(G)SSL - learning about Release variations

tata · Aug 7, 2016

i'd need a little help, can you please tell me whats the reason of these variations?

in particular the cap polarity and the auto release resistors / caps orders? (.47 or 6u8 in paralell w/ 91k, for eg.)
Bottom left and top right seem to be the same to me - but the GSSL and Gustav's versions are different (right?)

thank you

Gustav · Aug 8, 2016

tata said:
in particular the cap polarity

Ill comment on the schematic drawn up by me.

Thats a complete brain fart - not sure what its doing now. I step through the release settings, the release changes - I haven't done more carefull measurements than checking recovery of the sine on the scope, but it always seemed to work as expected...

I will update the errata with a correction of the orientation. .

Gustav

tata · Aug 8, 2016

Gustav,

please check the 91k/6u8 and 750k/.47 caps too.
should be 91k/.47 and 750k/6u8 (original and SB-4000 are the same)

i saw the new errata, is that the "rockandroll" mod (fix?)

Gustav · Aug 8, 2016

For the actual release times, I never understood the "nominal" values, but maybe this can get us both there, if someone can answer.

We are looking for the time constant, which is simply given as TC = R x C

In this case, it is the discharge time for the parallel RC (would be charge time in a series connection).

Doing the math for 1TC.

0.084 (for the 0.1 setting)
0.127 (for the 0.3 setting)
0.263 (for the 0.6 setting)
0.564 (for the 1.2 setting)

For the auto-relase, the parallel pairs are

0.352/0.618 with the GssL stack values
0.042/5.1 with the SSL stack values

The numbers dont fit the nominal values, but a time constant does not denote full discharge (or in case of the series connection, full charge).

The time constant denotes a discharge to about 36% (or a charge to about 64% for a series connection). For each time constant, the gap is closed exponentially.

So TC can be pseudo-indexed like.

100 x 0.36 = 36% (1 tau)
0.36 x 0.36 = 12.96% (2 tau)
12.96 x 0.36 = 4.67% (3 tau)
4.67 x 0.36 = 1.68% (4 tau)
1.68 x 0.36 = 0.6% (5 tau)

Perhaps the release times are given at a discharge to 10%? That would fit pretty well with everything but the fastest release.

Gustav

Gustav · Aug 8, 2016

tata said:
i saw the new errata, is that the "rockandroll" mod (fix?)

The added errata is not a modification to the ratio circuit, just a lowering of the gain in the SC for two reasons.

1. Some builders think they should set their DAW fader to 0dB when looping outboard.
2. Some builders compare it to the GssL, and the threshold difference is freaky to some of those people.

Without the mod, you have to hit the unit with 12dBU to hit the threshold with threshold set fully CCW. I think going out hotter than that is as rock and roll as driving 100mph in second gear, but I'd rather bump it 6dB than worry about peoples personal work-flows

So, back to the release times

Gustav

PRR · Aug 8, 2016

> 0.84 (for the 0.1 setting)

180,000 and 0.47uFd seems to be 0.0846 , unless my abacus is sticking.

I would call 84mS and 100mS "the same" for this purpose.

I don't see a lot of serious timing differences. The polarity difference DOES need to be cleared-up.

tata · Aug 8, 2016

Gustav,

in case if you think fixing the VECA schemo too would be useful (not only the errata), would you please consider rearranging the labels too - for easier reading? they overlap badly at many places.

tata · Aug 8, 2016

PRR said:
The polarity difference DOES need to be cleared-up.

what is the no1 reason to use tantalums there, please? using film caps is a totally no-go in this situation?

i feel somewhat lucky, fixing the polarity in this version of my VECA build should be easy enough - i use a potentiometer and 1 cap for manual release. 2 more caps on the Auto.

PRR · Aug 8, 2016

Tantalums had a period of fashion.

They were also lower leakage than the aluminum caps of the day.

Either was much cheaper than film caps in large sizes. When you make a million (even a thousand), you have to watch your pennies. In DIY economics, not so much.

Michael Tibes · Aug 10, 2016

Here's a picture of an original card. The polarity of the caps in question is positive to gnd, the resistor parallel to the 0,47u measures 93k, the one parallel to the 6u8 measures 700k and rising, charging the cap with the meter - so 750k seems real. The 6u8 / 750k combination is connected to gnd. So it seems SSL drew a correct schematic. No tantals here obviously. This is a pretty old card, copyright notice says 1979 and it has a bunch of mods like an IC added with wires. Maybe there were tantals in later versions?
It seems counter-intuitive without diving into the schematic, but the control voltage charging the caps seems to be negative here?

Michael

gyraf · Aug 11, 2016

yes, voltage at timing circuit is negative-going

Original schematics at my gyraf diy page

Jakob E.

tata · Aug 12, 2016

Michael : thanks for the photo - to me its interesting why some diode polarities not matching the silkscreen. (D2, D3 - D13 seems ok) do you have any idea why?

could somebody please tell how the "auto" mode works, and what difference it makes to have the resistors in reverse order?

PRR · Aug 13, 2016

> how the "auto" mode works

Slow and fast rates combined.

> what difference it makes to have the resistors in reverse order?

Parts in series, order does not matter.

tata · Aug 13, 2016

PRR said:
Parts in series, order does not matter.

i understand this w/ simple resistors. but with caps in paralell?

MrDude · Aug 13, 2016

Gustav · Aug 13, 2016

tata said:
PRR said:

Parts in series, order does not matter.

Click to expand...

i understand this w/ simple resistors. but with caps in paralell?

Gustav said:
0.352/0.618 with the GssL stack values
0.042/5.1 with the SSL stack values

I would say it changes how fast and how slow the two times in the auto step are, so if you want it to 1:1 to the SSL times, you do it like that. Either PRR didn't go back to check with the schem, or I am wrong, but regardless of the correct answer to how the timing is set, "like the original" is not always "better". You can add more times as you see fit, or you can check which one you prefer

Gustav

Gustav

PRR · Aug 13, 2016

> PRR didn't go back to check with the schem

It really would be better if someone would Go Back To Authentic Sources, get the Right Plan, and delete the mistaken versions posted in this thread.

The business about "series" will apply to R-C *IF* you keep the C with their R.

Now that you poke me, I see someone didn't.

Here is my reading. The series-string order does not matter, but the R-C pairings differ.

In ignorance, we can't guess which was the design plan.

Mapping out the R-C constants, the left plans (either way) have the "same" (2:1) time-constants in each network. That's not enough different from a single network to be worth the extra R-C (not to mention confusion down the decades!). The right plans (either way) show a healthy ~~100:1 spread of time constants. The 0.04S (40mS) TC takes the quick peaks without "ducking". If you keep beating on it, the 5S TC pulls average gain down so there's still quick dynamics, not everthing smooshed flat.

I say the right two plans are equivalent and probably appropriate for music limiting.

tata · Aug 13, 2016

PRR said:
.. which was the design plan.

according to Michael's inverstigation, the original (and SB4000) schemo match the original SSL build.

Gustav · Aug 13, 2016

PRR said:
Now that you poke me, I see someone didn't.

Certianly wasn't mean as a poke, but glad to have everything cleared up!

Gustav

(G)SSL - learning about Release variations

Help Support GroupDIY Audio Forum:

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Attachments

Well-known member

Well-known member

Well-known member

Well-known member

Member

Well-known member

Well-known member

Attachments

Well-known member

Well-known member

Similar threads