LA4 LED-LDR problem?

[mitsos]

I have an LA4 that I've recapped and swapped the audio 4136 for 2x OPA2604. Sounds fine and compresses "to spec", but the GR meter is WAY off. I've calibrated it checking the I/O voltage, and the compression is working 100%, it's just the metering that is off. For example it might be compressing 5dB but showing 20, that kind of thing. In fact, the needle almost always shoots way to the left, even with small amounts of GR.  This is in GR mode, VU works fine. 

My guess is the two opto units (LEDs or LDRs) are not well matched to one another. I suggested to the owner that we swap the 2 optos for a dual vactrol which will be cheaper (or at least faster to get done). The LM301 seems to be working, I think.  If I were to test the optos, I guess I would have to remove them, apply the same current to the LEDs and measure resistance in each of the LDRs? And then vary the current and measure again?

thanks! 

 

[Kingston]

Did you scope the thing yet? Could be that the much faster new opamps are oscillating in certain situations where the slow old ones were happy. The schematic shows very little or no rail decoupling. You won't hear it but the gain reduction meter might see a crazy loud radio frequency bursts etc. weirdness.

 

[mitsos]

It was doing this before the opamp swap.  I made an "adapter" out of veroboard, I'm pretty sure I included 0.1uF caps from rail to ground. But yeah, I should scope it anyway.

 

[JohnRoberts]

If you want to redesign it for an accurate GR measurement you could inject a a fixed HF (above audio) carrier into the gain element and directly measure the attenuations of that HF signal. Alternartey you could superimpose a DC on the gain element and read that DC change, but you would need to filter that DC out of the audio and probably get a slower GR meter response.

Or don't worry about it? How does it sound?

JR

 

[mitsos]

I think I see what you're saying but I have no idea how to implement either of those properly, plus this thing isn't mine, so I need to keep it as original as possible (I know the vactrol won't be original, but I don't have time to go through a bunch of LEDs+LDRs to match them).  The idea is not super-accuracy of the GR meter, but as it stands right now, the GR meter is useless, it doesn't show anywhere near proper GR. 

While the comp itself sounds great (I know opinions vary on these comps, but I like it), the owner (a friend) wants the visual of the meter.
It basically goes to -20 no matter how small the actual GR.  I've never used an LA4 before but this can't be correct.  It makes me think either the meter LED is brighter than the GR led (could be one was swapped for a different Vdrop LED), or the meter LDR is more sensitive than the GR LDR.    I've built a few VCA and FET comps and those are generally accurate "enough."  This guy is all over the place.

Do you guys see any other likely causes?

 

[SSLtech]

Isn't there a GR tracking adjustment trimpot (I forget... it's been a while since I built mine!)

If not, it should be a trivial matter to swap a couple of resistors for trim pots.

 

[RuudNL]

IMHO the gain reduction calibration is a matter of adjusting R49 (trackning) and R51 (zero).

 

[mitsos]

yes I went through the procedure, it's similar to the 1176 tracking adjustment but doesn't specify a ratio, while the 1176 IIRC specifies 20:1. I don't have my notes here, but going from memory, what happens is, let's say you adust tracking while in 4:1. OK, done.  Then you redo the procedure at another ratio, again it does not match, so you have to re-adjust, which then negates your previous adjustment. And so on...  And then no matter what, in actual use it shows way more GR than is actually happening.

 

[SSLtech]

The ratio is actually irrelevant.

The LEDs are in series, so current is always equal, and luminance is predictably proportional for both... so that side is covered. Whatever ratio is being used to generate the luminance simply doesn't matter.

Tracking and 'rest' (0VU) adjustments interact, but the tracking adjustment should more than cover a range of scales... unless the CURVATURE (nonlinearity) of the optos is too far off...

Now while in theory this CAN happen, I've never once seen it among about a hundred T4's and other NSL-based designs that I've built over the years... I've ALWAYS been able to adjust for adequate tracking using R25 (in an LA-2a) or R49 (in an LA-4). That said, you may indeed have a CDS cell that's just chemically 'poisoned'. -Replace the pair and try again. -You're using steady tones at zero VU with the threshold in the fully-counterclockwise position (in the +4dB output meter setting) and switching between +4 and GR while matching using the 'tracking' control, and varying the 'threshold' control... with regular checks on the 0VU 'rest' position every time the 'tracking' control is adjusted, -right? -You're not trying to 'eyeball' it using program material, or anything like that?

I've probably set up about twenty LA-4's and never had any problem adjusting tracking to within say a dB error over say 10dB of GR range. (Asking anything beyond that is asking a little too much.)

 

[JohnRoberts]

Back in the '70s I spent a bunch of bench time trying to accurately control non-VCA gain elements. The best result I got was using a center tapped LDR. I could ground the center tap and use one half in a servo loop to control the other half reasonably accurately. I experimented with dual JFETs with less tracking success.  I ended up bailing on these gain elements for premium audio paths and just used VCAs, while VCA SOTA back in the '70s was not as good as now, their gain laws were nicely predictable.

JR

 

[SSLtech]

Yes, VCAs are inherently -gorgeously- predictable.

So to produce this exact display on something like an SSL limiter, there's no tracking adjustment needed with the Expat VU meter display adaptor for example... The control voltage is so delightfully predictable that you just set the zero and let it do its gorgeously accurate "thing"!

I love it.

While optos aren't as inherently predictable, the behavioral error is -normally- correctable to within a 'tolerable range' with a couple of variable resistors, in these designs.

 

[mitsos]

Keith, yes that's what I did. I posted the part of the manual which I followed exactly using about a 1V RMS signal. I did the calibration at my friend's home studio because he lives relatively close to the studio this thing belongs to, and I wanted to return it that day. But the weird results made me bring it home to test again. 

Here's the weird thing, I just re-checked it and it's spot on (I did not have to touch the 0 adj or tracking pots). The LA4 was left to "warm up" both now and last time.  Following the same GR calibration procedure I can now switch through any ratio (except for 2:1, which has a different knee, I believe) and the needle stays parked at -5dB like it's supposed to.

So either I messed up the first time (I doubt it because A)I went through it numerous times that day because I couldn't believe what I was seeing, and B)the thing is tracking perfectly now without me tweaking the adjustment knobs, I had done that last time).  The only difference between the tests is that my friend's AC supply is straight from the street (with a ground rod but still it's very iffy), while I have a balanced power trafo here. 

Could bad AC have caused something like this? I had all kinds of problems here myself before the balanced power install.  I live in a "developing" country so AC power quality is pretty much a crapshoot.

Thanks to all, I'm happy I can get this off my bench!

 

[jdiamantis]

Maybe wonky trim pots?

You re-re-re-calibrating exercised/cleaned/fixed them.

Seen it before.


jD

 

[mitsos]

Maybe wonky trim pots?
You re-re-re-calibrating exercised/cleaned/fixed them.

could it be the other way around?  re-re-re-calibrating wonky trimpots made them not want to settle in until left alone for a few days? Because this last time I didn't touch the trimpots at all.  Answer is probably change the pots, but I'll wait and see if the thing acts up again.

 

[JohnRoberts]

Maybe wonky trim pots?
You re-re-re-calibrating exercised/cleaned/fixed them.

could it be the other way around?  re-re-re-calibrating wonky trimpots made them not want to settle in until left alone for a few days? Because this last time I didn't touch the trimpots at all.  Answer is probably change the pots, but I'll wait and see if the thing acts up again.

Unlikely. Open, unsealed trimpots can get dirty, so exercising the trimpot could clear a layer of crud off the trimpot resistive element surface.

The only possible scenario I can imagine where sitting could improve the trimpot wiper contact is a thin layer of surface contamination that eventually gets displaced.

In any case consider replacing any unreliable unsealed trimpot with a sealed one.

JR