SSL Clone compressor ratio measurements

[Greg]

[quote author="matthias"]thanks,

can you give some details how you measured the ratios??

mat[/quote]

If you read back further in this thread, SSLTech gives a good explanation how to measure ratios.

 

[mesmer]

So the closest ratio to the design comes fromm the 2180 la? Better than the 2181? Is the discrepency in ratios soley caused by the vca chosen, or do the percentage of the resistors have a roll to play?
Peter

 

[mesmer]

The 2181 specs a helluva lot better. How close does it come to meeting the ratios?
Peter

 

[Mundox]

I have finished another one of these beauties but this time I am using the 2181LCs and I decided to measure my ratios. It looks fine for 2:1 and 4:1 but when it came down to the 10:1 I am losing half a db for every 10dB raise at the input above threshold! Ss I went back to see if this behaviour is common to the 4:1 setting. In fact above reductions of 10dB the compression starts to kick in harder!? Would that resistor in series with 1k like Greg suggested solve it? (I should try I guess) :sad:

 

[Mundox]

So I tested my other clone which had the 2150As. It's almost the same! And in fact that one was tried by a very established producer and he loved it to bits (now he owns one ) So I decided never to numerically test my DIY gear again!
And they lived happly ever after...
:green:

note: changing the 1k resistors to 1.2k eased the response a bit, and stabilized the 10:1 setting as a brick wall (instead of a decrease)

 

[gyraf]

..So I decided never to numerically test my DIY gear again!..

Welcome to the club..!

Not that I don't do tests (can't do much without these in electronics), but letting your ears (or your customer's ears) decide what rocks and what not..

:grin:

Jakob E.

 

[Emperor-TK]

EUREKA! This thread helped me solve my GSSL problem! My ear told me there was something wrong with my build. There was distortion, particularly on the lower frequencies. Also, the compressor just wasn't responding right. It just sounded bad, and I couldn't quite put my finger on it. It was a choked and strained sound.

After buying Audiotester, I went to work. I benchedmarked the GSSL vs. my RNC and Behringer compressors using the FFT and looking at the relative harmonic to fundamental ratios. The GSSL was definitely putting out more distortion than the others. Also, the RNC in "normal" mode has a similar amount of low end distortion. And I already knew that I hated that compressor for any low end material (except in "super nice" mode). So my ears weren't lying. Still though, my GSSL didn't sound as good as the RNC, which has too much low end distortion, but doesn't sound choked.

The solution was in the VCA inputs, as discussed in this thread. I also used Audiotester to measure the compressors response as described by SSLtech. 2:1 gave me 3.5:1, 4:1 gave me infinity, and 10:1 gave me an inverse response (greater than infinity). Originally, I had used the DBX c1252h2 VCA's. I changed those out to THAT2180LA's (thanks Zach) and the distortion went away. I played around with the resistor feeding the negative input of the TL072 just before the CV inputs (as suggested by Steffen). I needed a 122K resistor here and I am now getting normal ratios. No more choked response!

Can anyone comment on what the pro's and cons might be of changing the input resistor to the op-amp vs. the input resistors to the VCA CV's? The one thing I can see is that the make-up gain won't be effected with the op-amp input resistor (which is good I think). On the down side, I now need to tweak the resistor feeding the meter so that it shows the correct amount of gain reduction, but that should be easy.

Anyway, I am VERY excited now. Cheers to Jakob, Steffen, SSLtech, Gus, and everyone that has shared their general expertise and experience on the GSSL.

:guinness: :sam: :guinness: :sam: :green: :thumb:
-Chris

 

[tmbg]

Chris, I just finished reading this thread as you suggested in my other SSL thread,

I used c1252h2s also, in both of my SSLs, and the first one seems to work fine, whereas the second one has all sorts of problems.

I'm going to measure the ratios of both tonight, and see what I can find.

 

[aurt]

I just noticed this in That's Application Notebook V.2.

http://www.thatcorp.com/datashts/dn127.pdf

On the 2181 datasheet it says that it uses a 6mV/dB CV constant. If the schemo you guys are using is for the (I think) more common, later 50mV/dB circuit, that would certainly account for the wacky ratios. Now that I think about it, no I need to dig back into my desk and make sure the 202 I rebuilt (PIA) is using the right circuit.

 

[zboy2854]

Chris, Steffen or anyone who can help,

I'm having a hard time figuring out which 100K resistor is the one to change out to smooth out the ratio. Is it this one:

Or is it this one?

Thanks in advance.

-Zach

 

[Emperor-TK]

Zach,

The resistor I replaced is the one pictured in your top diagram. This one will require you to tweak the resitor feeding the meter as well for accurate gain reduction metering. This worked well for me, but you also might want to try changing the 100 ohm resitor to the right of the TL072. This is the last component after the opamp that feeds the CV inputs to the audio VCA's. As I understand it, the first method will allow you to keep full make-up gain, while the second method will avoid having to recallibrate the meter, but you will loose a little of your make-up.

What I did was to temporailly replace the resistors with an off-board potentiometer and dial in the correct value. I used Audiotester to get the right values, then checked what resistance the pot was at.

-Chris

 

[zboy2854]

Got it, thanks Chris! :thumb:

 

[drpat]

...

 

[gyraf]

ProFusion, UK

 

[daArry]

I used Audiotester to get the right values...

Hey, Could u possibly explain the proceedure for doing just this...?

Thanks

 

[Gurusound]

Sorry about my question, the vca can it be the IR2151 :sad: , Thanks.

 

[gyraf]

Show us a datasheet for IR2151, and we can tell you...

Jakob E.

 

[chrissugar]

By the way there is a final consensus about which is the best (most correct) procedure for the ratio adjustment. It is the one proposed by Keith or by Steffen. I tried both and they work but none of them give the exact ratios for all the possitions. If you tweak to have exactly 2:1 then the higher ratios will be slightly off or if you tweak for 10:1 the lower ratios will be a little off.
I used in my comps 2180LB VCAs.

chrissugar

 

[chrissugar]

OK, if nobody has a response to this I decided today in the morning to do some tests.

For more consistent results I opened three of my SSL comps and tried the solution proposed by Keith and the one proposed by Steffen on all of them.
After a lot of tweaking and measurements the verdict is this, Steffens mod produce the closest results to the original ratios. The difference is not big between the two mods but it is there.
If you calibrate first for the 10:1 ratio and then measure the 2:1 and 4:1 Steffens mod will be closer to the original values (1.95:1).
I moded three SSL comps with the same value, (120K gave me the optimal ratios) and the results were identical in all. Probably it counts that when i built them, I measured for each component identical values for all the boards I made (all the resistors and the other components)
I have to mention that I used THAT2180LB for the two VCAs and sidechain in each box.

chrissugar

 

[martthie_08]

yes, 120k worked best for me as well! (tried 100k, 115k, 120k and 127k).

For measurement, I put a 400Hz Sine to both channels and starting from -20dB went up to 0dB in 4dB steps, for "10:1" I started at -25dB and went up in 5dB increments. The measurements were mostly off +/-7%, but when I summed it up and divided it by the number of measurements, I came extremely close to the correct values (1.99:1 / 3.88:1 / 10.0:1).

I was using macthescope, there is a limited trial version, for those of you looking to doing analysis on a mac. My GSSL has THAT2180LA on all three positions and I left out the 68k / 10k resistors and the distortion trim, as the 2180's are factory trimmed. Another thing I did was add a 47K resistor in series with the + leg of the threshold potentiometer, because it went into heavy compression way too early...