Mixing 3 Line Level Inputs to 2.

[comlogical]

Greetings,
I am looking to take the 5.1 analog output from a PC Sound card and "mix" the center channel into the front channels before being sent into a four channel amplifier Integrated Circuit.

The current setup is as follows (excluding the rear satellites and sub channels):
Each Front Channel (L/R) has a 330ohm resistor in series.
The Center Channel has one 330ohm resistors between the "output" and each Front Channel for a total of two resistors.

FAILED ASCII ART ATTEMPT

L-----/\/\/330/\/\/\------------+--------- Input1-----+
..........................................|
R-----/\/\/330/\/\/\-------+--------------Input2------+
....................................|.....|
......-----/\/\/330/\/\/\----------
C---|.............................|
......-----/\/\/330/\/\/\----

Some of you more experienced individuals may already see where I'm going with this... Crosstalk. I find that whenever I have my Center Channel resistor setup tied into the circuit my Left and Right channels bleed audio into one another. Upon removing one or both of the "mixing" resistors from the circuit, the crosstalk goes away (hence, no crosstalk from the IC).

So the obvious question is... how can I mix my center channel into my right and left without creating a situation where they can create crosstalk between each other?

 

[alexan_e]

Check 3.0 - Active Mixing
**broken link removed**

Alex

 

[comlogical]

Read and acknowledged post. It'll be a while until I can make the changes and test them.

 

[comlogical]

After reading through the suggested link (multiple times) I built their concept with my modifications as outlined below in ASCII Art.
I built the circuit using a TL082, a dual JFET Op-Amp.

Just to reiterate, the objective is to take three separate audio signals, mix one of the three ("common audio channel") into the other two, and have two outputs. The final goal is to take five separate audio signals with a few "common audio channels" and have four outputs.

Building the below circuit with the "inverting" op-amps yielded the same results as before. To me, I still see a path for audio channels to mix (was crosstalk the wrong term to use before?). The red path highlights, in this case, one of the two "crossing" paths going on which I believe is giving me the undesired effect.

Did I miss the point? To me, (not knowing the proper terminology) I need a way to split the "common audio channel" and isolate these two "new" signals from each other so that when I tie these two "common audio channels" into their partnered audio channel (Left, Right) I will not get the "crosstalk" between the two sums (L+C, R+C).

..................GND<--:\
.............................:.\__________out
.............................:./......|
........................+--:/.......|
.........................|............|
O--/\/\/\--+------+--/\/\/\--
...............|
...............|
---/\/\/\---
|
O...<--Common Audio Channel
|
---/\/\/\--+
...............|
...............|
O--/\/\/\--+------+--/\/\/\--
.........................|............|
........................+--:\.......|
.............................:.\___|______out
.............................:./
..................GND<--:/

Though not indicated by the picture, polarity was observed on the op-amp as according to the referred link. Testing did result in audio playback.

 

[alexan_e]

Compare the results in the following passive and active mixing examples, left channel is 1KHz, right is 2KHz and center is 20KHz, all 2Vp-p, I have used high source impedance.
The 20KHz is clearly visible when it is mixed with the lower frequencies



EDIT: the probes are named left and center instead of left and right.

Alex

 

[comlogical]

Thanks Alex for your baseline. I unfortunately don't have the equipment nor the understanding to make full use of what you have detailed.
I was able to take advantage of the overall schematic concept to ensure that I was on the right track.

I attempted the circuit using both my original 330Ohm resistors and the suggested 10K (just in case). Both produced the same results except for the obvious added noise produced by the 10Ks.

At this stage I am not convinced that this is the solution. I am led to believe that either...
a.) I have the wrong type of Op-Amp selected
b.) This concept will not work in my application
c.) A nitpicky failure in proper circuit building that has been overlooked so far

I am still getting the channels bleeding over to the next. I have tried a couple variations on the design, but in each design I see a feasible alternative path for my "solo" channels to bleed over without proper restraint (whatever the proper restraint would be I am unsure at the moment).

Finally, I tried (like a shot in the dark) last night putting an audio transformer into the mix for the "shared" channel to try to "isolate" the two halves of the circuit, but yielded the same results.

Not sure right now... :-?

 

[FvM]

Compare the results in the following passive and active mixing examples, left channel is 1KHz, right is 2KHz and center is 20KHz, all 2Vp-p, I have used high source impedance.

I understand, that you wanted to show that the "high source impedance" causes a crosstalk between left and right signal in case of the passive mixer, but not for the active one. A 10k output resistance sounds very unlikely however, a few 100 ohm can be expected as a maximum value. So in my opinion, a passive "10k" mixer shouldn't produce much crosstalk.

I would rather suggest a more trivial explanation of observed crosstalk, simply assuming that the "C" output already contains a mixture of "L" + "R" signals. I would be necessary to know, how you determine "crosstalk" and what's the nature of the "C" signal.

 

[alexan_e]

Is this what you have tried

And you get the left channel into the right channel input (and vice versa) or do you get the center in the left/right inputs (I mean the crosstalk)?
Have you tried without connecting the center signal (leave the input empty), do you still get a crosstalk between left and right?

FvM the circuit in the link I have provided in the first post was using 10K resistors (**broken link removed**), I didn't change them so in order to show the crosstalk in the passive circuit I had to use a high (input impedance) resistor.
The 10K is not a realistic value of course but it was just to show the difference.

Alex

 

[comlogical]

For clarification...
Alex, the schematic image you have posted is one of the trials I tested with. In addition I tried your previous schematic which was slightly different in that the "Common Channel" had a series parallel set of resistors. I tried it just to see if it changed anything.

To be specific, during my tests I am actually mixing the center, left, and sub channels. The sub channel is my "Common Channel". I can then independently test the center channel and left channel through the sound chip manufacturer provided software (Realtek). In addition, I can test the Left Channel by playing a stereo audio file. Through these combinations as well as previous tests during the early phases of testing the amplifier integrated circuit I can determine that my audio play back tests are true to each channel and do not blend. [Addresses in part, FvM]

It is through the introduction of trying to mix a "Common Channel", whether it be the sub, center or even the rear if one wanted to, that the audio signals begin to cross. But, it isn't the audio channel itself that is causing the problem. By simply lifting the "Common Channel" from my circuit the problem continues. It is when a linking connection, be it from a set of resistor like the above circuits or a low resistance track between each channel and the common channel, is removed or broken (think, lift one resistor from the Common Channel in the circuit) that the problem "disappears". [Addresses in part, Alex]

Did that answer and clarify your questions thus far gentlemen?

[Edit]
One more thing. The "Common Channel" does mix into the other two channels. This is easily tested through the sound chip manufacturer provided software to test the individual channel.

[Edit Edit]
One, One more thing. The op-amps were only introduced as a suggestion to try to eliminate the problem. What I feel is a "must" for this project is to utilize the Amplifier IC that I have already selected, a PC sound card as my audio inputs, and attempt to mix channels together. And for further clarification, "crosstalk" occurred with and without the Amplifier IC in conjunction with the op-amps as well as stand alone without the op-amps.

 

[FvM]

Did that answer and clarify your questions thus far gentlemen?

Only in part.
"Lifting" the common channel is obviously a bad test, because it creates a cross talk path not present in the circuit before, at least for the passive mixer. Shorting the common channel to ground would be a more approriate test.

I assume, that you understand, that a passive mixer creates crosstalk with non-zero source impedance. An active mixer however shouldn't. Otherwise it's not working correctly. Details can be determined by real measurements.

 

[comlogical]

Hold off on your thoughts for a bit. I need to start over on my circuit because I have made so many alterations during the testing that I have lost some certainty in my contacts. I am going to rebuild the circuit to ensure that each test that I previously made was "pure". During this I hope to get a functional circuit and find what was wrong with my previous setup.

Thanks for your help so far.

 

[comlogical]

I assume, that you understand, that a passive mixer creates crosstalk with non-zero source impedance. An active mixer however shouldn't. Otherwise it's not working correctly. Details can be determined by real measurements.

I understand that in theory, an active mixer creates a "canceling" current that prevents crosstalk, but I don't understand it at the electron flow level.

According to the referred link (**broken link removed**) the purpose of an active mixer is to prevent crosstalk through this "Virtual Earth" zone. It summarizes earlier that within passive mixers there is the possibility of crosstalk. The article, however, does not go into the details of what the crosstalk is? What are its characteristics? Why does it matter?
I apparently don't fully understand because I don't see how the three mixing inputs (in their example) can affect each other in such a way that they would degrade the final output product.


I was able to do a bit of experimenting this weekend to ensure that my connections were clean. In addition, I retested my Amplifer IC without the Op-Amp and found that everything operates perfectly. With my setup as follows:

Op-amp1 -- Input1 --{Amplifier1
Op-amp2 -- Input2 --{Amplifier1

I found that as long as I did NOT connect BOTH op-amps into the circuit I had no crosstalk, whether none or one. When both op-amps were connected the same crosstalk that occurred in the last test setups returned (channel 1 bleeding into speaker 2, sound distorted). Be mindful that this is a dual Op-Amp chip and at the time of the experiment it was grounded and fed with power. This just leads me to experiment (when I have time) with just the op-amp to verify that...
a.) I'm employing it properly
b.) the chip has not been damaged
c.) the feeding power is acceptable


I suppose if anyone has any additional articles that you could refer me to I would appreciate it. Through having a better understanding of the principles of active mixers I may come to acknowledge that this is the solution to mixing three input channels into two outputs. As for now, my ignorance tells me that active mixing is not the right solution or the op-amp chip is not working.

 

[comlogical]

Purchased a second dual Op-Amp IC to see if the Op-Amp is to blame for the crossover.
I intend on (time permitting) using one Op-Amp per IC.