GSSL turbo mod issue

Hi,

I just implemented the Turbo mod to my SSL, but it doesn't seem to do what it does.
I created a test signal with same sinewave in both channels, with occasional bursts of noise in one channel only. Processed it with my turbo-modded SSL, but the burst of noise in one channel does still compress the signal even in the other channel. Tripple checked my wiring and everything already.

Should I look more for solder bridges and cold joints? (did and did not find anything). Or perhaps I just misunderstood, what Turbo mod is supposed to do? It perhaps does something even now, when i use it on music, but it's hard to AB compare (and avoid placebo effect) because turning Turbo mode on does offset the threshold (very annoying).

I attached the results of my measurement. Source test signal on top, SSL processed signal on the bottom. First part is bypass, then turbo on, then turbo off, then turbo off with adjusted threshold. That's not how the turbo-processed signal should look like, right? (or wrong?)

Please help. Thanks

Ok, so I checked the schematics and thought a bit. And it seems that the Turbo Mod is not really a whole 2nd sidechain, as advertised? Seems to me, that it duplicates only a part of sidechain - the rectifier, the ratio par (scaling+gain) and the dummy VCA. But not the attack/release part (it isn't even connected to those switches). So I assume that the L and R part is then summed again somewhere between "ratio" part and "attack/release" part? There is no turbomod schametic, but it feeds the control signal into some 100R resistor on the original PCB. Thing is, that I cannot see any 100R between Ratio and attack/release part of the original schematics. http://gyraf.dk/gy_pd/ssl/ssl.pdf

How is the turbo mod supposed to work?

And what it REALLY does? What's the point of having separate VCAs and ratio parts, if we sum it again later and feed both audio VCAs with the same signal?

In short... (because I'm running out of time here...)

The GSSL sums both inputs when creating a sidechain. That "summed" control voltage then controls both the right channel VCA and the left channel VCA. This means that if you get a loud event in the right channel, it will compress for left and right.

The Turbo splits it, so that it acts more like a pair of mono channels.

At least... that's how it was explained to me. I'll ask Keith to comment.

Thanks for a reply, Rochey. I understand how the L+R sidechain signals are not summed (by lifting the 47k from the pad on one side). I can also see how it is separate until the "ratio" part of the side chain. But I don't see how the attack/release "slew limiter" part could be applied to the 2nd sidechain signal separately if that part of sidechain is obviously not duplicated on the turbo board - no release caps on the turbo board, nor any connections between turbo board and the release/attack rotary switches...

Anyway, thing is, that my gssl with turbo mod is just not working as dual mono, as you can see on the screenshots. Perhaps what you say is correct and I just made some mistake while soldering/wiring + totally misunderstood the GSSL schematic... Then it would be great if you could provide more information, so I can troubleshoot. (very hard with no schematic and no explanation of how it really works.)

Thanks!

..The attack/release isn't applied "separately"... it's common to both. -Just as it is in the "real" SSL unit, and not as it is in the GSSL, where the two channels are summed.

Try testing it as per the video and see if it agrees. -The function of the turbo is to make the unit behave like an SSL (which the original does not do). There's no way to have separate time constants for the two units unless you build two complete mono units.

I'm wondering if you've misunderstood what the board does?

I don't know this "mod". But typically in a Stereo limiter you WANT both sides to attenuate the same. If you don't, the stereo image shifts toward the less-loud side. Put the singer in the middle, smack the guitar far-left, and the singer tilts to the right.

PRR: Ah, I see. Makes sense (and also makes me think, if it couldn't be a nice effect if used mildly, crossfeeding between channels only 80% instead of making it 100% mono.. The way music is processed nowdays had nothing to do with realism anyway.)

SSLtech: Good, thanks for the info. I wasn't sure if I made a mistake, or if it's supposed to behave like this.
There is some incorrect information about this mod on the forums (claiming dual mono, more separation etc).

Thinking about it, seems to me that the real difference between the TURBO and GSSL is that in GSSL sidechain does sum the signal before rectifying and thus the stereo information cancels out. If you feed audio in left channel and same audio but with reversed polarity in right channel, the GSSL will not react to it at all, no matter how low the threshold will be.
In turbo mod, you rectify first, so there is no way that anything will cancel out.

Lets put it into maths,
"x" = left input, "y" = right input
"Rs" = ratio's multiplying coeficient.
+ while it seems that ratio does not only scale the control signal but also offset it, let's have "Ro" as ratio's offset constant
(i will use the || signs for absolute value/rectifying)

for GSSL @ point C:
| x + y | * Rs + Ro

for turbo mod @ point C:
(|x| * Rs + Ro) + (|y| * Rs + Ro) = Rs * (|x| + |y|) + 2*Ro

So now I don't understand one thing. Why does the turbo mod duplicate the ratio part as well? If we would sum the signals before the "ratio" part, we would get:

Rs * (|x| + |y|) + Ro

Which seems MUCH better to me.
1) soundwise there is no difference. the difference is only a constant (you could adjust with threshold pot, but
2) thing is that in real life you would have to adjust the threshold much less than with official turbo mod. In the GSSL original mode, there is also the Ro constant only once. With the official turbo mod, we have the Ro constant there twice, so when we switch between the modes, we always have to radically re-adjust the threshold pot. This way it may be circuit-wise truthful to the original SSL, but who cares if it sounds the same? If we would sum the signals in turbo mod BEFORE the ratio part of the circuit, we would get the Ro constant added only once, so the control signal would have very similar magnitude to the original GSSL mode, and switching between them wouldn't result in the radical threshold shift and comparing them would be much easier.
Sounds like a win/win to me, but you guys are much much more experienced so i'm probably wrong and I would be very grateful if you tell me where I'm wrong, so I can learn.
Thanks a lot!

clorax hurd said:

Thinking about it, seems to me that the real difference between the TURBO and GSSL is that in GSSL sidechain does sum the signal before rectifying and thus the stereo information cancels out. If you feed audio in left channel and same audio but with reversed polarity in right channel, the GSSL will not react to it at all, no matter how low the threshold will be.
In turbo mod, you rectify first, so there is no way that anything will cancel out.

Lets put it into maths,
"x" = left input, "y" = right input
"Rs" = ratio's multiplying coeficient.
+ while it seems that ratio does not only scale the control signal but also offset it, let's have "Ro" as ratio's offset constant
(i will use the || signs for absolute value/rectifying)

for GSSL @ point C:
| x + y | * Rs + Ro

for turbo mod @ point C:
(|x| * Rs + Ro) + (|y| * Rs + Ro) = Rs * (|x| + |y|) + 2*Ro

So now I don't understand one thing. Why does the turbo mod duplicate the ratio part as well? If we would sum the signals before the "ratio" part, we would get:

Rs * (|x| + |y|) + Ro

Which seems MUCH better to me.
1) soundwise there is no difference. the difference is only a constant (you could adjust with threshold pot, but
2) thing is that in real life you would have to adjust the threshold much less than with official turbo mod. In the GSSL original mode, there is also the Ro constant only once. With the official turbo mod, we have the Ro constant there twice, so when we switch between the modes, we always have to radically re-adjust the threshold pot. This way it may be circuit-wise truthful to the original SSL, but who cares if it sounds the same? If we would sum the signals in turbo mod BEFORE the ratio part of the circuit, we would get the Ro constant added only once, so the control signal would have very similar magnitude to the original GSSL mode, and switching between them wouldn't result in the radical threshold shift and comparing them would be much easier.
Sounds like a win/win to me, but you guys are much much more experienced so i'm probably wrong and I would be very grateful if you tell me where I'm wrong, so I can learn.
Thanks a lot!

Click to expand...

Well, there's quite a bit of math there, and an apparent conclusion (which I've taken the liberty of highlighting in bold)... I haven't gone through the math, but I disagree with the conclusion... perhaps I could suggest that you try testing using a slightly different method before drawing conclusions?

Have you watched the video where the different behaviors are shown?

https://www.youtube.com/watch?v=EQD8GPhFGYE

The differences in compression nature are centered around the 'simple summing' characteristic of the GSSL that means in-phase material tends to dominate the detector. (in phase gains in summing) and that single-channel signals do not gain, and opposing-phase signals actually diminish... the REAL (original SSL) unit ignores ALL of these factors and limits both channels according to either channel or both channels exceeding thresholds, and not arbitrarily whether the phase relationship of any contributing signal within the program material sums above or below a threshold...

There are occasions when either approach may be judged to be preferable, so I happily build them to be switchable... in fact I just built another one last month exactly this way.

Definitely I'd encourage experimenting practically in preference to reasoning things out on paper and reaching apparent conclusions based on theory alone.

-As to the ratio being necessary, the 'Turbo' is not JUST a sidechain, it's a detector and rectifier also. The detector stage has the ratio and threshold DC bias integrated, so that section must also be duplicated and included in the second channel too.

Thanks for the reply and yes, I watched the video on youtube.

I don't understand why you disagree. I mean, why it should matter if you first add the same gain to two signals and then sum them, instead of just summing them first and then adding some gain to the sum? Theoretically it's the same and in practice I don't believe that you'd hear the difference if it's done well (which it is in here). + this is not audio signal, but control signal, so some 0.005% coloration nuances are of even less importance.
I really can't see how there could be difference. Please explain.

(Now I'm not comparing the GSSL and Turbo mod. I totally understand what's going on there already.)
I just claim, that we don't need to duplicate the ratio part when doing the turbo mod.
+ I suspect that if we don't it will be even better, because it would make the control even more user friendly, by not having to adjust the threshold everytime after switching between the modes.
You said that you just made another one this month and made it switchable. Do you still have it? Did you notice, that when you switch from GSSL mode to Turbo mode, you get this RADICAL boost in gain reduction and you have to adjust with threshold pot?
Isn't this happening because in turbo mode, the DC offset in ratio part is added two times, compared to adding it only one time in GSSL? And that DC offset is there just for compensation. It doesn't change stuff soundwise. It's just there to make the interface more user friendly, so that the GR wouldn't jump too much when switching between ratios. In the same manner we should think about not adding the DC 2nd time in the turbo board, so that the GR doesn't jump when switching between the modes.

What would convince you? Perhaps if I recreate both options (from sidechain beginning until point C) in LTspice and show that it does exactly the same thing? (i mean same thing sound wise. i believe that by duplicating the ratio part you're really just offseting threshold/GR from the GSSL mode and nothing else / nothing that would mean something soundwise.)...

...but you're incorrect on so many levels.

Firstly it is impossible to do without duplicating the ratio section. That is integral. -If you think you can do it without that section and have it work correctly, then go ahead... but it's simply impossible. -That's all there is to it.

-And as for "Better" without the ratio section... since it won't work, I don't see how it can be "better". 

Seriously, you may have to make the journey on your own before you understand why. -But I can save you some time and tell you that it's necessary simply in order to function. -"Better" or "worse" isn't a factor.

Isn't this happening because in turbo mode, the DC offset in ratio part is added two times, compared to adding it only one time in GSSL? And that DC offset is there just for compensation. It doesn't change stuff soundwise. It's just there to make the interface more user friendly, so that the GR wouldn't jump too much when switching between ratios. In the same manner we should think about not adding the DC 2nd time in the turbo board, so that the GR doesn't jump when switching between the modes.

Click to expand...

No.

It's happening because the signal isn't being summed. You're massively over-thinking this.

Look... just apply signal to ONE input and one input alone... -in EITHER mode (turbo or GSSL), nothing changes. That shows that there's no DC influence.

It's just the signal being summed in GSSl is -USUALLY louder.

Flip the polarity of one input, and you[ll find that it compresses a lot LESS in GSSL mode... so there's no way to compensate.

Just trust me... I've been down this path, and understand what's happening. -You seem to be on completely the wrong track, and perhaps assuming rather than observing.