diy 12+4 in 2 out summing / mixdown console

[Joechris]

Well i try, screens and chassis tied to audio ground at one point only. 4 power busses should work fine, +- and audio ground and screen chassis ground. Beefy ground busses. Atleast thats how its done in my console.
Dont think star ground is nessesary...others migth dissagree. Separate power for the computer and led stuf...mabye do not conect the negative data/led power to audio ground...hopes somebody that knows this stuff chimes in

 

[audiomixer]

I finally have resumed work on this, had to take a little detour regarding the opamp / DOA question that was in my head..... or still is, depending on the point of view. I will start with my original (opamp based) design and might go on to something esle in the future.....

now I got some of the components for the PC control part (trackball) so I started fooling around with that and came up with some modifications on the center part of the console. I had to add a rotary encoder to  act as a scroll wheel, works great right out of the box. the mouse buttons have leds now and I tested the general layout / feel / ergonomy on a 1:1 scale cardboard model I had manufactured...

here are some pics of my quest, more on my build blog on http://audiomixer.wordpress.com...

cheers,

-Michael

 

[abbey road d enfer]

Dont think star ground is nessesary...others migth dissagree.

Star ground in a mixer is counterproductive. Remember: "Star ground is an electrician thing for safety", repeat every morning.
Grounding in any audio piece of equipment must be such that it minimizes circulation voltages between adjacent stages. So hierarchical ground is the proper answer. easy when you have just one input going to one output, as in a HiFi amp, but in a mixer, you have two dimensions: depth and width. You must minimize ground resistance along the channel, which is complicated by the fact that signal may go round and around (inserts, tape monitor), but it is quite workable by making sure that no return currents are allowed to flow where they don't belong. You may want to draw a simplified diagram with ground connections replaced with resistors, then you'll figure out.
You must also minimize longitudinal resistance; clearly, connecting the channels to a central star-point via a length of cable is not the best way to do this. You must have a solid longitudinal ground connection running all the width of the mixer, each channel connected as solidly as possible. A good strong copper bar is welcome.

 

[audiomixer]

I had another go on the basic channel (I build four channels on a breadboard) and I took great care to disassociate power return path and 0V reference voltage up to the power supply connector.... this works nicely enough but I still have to do some measurements. the star topology was to deliver equal 0V reference to all subboards. I don't see why this is for safety only(?). this will not be a large production size 96 channel board with several meters between modules, but still I think to have equal length of wire to all boards insures a more equal 0V reference at each board. having the 0V reference on the whimsy 0.1" header and AWG28 of the bus is asking for trouble... affing a mayor buss bar is one oth the recomendations I always hear, but in my opinion it just reduces the scale of the problem, it doesn't resolve it. some modules will still be 'further away' from 0V (@ at the main power supply ground point) then others.

up to now I have learned the following in my experiments:

- have LEDs, relays, MCU, USB stuff and other  on a separate power supply line.
- connect all grounds and returns a one place
- have dedicated power return for bypass caps
- have dedicated 0V ground reference for audio
- have equal length 0V ground wires to all boards (try measure the voltage drop at a leter stage)

strangely enough I have only very little capacitance on the boards themselves - i might try to add some local electrolytic, but I am not sure if they will help or make things worse.... have to try.

I was wondering if you guys have measurements of the noise floor of your desks... master full open, no routed faders, 16 routed etc...

- cheers,

michael

 

[abbey road d enfer]

I had another go on the basic channel (I build four channels on a breadboard) and I took great care to disassociate power return path and 0V reference voltage up to the power supply connector.... this works nicely enough but I still have to do some measurements. the star topology was to deliver equal 0V reference to all subboards.

You are seeing only one side of the problem. What about the currents that run from the module to the star point? They will produce voltages that will not cancel because they come from different channels or sections receiving/processing different signals. The reference path for the summing bus will see all these voltages and amplify them. Balanced bus help in that matter, but they are not the absolute answer.
Then there is the matter of inductively generated voltages in the ground connections. These voltages are determined by the circulation of magnetic field along the path of the conductor (circulation is the integral of the vectorial product B.dl). By making the conductor length equal, you don't change the fact that since they don't go to the same place, there is very likely a differential voltage between the extremities. The only way to minimize longitudinal voltages is to make the reference conductor the shortest and the more solid as possible. Brute force is always a winner.

I don't see why this is for safety only(?).

Star grounding is a way to make sure that every part has an earth path whatever happens to the other parts. It's for electricians. 

 

[audiomixer]

I has taken me a while to think of an answer, and I wanted to document my view on this with some measurements or simulations, but i haven't found the time to do it properly. so it will stay as is: your view against mine ;-).

it is true that I might not get rid of all ground related issues in this way, but I will try none the less. I wanted to measure the ground current that flows on the different return path, but my workshop is just not equipped well enough. It's true that most of the current will flow from Vcc to Vee, not to ground. what I did is to have a look at what goes on on the ground and 0V reference traces (i did separate them on the strip-board test channels I made) but it's difficult to see a real difference. what I can say is there is much less noise on both ground and 0V reference on the channels that are not actively driven, (two channels are on the left, two on the right side) even if they are just a few centimeters away and on the same board. so the isolation with a common (star) ground point on the board itself does work - at least to to some extent.

would be interesting to see what currents flow on the ground 0V reference. I need a current microscope probe...

-cheers,

Michael

 

[abbey road d enfer]

what I can say is there is much less noise on both ground and 0V reference on the channels that are not actively driven, (two channels are on the left, two on the right side) even if they are just a few centimeters away and on the same board. so the isolation with a common (star) ground point on the board itself does work - at least to to some extent.

What isolation? You're talking about a mixer, isn't it? The individual channel grounds will need to be joined together, unless you want to have them floating with different voltages in respect to the summing amp reference. Maybe you'll have fully differential mix bus, that would be a solution, but experience tells me (an I'm sure JR follows me on this, as all the successful console designers I know), there is nothing like a solid common audio ground for the elimination of longitudinal noise.

 

[audiomixer]

all power return and 0V reference will finally meet at a central point, naturally. the small board I made has a single power connector (well two to be honest, so I can add a second board powered in daisy chain - duh!). the connector sits right in the middle and V+, V-, Ground and 0V reference is distributed to the (identical) circuits left hand and right hand. both power return and 0v reference meet at the connector, just a few centimeters away, as they would at the power supply main earthing point - think of the larger amplifier ground wiring diagrams with all ground or 0V reference going to one central point just at the filter caps.

now there are only a few centimeters, but the idle side (was testing with one channel active at that moment) has almost nothing going on on the 0V reference ("clean") and power return ("dirty") while the other side has quite some noise on the ground and 0V reference. of course they are all at the same level in term of ohms and fraction of ohms and thus volts, cable lengths and therefore trace resistance are ridiculously short in respect with a mixer, but the difference is visible non the less. the current follows a dedicated path and is confined in that path.

I don't want to run the 0V reference and the power return on a 28 AWG flat cable trough all channels, I'll have a separate connector with dedicated 0V reference and power return anyway to all modules. By the way I looked for some decent sq.area flat ribbon cable with IDC connectors but this seems to be almost impossible to find.

balanced and dedicated 0v reference for the mix buss are to be checked also, might make it even better, but the implantation is again much more complex. aarghh! more complex, more mistakes possible.

cheers,

- Michael

 

[abbey road d enfer]

all power return and 0V reference will finally meet at a central point, naturally. the small board I made has a single power connector (well two to be honest, so I can add a second board powered in daisy chain - duh!). the connector sits right in the middle and V+, V-, Ground and 0V reference is distributed to the (identical) circuits left hand and right hand. both power return and 0v reference meet at the connector, just a few centimeters away, as they would at the power supply main earthing point - think of the larger amplifier ground wiring diagrams with all ground or 0V reference going to one central point just at the filter caps.

You just cannot compare a guitar amp, which is a single continuous channel, with a mixer that has several ladder paths AND a super-extra-sensitive stage, the summing amp.

now there are only a few centimeters, but the idle side (was testing with one channel active at that moment) has almost nothing going on on the 0V reference ("clean") and power return ("dirty") while the other side has quite some noise on the ground and 0V reference. of course they are all at the same level in term of ohms and fraction of ohms and thus volts, cable lengths and therefore trace resistance are ridiculously short in respect with a mixer, but the difference is visible non the less. the current follows a dedicated path and is confined in that path.

Exactly; all your channels will have a different "ground", i.e. a different potential. This potential is actually in series with the output signal of the channel, so it will be amplified by the summing amp, unless you use differential bus.

I don't want to run the 0V reference and the power return on a 28 AWG flat cable trough all channels, I'll have a separate connector with dedicated 0V reference and power return anyway to all modules. By the way I looked for some decent sq.area flat ribbon cable with IDC connectors but this seems to be almost impossible to find. 

You'll never find IDC ribbon cable with enough copper section to make a half-decent mixer ground. In the 70's, many mixer manufacturers used Molex connectors with AWG22 wire; that was acceptable for mid-sized mixers (16 ch.), but then they started using ribbon cable, which was becoming more and more economically justified, because of the wide usage in computers. Ribbon cable is AWG26, which is 3 times smaller in section, and they found they were unacceptable, so they had to provide a separate connection for the reference ground.
But thin gauge is good for X-talk and bus capacitance.

 

[JohnRoberts]

Sorry I don't spend much time in this area of the forum...

In consoles ground is ground and audio is audio + and audio - or audio hi relative to audio low.  Audio low can be the same thing as a local audio ground, but it is helpful to think of all audio signals as the audio + relative to the audio -  wherever that audio + is coming from or going to.

In fact you can have several audio - nodes hard connected to the same larger ground system. While these different ground nodes can and will have small voltage differences between them, the local audio can still be clean relative to the local ground node.

To maintain audio signal integrity, use differential stages to accurately reference between different local grounds. In the example of a sum bus, sum the audio + into and audio + sum and the audio - or local channel grounds to an audio - sum. Then subtract the audio - from the audio + to extract clean audio and reference it to the master section local ground. This summed audio will accurately contain just the audio + relative to audio - and ignore all the ground to ground differences.

I hope this makes sense. Think differential.....

JR

 

[audiomixer]

I am aware of the benefit of a symmetric mix bus, just wasn't quite sure if I want to implement it... but seems the technically best option. right now I am testing different sum mix technologies - from simple opamp to transistor enhanced to a differential bus, but I am rather slow, had some minor difficulties with my implementation. should have double checked that there is no V+ on one of the opamp inputs, they don't like and emit smoke in protest...... must say that I had to find out how big difference the capacitive crosstalk and therefor the right routing makes. I had the mix bus running parallel to a ground trace (capacitor return path to supply) and could simply not achieve any decent noise performance. rerouted that and is now several dB better.well, still trying to get some hard numbers on all that.
pc power supply doesn't help either.

regarding ground, decoupling and supply rails I kind of have made up my mind and want to implement that according to some of the threads here, with two electorlytics to ground and a small 100n mkt between the rails. we will see if thats any better.

- michael

 

[JohnRoberts]

This mix bus was just the poster boy example for the benefit of differential signal handling.  Another common example is the typical channel post fader gain stage (often +10dB), where differential treatment can properly reference signals forward and backwards, even improving things like fader kill specs. 


JR

 

[alexc]

Thanks John for the interesting discussion regarding differential treatments.

Could you please elaborate on the example of the post fader gain stage ?

Do you mean that the wiper of the pot is fed into the + terminal of a differential stage with the - terminal connected to local ground?

 

[JohnRoberts]

Thanks John for the interesting discussion regarding differential treatments.

Could you please elaborate on the example of the post fader gain stage ?

Do you mean that the wiper of the pot is fed into the + terminal of a differential stage with the - terminal connected to local ground?

Again this is just another example but things to think about in console design and layouts is follow the current, and remember that wires have resistance.

The differential for the fader gain stage can generally only be perfect for one gain setting. It is common to dial it in for max fader kill, so yes in the non-inverting case, the feedback gain resistor string gets grounded to the fader ground connection. Inverting topology has higher noise gain for the same closed loop gain so is not commonly used, while the difference between 3x and 4x is not a show stopper if there is some good reason to make it inverting.

Trying to make everything inside a console perfect, can make you a little nuts, so you need to set up a hierarchy of what is most important and optimize in that order. While we all find stuff out after the first build, that we didn't anticipate.

JR

 

[Pliplo]

Great project! Congratz! Any progress?

 

[audiomixer]

Hi all,

I did make some progress, finally got to it and made the first board, switch and logic board interfacing to the micro controller board. I am waiting for it to arrive, for starters I just made a single board, but I have to learn some more before I get into real production. smd skills, general layout and the like.

we will see if this turn out right when I have some time to solder some components ;-)

cheers,

Michael

 

[audiomixer]

Done the first bit of kit tonight! exciting!

these are just a few switches and leds, but they are SMD (my very first), on my pcb design, pcb done by myself, new SMD oven, new toner transfer done (quite) nicely... so I'am kind of happy about it.
I had to botch one smd led (two led under each switch...) that wanted to swimm away in the oven, but otherwise everything is fine. I am experimenting with solder mask, that would give it the edge of course.

I finally got my arduino centipede I/O extension boards today, I forgot how long 7 to 21 days shipment can be. I will get my channel switch boards this week or beginning of next week, the PCB fab has images of the making of the boards (of mine, I can track what they are doing!) online. thats kind of cool ;-)

so the next few weeks should be about switches, leds, arduino, and writing preliminary code. not my favorite, but at least visible progress.

more pics to come on my blog audiomixer.wordpress.com

time for a beer for me - cheers,

michael