Optimal Linear Stereo Matrix for three Speakers

Thread Starter

Mashie Saldana

Joined Feb 12, 2016
9
Hello,

First post on this forum, so hi everyone.

I have not built any electronic projects for nearly 20 years so to say that I'm rusty on the theory side of things is not an understatement.

The idea is that the two stereo inputs needs to be mixed like PLII.

I have been trying to see if this passive surround sound decoder design http://www.dfad.com.au/links/DFAD_PASSIVE_SURROUND.pdf could be tweaked for the above number of channels and values, however I got stuck figuring out how to subtract the R channel from the L channel.

Any help is appreciated. This circuit will sit between a stereo pre-amplifier and a power amplifier.
 
Last edited:

AnalogKid

Joined Aug 1, 2013
10,987
The "passive" decoder has 32 opamps in each channel, a 7 kHz lowpass filter, and induces noticeable ringing on transients; hardly passive. Also, the inventor admits it works better on film soundtracks than on recorded music.

As for the matrix, I like the one you found better than the one in the DFAD project. The energy equation is the equivalent of saying the total RMS power (redundant, I know) of the two before channels and three after channels must be equal if you are trying to preserve the feel of the original audio. Fore every watt going into the center channel, the left and right channels must be turned down proportionally. Makes sense. As for whether his matrix values are better than others, don't care; that's your department.

A common problem with center-channel systems is that they reduce the perceived stereo image. Mono is 100% L and 100% R summed into one channel. Summing a percentage of L and R into the center channel makes the overall system partially mono. This mimics what we actually hear, but some people don't like it. His solution is to subtract a part of what was added to the center from what the opposite ear hears. This is a cute psycho-acoustic trick.

To get there, each audio signal is buffered and then inverted. In this way, both the in-phase and out-of-phase signals are available for adding, subtracting, mixing. We'll call the in phase signals L+ and R+, and the inverted signals L- and R-. Conceptually (practice comes later) L+ and R+ go through resistive dividers that attenuate them by 16.67% and 52.7%, and L- and R- go through dividers that attenuate them by 83.33%. These six signals now go to the inputs of three 2-input summing amplifiers.

Practically, it is easier than that. Rather than create the signal ratios, then sum those ratios equally, you can set the ratios directly in the cumming amplifiers:

https://en.wikipedia.org/wiki/Operational_amplifier_applications#Summing_amplifier

Now, a tricky part. Because classic summing amplifiers are inverters, you have to pre-invert everything if you want the phase of the system outputs to be the same as the inputs. Golden-eared nutjob crazies worry about this, but other than that I have no opinion.

If you've got this so far, we move on to circuits.

ak
 

Thread Starter

Mashie Saldana

Joined Feb 12, 2016
9
I'm with you. Your solution sounds less complicated than expected so far.

I will need to read up a bit more on summing amplifiers tonight, especially how to do the ratios.

To keep everything in phase is preferred or the calibration function in the pre-amp will moan the speakers are plugged in backwards...
 

#12

Joined Nov 30, 2010
18,224
AK has it correct. All this proportionality and subtracting can be done with resistors and op-amps. Invert, sum, adjust proportions with the resistors. I call that, "simple" because it isn't difficult to find op-amps with plenty of frequency range for this job. ps, the TL-07x series is lower noise than the TL-08x series shown in the reference material.
 

crutschow

Joined Mar 14, 2008
34,285
Keeping the output phase the same as the input phase is not necessary. That's the same as reversing all the leads to the speakers.
All you need is that all the output phases match.
 

AnalogKid

Joined Aug 1, 2013
10,987
According to certain golden-eared nutjob wacko rabid compulsive crazy persons who otherwise seem perfectly normal, it is OMG beyond necessary. The idea is that if a positive voltage slope at the microphone output was caused by the compression part of an acoustic wave in air, then a positive voltage slope out of an amplifier should cause a compression front to the ear. Most musical instruments have a relatively slow attack, and matched phase reversal at all speakers is unnoticed. But for something with a fast attack, such as the crack of canon in the 1812 Overture, it is noticeable to a surprisingly large percentage of non-whackjob people. Before my MIL phase, I helped install a classical music FM radio station and learned about real audio, then designed instrumentation for a psycho-acoustics research lab and learned about real hearing.

ak
 
Last edited:

AnalogKid

Joined Aug 1, 2013
10,987
A fixed unity-gain matrix box is pretty straightforward. Seven opamps all running at unity gain or less, a bunch of 1% or 0.1% resistors, bipolar power supply, decoupling, etc. Lotsa high quality audiophile grade opamps to choose from. Where are you located, what kinds of components do you have access to, and are you doing this on perfboard or going for a pc board layout?

ak
 

Thread Starter

Mashie Saldana

Joined Feb 12, 2016
9
A fixed unity-gain matrix box is pretty straightforward. Seven opamps all running at unity gain or less, a bunch of 1% or 0.1% resistors, bipolar power supply, decoupling, etc. Lotsa high quality audiophile grade opamps to choose from. Where are you located, what kinds of components do you have access to, and are you doing this on perfboard or going for a pc board layout?

ak
I will go for a perf board as proof of concept and if it works get it on a pc board. I'm in the UK and will need to order everything as my old box of goodies is long gone.
 

crutschow

Joined Mar 14, 2008
34,285
According to certain golden-eared nutjob wacko rabid compulsive crazy persons who otherwise seem perfectly normal, it is OMG beyond necessary. The idea is that if a positive voltage slope at the microphone output was caused by the compression part of an acoustic wave in air, then a positive voltage slope out of an amplifier should cause a compression front to the ear. Most musical instruments have a relatively slow attack, and matched phase reversal at all speakers is unnoticed. But for something with a fast attack, such as the crack of canon in the 1812 Overture, it is noticeable to a surprisingly large percentage of non-whackjob people. Before my MIL phase, I helped install a classical music FM radio station and learned about real audio, then designed instrumentation for a psycho-acoustic research lab and learned about real hearing.
But that would require that the phase of the signal be tracked all the way from the studio mic to the speaker output. :eek:
 

AnalogKid

Joined Aug 1, 2013
10,987
But that would require that the phase of the signal be tracked all the way from the studio mic to the speaker output. :eek:
Not nearly as hard as it sounds. The defacto standard for all audio circuits is non-inverting from input to output. Some sound-reinforcement audio consoles have phase invert switches on a set of outputs because it's easier to do that than rewire the house speakers that might be 30 feet in the air. If the microphone-to-preamp phase is correct (positive pressure slope = positive voltage slope), everything up to the speaker terminals is world-wide consistent. Ish.

ak
 
Last edited:

KL7AJ

Joined Nov 4, 2008
2,229
Not nearly as hard as it sounds. The defacto standard for all audio circuits is non-inverting from input to output. Some sound-reinforcement audio consoles have phase invert switches on a set of outputs because it's easier to do that than rewire the house speakers that might be 30 feet in the air. If the microphone-to-preamp phase is correct (positive pressure = positive voltage slope), everything up to the speaker terminals is world-wide consistent. Ish.

ak
I don't know if this gives you any more spatial resolution than a normal "center channel" setup
 

AnalogKid

Joined Aug 1, 2013
10,987
Other than U1 and U4 being inverters, you are exactly where I thought this would go. Given the strange attenuation ratios, there is no way to reduce the number of different resistor values beyond what you have. For example, change R1, R16, and R19 to 10K each would mess up R2, R3, R17, etc. The only thing to consider is the accuracy of the attenuator ratios. With off the shelf resistors you can't hit the exact ratios in the article, and I don' think it matters. But if you are of the whackjob mindset, decrease R1, R16, and R19 by around 100 ohms each, put a 200 ohm 10-turn trimpot in series, and tweak away.

Power - the power req is so low that you can go for convenience. If you don't have something like a little 12 V 100 mA transformer, start with an AC output wall wart, like what was common with external modems in the 80's and 90's. Purists will disagree with the next part - use half-wave rectifiers to get + and - DC outputs. Yes, the ripple amplitude will be greater, but so what? You're pulling less than 50 mA, and it doesn't take much capacitance to flatten that out before the regulator. Plus, the primary ripple freq will be 60 Hz, not 120 Hz, farther away from the center of the audio band. If you're still worried about power supply hum, add a pre-regulator stage (hum clipper). At these power levels we're talking pennies.

ak
 

Thread Starter

Mashie Saldana

Joined Feb 12, 2016
9
Hi ak, do you happen to have a link to such a power setup that I could have a closer look at?

I have been messing around with the resistor values today and by mixing and matching two in series instead of single ones I can hit the intended values. But as you pointed out, it is probably overkill.
 

AnalogKid

Joined Aug 1, 2013
10,987
I have been messing around with the resistor values today and by mixing and matching two in series instead of single ones I can hit the intended values. But as you pointed out, it is probably overkill.
And even though you can string together three or four resistors in series to hit exactly the right value, it still will have the tolerance of the worst resistor in the string. So with all 1% resistors you can have an overall resistance of 10,123 ohms, but it still will be +/-101 ohms.

http://www.play-hookey.com/ac_theory/power_supply/ps_rectifiers.html
5th image from the top.

ak
 

Thread Starter

Mashie Saldana

Joined Feb 12, 2016
9
And even though you can string together three or four resistors in series to hit exactly the right value, it still will have the tolerance of the worst resistor in the string. So with all 1% resistors you can have an overall resistance of 10,123 ohms, but it still will be +/-101 ohms.

http://www.play-hookey.com/ac_theory/power_supply/ps_rectifiers.html
5th image from the top.

ak
I'm taking the tolerances into account, once the target is within 0.1% of what I want I call it good.
I will have a read about the rectifiers, thanks.
 

#12

Joined Nov 30, 2010
18,224
When you consider that humans can barely detect a difference of 3db and that represents a 50% change in power, you have to realize 0.1% precision isn't necessary. Personally, I'd settle for 1% and still get called dirty names by a, "real" engineer.
 

wayneh

Joined Sep 9, 2010
17,496
Yeah, that reminds me of the corporate bean counters controlling where we had to buy pencils, while the company was spiraling the drain. It's easier to fiddle while Rome burns, to focus on something you think you can control and let the rest burn.
 
Top