Altec 436C: to recap ps or no? ...plus strange behavior

I just finished checking out an Altec 436C on the bench, and found the voltages to be about what they're spec'ed to be in the published schematic (biggest variation about 10% under).

My only complaint in its behavior. Bear with me here for a second:
In its compression action, it has a weird recovery/release lag. Say I've got an organ plugged into it and I play two sustained notes in a row with a short rest between. On the first note, the dB reduction meter swings to 10 and stays there. I release the note, and the meter starts to swing back to 0 dB reduction. No problem so far. However, on playing the second note (say, when the needle's made it to about 5 dB), the note starts very quiet (almost off), and gets LOUDER as the needle swings back up to 10dB reduction! That's not right... it sounds like the note's being played with a slow attack envelope. This seems to happen more with fast release times.

So... my current question is, would recapping the PS "stiffen" its voltages and fix this weird mushy recovery lag, or is this inherent to the circuit?

Full docs and schem here:
http://www.dvq.com/hifi/images/436c.pdf

Cheers,
Leigh

edit: or maybe just replacing C4, which is part of the sidechain rc?

Actually, if the release time is set too fast, it will do exactly as you describe. I'm familiar with this because I'm working on a 436A right now, doing a number of mods that include variable attack and release. I found that if the time constant is too short, the audio will cut out altogether, giving a "choppy" effect.

If the B+ voltages are where they should be and there's no ripple, there's no need to replace the PS caps. But having said that, it certainly wouldn't hurt to replace C4. It may have become leaky or drifted low over time, which case it would definitely affect the action of the sidechain.

By the way, thanks for the link to a nice clean scan of the 436C manual. I haven't seen that particular PDF before.

doing a number of mods

Click to expand...

would be nice to know more...

[quote author="NewYorkDave"]Actually, if the release time is set too fast, it will do exactly as you describe. I'm familiar with this because I'm working on a 436A right now, doing a number of mods that include variable attack and release. I found that if the time constant is too short, the audio will cut out altogether, giving a "choppy" effect.[/quote]

Well, I haven't modded it to have a shorter release time, but maybe if C4 has aged badly, there's a natural "mod" going on there. That choppyness is no good, and I can't believe that the original designers would have intended it to work as such.

[quote author="NewYorkDave"]If the B+ voltages are where they should be and there's no ripple, there's no need to replace the PS caps.[/quote]

What would be acceptable ripple? A few millivolts AC too much?

[quote author="NewYorkDave"]By the way, thanks for the link to a nice clean scan of the 436C manual. I haven't seen that particular PDF before.[/quote]

You're welcome - whoever did that retyped all the text!

Cheers,
Leigh

A little more detail on the B+ ripple:

The center of the output transformer is spec'ed to sit at 152 V. Mine measures at 162 V (yes, higher than spec) with ~3 mV AC.

This is with the output (terminals 1 & 4, with 2 & 3 strapped for a 600 ohm output) loaded with a 620 ohms resistor. I leave that on there all the time, it was a Dorsey trick to reduce output transformer ringing when sending to a modern, high impedance input. It also reduced the output level a hair, but I really should see about putting a real attenuator on there as well.

Thanks,
Leigh

3mV is nothing. The push-pull action of the amplifier will cancel out small amounts of ripple, anyway.

As for the B+ being higher than spec, check the following in this order:
1. Your AC line voltage and your heater voltage. Are they higher than spec? Line voltage today tends to run a bit hotter than it did in the '60s. If this is the cause, don't worry about it. That 10 volts won't make much difference.

2. The voltage drop across the cathode resistor of the output tube. If the tube is becoming weak, the plate current will be low, the cathode voltage will be low and the plate voltage will be high. If the tube is weak, replace it.

As for the choppiness in the audio, the most likely culprit by far is C4. Replace it if you haven't done so already.

Oh, the "Dorsey Trick" is nothing new. When running a transformer-coupled piece into a "bridging" load, it's often necessary to add a termination for flattest response. This has been known for a very long time.

As for an output attenuator, stay tuned. I'm adding one to the 436 I'm working on right now.

Thanks Dave for the advice!

C4 is getting swapped out as soon as I find a replacement.

Leigh

Checked line voltage, and it's 123 (vs 117 as listed on the schematic), so I think that higher voltage on the B+ is just from that.

[quote author="NewYorkDave"]As for the choppiness in the audio, the most likely culprit by far is C4. Replace it if you haven't done so already.[/quote]

Tracked down a replacement cap today, but unfortunately the choppy behavior persists. Also tried swapping all the tubes one by one, as well as disconnecting the meter, all to no avail.

So I'm wondering now whether to keep digging for a faulty part, or to accept that this is simply the behavior of the beast. Can you reproduce the weird recovery thing with everything at stock values?

Just found this thread, which talks about the linkage b/t distortion and time constants in variable-mu circuits, which I'll have to give a close read:
http://www.groupdiy.com/index.php?topic=4955

Also, found a good explanation of the sidechain circuit and a power supply discussion, both from R.A.T..

I'll be interested to see what your attenuator looks like...

Cheers,
Leigh

Leigh-

Ive had a bunch of 436/438's over the years and what you are describing is fairly normal. These limiters are actually fairly worthless so far as Im concerned if you leave them stock. NYD is actually working on my very last 436 and the both of us thought out together a few things to make the unit more usable. It was never designed as a piece for recording studios and was never designed to interface at the levels that have become standard in recording studios today. If you dont do a bunch of mods you kinda have the square peg in the circle hole. Anyhow, did you check all the resistor values around the attack? Its definitely the attack that is doing that. You may have a resistor that has drifted a bit, those things get pretty hot inside sometimes so it wouldnt be so far off. I would off the bat start screwing around with changing the attack and release times into something a bit more usable. The units dont necessarily sound bad (one old 438 I had would fatten a snare drum like nothing Ive ever heard before or since, real glad I sold it...) however I have no clue how computer world people deal with these things, I have a totally analog transistor studio and could never efficiently use the thing. If I have to futz with gear to interface it, I generally just unplug it and forget about it, that altec crap is the king of futzing after you patch. Stuff was designed for supermarket pa systems and definitely behaves like it. Cleanup on aisle 7.

dave

The 436 does seem very limited (pun intended) to those of us whose expectations have been shaped by more modern compressors. It's a simple device designed to tackle a simple task: to take the fairly weak output from a mic preamp, smooth out any gross variations in volume and amplify the signal up to a level sufficent to feed a telephone line, studio-transmitter link or power amplifier in a PA system.

Although the manual makes no reference to doing so, you could run a mic directly into the 436, it has so much gain. And because of its high gain, and high output level, input and output attenuators are most definitely needed to interface it to typical +4dBU line-level system. On Dave's 436, I added a stepped bridged-T attenuator after the output transformer that gives 0 to -27dB (plus "off") in 3dB steps. I also added a switchable 12dB attenuator after the input transformer, just to give some measure of input level control.

But even with the input attenuator switched in, the average input level still needs to be below -10dBU or so if you desire full range on the threshold control--that is, 0dB compression when the control is turned up all the way. So, with a +4dBU (nominal) input, a 10 or 20dB pad may be needed even when the internal attenuator is switched in. I didn't want to make the switchable attenuator any greater than 12dB, figuring that any really big mismatches could be better handled by an external attenuator before the input terminals, preventing overload of the transformer.

If you play around with the sidechain time constants in your 436C, you'll soon find out why Altec choose those particular values. You can't reduce them much below stock values without causing instability at certain settings of the controls. I was only able to reduce R9 from 270K to 180K, and R12 from 33K to 27K. This equals nominal minimum attack and release times of approximately 30mS and 200mS, respectively--still pretty slow compared to what a modern compressor can manage! I also implemented a variable attack control by including a 100K linear pot in series with R12. Of course, this pot can only slow down the attack by an additional 0 to 100mS.

For what it's worth, adding a balance control to the 6BC8 variable-gain stage seems to help in minimizing the "roughness" at short attack and release settings. I used a 250-ohm trimpot wired between the cathodes, with the wiper going to the GR meter. I tried 100 ohms originally and found that it didn't give enough of a range to keep the "sweet spot" near the middle of the pot, where it should be.

The balance adjustment is made by applying a common-mode signal to the grids, adjusting the threshold control to some median value of compression (say, 10dB) and trimming the pot for minimum signal at the 436's output terminals--or, if preferred, minimum negative DC voltage at C4. An easy method for providing a common-mode test signal is to connect a 1uF, 100V cap between either side of the heater supply and the junction of R9 and R12.

Also, I replaced R8 with a 100-ohm trimpot to allow calibration of the GR meter.

Excellent! Thanks a million... I've gotten my head around what's going on in the guts of this unit, thanks in a large part to your patient explanations.

Leigh

omg, listen to dave, got my 436 back from him today and its insane. Gave him a brick and got back a bar of gold. Also listened to his single bottle preamp which was super super cool.

dave

Thanks, Dave. Soon, I'll start a new thread with pics and details of the mods. Stay tuned.