average to peak detector for SSL clone , schemo ?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

andre tchmil

Well-known member
Joined
Jun 4, 2004
Messages
1,654
Location
land of chicon
Is there a way to incorporate an average/ peak detector in the ssl.
My Urei 7710 has this feature, variable with a pot and I love it.
I'm building a 3U racked one, so plenty of place to add an extra pot.
:green:
 
Well, the SSL is peak detecting as it stands. Analog devices and others make precision RMS-to-DC convertors. I think that you would have to employ one of each, but the problem lies in the fact that such RMS-to-DC convertors are usually deployed in feed-forward circuits. The SSL uses a relatively complicated rectifier circuit around the ratio selection, so it's not so easy to do.

Since what you would end up with would not be an SSL clone, the easiest way would be to send the signal to two compressors and blend the outputs... one would be RMS feed-forward and the other would be the SSL in feed-back.

The GML circuit is feed-forward and it also uses a 'crest factor' control which balances between RMS and peak sensitivity. It too is an all feed-forward device.

Since there was some disagreement about the merits of feed-back topology, including no less than George Massenburg saying that his opinions was that feed-forward just works better, and my observation that the majority of drooled over and coveted comperssors are feed-back (LA2a, SSL buss comp, 1176 etc) and sound better, then there was a need to agree to differ. Certainly the differences in operation are difficult to reconcile, so essentially, you'd have a hard time adding an RMS option to the SSL, and -while not impossible- the engineering changes would possibly have such an effect on the peak operation behavioiur that it might not have an 'SSL' mode any more.

DBX and certain later JBL-designed UREI compressors have been all-feed-forward, so crest factor or peak/RMS controls have been easy to integrate. Not so simple in the SSL I'm afraid...

Keith
 
If you're building 2 detector _FB_ compressor, don't go for so-called rms
detector in "slow" section. First, rms is root-mean-square, and
in electronical therms, you cannot get here adequate response, except
heatting method; also, great qty of "ripple" if detector is fast and
very slow response if detector slow making use of this stuff
good only in log detector log time constant vca-based compressors;
average detector more usable in non-linear (vc db) time constant
compressorwith linear detector like varymu or sslmixbus.
(this is btw theirs musicality) :wink:
You can go for enchanced 7110 just make fast detector on kinda
balanced stuff like ua733 (possible with buffered outs too :))
with two diodes and small (47-100n)cap and 1-4 megs release R
and for slow det. there' s tone of appnotes you can play with;
then, you have to decide what way you want to mix them - like in 7110
or like in whatcompressor or...something different ;-)
Crestfactor in 7110 is really not xfctor, it is threshold for fast detector.

I played with capacitance multiplyer sircuits, tryed opto's
in release controll....the best sound for me is voltage controlled
current source for release, and how do you controll it...
depends on what you have in your mind.
Hope it helps,
IJ.
 
some time ago I thought of another method, to improve the G-SSL sidechain/detector...

actual the sidechain-signal gets summed and leads into the sidechain vca...

in the SSL FXG 384 Comp each signal gets compared and the loudest signal on each channel drives the sidechain...

so this should provide a better stereo-imaging, and avoid phase problems..

there's an interesting dbx note on that :

http://www.dbxpro.com/ftp_mirror/PDFs/WhitePapers/DetectionandDynamicsDuo.pdf


what do you think?

would it be worth to implement it..??
or would the difference be inaudible??

mat
 
Yes, better is two (l and r) detectors, actually four for dualdetector sidechain in stereo,and then shared timeconstant circuits,one for
peak/one for avg, then summing(actually, comparator logic) circuit; difference beetween "mono before detector" and
"two detector then timeconst." is very audible.
 
The difference is audible, and leads to entirely different behaviour on mono and wide stereo signals between the two compressors.

I have a couple of mods to do this, but none fully documented sadly...

Essentially it can be done with two precision rectifiers feeding the same VCA, saving having to use the second VCA... and a further variation is to allow the two methods to be 'blended'. -Either way produces great distortion of a stereo signal in the sidechain, but it is the exact distortion needed to correct for the imaging error in Jakob's version as compared to the original.

There are also other mods... I'll post them when I've completely built them, they're on my bench at the moment, while everything is on hold with this Bloo stuff...

Keith
 
Essentially it can be done with two precision rectifiers feeding the same VCA, saving having to use the second VCA...
It will work not very good on fast peaks:))
I'll avoid using "diodes-in-feedback" method for fast detector too;
sidechain distortion and phaseshift in not something insignificant...
 
[quote author="ijr"]It will work not very good on fast peaks:))
I'll avoid using "diodes-in-feedback" method for fast detector too;
sidechain distortion and phaseshift in not something insignificant...[/quote]Perhaps, but the result will be a lot more like the true SSL behaviour than the Gyraf version.

In this case the distortion is not a random matter, it's actually the same method that SSL use with the "diodes-in-feedback" as well. The only difference is summing before one VCA as opposed to summing after two (or four on the original!...but two are silent on stereo operation and are only there for quad) VCAs. The result should be identical whether you turn down then sum, or sum then turn down. In both cases the sampled signal at the junction of the detector outputs is a distorted version of the inputs where two conflicting signals meet.

The result is the same behaviour characteristic as the original SSL.

Keith
 
One question about feedback and feedforward.

Routing signal from the output of a VCA compressor back to RMS chip instantly turns the unit into a feedback VCA comp?

I have a DBX 266XL here and wondering what I would expect if I add a feedback switch on it. It has DBXRMS1 chip on the side chain.
 
..try it and report back?

FF and FBCK architectures gives qualitatively different texture of attack/release timings, difference is quite usable if you can control it

..this thread 17 years old btw - I've since made the G24 which is both FF and FBCK seamlessly blended: I just fade between input- and output signal of the control element..

/Jakob E.
 
The big difference between FF and FB is that, assuming both the detector and VCA operate at the same V/dB, the former has a ratio of infinity:1, when the latter achieves 2:1.
That's why a FB structure can never be a true limiter, when a FF can reach negative compression ratios (for what it's worth).
So, just switching the pick-up point for the detector does not make the transition smooth. The side-chain gain must be much higher for FB type.
Change of compression ratio is done by changing the side-chain gain.
For a FF type, the gain varies from 0 to 1 (for infinity:1 limiting), for a FB type, it should go from zero to infinity, which is not possible. In practice the side-chain gain is restricted to about 20.
The side chain gain is the gain in V/dB above threshold.
For the typical Blackmer VCA and rms detector, the native law is 6.02mV/dB.
 

Latest posts

Back
Top