OPTO specs believable??

[ruffrecords]

I am developing yet another OPTO based compressor - this time a stereo linked bus compressor. But I am having a lot of trouble with OPTOs. I was originally going to use Vactrols but you can only get Chinese "clones" now which I do not trust. So I moved to the Silonix NSL-32 range. I originally chose the NSL-32R3 for is fast turn on and turn off times which meant I could set attack and decay times mostly independent of the OPTO properties and its nice low on resistance of 150 ohms at a modest 5mA LED current. However, these did not work too well and there was a large difference in maximum available attenuation between the two I was using (more than 10dB).

So I looked at a whole bunch of other designs and discovered that the most popular choice was the plain NSL-32. This has a nice short on time but its off time is 500mS so this is not going to make a fast compressor. It has a reasonable on resistance of 500 ohms at 20mA LED current. I also discovered that one manufacturer uses specially selected one which seems to imply there is a lot of variation between devices. I decided to buy ten and characterize them initially by on resistance. To do this I made a simple test rig:



The LED is driven from a bench supply via a 1K resistor. Placing a voltmeter directly across the 1K means the voltmeter reading is mA of LED current. Another meter set to ohms is connected directly across the photocell.

To begin with I turned the bench supply down to zero, turned it on and slowly raised its output voltage and watched the meters expecting not a lot to happen until a few mA was flowing. So I was very surprised to see the resistance had dropped dramatically at 1mA and by the time I got to 2mA current the resistance was not far from the specified on resistance which was not supposed to occur until the current reached ten times this value. So I tested all ten at 2mA. The resultants were as follows:

1. 720R
2. 950R
3. 490R
4. 620R
5. 747R
6. 560R
7. 660R
8. 810R
9. 720R
10. 840R

So the values with 2mA LED range from 490R to 950R with an average value of 712R (there is a nice cluster in the high 600s to low 700s).

The really strange thing is this values are obtained at one tenth the current specified in the datasheet.

So, as a check I tested a single NSL-32R3 which should have an on resistance of 150R at 5mA. It turned out to have an on resistance of 84R at 2mA LED current.

Bottom line is it seems you cannot trust OPTO specs at all.

Edit: I have now found three different datasheets for the NSL-32 and each gives a different value for the on resistance. Ho hum.

Cheers

ian

 

[L´Andratté]

Ian, that´s what I experience too. I think those specs are nominal only. The variance is big.
You simply have to select to get comparable behavior. I think it will be difficult for stereo, I tried and ended up with the M/S approach, which makes the imbalance not so bad. But I didn´t use tubes.

In the synth scene there is very much a diy approach to predictable resistive optocouplers.
F.i. this

Also, best price atm for originals I know of if you need a lot to select ): still not cheap...

Iirc there´s some trimming for stereo talk in the old Forssell opto thread, if it´s still around?

Another approach would be to use more than one opto per signal to average the differences, but I didn´t try that.

 

[gyraf]

After we lost the Exelitas and Vactec production lines to WEEE/ROHS,the LDR consistency has been truly horrible, bordering the random.

On the other hand, if you have the time and patience to screen large batches of currently-available parts, you can come up with some that has truly unique properties:

For many years it was impossible to make my "passive" G24 compressor, it was born just a "mind project" because of the obvious lack of available optocouplers that needed both very-low resistance and very high speed release as well as decent temperature dependency - but all these parameters turned out to show (wide) uncorrelated Gaussian distributions in Chinese current-production photocells. As there's always tails in good Gaussians, some 0.8 to 1.5% of a random batch of "VTL5C1" turns out to get you exactly where you wish to go...

/Jakob E.

 

[L´Andratté]

Note that the link I posted above goes to newly fabricated Advanced Photonix NS32R2, which presumably use original formulation*, which is not forbidden outside EU, and inside only to sell as commercial products (*which the interweb told me, fwiw, so could be the opposite)...

 

[scott2000]

Iirc there´s some trimming for stereo talk in the old Forssell opto thread, if it´s still around?

I think that may have been more geared towards getting the meters calibrated to show gain reduction but there's a lot of info... would be nice to see it....

One member mentioned this which may have some relevance here but it's from my notes and you'd have to look at schematics to have context and it's unclear if it's useful as nobody ever responded iirc..

"I dont know why Fred choose such low resistance for the series resistors(input-to-output), but it will make matching/selecting nececery if you build more than one chanel since it will require the LDR to work in a region where it is very close to its limmit ( lowest possible resistance). try increase the 4k99 and 1k resistors to 49k and 10k and the difference between different indevidual opto's will be much smaller...I'm sure it might affect other things like the responce curve, noise and how you set up the gain recovery amp due to different impedance, but that's what I'd try.."

 

[ruffrecords]

I think that may have been more geared towards getting the meters calibrated to show gain reduction but there's a lot of info... would be nice to see it....

One member mentioned this which may have some relevance here but it's from my notes and you'd have to look at schematics to have context and it's unclear if it's useful as nobody ever responded iirc..

"I dont know why Fred choose such low resistance for the series resistors(input-to-output), but it will make matching/selecting nececery if you build more than one chanel since it will require the LDR to work in a region where it is very close to its limmit ( lowest possible resistance). try increase the 4k99 and 1k resistors to 49k and 10k and the difference between different indevidual opto's will be much smaller...I'm sure it might affect other things like the responce curve, noise and how you set up the gain recovery amp due to different impedance, but that's what I'd try.."

I too noticed that the series resistors in most design were at least double what I was planning to use (11K). I am not sure if this is a good idea or not. On the one hand, you don't really need more than 20dB of attenuation so as long as the opto will reach less than 1K at a reasonable LED current there is no real need to make the series resistor any bigger (for example the NSL-32 reaches about 500 ohms at 2mA LED current). Also, if you make it larger you then operate the opto in the lower LED current region where there is likely to be even more variation in opto properties. Having said that, I have noticed that in designs where the opto is driven directly by the audio (as apposed a a rectified version of it) there is measurable breakthrough at high LED currents which manifests itself as several percent distortion so reducing LED current is possibly a 'good thing'. In my very first design I used a 668K series resistor which worked very well.

Clearly more experimentation is required.

Cheers

Ian

 

[Chandra]

Probably the LED output is too bright at that current?
You may to experiment with white and red LEDs for you to decide.
Regards.

 

[13engrsapper]

What about making your own out of a suitable LDR and light source? then you have a bit more control perhaps?

 

[rock soderstrom]

What about making your own out of a suitable LDR and light source? then you have a bit more control perhaps?

That's true, but it also makes it more difficult for others to replicate the Opto compressor. An opto cell with known, consistent specifications would make this easier.

 

[ruffrecords]

What about making your own out of a suitable LDR and light source? then you have a bit more control perhaps?

That is an option. I have corresponded extensively on this topic with another GDIY member who ended up making his own opto cells, even selecting the best colour LED to use. Driving the LED directly with the audio is attractive from the point of view of simplifying the design but the breakthrough issue is a serious problem. Fortunately light is not attenuated much over short distances so a possible solution to the breakthrough problem is a cell made with an LED and an opto sensor separated by an inch of tubing.

Cheers

Ian

 

[gyraf]

Do you believe that the breakthrough is (electrostatic) interference? Was always my understanding that it was about the LDR being extremely-fast in its low-resistance zone, thus leading to modulation distortion at high GR's, even if low GR's were fine because of ldr delay

I've been toying with ideas about using some phosphorous "after-glow" material placed between led and ldr to even this out a bit. Perhaps even just chop off a corner of a "real" T4B EL-panel, find the right led wavelength to exite it, and off you go...

/Jakob E.

 

[JohnRoberts]

Back in the 70s I evaluated LDRs on my bench for dynamics processing and indeed they were not very consistent. I sourced one with a center tapped resistance element and tried to use that center tap to linearize the resistance change. Grounding the center tap and putting 1/2 inside the feed back loop of an op amp would allow more precise control of on resistance.

The unavoidable slow release is the unintended magic that makes some of these opto based dynamics processors not suck as much as they would otherwise.

JR

PS: Years later when Peavey designed an opto limiter (VCL-2) I only steered the project from a distance. I requested the LDR gain element because of its characteristic sound, not for precision.

 

[gyraf]

to linearize the resistance change

I feel your pain - I was trying to do something like this 35 years ago, in an attempt to linearize the CV/GR coherence (here with a LM13600 OTA as reference)


No, it never worked well - only later I realized that need for exact ratios is BS

/Jakob E.

 

[NovaPA]

Recently, in order to check the error of homemade optrodes (photoresistor + LED in black heat shrink), I applied a small stable voltage to the LED, connected a multimeter to the resistance measure and left it like that all night.
Overnight the value biased by about 10-15%

 

[ruffrecords]

Do you believe that the breakthrough is (electrostatic) interference? Was always my understanding that it was about the LDR being extremely-fast in its low-resistance zone, thus leading to modulation distortion at high GR's, even if low GR's were fine because of ldr delay

To be honest I do not really know. I do know it happens with Chinese Vactrols and also with the NSL-32 range and the guy who made his own apparently did not have the problem. It would be interesting to feed some dc through the one or more different types of opto sensor and drive its LED from a sig gen and see exactly what comes out the sensor.

Cheers

Ian

 

[jaddie]

Don't want to derail the thread, but could someone give me the "elevator speech" of why you want to use the OPTO solution in these days of really good VCAs?

I built stuff with Vactrols a lifetime ago. They were trouble then too. There was a company in the late 1970s that built a mixer that used Vactrols as stereo channel faders driven by unconditioned(!) DC voltage from a linear pot. No, it didn't work well.

 

[jaddie]

For entertainment, here's a link to the manual of the console Mono version. It was awful, but the stereo version was worse.

See page 28 for the "electronic attenuator" schematic.

https://www.steampoweredradio.com/pdf/ramko/manuals/ramko dc8m audio console manual.pdf

We don't even need to talk about the CMOS-based touch switches. OMG what a mess!

 

[ruffrecords]

Don't want to derail the thread, but could someone give me the "elevator speech" of why you want to use the OPTO solution in these days of really good VCAs?

Because I design and build tube mixers and I do not want any sand in the signal path.

Cheers

Ian

 

[jaddie]

Because I design and build tube mixers and I do not want any sand in the signal path.

Cheers

Ian

Ok, I understand the overall desire. I'm a little puzzled by the desire to keep "sand" (what is that, distortion?) out of the signal path when the distortion characteristics of optos are so poor. Page 10 of this pdf, for example, details it. Are the new ones better? Because every modern VCA certainly is.
http://denethor.wlu.ca/pc300/optoisolators/analogoptoisolatorintroduction.pdf

So...operating them with very low audio voltage with lots of makeup gain?
Or maybe that's "color"?

 

[JohnRoberts]

Ok, I understand the overall desire. I'm a little puzzled by the desire to keep "sand" (what is that, distortion?) out of the signal path when the distortion characteristics of optos are so poor. Page 10 of this pdf, for example, details it. Are the new ones better? Because every modern VCA certainly is.
http://denethor.wlu.ca/pc300/optoisolators/analogoptoisolatorintroduction.pdf

So...operating them with very low audio voltage with lots of makeup gain?
Or maybe that's "color"?

I'm no Ian... but I suspect by "sand" he is referring to silicon, as in solid state electronics..... or not.

JR