mono to stereo summing

[SSLtech]

Into virtual earth summing amps?

No.

Anything else? depends on impedances into and at the summing node.

Besides.... -crosstalk? -It's mono!!!

Keith

 

[SSLtech]

Make the summing node virtual-earth, and there can be NO crosstalk is the simple solution. That's the big selling point which drove all manufacturers to virtual earth (otherwise occasionally referred to as "current summing".)

Keef

 

[Guest]

A first, it will need resistors. And a signal loss on them will occur depending on how many channels do you mix. I suppose your stereo ins have high input impedance, right? In such case you will have no loss if to "mix" one device, 6 dB loss if you mix 2 devices, and so on.
If your mono gear is fine with driving 1K load you may go with 2K resistors, otherwise choose values depending on your mono gear outs.
If your gear has an adequate level margin you may not need any buffers, otherwise you will need them.

Use Ohm's law for calculations in such a simple "passive" case.

1) Are they happy with 1 dB load?
2) Are they happy with N times higher output voltage, where N is number of summed devices?


The further design will depend on this answer.

 

[SSLtech]

Butta,

Another advantage of Virtual earth summing amps is that they automatically maintain same gain, irrespective of how many signals you assign/deassign to the bus.

Gain is a big deal.

Passive example:

Say you have eleven sources and they all have nice low (assume zero-ohm) source impedances. Sum ten of them them through 10k resistors (to keep the math easy). Now add the eleventh and see what the gain is. With ten sources silent and the eleventh source with a 1-volt signal on it, the signal goes through a 10k resistor to the summing node... but then at the node it sees ten x 10kΩ resistors going to ground... an effective 1kΩ shunt... which attenuates the signal by a factor of ten. Now your 1Volt signal is reduced to 100mV.

It's the same no matter WHICH path sources the signal... the other ten paths in parallel "beat it down".

Now if ALL paths have 1Volt on them, it sums to 1V...

Now... consider the input impedance at the summing node: If the makeup amp has ANY impedance fluctuation across the frequency spectrum, the system gain varies as a result.

Active example:

An effective zero-ohm node is achieved by using an inverting input on a (sufficiently fast) op-amp, with the non-inverting input grounded. Use a 10kΩ feedback resistor to the same common summing node in the above example, and you get unity gain AT ALL TIMES.

10K input resistor, 10k feedback resistor, unity gain.

"Ah yes" I hear, "but what about the other (equivalent) 1kΩ path to ground attenuating the signal by a factor of ten???" -Well, it also attenuates the feedback signal to ground BY THE SAME EXACT AMOUNT, so the signal is maintained at unity gain.

Add more paths onto the node (more shunts to ground) and the ratio of feedback attenuation (gain) to input signal attenuation (loss) is maintained CONSTANT.

Thus, V.E. summing amps allow you to add/deselect signals to (or from) the bus, with NO effect on the signal level.

Try that with any other topology, and you'll have variable gain as you add/remove things to/from the bus.

Keith

 

[Guest]

[quote author="buttachunk"][quote author="Wavebourn"]If your mono gear is fine with driving 1K load you may go with 2K resistors, otherwise choose values depending on your mono gear outs.[/quote]

thanks.
yes, these will drive 600r.


[quote author="Wavebourn"]If your gear has an adequate level margin you may not need any buffers, otherwise you will need them.[/quote]

i don't need one per channel, but will need one per bus (3). it will be driving something like a G-SSL.[/quote]

Sorry, I am not familiar with G-SSL and your mono devices. If to divide output voltage of your mono device by number of them, will it be enough for a stereo input of G-SSL?

 

[SSLtech]

Butta...

Tap an input to the GSSL so that the summing resistors feed DIRECTLY to the inverting node on the input op-amp, ground the non-inverting input and you'll have a virtual earth summing node, with no additional circuitry needed.

:thumb:

Keef

 

[SSLtech]

[quote author="buttachunk"]may build a mixer with a G-SSL built-in...[/quote]
Wow!

Novel idea...

A mixer with a buss compressor built in... Let me see who was first to do that...

Oh yeah... SSL!

:green:

Seriously though... it's the BEST use you can put a GSSL to, and if you "gang up" the sidechains (like in the original SSL buss comp, as distinct from the GSSL) then it becomes a VERY good box indeed. Adding a simple voltage control input from a linear fader, and you have a master fader for your mixer... JUST like the SSL.

..Then it's not much more than a switch, a pot, a capacitor and an op-amp, and you've got "auto-fade" as well...

-Hell, before you know it, I'll have talked you into DIY-ing an SSL console! :wink:

Keef

 

[Guest]

Keef, what about phase, or you use an inverting input by default?

 

[SSLtech]

[quote author="Wavebourn"]Keef, what about phase, or you use an inverting input by default?[/quote]When you say 'phase' do you mean polarity, or actual phase linearity?

Polarity on a GSSL is easy to change, because the output is differential... swap the wires and all is right with the world again!

Since the impedance-to-ground at a virtual earth has to be effectively zero ohms (hence 'virtual earth'), then the voltage between that node and ground HAS to be zero AT ALL TIMES. (Now we can't have any volts across two points with no resistance between them now, can we? -same goes for impedance.)

An op-amp deployed in a stable configuration will always shift its output until the two inputs (inverting and non-inverting) have the SAME potential on them. -So if you ground one input (which usually happens to be the NON-inverting input) then the other will always be 'pushed' towards ground by the output, and its feedback path.

SO...

Consider this: an inverting op-amp with a 10kΩ feedback resistor to the inverting input, and the non-inverting input grounded. You shove +1 volt into the inverting input via 10kΩ source resistor and you get negative 1V out. Now look at the resistance "chain" between input and output, ignoring the op-amp... you have +1V at one end, -1V at the other end, and two 10kΩ resistors in series in between... What's the potential at the mid-point? -Zero volts.

Double the voltage in (+2V) you get double the voltage out (-2V) same big pile o' nuthin at the middle.

Now, change the gain... use a 5kΩ feedback resistor... 10kΩ input, 5kΩ feedback.

+1V in, negative HALF a volt out, because the gain is Rfb/Rin... 10k/5k = ½, or 0.5

Draw the same resistor chain... one volt at one end, half a volt at the other... this time 1 volt is next to the 10k resistor, 0.5V at the other end of the 5k resistor... still the same old zero at the center point.

It stays that way until the output of the op-amp can no longer swing far enough to match the input... for example if you had a gain of ten (100k Rfb and 10k Rin would give you that) and you fed +10V into the input. The output of the op-amp would have to swing to -100V, which it can't do, if -for example- your rails are ±15V. -So an error develops, and this can be measured as a voltage at the inverting input node. Measuring voltage between the two inputs of an op-amp is an EXCELLENT basic troubleshooting step, by the way. -Other reasons that the op-amp can't get its output to "match" the inputs might be inability to slew fast enough... gain/bandwidth limitations etc.

Remember this: while an overall summing SYSTEM may well have unity gain, the actual op-amp itself might have quite a HIGH gain indeed.

An earlier example revisited: Eleven 10kΩ summing resistors and a 10kΩ feedback resistor. 1 volt in, always the same-sized 1-volt signal out... just inverted, but easily put back to normal (so ignored for now). However, the op-amp's gain is automatically compensating for the loss. Each source sees ELEVEN other resistors, 'bleeding' gain off the summing node (the ten other sources, and the feedback resistance) causing a loss-reduction by a factor of eleven. However, the op-amp feedback resistor also sees a feedback reduction by a factor of eleven (the eleven 'Rin' source resistances) and so it has a gain of eleven, to keep the system gain overall zero, but for noise and distortion purposes, the op-amp's gain is eleven. More if you add more sources, less if you remove them.

"SO", you may think... "virtual earth summing amps can be a curse."

Many people have this thought because the noise and distortion can go up without them realising it. The gain is always applied 'invisibly', so they don't see the problems creeping in.

Faulty thinking. -Just use good practice and it's a non-issue.

If you were using a non-virtual-earth summing amplifier, the loss would be ten, and crosstalk can indeed inter-bleed. However, the loss is always comparable in numbers to the loss using virtual earth (usually being -at best case, ans only then if the input impedance of the makeup amplifier is infinite, at which point the crosstalk is at its worst- one less than the loss factor of the virtual earth summing amp) , and you therefore have to still crank up about the same amount of makeup-gain.

One 'advantage' -if it can be called that' of non virtual-earth makeup gain amps, is that every time you add another signal to the buss, the loss makes you crank up the gain on the makeup amplifier, and that keeps you AWARE of the 'cost' of adding that additional source. This often prompts people to de-assign 'dead' or inactive sources, but the same practice using virtual-earth summing amplifiers does exactly the same thing, -just invisibly. Since the 'cost of doing business' with virtual-earth summing amps is 'invisible', people often leave everything assigned, and that's the worst way to use them.

However, Virtual-earth does one thing that nothing else can do: -Once the overall gain is set, it STAYS that way, no matter how many things are assigned or de-assigned.

For that reason, a tone at 0dB on a fader, set to 0dB, panned center, can deflect the mix buss meters to zero VU. Add a few more sources, and it's still RIGHT there at zero VU. Try that with other makeup amplifiers, and you'll find that "zero isn't zero any more..."

Those are a couple of basic arguments for and against. Me? -Virtual earth ALL THE WAY. I can't stand having the gain move all over the place, and I like 0dB to mean 0dB.

-Always.
[quote author="buttachunk"][quote author="SSLtech"]if you "gang up" the sidechains (like in the original SSL buss comp, as distinct from the GSSL) then it becomes a VERY good box indeed.[/quote]

two sidechains summed, instead of two channels summed into one sidechain ?[/quote]

Yes. See the "ultimate SSL buss comp" thread.

Essentially the GSSL does NOT compress in the same way that the original SSL does. Jakob has stated very clearly that he modified his console compressor to compress differently, and then brought that mod forward into the DIY GSSL. My 'ultimate' GSSL reverted back to the original version, and then subsequently I added a switch to allow either mode to be used.

This is CRUCIAL when considering feeding mono sources into the compressor... the GSSL sums the two inputs and compresses the result, thus mono signals will be compessed as though they were 6dB louder. The original SSL compresses mono signals the SAME as either side on its own.

A very quick demonstration:

Take a GSSL and feed a 1klHz signal into one side. set it at a 10:1 ratio with attack and release somewhere in the middle, and get it to compress about 4dB. Now send the signal to BOTH sides in parallel (use a "Y-cable", a mult or something similar) and you'll see the gain reduction jump to about 10dB. Do that on an SSL, and the gain reduction will stay at 4dB.

Second demonstration:

Take a mono 1kHz signal and pan it hard left. Set the compression as above, 10:1 ratio, nominal attack and release, threshold set for about 4dB gain reduction. Now slowly pan the signal from left to right. The Gyraf compressor will compress about 2dB HARDER in the middle. Repeating the test with an SSL G384, the SSL will compress about 4dB LESS in the middle than at the sides.

If that doesn't say that the GSSL isn't the same as a "real" SSL, then I don't know what does. -For clarification, the modification that I did for the dual-sidechains makes my compressor behave EXACTLY like a GSSL in one mode, and EXACTLY like an SSL G384 in the other.

Keith

 

[Guest]

[quote author="SSLtech"][quote author="Wavebourn"]Keef, what about phase, or you use an inverting input by default?[/quote]When you say 'phase' do you mean polarity, or actual phase linearity?
[/quote]

There is also such terms as "In Phase" and "Counter Phase" that means non inverted and inverted polarity of a signal.

Polarity on a GSSL is easy to change, because the output is differential... swap the wires and all is right with the world again!

So, it is not a problem... As I said, I am not familiar with GSSL.

 

[SSLtech]

[quote author="Wavebourn"][quote author="SSLtech"][quote author="Wavebourn"]Keef, what about phase, or you use an inverting input by default?[/quote]When you say 'phase' do you mean polarity, or actual phase linearity?
[/quote]

There is also such terms as "In Phase" and "Counter Phase" that means non inverted and inverted polarity of a signal.[/quote]
The word is often used thus, which is why I asked.

To my way of thinking, the word 'phase' sould ONLY be used when discussing the "point in time of a cycle" which is the absolute and literal meaning of the word 'phase'. -Any time polarity of a complex waveform is inverted, it should NEVER be called a 'phase reversal' or anything similar. If no time-shift has taken place, -even with a simple sine-wave- it should strictly be called a POLARITY inversion or polarity reversal, never a 'phase reversal' or 'phase inversion'.

Again, hence the request for clarification.

Semantics I know, but in fact phase response at the summing node DOES benefit GREATLY from being linear, and I wasn't certain if you were raising that -very salient- point.

Keith

 

[SSLtech]

Or you can use a class-A op-amp... Not all op-amps have to be class AB.

Differential input, with negative feedback, and usably high gain. Those are going to be key.

If you're going to sweat the sonics in terms of amplification class, make it modular, and swap in what you feel the need to try.

Keith

 

[Guest]

[quote author="SSLtech"][quote author="Wavebourn"][quote author="SSLtech"][quote author="Wavebourn"]Keef, what about phase, or you use an inverting input by default?[/quote]When you say 'phase' do you mean polarity, or actual phase linearity?
[/quote]

There is also such terms as "In Phase" and "Counter Phase" that means non inverted and inverted polarity of a signal.[/quote]
The word is often used thus, which is why I asked.

To my way of thinking, the word 'phase' sould ONLY be used when discussing the "point in time of a cycle" which is the absolute and literal meaning of the word 'phase'. -Any time polarity of a complex waveform is inverted, it should NEVER be called a 'phase reversal' or anything similar. If no time-shift has taken place, -even with a simple sine-wave- it should strictly be called a POLARITY inversion or polarity reversal, never a 'phase reversal' or 'phase inversion'.

[/quote]

Ok, I'll use this semantic strictness talking to you. :sam:

Again, hence the request for clarification.

Semantics I know, but in fact phase response at the summing node DOES benefit GREATLY from being linear, and I wasn't certain if you were raising that -very salient- point.

Not in this topic. Since you've provided more optimal solution that was asked for initially I pass. :grin:

Speaking on inverting VS non-inverting I am with you, because I believe that parallel feedback causes less distortions (the signal and the feedback are applied to the same point so transfer functions for them are identical).

 

[Guest]

[quote author="buttachunk"]

so a solution is; force-biased Class-A... ?[/quote]


...like one common base transistor stage.

 

[Guest]

[quote author="buttachunk"][quote author="Wavebourn"]common base transistor stage.[/quote]

Anatoliy,

right on. :guinness:[/quote]

:sam:

Be sure to select a collector resistor in regard to an emitter resistor such a way it will not exceed available voltage; and a working point for symmetrical saturation.
A common gate, though, may be an option...

 

[SSLtech]

I considered the common-base yesterday evening. Nice low -but never zero- input impedance... Crosstalk benefits, but self-correcting gain doesn't happen, since it's still voltage and not current-driven.

Keith

 

[Guest]

[quote author="SSLtech"]I considered the common-base yesterday evening. Nice low -but never zero- input impedance... Crosstalk benefits, but self-correcting gain doesn't happen, since it's still voltage and not current-driven.
[/quote]

The difference between one transistor and one opamp is a feedback loop gain, i.e. in both cases the same artefacts present, but in different degrees: input resistance of opamp will be less on lower frequencies, but never zero; one transistor's input resistance will be less frequency dependent (now we can speak of phases! :grin: ), and distortion patterns will be different, i.e. a transistor will sound more Neve'ish while an op-amp will sound more Mackie'ish.

i.e. it is the matter of a taste, I guess. But if to use it always with a GSSL I would personally prefer to utilize already existing opamp rather than add one more stage in signal path.