LV as my “recording interface”. It's connected via lightpipe to my the VSL2020 card in my computer. This only has 2 ADAT I/Os, so I can get up to 16 simultaneous tracks in and/or out at a time, which is more than I've used in years.

So I'll soon have an 8 space rack. 3 for the D2424LV, 3 for my computer, and 2 for patchbays. One patchbay will be for the inputs to the recorder. The other will have the first 16 outputs of the recorder on the top half-normalled to this mixer's “pigtail” on the bottom. The other 4 inputs on the mixer will be for the stereo outputs from my computer and then a spare set for anything else we might want to patch in. The outputs will go to the monitors I discussed earlier and everything will happy. If desired, I could run one of the outputs back inputs on the computer (either via the D2424LV, or the analog ins on the card) to try the whole analog summing thing that everybody seems to think is so cool. I don't really buy the hype, but again I've got the option.

Overall, this was a pretty simple build. A whole lot of soldering connecting that pigtail, but otherwise not a particularly complex circuit.

[JohnH]

Very nice indeed. And I like your old mixer too but I sympathise with the issues of servicing such a large desk of that age, once the pots and contacts start to go flaky.

Now here’s a comment about your circuit - wondering why not an inverting arrangement?

There’s two ways that I’m aware of to make an opamp act as a mixer, and the other way is to basically ground the + input, and have all inputs coming to the – input, with the feedback loop with the gain pot from output to -. That’s the arrangement I’ve seen more often in mixers.



The two main differences as compared to yours is that the – input then sits at 0 volts, a virtual earth, and so there is less cross talk between channels. Also, as you add more connections to the input, the levels from other signals are not reduced. The gain pot would then go from 0 to max.

cheers

John

[ashcatlt]

Well, mostly I have an irrational desire to maintain absolute polarity.

I have been a bit concerned that I might get some crosstalk at higher gain levels thanks to there being only 1360 Ω between the feedback paths. This will be like negative crosstalk, won't it? Might cause some stereo widening?

I've always heard that non-inverting gets you higher input impedance with less noise, but I don't really know how that works out here. I don't see needing gain < 1.

[JohnH]

I would think any cross-talk will be in a positive phase - a partial mixing of all the inputs getting back into each seperate channel, depending on the ratio of its output impedance and the mixer input impedance. The mixers input impedance seen by each input device will depend on how many are connected - 1M for a single input, 4k for two inputs, decreasing down towards 2.1k for all 20 inputs connected, assuming all devices have a low output impedance. Thats another thing to check - if these are for line inputs, 2k may be too low for the input resistors? (unless your connected devices are like 200Ohm or less output Z)

Interesting about the lower noise on a non-inverting amp. I dont understand noise properly, but it could be right. It seems to me that your arrangement makes the opamp work less hard than in an inverting configuration, since it is not having to send a lot of current back through the FB loop to keep the virtual earth at 0v.

Main thing is - if it works, it works.
J

[ashcatlt]

The noise thing that I've heard about is just that a high in-Z in an inverting setup requires a large series resistance, which apparently means more thermal noise or whatever.

The recorder says it's out-Z is <600 Ω, and things I read led me too believe that the 2K would be generally safe. I guess it doesn't exactly meet the (at least) 10 to 1 Rule (of thumb), but... Darn I hope it works okay in practice. I think I simmed it up with cable capacitance on the input.

[ashcatlt]

Initial test of the finished product shows positive results.

I plugged 16 channels from the Fostex into the mixer. Set up a stereo track in Reaper and had it send to all 8 pairs of ADAT outputs through the Fostex. I loaded up recording of a song which I know very well having listened to it in all sorts of environments for like 20 years. Set up a second stereo track to record back from the output of the mixer.

The original file was peaking at -0.6dbfs.

With Gain all the way down and Volume all the way up (unity at both stages) The new file came up at -2.6dbfs.

With Gain and Volume all the way up, it went (tried to go) WAY over 0dbfs. I say "tried" because 0 is the end of the scale. Based on my calculations and simulations I'd guess it would have gone to something like +18 or 20dbfs if it could.

With Gain all the way up and Volume rolled down to where the input matched as closely as possible the original file level, there was noticeable distortion which must have been coming from the opamp stage. It's not the worst sounding opamp distortion I've ever heard, but...

Then I set the Volume all the way up and adjusted the Gain to where it came out about the same level as the original file. This happened right around 3 on the knob. There is no audible distortion here. There is a subtle difference in the frequency response and possible the stereo image. It's hard to say which I prefer, especially since my (not blind) A/B listening tests were done by running back through the mixer in question.

So I tried lining up the recorded wave file with the original, normalizing both to match levels as closely as possible, and inverting polarity on one of them. This "null test" was not completely successful. I could not get the signals to cancel completely, but can get a very drastic broadband volume reduction. The frequency analysis plot of the difference signal looks identical to that of either of the originals, so I'm thinking that frequency and phase response must be fairly flat at this gain level. It looks and sounds like I might be getting some small amount of compression or something out of this, though.

It occurs to me that I'm going to want some form of metering - even if that simply means an "over" LED indicator. Since this thing will be interfacing with solid state power amps feeding fairly expensive speakers I should probably have some way of knowing that I'm approaching clipping levels. I'm thinking I might even try to build a simple "diode ladder" type soft limiter in there somewhere.

At this point, though, I'm confident that it will work as intended. Now I just have to figure out what to do with the Ghost!

[ashcatlt]

Okay, so if anybody's still got eyes on this thread, I'm having a little trouble with the last point I made:

What is a good, generally safe, maximum voltage to feed to a typical solid state power amp to avoid clipping the amp itself?

[sumgai]

ash,

Jul 28, 2011 13:01:30 GMT -5 ashcatlt said:

Okay, so if anybody's still got eyes on this thread, I'm having a little trouble with the last point I made:

What is a good, generally safe, maximum voltage to feed to a typical solid state power amp to avoid clipping the amp itself?

The problem is, there's no set standard. It depend on what the amp company built in as a first stage - is it like a pre-amp (with gain), is it strictly a buffer, is it merely a phase-splitter (for push-pull operation), what? Most amps have at least a buffer and the phase-splitter, only a very few that I know about have a gain stage in addition to the other stuff.

In most cases, a signal level of 3 to 5 volts RMS can easily get to full output power, but sometimes it does take more. If 2 or 3 volts RMS seems to sound distorted, then there's likely a small amount of gain inside, and you know not to push so much into it.

Something to keep in mind is that most meters that read AC volts aren't really True RMS meters. And even those that do claim to be such, they don't have very much bandwidth, most likely only 1 or 2KHz for the average sub-$200 meter. All of which means, test and measure your signal at no more than 1KHz. (Unless your meter states unequivocally that it is accurate to a higher frequency - and you trust that statement.)

HTH





sumgai