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

[RuudNL]

The distortion is definitely NOT in the diode attenuator. I tried to apply a control voltage from a regulated power supply and the distortion was very low over an attenuation range of 0 to 20 dB.
I am convinced that the 'LF distortion' is caused by a component superimposed on the CV.
In the meantime, I am following this thread with interest...

 

[SaMpLeGoD]

Hello all!

I have an original 33609 that I need to regulate. My question is: the trimpots of the unit builded here are numbered in the same way as the oiginal? can I use the regulation method posted here to regulate the original unit?
The major problem my unit have is that there's a little bit more gain in one channel than the other... I think is a simple fix.
Thanks a lot!

Eddie

 

[Rob Flinn]

Eddie the gain is just one of the preset pots. 

Switch the unit in, but the comp & lim off.  Put a signal through the unit & trim the gain pot RV 1 for unity gain.  Do the same on the 2nd channel.

 

[tommypiper]

I'm wondering if tighter matching of the "bridge diodes" or variolosser would reduce the LF distortion? In my own experience with other limiters using the dynamic diode resistance of rectifiers to implement gain reduction has shown that matching of these elements at stratigic bias points reduces distortion (and feed-through thump).

jDiamantis

Sounds very plausible to me.  Care to elaborate?  What unit?  How did you test?  How did you match?  What bias points?  How?

 

[tommypiper]

The distortion is definitely NOT in the diode attenuator. I tried to apply a control voltage from a regulated power supply and the distortion was very low over an attenuation range of 0 to 20 dB.

Well, wait a minute.  All that shows is that the attenuator was attenuating an already-clean DC signal.  What you want to do for a real test is feed the diode bridge music, and then check the output for a clean DC signal.  Right?  Because you're rectifiying the side chain signal, right?  (I haven't looked at the schematic, but this is how some tube compressors work and seems to be the same principal here.)

I am convinced that the 'LF distortion' is caused by a component superimposed on the CV.

What component are you referring to?  A physical part on the circuit board, like a capacitor?  Or do you mean an electrical signal component created by the circuit somewhere...?  Please can you clarify? 

 

[Igor]

In other words, CV breakthrough in VCA
Even when input is zero, and about 3V P-P @1Khz triangle wave CV feed into "link",
worst case of CV breakthrough at 10db attenuation range will be -40 db
(with diodes matched within 2%, i.e. worst case from same batch of diodes).
I think the LF distortion some guys talked about is the natural, inherrent part of 33609, which comes from fast recovery time constant settings....anyway, I'm very curious about this issue.

 

[Igor]

3) Igor, can I still send the boards to you? Would you look at them? I packed them in the box many months ago and it's been sitting here all this time, since I didn't have time to work out the shipping. I can ship the package as early as the end of this week.

Of course.  I am on very tough timetable, but will try to see what is possible here....
Feel free to pm.

 

[SaMpLeGoD]

Eddie the gain is just one of the preset pots. 

Switch the unit in, but the comp & lim off.   Put a signal through the unit & trim the gain pot RV 1 for unity gain.  Do the same on the 2nd channel.

Thanks a lot Rob! It's a wonderful comp!!!
Cheers,

Eddie

 

[tomcat]

Regarding the meters:

Do i need a meter with 1mA movement or the standard 1.734V VU?

Was asked this when ordering.  ??? In my opinion the 19AW and the AL19AW are both 1mA, just the ballistics are different on the 19AW (and of course the price).

I think they just want to sell a PPM meter instead of the Audio Level meter

 

[Igor]

1ma, try to get faster one to have right indication.
(slower vu's showing actually less gain reduction on fast peaks).

 

[tomcat]

Thanks Igor, BOM on 1st page ist just for 1 channel, correct?

And the (very) correct part nos for the elmas 24x1 is:

04-1103-20 (this is the pcb mount version),  04-1103 is the version with solder eyelets, both are endless rotating, you need a stop screw (4124-21) if you dont like this or if you need a e.g. 21 pos switch

04-1133-20 (pcb mount version) is not endless rotating, here you dont need the stop screw.

 

[Igor]

We discussed it already, 04-1133-20
(pcb mount version) with stop screw, for comp.threshold stoped at 16 pos.

 

[jdiamantis]

In other words, CV breakthrough in VCA
Even when input is zero, and about 3V P-P @1Khz triangle wave CV feed into "link",
worst case of CV breakthrough at 10db attenuation range will be -40 db
(with diodes matched within 2%, i.e. worst case from same batch of diodes).
I think the LF distortion some guys talked about is the natural, inherrent part of 33609, which comes from fast recovery time constant settings....anyway, I'm very curious about this issue.

While this is essentially correct I would also test with a 50hz sine wave at the CV feed point, checking breakthrough at 10db and 20db attenuation. This will flesh out any intermod products that may be generated in addition to those generated by fast release times, which will track low frequency signals and make the attenuator act like a multiplier.

As the CV law is close to log, the resulting release with a simple R/C will be steeper or quicker initially (more so than a linear circuit ca. an FET attenuator), slowing down as the charge cap reaches zero. IIRC, the release time constant was in the neighborhood of 50ms. Perhaps a doubling of this might reduce the distortion enough to not be an issue. The only other alternative would be to build an anti-log release which increases complexity somewhat.

jD

 

[Igor]

Ye, the test signal should have an offset (to be under 0volts), in other words,
if AC is 3V peak-to-peak, offset should be +1.5V,
in this case the attenuation range is 8db.

IMHO solving the "low distortion" issue will make different sounding compressor...OK
Anti-log release? I think it is possible, but with some op-amps and kinda
-5....15V power supply needed. The best solution was
hold circuit offered few pages ago by RuudNL (IIRC).

 

[promixe]

Hi All,

I've sent the 33609 channel to Igor a couple of weeks ago to check up on the "LF Distortion" issue, hopefully he will receive the package soon and let us know what he finds!

In the mean time I have a few 'theoretical' questions about 33609 operation some of you could answer:

1) What is the reason to tie the compressor attack timing cap  (C13) negative pin as close as possible to PSU ground pin, rather than just to a local ground nearby?

2) How exactly is the gain set in BA340? I see there is R40 (1k3) in parallel with selectable gain switch resistor. The manual says that +20dB gain is equivalent to 1k6 gainsetting resistor, but I don't calculate anywhere near 1k6 around R40 & gain switch Rs at any gain setting... Am I overlooking something internal to the BA340 to take into account?

3) Kind of a continuance of the question #2: The schematic shows there is +17dB make up gain between T1 and T3, I assume is handled by BA340 when the makeup switch is at 0dB (off). This default +17dB setting added to +20dB max makeup gain selection = +37dB total gain from BA340, which is specked to go up to 40dB. BUT, Geoff T is saying that BA340 is making up around 32dB gain AND THEN adds another 20dB on top of that bringing the total gain 12dB over its spec (thus making it a bit noisy at full +20dB makeup setting). How am I not following the schematic to see what he's talking about?



PS: Check this out, Geoff T is asked the "LF Distortion" question =)) - http://www.auroraaudio.net/dcforum/DCForumID1/792.html

 

[RuudNL]

I have a feeling that the question on the aurora forum might have been posted bij a DIY builder also active on this forum... 

 

[promixe]

Hi All! Igor received my channel a while ago and had a chance to check it out and compare with known working units. Here is what he wrote back to me (loose translation =):

..Regarding the LF distortion, here is a normal channel, 20Hz signal, 5-6dB limiting - http://i251.photobucket.com/albums/gg291/diy33609/P1030352.jpg

Here is what I'm thinking: if the diode bridge is not balanced you could get this behavior - http://s251.photobucket.com/albums/gg291/diy33609/?action=view&current=P1030358-2.flv

The way this was done was feeding 50Hz +10dB into the compressor input with 5-6dB of limiting applied and then one of the diodes (in the bridge) was heated with soldering iron - http://i251.photobucket.com/albums/gg291/diy33609/P1030359.jpg

Then the limiter starts freaking out, especially in LF region. I had this happen with one of the channels with diodes not matched all that well. It was my very first 33609, after fixing the matching the customer was happy, compared the unit with the original and haven't heard any LF distortion.

One more thing, if one needs to completely get rid of the low end distortion, the only way is to increase C14 - http://i251.photobucket.com/albums/gg291/diy33609/P1030348.jpg.
For instance (with C14 mod) - http://i251.photobucket.com/albums/gg291/diy33609/P1030350.jpg
And here is the normal behavior of the "normal" channel (without mod) - http://i251.photobucket.com/albums/gg291/diy33609/P1030352.jpg

And another thing, at D9 cathode - http://i251.photobucket.com/albums/gg291/diy33609/P1030356.jpg, sometimes you can toss C24 (330pF), then it looks better - http://i251.photobucket.com/albums/gg291/diy33609/P1030357.jpg.....

Overall Igor says my channel behaves withing the usual specs and is considered normal (in other words, nothing's wrong with it). He also says that he prefers Belclere TF10015/10016 xformers (I assume over Carnhills), when listening AND on the analyzer.

While it's good to hear that my channel is all good to go, I'm still quite puzzled as to why I hear a huge distortion difference between my units and the original at the same settings on the same material. And why others cannot hear anything...   More to follow... =)

PS: Thanks to Igor for taking his time to check out and participate in our LF distortion saga! =)))

 

[tommypiper]

Good info.  I didn't understand the D9 C24 comment.  Is he saying remove C24?  (And I also don't understand that waveform pic.)

Looking forward to your further comments about why you can hear it when others can't...

 

[promixe]

I didn't understand the D9 C24 comment.  Is he saying remove C24?

Yes - he said to get rid of the C24 cap to somewhat improve the limiter response. The scope picture, as I understand it, has a momentary impulse on one channel (you can see it ringing), and the CV response superimposed on the other scope channel (as you feed the pulse into limiter, you see how it responds generating a ducking CV)...

I've found some interesting info while reading the 2254E service manual. It says that the fast attack setting in the limiter sidechain should only be used to drive radio transmitters ("and should be used with care in recording"...), as it has time constant of 100uS (microseconds), while the "slow" setting is a 5mS attack and should be used on music related material. Having read that, I still don't get why the original 33609 I had for comparison didn't distort _audibly_ nearly as much as my own-built unit... If I only read the above Neve info and didn't have the original to compare, I would've settled with LF distortion. Hey, if Neve says to not use these modes on music, then why bother... But the original sounds way smoother when used with "fast" limiter attack and fast recovery, eventhough it too distorts!

Again, here is a sound example so everyone can hear and understand what I'm talking about:
http://dev.ax-design.com/electronics/33609/33609-test2.mp3 - MY unit (bottom)
http://dev.ax-design.com/electronics/33609/33609-test3.mp3 - ORIGINAL unit (top, the "no letter" revision)

Settings used on these samples are seen in the picture above..

 

[tommypiper]

Yeah, the audio samples are night and day.   It only takes a second of each to hear the difference! NOT subtle.

Just a dumb question, is it possible you were running the signal differently through the two units?  Different gain structure, anything?  Were they set up differently internally?

For example, I don't understand the two switches on the right on the original.  It says compress, In.  Then it says Bypass, In.  Does that mean the other position is Bypass, Off?  Meaning the bypass was "In," or active (in bypass)?  In other words, was the compressor in bypass on the original?  Can you absolutely be sure how those switches were set up?