GSSL Mix bus Improve THD low frequency mod

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

xarolium

Well-known member
Joined
Feb 5, 2007
Messages
98
Hi there,

I know what you mean: "An other SSL Bus comp topic" but this one should help a lot of us

I made some measure with REW on SSL Bus comp clone on the THD at 1KHz and at 100Hz.
Here is the result: (the comp is set like that: Threshold = -15dB, Ratio = 10, Attack = Fastest, Release =  Fastest, Sidechain = OFF, MakeUp = 0dB, Turbo Mode = OFF)

1KHz THD measurement is 0.65% (think is huge)
95158400_10157356582859422_504413388877594624_n.jpg


100Hz THD measurement is 5.52% (Incredible Huge)
95460695_10157356582849422_1492287142909444096_n.jpg


So i started checking all components, PSU values, Components way... Everything you can check i did.
And i found something interesting. The Full wave rectifier circuit doesn't work as good when frequency is low. The Voltage control is distorted when Frequency is low.

Both scope shows the voltage that control the VCAs at 1KHz and at 100Hz with the comp  set like that: Threshold = -15dB, Ratio = 10, Attack = Fastest, Release =  Fastest, Sidechain = OFF, MakeUp = 0dB, Turbo Mode = OFF)

Voltage control at 1KHz. DC looks quite flat and clean
95749200_10157356583174422_1346295792859611136_o.jpg


Voltage control at 100Hz. DC looks distorted that is why we got to much 3rd harmonic
95704605_10157356583344422_6583174748589522944_o.jpg


I'm thinking of a solution by  using fast schottky diodes and a fast opamp and match input resistor perfectly
What do you think ?
Do you experiment with low frequency THD ?

 
The reason the distortion is so high at 100Hz is because you have the fastest attack and release times selected. This means the side chain can almost follow the waveform. Try again with a release time of at least 100ms.


Cheers

Ian
 
ruffrecords said:
The reason the distortion is so high at 100Hz is because you have the fastest attack and release times selected. This means the side chain can almost follow the waveform. Try again with a release time of at least 100ms.


Cheers

Ian

Thank you for your reply
The fastest Release is 100mS, with the slowest release 1.2s i have 0.92% of THD with Sine 100Hz and Threshold all the way down.
I make simulation on ISIS proteus of the Full wave rectifier and if Resistor are not perfectly matched, the rectifier doesn 't work properly and add distortion.
Is there somebody that experiment to match rectifier's resistors ?
Cause in this state is like unsuable with bass and fast release, the comp add to much 3rd Harmonic
 
Can you post the schematic of the part containing the rectifier and the resistors you mention.

The scope waveform still looks to me like like a regular full wave rectified signal with a little ripple on it.

Cheers

Ian
 
ruffrecords said:
Can you post the schematic of the part containing the rectifier and the resistors you mention.

The scope waveform still looks to me like like a regular full wave rectified signal with a little ripple on it.

Cheers

Ian

I'm talking about this part and i found something interesting if i decrease the 1Mohm resistor to 20K i got lot of less THD and more usable but less GR too. Bad idea cause GR is too low
95903526_10157357258344422_5628596611828416512_o.jpg

 
i finally find a way to reduce the THD to less than 1% when full comp that allow me to make it pumping with not a lot of noticeable distortion. I filtered the Voltage control of the VCA with a 100uF cap. Try it and let me know
Sounding result is a lot better than before when huge compression is apply.  This mods slow down a little bit the attack but by readjusting resistors value i will set the correct attack.
  Bad idea too cause minimum attack time can't go lower than 10mS  :(

Here is the result
95684224_10157359017014422_8369919918690271232_n.jpg
 
Changng  the 1M resistor changes the Threshold.

Have you tried the GSSL THD mod in this thread

https://groupdiy.com/index.php?topic=68871.0
 
xarolium said:
I'm talking about this part and i found something interesting if i decrease the 1Mohm resistor to 20K i got lot of less THD and more usable but less GR too.
You are just experiencing something almost everybody knows. THD at LF increases in all compressors, because the time constant of the rectifier allows it to follow the londer waveform of low frequency input signal.
By reducing the resistor from 1Meg to 20k, you reduce significantly the gain, so no wonder there is less ripple on the control voltage, because the overall control voltage is actually smaller.
 
xarolium said:
i finally find a way to reduce the THD to less than 1% when full comp that allow me to make it pumping with not a lot of noticeable distortion. I filtered the Voltage control of the VCA with a 100uF cap. Try it and let me know
Sounding result is a lot better than before when huge compression is apply.  This mods slow down a little bit the attack but by readjusting resistors value i will set the correct attack.
  It is well known that increasing the release time decreases LF THD. Actually you have also increased the release time.
IMO there are slightly better ways of doing it.
 
Script said:
Changng  the 1M resistor changes the Threshold.

Have you tried the GSSL THD mod in this thread

https://groupdiy.com/index.php?topic=68871.0

I will for sure but as my GGSL is based on DIY racked PCB that use resistor in serie on Atttack time and release i need the schematic to implement this "diodes" mods. I asked him and waiting for the schematic

abbey road d enfer said:
  It is well known that increasing the release time decreases LF THD. Actually you have also increased the release time.
IMO there are slightly better ways of doing it.

You're right. I kept the 1Mohm value and try to reduce THD by filtering the VCA control voltage. I think there is mistake in the design cause at low frequency with fast attack and release the VCA control voltage is very distorted that why there lot of harmonic. I compared original SSL Gcomp with the gyraf design and i found some difference.  let see attached picture
95838276_10157361877114422_8275032700868362240_o.jpg
 
What is the differences exactly that you see?

The red-encircled 100nf is on my GSSL rev7 schematic. It's a coupling capacitor to keep the amplifier's  VC / EC  input happy.
--------
You mentioned changing something to 100uf. Where? Could you show on the GSSL schematic?
 
Script said:
What is the differences exactly that you see?

The red-encircled 100nf is on my GSSL rev7 schematic. It's a coupling capacitor to keep the amplifier's  VC / EC  input happy.
--------
You mentioned changing something to 100uf. Where? Could you show on the GSSL schematic?

On GSSL Rev7 the 100nF capacitor is not present between the 1K resistor and the Voltage control input of the VCA. There is on before but it's not this one.
The second 470ohm resistor is used to set the same impedance on both input of the op amp but i think it was omit cause on the GSSL we used a JFET op amp (tl072).

About the 100uF, it was a bad idea cause it added time to the attack. The minimum attack i could have with this mod was 10mS, far too long for what i need. I'm working on the solution describe on this thread https://groupdiy.com/index.php?topic=68871.0.
 
xarolium said:
The second 470ohm resistor is used to set the same impedance on both input of the op amp but i think it was omit cause on the GSSL we used a JFET op amp (tl072).
It is absolute nonsense. A resistor added in the non inverting input cancels the offset due to current unbalance. If the output of the VCA was very low impedance, the resistor in the non-inverting input woould need to be equal to that between output and inverting input, i.. 33k. This would add significant noise, so it would need to be decoupled.
But the output of the VCA is actually a current output, with a very high output impedance, so the non-inverting input is just as well fully grounded.
Anyway there would remain two signiificant sources of offset: the opamp's own offset voltage, and the VCA's residual output DC current. There is no common remedy to that issue, only two recommandations: the input signal must be AC-coupled via a suitable capacitor, and the output too.
I have experimented with injecting trimmable DC at the input, it gave unsatisfactory results because temperature variations affected the proper offset zeroing.
 
abbey road d enfer said:
It is absolute nonsense. A resistor added in the non inverting input cancels the offset due to current unbalance. If the output of the VCA was very low impedance, the resistor in the non-inverting input woould need to be equal to that between output and inverting input, i.. 33k. This would add significant noise, so it would need to be decoupled.
But the output of the VCA is actually a current output, with a very high output impedance, so the non-inverting input is just as well fully grounded.
Anyway there would remain two signiificant sources of offset: the opamp's own offset voltage, and the VCA's residual output DC current. There is no common remedy to that issue, only two recommandations: the input signal must be AC-coupled via a suitable capacitor, and the output too.
I have experimented with injecting trimmable DC at the input, it gave unsatisfactory results because temperature variations affected the proper offset zeroing.

Abbey road enfer, you seems to get a lot more knowledge than i have on this question. What about original SSL with the 470ohm resistor on the negative input ?
And do you manage to get less THD with low frequency on the G-comp ? what do you think on this mod https://groupdiy.com/index.php?topic=68871.0 ?

Thank you for your help
 
xarolium said:
Abbey road enfer, you seems to get a lot more knowledge than i have on this question. What about original SSL with the 470ohm resistor on the negative input ?
I have no idea. Just to be clear, I've never built or even used a GSSL, but I've designed many compressors in my professional tenure and I've fried a few brain cells trying to solve the "squaring the circle" that LF distortion in comps is. I don't see any advantage in putting a 470r there, not any significant disadvantage because the noise contribution of a 470r is negligible in this position.

And do you manage to get less THD with low frequency on the G-comp ?
I think it needs a significant rearrangement of the rectifier/side-chain.

[/quote] what do you think on this mod https://groupdiy.com/index.php?topic=68871.0 ?
[/quote] That is one of the possible rearrangements. Due to the varying dynamic resistance of diodes, small variations of the rectifiued voltage are considerably slowed down, while large variations are faster. That is a commonly used technique.
RuudNL is a good thinker; I believe his mod is valuable. As he mentioned, that would change the threshold and even the ratio somewhat, but would certainly retain the global sonic signature of the GSSL.
 
abbey road d enfer said:
I think it needs a significant rearrangement of the rectifier/side-chain.

Do you have any suggestion on how to rearrange the full wave rectifier/sidechain ? what about the use of a true RMS to DC converter like the AD536 used in the SSL4000 channel comp instead of a VCA?
 
xarolium said:
Do you have any suggestion on how to rearrange the full wave rectifier/sidechain ? what about the use of a true RMS to DC converter like the AD536 used in the SSL4000 channel comp instead of a VCA?
SSl's use of a VCA is clever since it results in a decilinear response (constant mV per dB). OTOH the AD536 is linear, so would need a logarithmic converter. One of the elements of the sonic signature of the SSL is that it's actually a feedback limiter, which control voltage drives the channel VCA.
The AD536 has intrinsic symmetrical attack and release times, so it must be completed with a separate Atk/Rel circuit.
I think that redesigning for AD536 would be a major task and probably change significantly the sonic signature.
Actually there is nothing wrong with the circuit, maybe a few adjustments are needed and maybe you are trying to fix an unsolvable problem. Is the increased LF THD really an operational problem?
 
abbey road d enfer said:
SSl's use of a VCA is clever since it results in a decilinear response (constant mV per dB). OTOH the AD536 is linear, so would need a logarithmic converter. One of the elements of the sonic signature of the SSL is that it's actually a feedback limiter, which control voltage drives the channel VCA.
The AD536 has intrinsic symmetrical attack and release times, so it must be completed with a separate Atk/Rel circuit.
I think that redesigning for AD536 would be a major task and probably change significantly the sonic signature.
Actually there is nothing wrong with the circuit, maybe a few adjustments are needed and maybe you are trying to fix an unsolvable problem. Is the increased LF THD really an operational problem?

Ok thank you for all this information. Low frequency THD is sounding awful to me it add to much 3rd harmonic as soon as GR is more than 4dB. On "polish mode" (ratio 2, 1 or 2 dB of GR) is sound great but sometime i want to use the compressor on drum buss and i can get on nice clean and punchy kick without distortion.

I finally managed to make the  mod describe on this topic https://groupdiy.com/index.php?topic=68871.0 and it reduce the thd to 0.31% with extrem GR. That great !
I made a sound test on different mix, track (bass, drum buss) and it improve a lot the sound of the compressor. Cleaner and brighter, more open. Attack curve is a little bit different when fastest but i never use this compressor with fast attack so it's ok to me. I really enjoy the sound of the compressor now, for me it's day and night.  Release curve was altered too but still enough smooth to me.
Thank you all for your help and knowledge.

Ps: for those who build the DIY racked MixBuzz² attached file are the mod of the front panel PCB. You need to cut two traces on the PCB but doable. ;) Don't forget to add a 0.22uF across the 3.3Mohm resistor

 

Attachments

  • DR_MIXBUZZ_FP_V1_MODDED.PDF
    204.3 KB · Views: 43
Back
Top