Hello forum!
Can confirm the negative compression action on the stock VECA. Get the same inverting behaviour when recording a ramping sine (and music!) through the veca at -3dBu RMS or louder. It's particularly obvious as signals gets louder, but it's also measurable around 0 dBu. Have only tested -3dBu and up. I see no reason to test softer signals. This is 30 volt unit, not a guitar stomp box! I have industry standard Lavry and Mytek converters which deal with signals in excess of 20dBu peak. Anything smaller means gaining to restore signal. I'd like to avoid that, thank you, unless it's absolutely warranted, and perfectly controlled, as in "passive" EQs, diode compressors, etc.
Changing R101 (audio path VCA inverter input) from 120k to 200k did make for a major improvement. Compression curve now acts more normal. As in: no negative compression. Though it seems 220k may be the ideal value!
.. It occurred to me to today to compare VECA, GSSL schematics, and the original SSL comp. It seems the R101 scaling is an issue left over from the making of a different circuit (GSSL) from the original SSL schematics.
TL;DR: the SSL circuit have the same input at both audio VCA and sidechain VCA, while GSSL/VECA does not.
Both original and GSSL can be seen here:
http://www.gyraf.dk/gy_pd/ssl/ssl.htm - Look at the amps ahead of audio VCA and sidechain VCA outputs at
the lower left(edit: right!) corner of the 82e27 board. Those signals are scaled equal with a common gain, as both have 100k feedback and input resistors. In the 82E26 part, both of these equal level signals enter DBX VCAs through equal 4k7. All perfectly balanced between VCA and SC inputs, as expected, at least IMHO.
In the GSSL, audio VCA amp have 100k feedback and 120k input(100/120=
0.833), while SC output have 56k feedback and 100k input. (56/100=
0.56). Then it gets messy.. The SC VCA is a 218x where the audio VCA signal goes to the input trough a 470 with 47 ohm to ground 47/(470+47)=0.0909, with some of the negative rail added from pin 5. The scale factor for the SC is 0,56*0,0909=
0.050904
All while the VCA inputs are DBX202 substitute circuits, not (nearly) straight to 218x inputs as the SC VCA, as seen in the lower left(edit: right!) hand corner of the GSSL schematic, and on the audio VCA inputs of the VECA. These inputs are quite different from the straight to 2180 connections, essentially using an opamp to give better driving capability, if I’m not totally mistaken. These starts with a scale factor of 0.833 and then heads off to the DBX substitution circuit with a non-inverting opamp using a 1k over 120 ohm input and 1k feedback. This gives me a scale of 0.107, but I’m not dead certain on that,
though that's what the circuit simulation says. If that is the case, the input to the audio VCA is scaled by 0.833*0.107=
0.089
The numbers
0.050904 and
0.089 surely are different! In the original SSL circuit, these ain’t different. Using a 200k instead of 120k on the input to the audio VCA gives; 100/200=0.5 *0.107 =
0.0535 – a lot closer to the scaling of 0.05 at the sidechain VCA input, compared to the current 0.089 scaling. I guess these input wants to see the same behaviour..
If this theory is correct, then 200k isn’t spot on either. Solving for X, and then another X gives.. 210199 ohms. I have no resistors socketed.. Tata: Please try it and see if it works!
If you hit spot on with that 210199 ohm value, I’m way deeper into electronics and circuits than I thought. <grin> Will try to do the fixed resistor + trimming pot thing and see what the optimal value turns out to be. As it is, with the 200k, the VECA sounds normal on all levels.
References to trimming the 120k resistor on the GSSL are around too, with results far closer to 120k than 200k. Those are however at a different negative DC offset value for a GSSL compared to VECA, as those SC VCA pin 3 inputs are at different level DC offset from pin5 on the VECA and GSSL SC VCAs - 5k1 over 10k047 on the VECA vs 3k9 over 10k047 on the GSSL. It may affect the result..? Lifting pin4 may affect it too, if distortion turns out to be trimmed way off through the ground connection on the SC pin 4 connection.
.. As with other things regarding the GSSL and VECA, it beats me why this haven’t been raised as an issue before. Perhaps it as as simple as the post 2733 in this
http://groupdiy.com/index.php?topic=47.2720 thread says: .. “ The 120k resistor with 2180la will only result in the correct ratios at the FASTEST attack time setting...which is where I'm assuming everyone sets it when performing their measurements.(I'd like confirmation of this)” - though Tata does mention fastest release/attack...
Best regards to all involved,
Andreas Nordenstam
