Aux send/return and mix design : grounding question

Hi,

I designed this circuit to be used as an interface for using Ableton Live as a multitrack guitar Looper. The principle is here :
http://www.wuala.com/ricothetroll/public/DI+aux.png/

Here are the schematic and the layout for one track :
http://www.wuala.com/ricothetroll/public/Live_Looper_Interface_Schematic.png/
http://www.wuala.com/ricothetroll/public/Live_Looper_Interface_Layout.png/

Here are the schematic and layout of the PSU I used :
http://www.wuala.com/ricothetroll/public/Live_Looper_Interface_PSU_Schematic.png/
http://www.wuala.com/ricothetroll/public/Live_Looper_Interface_PSU_Layout.png/
I used an external toroidal power transformer, that's connected to the device but a 4 ways speakon (2x15V AC + Center Tap + Earth). The center tap is connected to the earth right next to the speakon chassis connector.

And here is a gutshot of the 5 tracks device :
http://www.wuala.com/ricothetroll/public/Live_Looper_Interface_Gutshot.JPG/

My problem is that when I plug a guitar and an amp into it - without even connecting the send/return - I get hum on the amp. I first tried without a ground connection on the output, but I had even more hum. The problem can be solved by disconnecting the shield of the cable that goes to the amp (at the amp's end) to cut the (supposed) ground loop, but I guess that's no "clean" solution.

Could I disconnect the center tap of the power transformer and use it for the signal ground, with Earth still connected to the chassis and chassis-isolated jack connectors ? Could it be a possibility to solve my hum problem ? Are there other - better - solutions to isolate my signal ground from earth, apart from the use of isolation transformers ?

By the way, any comments on the design, layout or enclosure implementation are welcome. I can give the Eagle files or copper layout if people here want to build one.

Best regards.

Eric

Well I did some tests and it seams quite impossible to eliminate hum without using isolation transformers at the "to amp" output (there is a ground loop anyway if 2 or more amps are connected).

Could I use those transformers :
http://be.farnell.com/oep-oxford-electrical-products/z1604/transformer-audio-pcb-600r/dp/1172421

Best regards.
Eric

Sure you could add a transformer but resistors are cheaper.

I am having trouble following the actual schematic of your output stage. It seems using some resistors to reference the guitar amp ground to your send/return box local ground could help reduce the hum.  There are many examples of simple differential amps, in your case since you are also combining two signals in that same stage the differential is only slightly more complicated.

To pull a few values from my butt, looking at your non-inverting output stage, you have two 3.3k resistors going to the + input of the output opamp. I would add a third resistor (say 3.3k again), from the + input to the output jack ground and therefore guitar amp ground reference. The - input to the opamp needs a similar network to balance out the differential. Again using a 3.3k feedback resistor from the - input to opamp output, and one half that value 1.6(5)k from the - input to the send/returns internal local audio ground.  Since 1.65k is not a nice round value, you could scale up or down the 3.3k to get values that work out more neatly. The ratio is important for the differential to work properly. 

Your return looks like it has different gain in the + and - leg, which will not be very good for CMR.

I expect the guitar should be quiet with it's ground telescoped out from the send/return ground. You may want to consider adding a compliance (say 100 ohm resistor and .1uF cap) between the ground on the output jack and the send/return local ground.

Your balanced send and return look a little crude, but should work if interfacing with good balanced gear.

JR

Hi John,
Thanx a lot for your answer !
I tried to tie a 3k3 resistor between IC1B non inverting input and the output jack ground connection but it has no effect on the hum. That was the opamp you were talking about, right ?
By the way you are right about the return, the 4k7 resistor should be 10k ! I must admit I don't know why  put that value here  ???
Best regards

i think you must re desing board layout with more consistent ground area.
again you must put 0.1 µF near supply pin of opamp

ricothetroll said:

Hi John,
Thanx a lot for your answer !
I tried to tie a 3k3 resistor between IC1B non inverting input and the output jack ground connection but it has no effect on the hum. That was the opamp you were talking about, right ?
By the way you are right about the return, the 4k7 resistor should be 10k ! I must admit I don't know why  put that value here  ???
Best regards

Click to expand...

You also need the symmetrical resistor network for the negative feedback path.

JR

You also need the symmetrical resistor network for the negative feedback path.

Click to expand...

so the ground path has the same impedance, that's it ?

Best regards.

i use this

Hi Brian,
Thanx for your answer !
I must admit that I don't see much differences between our two schematics, as far as grounding is concerned. The differences may be in the PSU wiring. Where do you get your ground bus from ?

ricothetroll said:

You also need the symmetrical resistor network for the negative feedback path.

Click to expand...

so the ground path has the same impedance, that's it ?

Best regards.

Click to expand...

No... actual impedance between grounds is not very important, it just needs to be large relative to mains wiring impedance to steer ground currents to the appropriate path.

Many opamp circuits can be configured as "differential" amplifiers. What that means is the amplifier can literally subtract two signals from each other and amplify the "difference"...  In your application, we have two different powered chassis. The guitar amp, and the send/return, so these two local grounds will be different. A fact of life.

Ideally we want to compute the difference between the ground on the send return board and the ground at the guitar amp input, and add (or subtract) that difference from the signal at the send/return output, so the signal arriving at the guitar amp is is just pure signal, relative to the guitar amp ground.

This will work even without a hundred ohm, (or whatever value) between the guitar amp ground and send return ground, but this compliance will keep stray currents from flowing between the two grounds if both connect back to safety grounds, and/or have stray rouge currents trying to find the easiest path to earth ground.

The signals being applied to the differential amp all are important. At the plus input with three equal value 3.3k resistors we are getting +1/3rd signal A, +1/3rd signal B, and +1/3rd guitar amp ground. The actual gain of these at the output of the differential amp depends on the feedback values too. The general equation for gain applied to the plus input is  1+(Rf/Rg), where Rf is resistor from output to - input, and Rg is resistor from - input to ground.  So with Rf= 3.3k and Rg= 1.65k, the gain for the three signals applied to the plus side is gain A= 1/3 x (1 + 2)= 1
Gain B= 1/3 x 3= 1
and gain for guitar amp ground is likewise = 1/3 x 3= 1, so the guitar amp ground is present in the output at 1:1. Note: this ground is not added but present at unity so ignored by the guitar amp input that is sitting at that same potential.

The gain for the send/return ground, coming into the - input is simply -Rf/Rg or -2x. So - 2 times the send return local ground voltage. The good news is the Signal A and Signal B, that are also in that output each contain +1x the send return ground each.

So the total result in the output is

+!x signal A (with +1x send/return ground)
+1x signal B (with another 1x send/return ground so +2x send return ground total)
+1x guitar amp ground
-2x send/return ground.
--------------------------
What is left is 1x signal A + 1x signal B, all relative to guitar amp ground, with send/return ground subtracted out.

The actual values of the resistors used don't matter but the ratios need to be correct, so 3.3k resistors can be any values, but all need to be the same value.  Rf and Rg can be any value as long as Rg=Rf/2

This circuit may not seem obvious but this simple differential amp, with it's ability to forward or back reference signals between different local  grounds is the glue that makes consoles and many products possible (without needing transformers everywhere).

JR







 

John, thank you very much for that great explanation ! That gives me a different point of view on opamp design !
Here are the modifications I made to my schematic :
http://www.wuala.com/ricothetroll/public/Live_Looper_Interface_Schematic_v3.png/
I separated the amp's ground (GNDA on the schem) from the local ground (GND) with 100R - 0.1u.
What should I change improve my balanced send/return circuits ? Brian, I see on your schematic that you added some zeners at the HiZ input, what are those for ? I'll add some caps near the opamps supply pins.
Best regards.
Eric
[EDIT] There was an error in the first schematic I posted (v2). Now corrected.

Ok some quick observations.

#1 I can't read the values easily, but as drawn your loop level pot is directly loading down the guitar pickup. These pickups are generally not capable of driving low impedances and will lose HF content. Since you have a unity gain input buffer right there, just connect the pot to the output of that first opamp.

#2 In your loop return input the the two legs have different paths. One side passes through two LPF and the extra stage delay, vs the other leg. This is generally not an issue at lower audio frequencies, but very HF CM noise rejection will not be very well balanced. Using that (and several two opamp topologies suffer from the path asymmetry). this will work, but I would use a smaller cap across IC2c to minimize that mismatch between + and - legs at HF.

#3 For the loop send, TL074 are only rated to drive 2k loads, so they will work OK for typical line inputs (10k) but will have trouble with 600 ohm terminations. Likewise if driving any single ended gear, one side could be driving ground (which is also less than 2k).

#4 output differential, better, but regarding drive capability, if the TL074 is only rated to drive 2k, you feedback network is using up your entire drive budget. At HF when the feedback cap is a low impedance, the opamp is driving the 1k resistor (Rg) to ground. Also the 1n (?) cap across the feedback resistor (Rf) will change the forward gain frequency response and the ground cancellation since it isn't symmetrical for both paths any more. If you want to use a small cap there for stability, use a similar small cap across the resistor coming from GNDA. The lead in that circuit will compensate for the lag in the feedback so ground referencing is intact. Note: these two poles need to be tuned for the same RC so if R is different C will need to be different in both places. Plus, these two RCs will look like a 2 pole roll off to the forward signal so pole frequency needs to be set well above the passband.

There are many variants on balanced sends and returns, and I don't want to get into comparing all the different topologies. What you have will work if applied carefully (don't short outputs, etc).

JR

 

Hi John,

I did the modifications you pointed and it's now working pretty well ! No hum appearing on the amps anymore, even with 3 ones connected.

To have a better view of the image it's better to download it with the link "Télécharger 'Live_Looper_Interface_Schematic_v3.png' sur votre ordinateur" (might appear in english on your computer). The quality is then far better !

About your observations :

#1 I'm using a 1M pot at the input, that sets the input impedance (I guess  : I often find out that some knowledges I think I have are just wrong or incomplete...)

#2 That's right ! I saw that configuration on the balanced output of the GSSL project, but I see now it's not right. I guess there's a compromise to be found between opamp stability and good matching between + and - outputs. Would 10p be right ? By the way, what are the chances for such an opamp configurations (gain=1) to become unstable ?

#3 Here is another thing I didn't consider. Nevertheless it works fine with the Motu 828 mk1 I intend to use. Is there a way to protect IC2B against a short to gnd caused by unbalanced connection ? The obvious answer would be to increase R10-R13 to 600R but then I'll have a 1/2 loss if connected to some 600R input (though 1k2 would still overload the TL074...).

#4 Makes sense, but as for #3, I just didn't think about it this way. I definitely have a lot to learn about audio design ! I used small Rf/Rg values for thermal noise considerations. I'll make Rf 33k and Rg 16k5 (that value is available at Farnell for 0.025€/p so it's just fine). Then I could use 220pF across Rf and 22p across the 3k3 resistor to ground (f~219kHz).

I thought this project would be a pretty easy one but finally there's a lot of things I didn't think about... And that's fine because I learned a lot then ! Thanx a lot for spending time answering me.

Best regards.

Eric

ricothetroll said:

Hi John,

I did the modifications you pointed and it's now working pretty well ! No hum appearing on the amps anymore, even with 3 ones connected.

To have a better view of the image it's better to download it with the link "Télécharger 'Live_Looper_Interface_Schematic_v3.png' sur votre ordinateur" (might appear in english on your computer). The quality is then far better !

About your observations :

#1 I'm using a 1M pot at the input, that sets the input impedance (I guess  : I often find out that some knowledges I think I have are just wrong or incomplete...)

Click to expand...

Im is high enough, but even that doesn't need to be loading the pickup. Total R to ground in parallel with input sets input Z.

#2 That's right ! I saw that configuration on the balanced output of the GSSL project, but I see now it's not right. I guess there's a compromise to be found between opamp stability and good matching between + and - outputs. Would 10p be right ? By the way, what are the chances for such an opamp configurations (gain=1) to become unstable ?

Click to expand...

5-10p is probably fine. In circuits where the output is captive, only loaded by other parts inside the design, it is unlikely to introduce difficult loading. Any amplifier connected to the outside world, need to be alert to RF coming in.  The bifet opamps have large input dynamic range (Vth) so won't saturate and rectify rectify until a couple volts of RF. A small feedback cap can help keep that input voltage down below the 2-3V that causes problems, in extreme RF environments. 

#3 Here is another thing I didn't consider. Nevertheless it works fine with the Motu 828 mk1 I intend to use. Is there a way to protect IC2B against a short to gnd caused by unbalanced connection ? The obvious answer would be to increase R10-R13 to 600R but then I'll have a 1/2 loss if connected to some 600R input (though 1k2 would still overload the TL074...).

Click to expand...

Welcome to the trials and tribulations of circuit design... everything involves tradeoffs.

#4 Makes sense, but as for #3, I just didn't think about it this way. I definitely have a lot to learn about audio design ! I used small Rf/Rg values for thermal noise considerations. I'll make Rf 33k and Rg 16k5 (that value is available at Farnell for 0.025€/p so it's just fine). Then I could use 220pF across Rf and 22p across the 3k3 resistor to ground (f~219kHz).

Click to expand...

Accuracy of the time constants match will affect rejection quality of ground noise at HF.

I thought this project would be a pretty easy one but finally there's a lot of things I didn't think about... And that's fine because I learned a lot then ! Thanx a lot for spending time answering me.

Best regards.

Eric

Click to expand...

It is cliche to say, but opamps can be deceptively easy to design with in general, but when you peel back all the layers, almost everything matters to some degree. I like to joke that consoles are the most difficult, "easy" circuits to design. Simple circuit functions, but a great deal to consider to make it all work nicely together. 

JR

ricothetroll said:

Hi Brian,
Thanx for your answer !
I must admit that I don't see much differences between our two schematics, as far as grounding is concerned. The differences may be in the PSU wiring. Where do you get your ground bus from ?

Click to expand...

The schematic posted describes a real machine used by me for recording an HiZ signal into a daw.
Not have hum, ground bus contain center psu transformer but its important to have an adeguate large ground area on pcb without loop .
Normally loop ground establish on by metal case.
For this I use a switch for ground lift.

Hi,

I tested the device in rehearsal conditions : it works great ! Thank you both again for helping me !

Here are the final, updated files, for people who would like to build it :
http://img855.imageshack.us/img855/2692/livelooperinterfacesche.png
http://img35.imageshack.us/img35/5262/livelooperinterfacelayo.png
http://img814.imageshack.us/img814/5936/livelooperinterfacecopp.pdf (reversed)
I actually kept my first PCB and updated it as I had some room around it to add the extra components.

Now I'm working on adding this circuit in the room I have left in the enclosure :
http://www.ucapps.de/floorboard.html
There will be two remote floorboards with 5 switches on each. But that'll be another story 

I can give the Eagle files if some people are interested in modifying the circuit or the layout.

Best regards.

Eric

Hi !

I've now been using that circuit for a while in rehearsal conditions. Though I'm already happy with it, I think it can still be improved.

It adds a significant noise contribution to the guitar signal. I've heard it by performing some A/B comparisons with, on one side the guitar going straight to the amp, and on the other side with the guitar through it. This noise contribution is insignificant when using an amp's clean channel, but it can be really annoying when using a drive channel. I have also noticed some RF pickup, that's about the same level as the electronic noise (by ear). Even it can be improved by putting a 22nF cap between the output and GNDA (the amp's ground, as seen on the schematic), it still can be distinctly heard.

I removed the TL074 (Send/Ret) from its socket and shorted pin14 to GND : noise is still the same. So I deduced that the noise contribution come mainly from the TL072. There are few chances for it to come from the input buffer, as it's quite a classic one used in a lot of other guitar oriented designs, by me and also by a lot other for years.

So I guess that the summer just before the output is the problem. But I must admit that I don't know how to make it less noisy, as I've used low value series resistors (3k3) and reasonable ones for the gain network (Rf=33k and Rg=16k5).

Any clues on how to improve my design would be appreciated !

Best regards.

Eric

Hi !

As my guitar player switched his amp to an Orange AD30HTC with some hi gain settings, the noise is now unacceptable !
I first though of adding an FX loop to the amp, and thus get rid of my circuit being put before the high gain stage, but each channel has its own long tail phase splitter AND the master pot wired after it (dual ganged), it appeared to be pretty much impossible to do it correctly. Just for the record, here is the schematic of the amp :
http://users.telenet.be/orangefg/OFG_SCHEM/AD30TC_mainpcb1.pdf
(I would still appreciate any suggestions for an FX loop even if it's not the way I chose to deal with my noise problem  )

So I decided to draw another circuit, trying to make it less more silent.

Here is my new schematic :
http://img718.imageshack.us/img718/2692/livelooperinterfacesche.png

I noticed that the most important noise source of my circuit is actually the line output of my Motu 828 mk1. There's a lot of hiss coming from there and also some stange noises that react with actions done on the computer itself (mouse movements, clicks, pretty much any action on the computer causes noise at the soundcard's output). So I decided to put an attenuation setting at the return path, and also to add a squelch circuit to mute the soundcard's return when no loop is played from the computer. I decided to use a relay because those do not add any extra noise as an active VCA would do, and also because the Omron G5V-H1 appeared to be silent enough.

I also replaced the former TL072 input buffer with a jfet buffer (5457), and put some NE5532 instead of TL072 everywhere. After some breadboarding and comparison with the original circuit, the hiss has been seriously reduced.

Before doing the PCB, I'd like to ask you guys your opinion on the design, and also if you have some ideas about how to further improve it, or if I did some nonsense things. Any remark would be appreciated ! I'll then put the PCB design here once I'll have tested it good !

Thanx in advance.

Best regards.

Eric