Official 33609 builder's tread. See 1st page for updates.

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
I have wired it this way
Looks correct.

just a few posts above prescott reported using the same lundahls and he had to change one resistor on the input. I guess this is what I am missing as well, but have no clue which one should be changed

I don't see why a resistor would have to be changed using a 2:1 lundhall vs 2:1 carnhill.
Could it be that your diode bridge is not to spec and adding attenuation when it should not?
 
Did someone in the meantime come up with a real solution for the LF distortion problem some people seem to have?

Since you've been posting about this since 2009 I'll assume you're familiar with all the tests and hypothesis that have been posted throughout this thread over the years.
My $.02 is that the circuit is fine but that there is variation from unit to unit (DIY and original NEVE versions).
I suspect Neve was building these with T/C components in the original builds that not all DIYers know to do properly (test and check). Or values have drifted over the years in the vintage units to explain differences (capacitors tend to drift up in value).
In this thread, the following have been proposed, which I think do explain the differences.
-  An in-balanced diode bridge will increase distortion (or other imbalancing flaws in the compression section between T1&T2)
-  slower sidechain response will reduce the distortion (promixe measured an original and found it slower than the DIY version), so to-spec capacitors in the sidechain may make the compressor fast enough to distort more than you like. So if you want to slow it down, increase the 10uf cap or tweak other resistors.

It's important to realize that compression is inherently distortion! You cannot compress a signal without distorting it. This is a simple observation but I think it is somehow being missed in this quest to reduce the distortion of the unit while under heavy signal compression.
 
dmp said:
It's important to realize that compression is inherently distortion!

True! But other compressors produce significant lower distortion, even at low frequencies and fast release times.
As you mentioned, I have been struggling with this LF distortion problem for years.
(I couldn't get this problem out of my head, so that is why I started experimenting again recently!)
I have already increased the release times, but still I don't completely trust this compressor.
The diode attenuator does not produce the distortion by itself. If I use a constant voltage as a control voltage, the distortion is very low.
So my conclusion is that the distortion is caused by a 'ripple' on the control voltage.
Of course more 'smoothing' (longer R-C time=longer release) will reduce the 'ripple' on the control voltage and thus reduce the LF distortion.
But I had hoped that someone would have had a brilliant idea!

I have the impression that member Promixe even gave up this project, because it seemed impossible to solve the LF distortion problem.
 
But other compressors produce significant lower distortion, even at low frequencies and fast release times.
Not all compressors are the same.  Some may have design features that would work better for LF applications.

The diode attenuator does not produce the distortion by itself. If I use a constant voltage as a control voltage, the distortion is very low.
So my conclusion is that the distortion is caused by a 'ripple' on the control voltage.

You've posted this before and I think it pretty much explains the issue. Also, note that a member looked at an original and saw the same ripple on the cv, just that it was not as much (and the times were slower, which goes hand in hand).  Getting a sidechain to respond fast without having a ripple on the CV is a design challenge.
I was pointing out that version to version build differences by DIYers may be increased because some may not have a balanced diode bridge.  If the bridge is not well balanced, distortion would be increased.

In the past few years I've built many different compressors (1176, LA3A, Varimu, SSL, Pico, LevelLoc, api525, LA2A)
They all have different personalities when compressing heavily (distorting). 
I lent my 33609 to a studio that had an original and they used them side by side for about a month. They loved it and said they couldn't tell the difference between them.  Now that it is rhetorical and my build may not pass the tests you've been doing or sound good on the material you are putting through it, but I am trying to explain why I don't think there's any problem with this diy 33609 when built properly.

 
For what it is worth, I measured the attenuation vs. control voltage today on a real 33609 compressor:

-1 dB : +1,60 V.
-2 dB: +2.09 V.
-3 dB: +2.34 V.
-4 dB: +2.66 V.
-5 dB: +2.93 V.
-8 dB: +3.50 V.
-10 dB: + 3.87 V.
-12 dB: + 4.17 V.
-15 dB: + 4.65 V.
-20 dB: +5.44 V.

Also I measured the peak-to-peak ripple on the control voltage.

Limiter, 50 mS, 40 Hz, 4 dB gain reduction: 0.32 V. p-p
Limiter, 100 mS, 40 Hz, 4 dB gain reduction: 0.20 V. p-p
Limiter, 200 mS, 40 Hz, 4 dB gain reduction: 0.11 V. p-p

Limiter, 50 mS, 100 Hz, 5 dB gain reduction: 0.18 V p-p
Limiter, 100 mS, 100 Hz, 5 dB gain reduction: 0.10 V p-p
Limiter, 200 mS, 100 Hz, 5 dB gain reduction: 0.05 V p-p

Limiter, 50 mS, 20 Hz, 5 dB gain reduction: 0.60 V p-p
 
Ilya said:
So how does that compare to the DIY version?

My DIY version measures exactly the same.
I have asked a friend of mine to compress 40 Hz on a real 33609 and to my surprise the Neve also produces an audible distortion when compressing 40 Hz with a release time of 100 mS. So this might be 'normal'.
In the meantime I have developed an alternative release network that gives the same release times, but with much lower distortion.
(Maybe I should sell my solution to Rupert! :D )

HERE is an example taken from an original Neve 33609 compressor, ratio 6:1, release 100 mS, gain reduction 6 dB, gain makeup 6 dB.
The first 10 seconds are 40 Hz without compression, followed with 40 Hz, 6 dB compressed with the settings mentioned above.

Here are some distortion figures, measured with a genuine Neve 33609, compressor section, 40 Hz, 8 dB gain reduction:
100 mS: 6% (!)
400 mS: 1,6%
800 mS: 0,94%
1500 mS: 0,79%
A1: 1,88%
A2: 1,34%

But with my alternative release circuit:
100 mS: 0,60 %
400 mS: 0,26%
800 mS and longer: 0,25%
 
RuudNL said:
But with my alternative release circuit:
100 mS: 0,60 %
400 mS: 0,26%
800 mS and longer: 0,25%

Ruud, I've been following your posts for years, and appreciate you looking into this issue, thanks.  Please share your alternative circuit, it will benefit the 33609 builder community.  Or sell PCBs, or whatever.  :)
 
At the moment I am in contact with the manufacturer of the 33609.
Maybe they are interested to implement this modification in the current production, so I can't give all the details at this stage.
I can give a hint however: you don't want the release to set in during the positive or negative halves of the waveform.
You should have a short delay (1/f)/2 before the release sets in.
So for 20 Hz the delay should be 1/20 = 50 mS/2 = 25 mS, because both halves of the waveform can trigger the compression or limiting. By not 'cutting' into the waveform, the (LF) distortion  gets significantly lower.
As soon as I can give more information I will publish it here!
 
RuudNL said:
At the moment I am in contact with the manufacturer of the 33609....

You mean AMS Neve?  Are you trying to sell them your circuit?  Good luck! 

I have no idea how to get what you're talking about.  But glad you are dropping hints.  :)

Maybe you can give us some bread crumbs or alternative circuit that is almost as good... ?
 
The guy who was making and selling them (Igor) disappeared from the forum.
There are still some guys making 2254s which is another Neve diode bridge compressor .
 
RuudNL said:
True! But other compressors produce significant lower distortion, even at low frequencies and fast release times.
As you mentioned, I have been struggling with this LF distortion problem for years.
(I couldn't get this problem out of my head, so that is why I started experimenting again recently!)
I have already increased the release times, but still I don't completely trust this compressor.
The diode attenuator does not produce the distortion by itself. If I use a constant voltage as a control voltage, the distortion is very low.
So my conclusion is that the distortion is caused by a 'ripple' on the control voltage.
Of course more 'smoothing' (longer R-C time=longer release) will reduce the 'ripple' on the control voltage and thus reduce the LF distortion.
But I had hoped that someone would have had a brilliant idea!

The LF distortion is caused by the ripple on the CV, but for not the reasons you think. If the bridge, and associated components are not matched extremely well, low frequency CV feedthrough components will appear at the output when the bridge is modulated with the LF audio and the CV, creating the distortion. This is prevalent in tube processors and some VCAs  as well, but the mechanism is different. (called limiter "thump" or pedestal low frequency components.)

Since there doesn't appear to be any way to trim the circuit, to achieve lowest LF distortion, components will need to be "selected" for minimum feed through. This includes not only the bridge, but buffering/amplifying circuitry that drives the transformer.

Delay/release CV circuits have been implemented before with these circuits, notably, CBS Volumax, Moseley TFL-280, in addition to close matching of component parts.

Good Luck.

jdiamantis

 
jdiamantis said:
The LF distortion is caused by the ripple on the CV, but for not the reasons you think. If the bridge, and associated components are not matched extremely well, low frequency CV feedthrough components will appear at the output when the bridge is modulated with the LF audio and the CV, creating the distortion. This is prevalent in tube processors and some VCAs  as well, but the mechanism is different. (called limiter "thump" or pedestal low frequency components.)

Since there doesn't appear to be any way to trim the circuit, to achieve lowest LF distortion, components will need to be "selected" for minimum feed through. This includes not only the bridge, but buffering/amplifying circuitry that drives the transformer.

Delay/release CV circuits have been implemented before with these circuits, notably, CBS Volumax, Moseley TFL-280, in addition to close matching of component parts.

Good Luck.

jdiamantis

OK... interesting... got some of that.  :)  So match the diodes extremely well.  What method do you suggest for doing that?

Why would the buffering and amp circuits that drive the CV need matching?  You mean push-pull circuits to prevent crossover distortion, or something else?

Wouldn't it be possible to simply filter out low frequencies in the sidechain before it reaches the diode bridge?  And thus prevent the distortion in the CV?  Or is it another mechanism?

thanks
 
Back
Top