What Compressor

Here's my latest build - a pair of 'What Compressor' builds in a 1RU.

The What Comp pcbs are Bauman Canil - they are implementations of the 'Joe Week' compressor which is based on the Meek units.

Single vactrol (so far!) VTL5C9 with fairly standard chip sidechain - also has a switchable sidechain filter of 1 frequency.

Other interesting pcb features are a 'Blend' control for mixind dry/compressed signal and DOAs for input buffer, mix amp/GR buffer and output buffer.

I used bits'n'pieces from long dead gear - some Haufe transformers, Yamaha DOAs and some (as yet unbuffered) old Sony meters for output level. A led is used to indicate compression amount. I may end up using the meter for GR - I'll check how it goes.

I also added relay bypass with bi-color leds.

Anyway - firing it up now for test. Will be interesting to hear how it compared with a DOAC, it's distantly related all tube vactrol compressor.

I plan to use this on my Kurzweil K2000 which is a little quiet for single instruments. It really benefits from a compressor with a decent make up amp if it's not too noisy.



PSU is JLM (5 rail is temporary - will replace with a 3 rail soon)

Inside pic

Excellent build!

I have sooo many questions!!
Where did you get the boards, if I may ask?
Are they available to order somewhere or you etched them?
Is the "blend" feature part of the board or it's offboard?

I was just planning to start doing Weak Joe compressors, ordered all components for it, but this intrigues me since What compressor has blend option!
Is there a big difference between What compressor and Weak Joe compressor?
As I've understood Weak Joe is supposed to be more advanced version based on What compressor, but the options you have on your What compressor are making me not so sure in that fact.



Luka

Thanks! Nice to have someone so interested

Here's the schematic for the pcb. It was done by Canil (Fabio Bauman) years ago when this forum was still new
They were available from The DIY Pill

http://www.thediypill.com/store/index.php?main_page=product_info&cPath=65&products_id=194

I know the site is still operating but not sure if the pcbs are available - if you are very interested, best to register and email to find out if they are. I'm not affiliated with the site, just a user

Sometimes they come up for sale in the Black Market too - you could ask around.

So yes - it is an enhanced Week Joe, which itself was based on Joe Meek.

The enhancements include options for variable Attack/Release, the Blend control, DOAs and transformer i/o.
You can also build without using any of these - fixed attack/release, no blend and ICs all round, unbalanced i/o.

It also has onboard psu - I prefer to use psu from JLM for my builds - I like to minimise ac wiring to least possible area of chassis.

Week Joe is more basic than this implementation.


It is also quite similar to the LA4A style of compressor. And JLM has a pcb called the 'MAC Opto Comp' which is a simpler version, although it sounds very good indeed. It does require to add input/output circuitry. It is just a side chain basically.

The blend is simply a pot across the 'april resistor' R27 in the schematic.  The 'april resistor' is discussed in the Week Joe documentation. It is a summing junction of buffered input signal and buffered attenuated signal into the sum/output buffer amp. The pot makes the ratio of dry/attenuated signal variable. And yes, it is all on board.


I'll let you know how it performs in the next couple of days as I put it thru test.

Cheers

I'm really interested how it will sound when you finish it!
Thanks for the schemo for the What V2 and the link to Fabio's shop!

I'm about to start building few Weak Joe compressors from mnats board (link). That's why I was so interested in your build.
Few moths ago I bought all the parts for 2 units, but I didn't start yet since I'm finishing my LA4s. They are ready to be racked and when I do that I'll start with WJs.
For Weak Joe, the only way is to etch the boards myself. There's noone selling them premade. And I already tried but it came not so good since it's double sided and my skills were not up to par. And I did it with toner transfer which is itself less precise than phototransfer.
Anyway, I found a local guy who will help me. I sent him all the files and he'll do the boards for me in about a week from now. That's why I'll stick to WeakJoe still, even though Fabio's boards are definitely superb.

I went through A/B compare between What V2 schematic and Weak Joe schematic. Seems as very few parts differ!
Differences are:
1. Weak Joe has external sidechain IO.
2. What V2 has a switch to bypass the filter. I must admit I don't undestand completely how this filter in the start of the detector path works. It has four parallel R+C's. Could someone help a newbie like me undestand this
3. What V2 has a parallel branch across the part where attack and release controls are. As I understand, it's meant to be there if you're not using attack and release controls. Right?
4. What V2 is using BC559 transistor and Weak Joe has BC327. I went researching and they seem to be similar. But I don't know if they are interchangeable.
5. They have different values of ratio pot. Weak Joe has 10K, WhatV2 has 5K. I wonder if it makes any difference. Probably different range of ratios, but it should perform similar.
6. April resistor! Weak Joe hasn't have it. That is supposed to be blend control. But I went to see if it could be added to Weak Joe, and I believe it is possible without redesigning it's pcb. It could be done offboard if connecting across WJ's resistor R25 out leg and external sidechain out connector. If using screened cables and april resistor mounted on blend potentiometer, I believe it can be done. I'll probably try when I get to a point of testing my WJs. But in WhatV2's schematic there is no value for this blent pot. What value should I use?
7. Gain pot and it's series resistor have different values. WhatV2 uses 100K+5K49 and Weak Joe uses 50K+1K. I wonder what is the difference...
8. WhatV2 has a parallel 100nF across 33uF on the output. That's probably my newbieness - I don't know what it does! Help anyone?

It may seem that there are many differences between What V2 and Weak Joe, but in fact they are almost the same.
But Fabio's board is superb to Weak Joe (mnats) self etch if it has option to use DOAs onboard and IO trasformers.
If not using DOAs and trafos, they should probably sound almost the same!

In a few months when I build my Weak Joe (it's a hobby for me... it takes time to have free time!), maybe we should compare how your What compressor and my WJ perform. Maybe use few same audio files and run them through those two compressors!

Well thats a detailed comparison to be sure

The side chain filter is a single frequency, switchable low pass filter. It filters out low frequencies from the side chain/detector circuit to prevent the compressor from responding too much to the bass frequencies.

The parallel caps and resistors set the frequency of the filter and the steepness of the pass band curve.

I'm pretty sure the blend 'april resistor' is discussed in the week joe - I'll dig out my docs and post them up.
In any case - yes, a useful and simple addition.

External sidechain i/o - useful but I've never actually used it!

As you note - yes there is a diff in the detector circuit - parallel section of attack/release.  Also discussed in the Week oe docs. A 2 time-constant thing if I remember. I'll dig it up again.

That's the main differences I think. Changes in pot values, transistor types etc will not amount to much diff I think.

And just getting into the testing now.

Cheers

So getting closer to dialling in this compressor.

I've had to tweak a lot of stuff to suit my needs. First up, some clarification on some of my previous comments regarding the lineage of this guy.

Reviewing the info I have, it seems that 'in the beginning' was the revered Joe Meeks.

Then came the  diy Weak Joe compressor (originated by ?)

- sidechain key in and out patch points
- hardwired 3dB/octave HPF
- single envelope generator path
- adjustable Attack and Release controls
- adjustable Ratio control
- no 'April resistor' ie. dry signal summing into the output makeup amp and therefore max GR > 14dB

After which, the diy What compressor V1.0 (David B Thomas)

- no sidechain key in and out patch points
- hardwired 3dB/octave HPF
- dual envelope generator path
- fixed Attack and Release
- no Ratio control
- 27K4 'April resistor' ie. dry signal summing into the output makeup amp and therefore max GR = 14dB


Then, the diy What compressor V2.0 (Fabio Bauman)

- no sidechain key in and out patch points
- switch modifiable 3dB/octave HPF
- option for single or dual envelope generator path
- option for fixed or adjustable Attack and Release controls
- option for fixed or adjustable Ratio control
- option for 27K4 April resistor' ie. dry signal summing into the output makeup amp and therefore max GR can be set for 14dB or greater
- option for Blend control for varying the April resistor and therefore the Max GR can be varied

To which I have added transformer i/o 1:1 all round, relay bypass, GR metering and some more modifications of various operating parameters.

-------------

I'll describe basically what I've modified, referring to the schematic rev 2 I posted of the What comp earlier.

- the sidechain hpf switch needs to have some resistance added in series with it. If you short out the switch contacts, shorting the filter stack, the gain of the sidechain goes thru the roof. So, some resistance in series is needed. I think 5K or so but I'm still messing here.

- threshhold is too bunched up in a tiny part of the threshold pot. I have reduced the gain of the threshold amp U2B by increasing the size of R9 from 2k2 to 10K. Also changed 10K log threshhold pot  to a linear

The reason this is happening seems to be the current gain of the transistor Q1 bc559 used is quite a bit higher than the original bc327 and so the sidechain amp gains, as is, make the sidechain way too sensitive.

- bias control is not adjustable down sufficiently. Changed R21 to 1K from 4K64 so that the threshhold of compression can be reduced to around -45dBu, down from approx -20dBu or so

- Release control too long. Changed Release pot from 1M log to 250K log to make max release time around 1s or so.

- changed April resistor/Blend option to be 23K in series with 250K linear pot. This ensures that when Blend pot is at minimum (least Dry signal mixed into output), and Threshhold is set for no GR, overall gain at minimum makeup is unity  And then, when Threshhold is set for max GR, the GR is approx -30dB (for a -10dBu input signal)

ie. ensures that when Blend is set for min dry signal mix, you can get a decent amount of max GR

- increased the makeup gain to 35dB up from approx 10dB to cater for much increased GR available with Blend control
  Changed Gain pot to 500K log from 10K log

- changed min gain setting to provide unity gain when no GR. R28 to 7K7 from 5K49

- changed output dc blocking cap from 33uF to 470uF to ensure flat response to 20Hz. Also removed the current limiting resistor R29 which is not necessary with transformer coupling


------------

Now I have lots of control - not sure if the Ratio control does anything more than the Blend or even the Threshhold   ie. just scales the amount GR amount up or down. Don't yet see any slope changes with the Ratio

I can get unity gain with no GR, max GR of around 35dB and the Blend takes that down to around 14dB by mixing in the dry signal.

Release control is now useful range instead of 5+s max
Threshhold control has some decent range on it instead of being like an on/off switch
Useful amount of makeup gain available
Blend works well
Filter defeat works

So yes - a lot of effort. Only recommended if you like to tweak alot  (Sir Tweakalot)

But the unit is very useful now :

- frequency response +/- 2 dB 20-20Khz
- quiet noise wise - 35dB of makeup leaves me with a noise floor of around 75dBu which is pretty good
- lots of controllable GR with Blend restoring some transient in the case of heavy handed GR (my fav kind

Not quite finished but close.

Cheers

Comparing this to the D-AOC all tube +vactrol compressor is interesting.

The sidechain in the DOAC uses no extra diodes for sidechain rectification/envelope generation.

There, the audio signal is amplified and fed right into the vactrol.

The ac signal combined with the forward drop/reverse breakdown of the vactrol led results in an averaged power level which causes the  illumination of the led. It's not clean dc but a complex ac signal with partial single rectification and dc offsets contributing to the average power in the led.

The circuit is often modified with parallel additional resistance, leds and also antiparallel leds which all add twists to the rectification and dc offsets and hence the average power level in the led.

A complex signal indeed fed to the vactrol led. The time constants of the vactrol essentially determine the compressor time constants.

The What compressor sidechain signal is much simpler, using additional buffered diodes to rectify and and conventional attack/release cap+resistors to shape the envelope of the signal into a fairly steady dc dependent on the averaged amplitude of the audio.

Both use vactrols in very different ways. The DOAC is much more like the LA2A kind of thing!


So there it is

I'm impressed in what you did!

(I've spent past days mounting dishwasher and cutting kichen furniture to fit it in... instead of electronics... damn!)

Thanks - getting there...

After some listening, I've backed down some of my changes - the GR is very strong, so :

- reducing the series Blend pot to 100K, reducing the max GR to some 25dB or so
- reducing the makeup gain pot to 250K in  line with the reduced max GR
- restored the Release pot to 1M. Seems OK there after all now I have strong GR

Still looking at the Threshold which remains quite 'twitchy' - changing the Threshold amp gain and pot is only somewhat effective. It works but still seems to be very 'bunched up'. So more work there.

  -> adjusting the bias did the trick here. I actually thought it was more about setting the voltage for the onset
  -> of compression but it is more about placing the onset of compression at a certain place on the current/resistance
  -> curve of the vactrol. Increasing the bias places the compressor operating point on a less steep slope part and
  -> so the threshold attains a wider range of control. Still, I went with reduced sensitivity in the threshold amp
  -> (x1 down from x5) and a smaller, linear threshhold pot (5K lin down from 10K log)

On the plus side, the sound of the no GR case is very clear - noise and distortion are low, freq response is all there.'

The GR is very noticable, so it is good to have the Blend and Ratio controls to tame it. I do think it could do with some taming of the Threshold to round it out.

The Blend really does work nicely, mixing a small amount of uber-squash into a mostly-dry signal is quite novel for me.

So now I have the controls responding the way I want and the sound of this thing is quite good.

Unique in that it can really smash a sound with large GR, and that can be tamed  with the Blend pot and the Ratio pot.

I'm not really sure they don't essentially do the same thing, but given that this guy can be so dramatic in GR, having both controls makes it quite easy to get a nice not too overtop compression characteristic.

So it can go from quite subtle to heavy pedal style compression.

Nothing at all like the DOAC which seems to be able to do massive GR very transparently. Perhaps the combination of heavy GR and very adjustable attack/release in the What comp makes it so obvious and effect, when the Blend and Ratio are  not taming the compression a great deal.

All in all, a good exercise - probably could do better with a commercial unit, but for the $375usd or so it's  not too bad.

And I used up a bunch of spare transformers, doas and meters!

Cheers

I'm hoping its ok if I resurrect this old thread.  I'm a little stuck.
I've had this project sitting around for years and recently picked it up.  Put it off because I was a bit intimidated by the lack of documentation.  Luckly with a bit of diligence and alexc's pics I was able to get in the ballpark.
Most of the info I got through DeeT page and old forum threads.  I had to compare schematics, translate part #'s make my own BOM because the old links are dead.
I did my best sourcing parts, I find this the most difficult aspect of a project like this.

The board is populated, it powered up ok and things looks pretty good as far as I can tell. See image.

I decided to go with the Chip opamps (MC33178 and TL072) because this is my first DIY in a while and I didn't want to be troubleshooting homemade Opamps as well as the circuit.  If I get the circuit behaving correctly I may populate the opamp boards I have. I double checked the chips are in the right was and nothing is hot or smoking..

Anyways the problems i am having:
Both channels are behaving the same way
They pass signal but somewhat poorly (quiter and a distorted at high spl they sound ok at lower volume)  It also passes distorted signal when it is turned off.  I'm not sure if that is normal.
the gain knob doesn't seem to do anything
the ratio turns the volume down
and threshold changes how bright the LED lights up but doesn't seem to affect the sound.
The Comp LED is on all the time

One thing i'd also like the mention jumpers JL1 through JL3 are not currently connected  because as I understand it they are only if using Lundahl Xfmr.  Is that correct?

Right now I have an XLR jack on the output even though the output is unbalanced I have pin 1 and 3 joined together at the PCB (as far as I know that should be ok for testing purposes)

Thanks in advance for help!

When the unit is working well, it is very clean - very low distortion even at high gains, as I configured mine.

First thing is to make sure the no-GR signal path works properly - that is basically the input buffer to summing buffer (no GR signal) to output buffer with gain control. That needs to be fully nailed down ie. no distortion, proper gain range on pot, proper freq response, noise floor.

Then it's a question of 'setting' the gain reduction part so that you get the response you want  ie. the ratio of gain-reduced signal to dry signal. That's tied up to the 'ratio' control and the 'april resistor' control and so on.

Not to mention your choice of yellow leds - they are 'dialled in' to set the dc thresholds. It's designed so that the yellow led gives a visual indication of what's happening inside the vactrol. They do seem to be 'on' but not brightly, in mine

The front panel led GR is related and similar and if working right, indicate the GR and shouldn't be on all the time.
You should be able to turn led on/off just with the  threshold.

------------

So - you really need to go thru step-by-step - there is so much range of adjustment available here that it is unlikely you will get a working unit without really going thru each of the blocks.

Good luck with it - it the end mine came out  with quite good basic audio performance and the capability of mixing signal+GR to achieve gently GR or brutal GR.

When I first did get the thing going, I had it like you describe - permanent GR. This thing is capable of extreme GR.
I find that a pc real-time-analyzer application makes the examination of GR much much easier. It's visual

For this sort of thing, I like 'TrueRTA'

Regards
Alex

Thank you for the prompt and informative reply!  Great info I will got through the circuit more carefully now.
I am hoping I can get this figured out!  I may be in over my head, still a bit of a noob.

Is there a way to test the bias trim pot while it is on the board?  And what measurements would i be looking for?

Any other measurements like this would also be helpful for troubleshooting.

Cheers

Referring to schematic revision 2.

Also, what test tools do you got? You'll need at least an audio signal probe and multimeter.

A CRO and an RTA application makes life a bit easier.

First thing to do is make sure you have good, clean audio after :

U1 - input buffer amp (pin 6)
U2 - sum buffer amp (pin 6)
U3 - output buffer amp (pin6) - here the 'gain' pot should vary the .. ''gain"

There is a link on the pcb, J1  which you 'unlink' to disable compression. I used to this to make sure the signal amp is working right before worrying about the compression.

So - unlink it, the feed some audio and trace thru those places I mentioned.

In terms of the relevent options,  I can see :

- you did not mount R25 (680ohms) as you use the 'Ratio' pot - correct
- you did not implement a 'blend' pot (variable 'april' resistor in parallel with R27) - fine
- you do have R27 populated with a fixed resistor - what value did you use here?
- you did not implement 'attack' and 'release' pot, so you would need to :
    i) have R16 populated - correct
    ii) jumper 'attack' pads - not done
    iii) jumper 'release' pads - not done  -> you actually may not need to jumper. Check it. The idea is that signal makes it thru.
    iv) not have RR1 - you do have this
    v) have R15 populated - correct
    vi) have C10 populated - not done
    vii) have D5 populated - correct

So I think you have issues there in your sidechain.


Also, XLR connectors are typically pin1 -> screen, pin 2 -> hot (+) and pin 3 -> cold (-).

For unbalanced operation into an xlr, one usually connects xlr pins 1+3 (screen and cold)  as the 'ground' and pin2 (hot) as the signal.

I can't see your inputs clearly, but with the oep, or any transformer, you don't want  C1 in the transformer debalanced output. There's no dc  to speak of so not needed - lose it!

Regarding the jumpers, with the oep - no you don't want jumpers X1, X2, X3. You do want primary pins 2,3 linked and it's prolly connected on the pcb. Worth taking a multimeter and checking the continuity between 2,3. Should be the same as when the multimeter probes are touched together

The other things for the oep are in 'input zobel' network - RC, CC and RL : you can leave them off for now. They are there to tailor the input response of the transformer. You would need to figure or know the values for an oep. I'm guessing you prolly don't have that right now, so doan worry for now. Possibly the 10K parallel would be worth trying later.

The other thing is to make sure your power rails are good : +/- 15V nominally, but could be anywhere from +/-12V to +/-17V really.

--
Once you can be sure you have good audio, then look at the sidechain amp in more detail.

Just to clarify a previous incorrect comment of mine :

- the bias led should be weakly on, not real strong
- the 'comp' yellow led mimics the led inside the vactrol - it should be able to be made to come on/off just with the threshold



Cheers

Thanks! That is super helpful. 

I realized a silly mistake I made.  I don't have the right opamp chips.  In the process of making my own BOM. I relied heavily upon DeeT's http://dt.prohosting.com/hacks/what.html info.  However after comparing the schematics between DeeT's and Fabio's the chip pinouts are not the same.  Bummer should have looked a bit closer at that.

In reading another thread it stated I should have used NE5534 for the single opamps O1, O2, 03 Instead of MC33178

I'm going to place an order and get to work. 
I appreciate the help! More to come!

Good news... the circuit is working now!  Having the right opamps is helpful.. 
You have been a great help.  I traced the signal with an audio probe and the signal appears to be clean after each opamp!
now i'm working on the sidechain.

I am having a few issues:
I want to make sure I can achieve unity gain. 
Unity Gain seems to be with the thresh, Ratio, April and gain pots all at their lowest levels. Is this similar to yours, or is something wonky?

The Ratio pot seems to boost the gain a lot (but doesn't really do much)!  To me it seems increasing the ratio would cause the volume to go down as the ratio increases.  (this brings up the question at the end about unity gain)
Threshold seems to reduce gain (a small amount) when ratio is at it's lowest level. But When it is at it's highest it doesn't seem to reduce the gain.
The April seems to also just crank the volume up.
The LED that mimics the vactrol works to indicate compression  but remain on (lightly) on all the time. (should I increase the resistance of resistor right before the LED (as the schematic had a ? by the value) or is the LED being on a symptom of some other problem?

I didn't remove (yet) the RC, CC and RL associated with the OEP input x-former.  (I didn't want to unsolder if it wasn't necessary.)  THe signal is 7-10db louder when going through the OEP.  Not sure if that is to be expected or if those components need to be adjusted like you suggested before.

Your assistance has been invaluable..
Thank you!

On the subject of gain staging ...

- first thing is the  ratio of any input transformer - you have oep with series:series connection, so 1:6.45 turns  ratio which is some 16.2dB increase. That's a lot of gain there. That is reduced some with the 10K series resistor into the first audio stage. I think if you have a hot input (> 4.5Vpp or so) and this could be an issue both for the transformer and the input buffer stage (+/- 15V = 30Vpp max swing at the most-est.)

- the first audio stage U1 is an inverting amp of gain =1 and passes the audio signal from the transformer out

- second audio stage U6 is an inverting buffer and passes a fraction of the audio signal from the transformer, depending on both the ratio pot and the vactrol resistance.

- The link J1 removes the vactrol resistance, so the gain of this stage varies only a little with the ratio pot (5K on the schematic) which is a series resistance into the buffer.

- With J1 in place, the signal into U6 is shunted to ground thru the vactrol resistance. I forget the light/dark values, but basically it reduces the audio a lot

- third audio stage is an inverting amp with variable gain set by the ratio of variable feed back resistance 5K to 15K, and the R27 28K  'april' resistor (in parallel with any other thing you add there, like a 'blend' pot), so around around x.17 to
x0.53.

- the input to the third audio stage is the sum of the output of the second stage and again part of the transformer signal
  ie. it is 'blends' the gain-reduced signal + part of the original signal, in a ratio set by  R26 and R27

SO - the third stage is a little complex and hard to calculate, easy to measure with a CRO.

In response to your wanting unity gain - you have to measure and tweak values to achieve that.

I would first of all get the side chain working properly, now you have established the signal amp (and gain structure!) before worrying about that.

ie. check you have clean audio at the wiper (centre pin) of threshhold pot and pin 7 of U2B.

After that, the sidechain audio is transformed to dc, so skip ahead to pin 6 of U5b - the last opamp in the sidechain and check you have some dc there. That sets up the yellow threshhold led and the vactrol + 'comp' led.

If all that is working, the 'comp' led should be off - on - variable in response to the Threshold knob and some audio.
If it doesn't go fully off, then that isn't right - either the 'bias' led is too 'bright' or the sidechain amp is too strong.

Then, you can start to look at overall gain with no GR and overall gain with GR.

Here's where you need to have a soundcard and RTA application - easy to see the signal input, the GR and the signal output.

Regarding mine - I have very different gain staging to you so my tweaks aren't useful here.
I also added a relay bypass so I can always have 'unity gain' on hand as a reference
And yes, as I said when I was building, this takes a lot of checking the circuit to figure it all out!


Good luck and check those side chain things I said before - they are not negotiatble

alexc said:

ie. check you have clean audio at the wiper (centre pin) of threshhold pot and pin 7 of U2B.

Click to expand...

I have lots of distortion at that point
what do you think of that Alex?

Now its working like a charm  ;D