rob_gould
Well-known member
Mine works fine with no distortion.
Official 33609 builder's tread. See 1st page for updates.
- Thread starter Igor
- Start date
Help Support GroupDIY Audio Forum:
RuudNL
Well-known member
That's what I first thought when I tested mine... The distortion is not so obvious on a lot of signals.
Did you already try to limit: http://dev.ax-design.com/electronics/33609/33609-test4a.mp3 with fast attack and 50 or 100 mS. release?
Also when you feed 40 Hz. 0 dB and you lower the threshold so that limiting occurs, you hear a change in sound. (Anyway: I hear it here!)
Did you already try to limit: http://dev.ax-design.com/electronics/33609/33609-test4a.mp3 with fast attack and 50 or 100 mS. release?
Also when you feed 40 Hz. 0 dB and you lower the threshold so that limiting occurs, you hear a change in sound. (Anyway: I hear it here!)
RuudNL
Well-known member
I tried to develop a simple way to measure the attack and release times without advanced equipment.
The idea is that you play a file though the compressor/limiter and record the result.
Any full-duplex soundcard can do that. You need a program like Cool Edit Pro or Audition to analyse the results.
The test file consists of:
1 - 5 KHz. 0 dB 10 seconds
2 - Silence for 10 seconds
3 - A short (100 mS.) 'blip' of 5 KHz 0 dB
4 - 10 seconds of 5 KHz. -20 dB
5 - 5 seconds of silence
6 - 5 KHz. 0 dB for 5 seconds
7 - 5 KHz. -20 dB for 10 seconds
You use step 1 to setup the compressor/limiter and adjust the gain reduction to say 10 dB.
Step 2 will give enough time to let the gain reduction fall back to zero.
Step 3 will trigger the compression/limiting, while during step 4 the signal will increase in level during the release time.
Steps 6 and 7 are useful to test the autorelease a1 and a2 after compression/limiting during a longer time.
To analyse the results:
Attack time: 'Cut' the re-recorded 100 mS. 0 dB part and save it to a new file. Normalize it and zoom in.
Trim it exactly where the sinewave begins. Find the point where the level is 1 dB above the final value.
This is the attack time.
Release time: 'Cut' the re-recorded -20 dB part and save it to a new file. Zoom in on the beginning and check that there is no trace of the 0 dB that was in front of it. If so: trim it off! Now the result is only the -20 dB part. Normalize it and find the point where the signal is 1 dB below the maximum (final) value. This is the release time.
The test file is here: http://www.ljfaahw.nl/Attack_Release.mp3. (800 KB)
The idea is that you play a file though the compressor/limiter and record the result.
Any full-duplex soundcard can do that. You need a program like Cool Edit Pro or Audition to analyse the results.
The test file consists of:
1 - 5 KHz. 0 dB 10 seconds
2 - Silence for 10 seconds
3 - A short (100 mS.) 'blip' of 5 KHz 0 dB
4 - 10 seconds of 5 KHz. -20 dB
5 - 5 seconds of silence
6 - 5 KHz. 0 dB for 5 seconds
7 - 5 KHz. -20 dB for 10 seconds
You use step 1 to setup the compressor/limiter and adjust the gain reduction to say 10 dB.
Step 2 will give enough time to let the gain reduction fall back to zero.
Step 3 will trigger the compression/limiting, while during step 4 the signal will increase in level during the release time.
Steps 6 and 7 are useful to test the autorelease a1 and a2 after compression/limiting during a longer time.
To analyse the results:
Attack time: 'Cut' the re-recorded 100 mS. 0 dB part and save it to a new file. Normalize it and zoom in.
Trim it exactly where the sinewave begins. Find the point where the level is 1 dB above the final value.
This is the attack time.
Release time: 'Cut' the re-recorded -20 dB part and save it to a new file. Zoom in on the beginning and check that there is no trace of the 0 dB that was in front of it. If so: trim it off! Now the result is only the -20 dB part. Normalize it and find the point where the signal is 1 dB below the maximum (final) value. This is the release time.
The test file is here: http://www.ljfaahw.nl/Attack_Release.mp3. (800 KB)
Hey guys, it seems like diode bridge is not matched or limiter's sidechain oscillates at peaks.
But maybe I am wrong...?????
Not good time for me at all (overloaded with non-audio projects), but in a free time will try
to find out kinda solution. I had one unit with oscillation in limiter's sidechain, but
it was fixed with change of C24.
Possible to try to change transistors tr12, 11, 7 to lower beta ones.
But maybe I am wrong...?????
Not good time for me at all (overloaded with non-audio projects), but in a free time will try
to find out kinda solution. I had one unit with oscillation in limiter's sidechain, but
it was fixed with change of C24.
Possible to try to change transistors tr12, 11, 7 to lower beta ones.
RuudNL
Well-known member
No trace of oscillation in the limit side chain here, even without C24.
IMHO the Hfe of at least TR11 and TR12 is not critical.
The voltage gain of TR11 is restricted by R22 and can never be more than 138. (20.7K/150 ohms), while TR12 is an emitter follower that has no gain at all. The only influence of the Hfe here is on the impedance of the base, roughly Hfe * (R28+R30).
The diodes in my version are matched better than 0.5 mV, so this should not be the problem.
When I apply a constant voltage to the diode attenuator, the distortion is never more than 0,2% over a range of 20 dB gain reduction. I am convinced that the distortion is caused by pulses superimposed on the control voltage. (See my previous post.)
By the way:
Is there somebody who can explain why the resistors for the limit recovery are different than for the compress recovery, at the same recovery time?
Limit recovery 100 mS=15K. (6K8+8K2), compress recovery 100 mS= 24K.
Because in both cases a 10 uF capacitor connected to an emitter follower is discharged...
Strange enough the auto recovery components are the same!
Or should they have thought: "it sounds good as it is, so why bother?"
IMHO the Hfe of at least TR11 and TR12 is not critical.
The voltage gain of TR11 is restricted by R22 and can never be more than 138. (20.7K/150 ohms), while TR12 is an emitter follower that has no gain at all. The only influence of the Hfe here is on the impedance of the base, roughly Hfe * (R28+R30).
The diodes in my version are matched better than 0.5 mV, so this should not be the problem.
When I apply a constant voltage to the diode attenuator, the distortion is never more than 0,2% over a range of 20 dB gain reduction. I am convinced that the distortion is caused by pulses superimposed on the control voltage. (See my previous post.)
By the way:
Is there somebody who can explain why the resistors for the limit recovery are different than for the compress recovery, at the same recovery time?
Limit recovery 100 mS=15K. (6K8+8K2), compress recovery 100 mS= 24K.
Because in both cases a 10 uF capacitor connected to an emitter follower is discharged...
Strange enough the auto recovery components are the same!
Or should they have thought: "it sounds good as it is, so why bother?"
nielsk
Well-known member
I think the recovery switch values were played with a bit over the several diferent revisions. I have a full REV C manual, and it has the 6k8 tied to the 47 uF cap (A wafer pos 5), and the other end at to the last (50mS on Limit & 100mS on Compress setting) of the B wafer resistor chain.
The Compress resistors on the B wafer are (from pos 1) 75k, 150k, 330k, N/C, & 18K
Limit are 8k2, 15k, 100k, N/C, 18k
The Compress resistors on the B wafer are (from pos 1) 75k, 150k, 330k, N/C, & 18K
Limit are 8k2, 15k, 100k, N/C, 18k
ultra-alex
Well-known member
hi there, i finished mine yesterday and today was the holy day of its first firing up...
i´am pretty proud that everything went fine (no cap explosion, no fuse action, no high voltage contact...)! ;D
so heres what i got... (in order of igors "rapid burn-in")
1. psu works properly, stable 24V volts.
2. one board lets me adjust the 250mV from BA340, input 0dbU output 0dbU (adjusted with RV3) ->okay!
3. the second board won´t let me adjust the BA340, although i got 247mV. RV3 is working so i also got input 0dBu output 0dBU. -> not okay!
now things get strange and today i only had freq.gen. and a multimeter to measure (no scope).
1. ratio and recovery switch seems working, but makeupgain switch (lorlin) starts working only @ + 16 dBu but then properly.
2. limiter seems to work, as far is as i can test. (and this site was down the whole day).
3. the elma threshold switch behaves really strange and seems working only on some resistors??
the voltages i measured are identical to promixe´s first post here, but my output trannys are not swapped (as it was in his case).
i know this is the most vague and unprofessional failure description of all times and i know i have to get a scope and some proper values for you to help. but the basic functions seem allright, so maybe i´am just missing something very obvious!? ???
my trafos are : all carnhill VTB9046 (intr1), VTB9045 (intr2) and VTB9056 (output).
enough for today.
thanks in advance, alex
i´am pretty proud that everything went fine (no cap explosion, no fuse action, no high voltage contact...)! ;D
so heres what i got... (in order of igors "rapid burn-in")
1. psu works properly, stable 24V volts.
2. one board lets me adjust the 250mV from BA340, input 0dbU output 0dbU (adjusted with RV3) ->okay!
3. the second board won´t let me adjust the BA340, although i got 247mV. RV3 is working so i also got input 0dBu output 0dBU. -> not okay!
now things get strange and today i only had freq.gen. and a multimeter to measure (no scope).
1. ratio and recovery switch seems working, but makeupgain switch (lorlin) starts working only @ + 16 dBu but then properly.
2. limiter seems to work, as far is as i can test. (and this site was down the whole day).
3. the elma threshold switch behaves really strange and seems working only on some resistors??
the voltages i measured are identical to promixe´s first post here, but my output trannys are not swapped (as it was in his case).
i know this is the most vague and unprofessional failure description of all times and i know i have to get a scope and some proper values for you to help. but the basic functions seem allright, so maybe i´am just missing something very obvious!? ???
my trafos are : all carnhill VTB9046 (intr1), VTB9045 (intr2) and VTB9056 (output).
enough for today.
thanks in advance, alex
radiance
Well-known member
ultra-alex said:
Must be a bad solder connection. Just redo ALL connections with some fresh solder make sure they're proper....
Stupid question, but I suppose you not in bypass when input 0db and output 0db, signal goes thru relays only.
OK. So, no change at all when you adjusting BA340's idle current? Check every connection on BA340, maybe 5k pot shorted as well.
now things get strange and today i only had freq.gen. and a multimeter to measure (no scope).
Try to get scope, it can help a lot. For future projects as well. There's a lot of PC scopes for $200 over the net.
Check every solder joint. Maybe wrong resistor's values as well.
Yap. Our DIY stuff works better than this site. I'd redirect this Q to Ethan and other moderators,
as well as makeup switch question...hm...ok, just kiddin'
Check every joint (not smokeable, but on pcb) and every resistor. Kiddin' again.
Both Lorlin and Elma are relatively impossible to kill switches.
You mean db voltages?
If you can adjust 0db in and 0db out and compressor is not in bypass, 99.9% trafos connected correct.
Alex, just some patience and attension, my RSA tells it will work
Good luck, btw, pics of comp thd and makeup switches can help to understand what goes here.
RuudNL
Well-known member
At the moment I am trying to find a solution for the LF distortion problem of the limiter at low frequencies.
IMHO the distortion occurs because the capacitor in the limiter release circuit has to be recharged between two cycles of the input signal. (At low frequencies.)
So: if you could prevent the discharge between two cycles, there would be no need to recharge the capacitor and the LF distortion would be gone!
At a frequency of 20 Hz. the duration of a cycle is 1/20 = 50 mS. However: the limiter circuit rectifies both sides of the waveform, so the charge peaks are 25 mS. apart at 20 Hz.
This means that if you could create a 'hold' for 25 mS, this would solve the problem.
The release time would be 25 mS. longer for all settings, because the release can only set in after the 'hold' time.
I think the solution could be as simple as a CD4528 and 2 transistors.
Of course you could say that it is no longer a 'real 33609' anymore with this modification, but I would prefer low LF distortion!
As soon as I have results, I will post it here.
IMHO the distortion occurs because the capacitor in the limiter release circuit has to be recharged between two cycles of the input signal. (At low frequencies.)
So: if you could prevent the discharge between two cycles, there would be no need to recharge the capacitor and the LF distortion would be gone!
At a frequency of 20 Hz. the duration of a cycle is 1/20 = 50 mS. However: the limiter circuit rectifies both sides of the waveform, so the charge peaks are 25 mS. apart at 20 Hz.
This means that if you could create a 'hold' for 25 mS, this would solve the problem.
The release time would be 25 mS. longer for all settings, because the release can only set in after the 'hold' time.
I think the solution could be as simple as a CD4528 and 2 transistors.
Of course you could say that it is no longer a 'real 33609' anymore with this modification, but I would prefer low LF distortion!
As soon as I have results, I will post it here.
radiance
Well-known member
Thanks Ruud!!
RuudNL
Well-known member
It works!
I have experimented with the 'hold circuit' and the distortion reduction on low frequencies is impressive.
This is what I did:
It works as follows: a retriggerable monostable multivibrator (4528) is triggered by the existing limit side-chain.
To match the voltages, it is necessary to split the 120K resistor (R19) in 47K+82K.
This wil reduce the voltage to the 4528 from 24 to 15V.
At the moment the 4528 gets triggered, pin 7 will become low and the connected transistor will not conduct.
25 mS. after the last trigger pulse, pin 7 will become 'high', the transistor will conduct and the capacitor in the limit side chain can discharge again. The transistor can be bypassed with a switch. When the switch is closed, everything works as in the original design.
(The compressor release ref. stays connected to D11.)
I have experimented with the 'hold circuit' and the distortion reduction on low frequencies is impressive.
This is what I did:
It works as follows: a retriggerable monostable multivibrator (4528) is triggered by the existing limit side-chain.
To match the voltages, it is necessary to split the 120K resistor (R19) in 47K+82K.
This wil reduce the voltage to the 4528 from 24 to 15V.
At the moment the 4528 gets triggered, pin 7 will become low and the connected transistor will not conduct.
25 mS. after the last trigger pulse, pin 7 will become 'high', the transistor will conduct and the capacitor in the limit side chain can discharge again. The transistor can be bypassed with a switch. When the switch is closed, everything works as in the original design.
(The compressor release ref. stays connected to D11.)
radiance
Well-known member
I still have to read myself into this project so I'm in no position to comment on the technical aspect of things but congrats anyway!!
ultra-alex
Well-known member
Igor said:...my RSA tells it will work
holy f****** ****, the beast is working!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
i haven´t really listened to it so far, but the neutrik tells me that i´am (at least) totally within specs.
i will post pic´s, first listening impressions and of course a more detailed description of what was going on,
when i calmed down...
need beers right now! cheers, alex
rob_gould
Well-known member
Congratulations Alex - good work
RuudNL
Well-known member
@ultra-alex: Great to hear that everything is working, congratulations! (Curious to see the pics!!!)
While you are in the stage of testing and listening: could you try to limit a 40 Hz. tone (4-5 dB limiting, 50 or 100 mS. release) and let us know if you can hear a change in sound when switching between limit in/out?
While you are in the stage of testing and listening: could you try to limit a 40 Hz. tone (4-5 dB limiting, 50 or 100 mS. release) and let us know if you can hear a change in sound when switching between limit in/out?
promixe
Well-known member
RuudNL said:
This is exactly what my conclusion was. I remember I swapped every component around that area in the circuit, but nothing got resolved.
Regarding your hold circuit[ - I'm very curious to try it out on my unit, however, in general I would want to keep the unit as "original" in its operation as possible. So, doing this mod, while eliminating the LF distortion issue doesn't give us an in depth view of why the unmodified units have the problem. I'm about to order the remaining parts for the 2nd 33609 - we'll see if it will have same LF problem.
Igor, what do you think of RuudNL's hold circuit?
RuudNL
Well-known member
I completely agree with you! I would also have preferred to keep the original design, however: the LF distortion really bothered me!
This is also why I suggested the switch option: with the switch (or jumper) closed (=transistor bypassed) everything is original.
The 4528 does not load the limiter side chain, so in fact the circuit is 'invisible' when bypassed.
This is also why I suggested the switch option: with the switch (or jumper) closed (=transistor bypassed) everything is original.
The 4528 does not load the limiter side chain, so in fact the circuit is 'invisible' when bypassed.
ultra-alex
Well-known member
still need some knobs and illumination, but...
THIS COMP SOUNDS WONDERFUL!! much respect and a big THANK YOU to igor! now i need to build a 2254C... ;D
the problems i had with the threshold and makeup gain steps was related to a wrong resistor value.
all my 75ohm were actually 470K. farnell send me the wrong values...
luckily there are only six 75ohm resistors in the whole unit. ;D
i have a higher THD ratio on one channel (0,4% vs. 0,14%). this is the board with the BA340 which i still can not adjust. i have 258mV.
ratio is increasing with higher makupgain settings, not directly correlated to the amount of gainreduction. so i think it has to be any problem with the amp!? have to figure that out still. but this ******* sounds great so far, it can only get even better.
rob_gould said:Congratulations Alex - good work
thanks for your help, rob!!
RuudNL said:
i can confirm the distortion when limiting low freq sine waves. if you go below 80Hz you can measure that THD(istortion) is rapidly increasing. although i think this isn´t a "real" problem cause it isn´t obvious when feeding "real" signals.
so far my drumbus and even whole mixes (with a lot of low end and compressed really hard) still sounds awesome and very musical.
i mainly record and mix rock´roll stuff, with already a lot of "distortion" going on. so this is just my personal opinion.
great that you found the culprit and a possible solution. respect!
i can not really express how happy i´am. this was my first DIY.
i better get back behind the desk before i write more ********.
thanks again, this is a great forum!
cheers, alex
Similar threads
