All-discrete GSSL signal path upgrade

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

SSLtech

Well-known member
Joined
Jun 3, 2004
Messages
5,447
Location
Florida (Previously UK)
The GSSL seems to have evolved into one of the most 'modular' DIY projects out there... Over the years there have been several 'add-on' projects which add functionality to the unit.

The first one I remember was the Sidechain high-pass filter; there are fixed and switchable versions that I've seen. Then there was the LED meter driver; I think the 'Turbo' may have been next (to make the inter-channel detection match the original SSL limiter), then the CRC (to address the hum susceptibility of the earlier boards... now partially incorporated in the latest GSSL revisions), and the VU meter GR display driver... there are probably others that I'm overlooking.

-But to date, all the builds have still incorporated the same basic signal path; based around the NE5534 input stage and NE5532 output stage. Nice, neutral ICs with decent line drive capability at decent levels.

For my latest project, I've done what I've wanted to try for a while... an option that allows for an all-discrete API-style approach which REPLACES the entire signal path from input to output XLR's, with just a single VCA between the two op-amps. -For my particular build I used the "A"-selected THAT trimmable VCAs, to ensure that the only thing in the signal chain that isn't a discrete op-amp or an API transformer, is my most favored VCA.

This is how it's turned out:

C10FFD1E-050F-4ED7-8956-229201E206D3_zpskbrmuhsh.jpg


EB23C4F6-F39F-471F-98E1-97E0A8D810E1_zpsgwi6yfpo.jpg


I laid out a 'test mule' based around an older 'Revision 7' board which I'd never built up... adding the Turbo (I won't build one without the Turbo nowadays!) the CRC (same) and -in this case- the VU driver, because I wanted to play with some nice metering options. (Plus I had a couple of SSL I/O module meters laying around... with VU and peak switchable ballistics, and I wanted to play with them!)

3E6A8EAC-8152-4B9B-A3B2-54A50269BFFE_zpsekokate2.jpg


...So I gradually built the unit up...

12985C4B-A0EF-4403-B2DA-88323EC5DAC4_zpswskxdclp.jpg


The unit basically houses the two pairs of discrete op-amps (one pair replacing the 5534 input stages, the second pair replacing the 5532 output stages and driving the transformers) and 'translates' the pinouts to 8-pin DIP sockets which are then connected to the main boards using either ribbon cables and crimp DIP plugs, or component headers soldered to ribbons, or individual wires if the builder prefers. Here's a pic showing the prototype, using ribbons soldered to 8-pin DIP headers...

67BE8051-FA21-4118-A035-9C08AFCFD515_zps8cpe2piv.jpg


...after a bit of tidying up:

58386107-A039-4452-86CD-CCC30909B0C5_zps6k57h4lj.jpg


oh, and another one showing some nice API 2520 op-amps as well!  ;D

F8F973DD-A660-4FC9-85BB-A9C45B0E5571_zpsbjihjspy.jpg


The unit finally assembled and ready for testing:

539B921E-C131-4180-BDB0-46318EB9F292_zpsfgtmhxtl.jpg


The board uses Mill-Max sockets, so I've been able to swap between a couple of different generations of API op-amps, as well as some nice 2520 clones (in this instance I built using the ML2520's from Whistle Rock Audio, but I've also plugged in a pair of Peter Purpose's 2520 clones, and all have worked stably and sounded great).

I also added a few components to make the unity gain trimmable; each channel has locations for a resistor and trim pot. The GSSL normally requires a 7.5kΩ gain set resistor, and most are built with this fitted. -You can just ignore the gain set components if you like and just use the GSSL's board locations, or remove/omit the 7.5kΩ feedback resistor for each channel on the GSSL board, and add a 5kΩ resistor and a 5kΩ trim pot (4.7k is fine for both, if 4.9 or 5k isn't readily available), then setting the trimpot mid-travel should be pretty close, but gives you the option of trimming to precision (I was fairly easily able to tweak mine within 0.01dB).

So now I've had a little while to play with it; my impressions.

None of the other 'bolt-on' projects that I've done for the GSSL have ever really been aimed at "improving" the sound, with the exception of the "ultimate GSSL" project from a few years ago... and -if you remember- while I had hoped for that to add something sonically to the signal path, I pretty quickly had to admit that it didn't really give any of the signal path improvement that I'd hoped for... even with a grand total of 34 VCAs, parallel input and output stages, and inverted signal drive etc etc etc... but despite the complexity of that build, the actual signal didn't actually sound any better. (all was not lost of course, because it led to the development of the 'Turbo' since the second sidechain did prove to be a significant operational improvement).

Anyhow hopefully you'll trust me to be objective, or at least honest enough to admit when I DON'T hear any difference... but what's my take on this?

Well, I like it. -I like it a LOT.

It's not "night and day"... but it's definitely "bigger". The low end has a pleasant 'bloom' to it. I'm presently hypothesizing that this may be something to do with the DC blocking capacitor before the primary. -Again, this is common API practice, but if anyone wants to build one and is fairly confident that their discrete op-amps are going to have a safely low DC offset at the output pin, it's easy to just install a link instead of a capacitor.

Everything fit nicely. The transformers I used were the EA2503's, and they're a lovely fit, as well as a strikingly faithful recreation of the API output transformer.

A couple of tiny wrinkles: the schematic I used on the Gyraf website contains a couple of errors, specifically regarding the 5534 output stages, and the two stages (pins 1, 2 & 3 and pins 7, 6 & 5) are transposed... so I had to wire the jumpers for the output stage with those pins swapped. Once that was done, everything was plug and play, other than one polarity swap with the output of channel 2... easy-peasy. -Will correct both of these with the next version of the board.

Output drive is nice and muscular, and there doesn't seem to be too much sonic 'penalty' if the outputs are strapped with a load resistor as a sort of 'torture test'. -I didn't try this with all different discrete op-amps, but the ML2520's (the ones I tried it with) held up and didn't appear to be particularly troubled. -The sound remained decently fat and mighty.

So, in summary: I like it. The difference is a bit like replacing "Margarine" (the IC ip-amps) with "Butter" (2520's). If anyone has a few 990's that they'd like to loan me, I'm curious to know how they behave. but initial impressions are universally positive.

Installation is intended to be plug and play. Other than the four 8-pin DIP sockets, the only connection to be made is a 0V link, which ties the ground plane and local decoupling to ground. -This can be done to any place on the unit, but ideal places would be places like pin 1 on the input socket, or any central star point, such as the 0V tie out of the transformer secondary. -Since each op-amp receives power down its associated 8-pin lead, there are test points for each op-amp's power points. -Two test points are located adjacent to each DOA, for safe and easy confirmation of power at each DOA. There's also local decoupling, since the length and cable type may vary between installations.

I'll post some more pics once I've played with it some more.

Keef
 
Very very nice...

Any plans to make a single board with the compressor circuitry, trafos, and all?
 
Mitsos - we talk occasionally of a near-finished unit that could do much of this. It would be one hell of an undertaking though. We've been talking about it ever since we got the pick and place machine! :)

OutofTune - thanks for the kind comment. get yours now, and help fund the next set of GSSL mods we have in the pipe!

 
Definitely would be interested in some all-in-one boards... the less wiring I have to do, the better!  And makes troubleshooting much easier too!  Anyway, nice job on this, and good luck going forward!
cheers!
 
I do like butter much more than margarine, despite all the health claims.........

This does look like interesting project. Would making it switchable between discreet and normal gssl chain be possible? I'm thinking a whole lot of relays and headaches would be required.
 
Ordered mine. This will be fun.
When I saw this I got stoked and initially thought of doing this to my SB4000 but.. no way. That thing sounds good as is.
This really will be its own thing.
You just gave me a good reason to etch.

 
Relay switching could work... though I doubt that it would be a silent switchover... the op-amps would likely do some very weird things when switched-"out"... Plus having heard it for a while, I'm not sure I'd ever use the IC's in a straight comparison...

An all-in-one board is a possiblility, but a lot of work. Would people want the VU driver built-in?

Keith
 
Nice job.. while I have not used a DOA since the ICs got so good a couple decades ago, VCA have long been the weak link for dynamics path performance. The latest generation THAT VCAs are a bunch better than previous generations, while I expect these improvements to be subtle in the context of a dynamics processor where the gain manipulations should dominate subjective sonic impressions.

I expect there to be subtle differences with how VCAs glue circuitry is implemented. Driving the gain control voltage differentially rather than single ended, seems more useful for large gain control changes (like VCA fader automation) than the generally modest compression gain changes in comps, but it could matter. If you drive VCA with single control voltage do you match impedance at both gain control ports?. IIRC even the source impedance of the drive at the CV inputs matters (affects distortion if high enough), with the double edged sword that buffering the CV drive injects noise at that port, effectively a multiplier. 

Sorry to ramble (I think too much about VCA stuff). Have you compared the modern THAT VCA to previous versions? It should be the weakest link in that path. With your modular approach you could, while I do not expect any potential benefit from going backwards in VCA technology, so it would mainly be of academic interest to compare. .

JR
 
Yes, I have a set of 202's here, and I've built versions of the GSSL before now using dbx 2001, 2000 gold. 202XT and 202C as well as 2881, 2180 2151 and I forget which others.

I thought about making a couple of relay-switchable parallel VCA options, but to be honest, the latest generation (in particular the 2181) is -as you say- SO much improved that I'd probably never use an 'alternate-switch' arangement.

the 'Ultimate' version which I built used differentially-driven multi-parallel sets of VCAs; sixteen per signal channel (with eight in each pole)... and at the end of all that hassle, the difference wasn't dramatic at all... in fact I find the 'A'-selected 2181 to be preferable in terms of after-trimming results. -I think that Wayne K was of a similar mind, but I've not asked him.
 
SSLtech said:
An all-in-one board is a possiblility, but a lot of work. Would people want the VU driver built-in?
Keith
Well, if it makes things easier maybe you and gyraf could redo the GSSL from the original files, remove the PSU to make room for the 2nd sidechain, and maybe just include DOA sockets around the I/O opamps, so people would have the choice, and they could just offboard the output trafos?  Space-wise it seems it could work and the control board would stay the same.

I haven't found much use for VU meters in these things, so my vote would be leave it out, or maybe stick a LED GR meter circuit in there? just thinking out loud...

 
Hey guys, I'm doing a GSSL build with the Cavendish and I have a silly question:

The values listed in the build guide for C9 and C10 are a minimum of 47uf at 20V+ as you guys suggest using as large as capacitors as possible for DC blocking. I picked up some 56uF at 50V, but the dimensions of these capacitors seem a little small for the area designated for C9 and C10. Does this really matter as long as I'm using a capacitor that is greater than 47uf at 20V+?
 
That'll be just fine.

In fact some APi designs completely bypassed the cap, but of course to be certain that you're not going to be passing DC through the primary of the transformer, you'd have to be resolutely sure that your op-amp is good first!

If using other op-amps which conform to the API layout (990's, or home-brewed designs etc) then they may or may not have a slight DC offset (which would necessitate a capacitor).

Also, if you're building an op-amp from a 'kit' and haven't tested it yet, it would also be VERY wise to test with a cap in circuit and verify no DC offset before bypassing the cap.

So the capacitor value is listed as a 'nominal' value. -The hole spacing and body size allows for up-sizing, if the builder's whim is to do so.

When you make the jumpers for the 5532, don't forget to transpose the two op-amp halves:

For the 5532 *ONLY* (the 5534 is just pin-to-pin)
Pin 1 transposes with pin 7
Pin 2 transposes with pin 6
Pin 3 transposes with pin 5

the power pins (4 and 8) are straight-through connected.

(The reasoning for this was the dual op-amp pins were wrongly numbered on the GSSL schematic that I was working from; it showed the output stage as using the *other* op-amp half compared to the op-amp half which was in fact used on the GSSL board which I connected it to. -The transpositions just swap pins between the two op-amp halves.)
 
A couple more new DIY guy questions to pester with which to pester you...

In the photos posted on this thread there are no resistors or radial capacitors in the output section of the GSSL. I didn't see that mentioned in the manual. Do I also need to leave out these resistors and capacitors? I also noticed pins 1 and 2 are not connected on the DIP8 socket at the output position on the main board. Are these not being used? Will it affect anything if they are connected to the Cavendish with the appropriate pins swapped?

Additionally, the manual mentions removing the 7.5K resistors on the main board and use the 5K pots and trimmer on the Cavendish. I cannot seem to locate any resistors with that value. Is this not already done with cables jumping the DIP8 sockets on the different boards?

These inquiries may be the result of either overlooking or not comprehending important details whilst reading the manual and this thread. I apologize and thank you for your patience.
 
Many compliments to the "Designers"

any audio sample "comparisation" between  "standard" Gssl and the  "Apized"  ?

peace as always

>R.
 

Latest posts

Back
Top