Yep - and another thing to consider for the people that believe that "the originals were all better" is that the EL-panel will change characteristics quite a lot over time...
Jakob E.
Jakob E.
LA-2A Theory of Operation - EL Panel Characteristics?
- Thread starter Joe Tritschler
- Start date
Help Support GroupDIY Audio Forum:
Jonte Knif
Well-known member
Hmm... There is still frequency dependence left in an EL compressor. Do you _really_ know it is insignificant?
Look at the curves in my link. The nominal compression is at 1kHz. First 5 are "comp" at -3 -6 -9 -12 and -15. The last is "limit" -15 just to show that because of less feedback the slope gets steeper. It is pretty extreme.
http://picasaweb.google.fi/jonte.knif/CompCurves#
I'm not sure if I was able to parse the meaning of up and down, but regarding the memory effects LDR not only has release memory, it also has "attack related memory", which is less known. Attack is slower after long period of no compression. Well, long meaning something like second or two. Is this "musical" too? Perhaps.
Attack time is also dependent on the amount of compression. A deep compression starts way faster than a mild one. Another factor of "musicality" I guess. Well, this was off topic. Just for those not so familiar with all the funny things going on there.
Thanks for further analysis PRR. Great.
Interesting stuff my brain is hurting sadly I can never get my head round all the figures 
only thing I can relate to is what CJ was saying about the LDR being slower than the light & any EL panel quirks get unheard, I would totally agree with that, a conclusion reached with my own humble experiments, so are there more experiments to be try???, can we get different sounds from LA2As by changing components or replaceing side chain amp???....
I have 2 homemade LA2As been fiddling with 1st one for lsat few years now but they just dont get used anymore I use my vari mus instead, its a shame .....I dont think Iv ever heard the true sound of the LA due to lack of cash for the right cell or real T4B till now ive always thought the compression sound thats so loved is all in the unobtainable to me Cell.....got the best cell I can find in mine but still not right.
At one stage I had a EL panel as light source & Neon bulb as light source aimed at the same Cell & a switch to switch either, the neon sounded better than the panel to my humble ears, I would have thought they sound the same, the light colour was different & the neon would compress with less signal I get that but how could it sound different ????
Please forgive my lack of theory glad that you guys are investigating the EL panels
cheers
only thing I can relate to is what CJ was saying about the LDR being slower than the light & any EL panel quirks get unheard, I would totally agree with that, a conclusion reached with my own humble experiments, so are there more experiments to be try???, can we get different sounds from LA2As by changing components or replaceing side chain amp???....I have 2 homemade LA2As been fiddling with 1st one for lsat few years now but they just dont get used anymore I use my vari mus instead, its a shame .....I dont think Iv ever heard the true sound of the LA due to lack of cash for the right cell or real T4B till now ive always thought the compression sound thats so loved is all in the unobtainable to me Cell.....got the best cell I can find in mine but still not right.
At one stage I had a EL panel as light source & Neon bulb as light source aimed at the same Cell & a switch to switch either, the neon sounded better than the panel to my humble ears, I would have thought they sound the same, the light colour was different & the neon would compress with less signal I get that but how could it sound different ????
Please forgive my lack of theory glad that you guys are investigating the EL panels
cheers
AnalogPackrat
Well-known member
Something else to consider is the emission spectrum of the EL panel (and whether it varies with input signal level or frequency content) as well as the spectral response of the CdS cell. The EL panel typically has a blue-green appearance (not sure if it is a continuous spectrum or a handful of emission lines), but I recall someone (probably CJ) stating that with higher the audio frequency, the panel started to turn purplish. This may matter since the CdS cell has a peak response in the green (usually around 560nm) and falls off in the red and blue. "Purple" (either red plus blue or deep violet) has no green spectral content.
Does anyone know anything about the EL output spectrum?
A P
Does anyone know anything about the EL output spectrum?
A P
Jonte Knif
Well-known member
I use modern "yellow/green". Level doesn't seem to affect the color very much.
At low frequencies the color is yellow/green. At about 1kHz it is pure green. At higher freqs it gets green/blue but not very saturated, kind of "weak". I don't know the name of the color in English. Kind of a certain semi precious stone.
And with constant voltage drive it gets more luminous towards high freq:s.
-Jonte
At low frequencies the color is yellow/green. At about 1kHz it is pure green. At higher freqs it gets green/blue but not very saturated, kind of "weak". I don't know the name of the color in English. Kind of a certain semi precious stone.
And with constant voltage drive it gets more luminous towards high freq:s.
-Jonte
> the panel is very fast
The old (~~1963) panels had a lot of after-glow. In the middle of the night, you could unplug a nightlight, take it to a darker room, and still see a faint glow.
I think the new ones are different.
Any suitable LDR should be much much slower.
The old (~~1963) panels had a lot of after-glow. In the middle of the night, you could unplug a nightlight, take it to a darker room, and still see a faint glow.
I think the new ones are different.
Any suitable LDR should be much much slower.
> consider is the emission spectrum
Yes, and testing that needs fancy sensors and calibration, plus information on the (perhaps non-linear) color sensitivity of the LDR.
But I think Jonte tested with his LDR as the "sensor". So any color-shift is included in his curves.
Yes, and testing that needs fancy sensors and calibration, plus information on the (perhaps non-linear) color sensitivity of the LDR.
But I think Jonte tested with his LDR as the "sensor". So any color-shift is included in his curves.
AnalogPackrat
Well-known member
PRR said:
Or simply reading the manufacturer's datasheet...
Silonex CdS Data
Yes, but he's still trying to find the source of various unusual characteristics found in his data such as frequency dependent compression response. Perhaps his choice of EL material and CdS cell are part of the root cause. Seems like he's not using the same stuff that would be found in a Teletronix (or UA) T4b. This may not be a trivial difference.
A P
Jonte Knif
Well-known member
Correct, like I said my element is not even a copy, but many people around use the same materials. These measurements surely should be done with an original or copy too. One could go even so far as to state that I have so far only shown my way to measure the element and the results will not be valid in any sense in general. Possible.
As I said in an earlier post it would be just great if somebody explained the physics behind the nonlinearities so that we could come up with a tentative theory to help us choose the right materials for a given application.
Or then we could just go one and measure.
-Jonte
As I said in an earlier post it would be just great if somebody explained the physics behind the nonlinearities so that we could come up with a tentative theory to help us choose the right materials for a given application.
Or then we could just go one and measure.
-Jonte
AnalogPackrat
Well-known member
I don't know much about the EL part. Here's something I found that at least shows the internal structure.
http://www.indiana.edu/~hightech/fpd/papers/ELDs.html
As CJ said, a lot of the temporal characteristics of the LA2a are due to the operation of the CdS photocells. From what I've seen and measured (which isn't nearly as much as what CJ did a few years ago), these things do not have tight tolerance on resistance or response time. They are likely measured and selected for use in T4bs. The attack and release characteristics of the LA2a are dominated by the photocells. CdS has asymmetric response--resistance from dark to light decreases much faster (more than an order of magnitude) than it increases from light to dark.
It has a memory of many seconds. If you keep one in the dark for a few minutes, give it a one second pulse of light, let it rest a couple of seconds and then give it another one second pulse, the two R vs t curves will be different.
As you can see from the datasheet I linked earlier, photocells are made to have different responses for different applications. Just look at some of the Perkin-Elmer white papers and datasheets for their (discontinued?) Vactrol series of LED-CdS optocouplers:
http://optoelectronics.perkinelmer.com/content/DataSheets/DTS_PhotocellAnalogOptoisolator.pdf
http://optoelectronics.perkinelmer.com/content/RelatedLinks/Articles/ATL_analogoptoisolatorB.pdf
There's a wealth of information out there... Also look for similar information from Silonex who also make old skool optocouplers. Maybe if you ask nicely CJ can remind us more about what he found in his rather extensive experimentation with the T4b.
A P
http://www.indiana.edu/~hightech/fpd/papers/ELDs.html
As CJ said, a lot of the temporal characteristics of the LA2a are due to the operation of the CdS photocells. From what I've seen and measured (which isn't nearly as much as what CJ did a few years ago), these things do not have tight tolerance on resistance or response time. They are likely measured and selected for use in T4bs. The attack and release characteristics of the LA2a are dominated by the photocells. CdS has asymmetric response--resistance from dark to light decreases much faster (more than an order of magnitude) than it increases from light to dark.
It has a memory of many seconds. If you keep one in the dark for a few minutes, give it a one second pulse of light, let it rest a couple of seconds and then give it another one second pulse, the two R vs t curves will be different.
As you can see from the datasheet I linked earlier, photocells are made to have different responses for different applications. Just look at some of the Perkin-Elmer white papers and datasheets for their (discontinued?) Vactrol series of LED-CdS optocouplers:
http://optoelectronics.perkinelmer.com/content/DataSheets/DTS_PhotocellAnalogOptoisolator.pdf
http://optoelectronics.perkinelmer.com/content/RelatedLinks/Articles/ATL_analogoptoisolatorB.pdf
There's a wealth of information out there... Also look for similar information from Silonex who also make old skool optocouplers. Maybe if you ask nicely CJ can remind us more about what he found in his rather extensive experimentation with the T4b.
A P
i have a bunch of charts and graphs, hostage at ePowerweb, but really, the most sound change you get will come from the cells, not so much the panel.
the cells have a multitude of curves to them, even samples from the same lot.
radio shack has these cheap cells that are really quick, faster than the stock Silonex,
so they give a different sound. the quicker the cell, the faster the recovery time.
so you get a difference both in attack and release.
the cells have a multitude of curves to them, even samples from the same lot.
radio shack has these cheap cells that are really quick, faster than the stock Silonex,
so they give a different sound. the quicker the cell, the faster the recovery time.
so you get a difference both in attack and release.
Joe Tritschler
Member
This is all fascinating stuff, guys. I do like PRR's philosophy...
> "Sounds good" is the best test.
And that they do. Many of 'em.
Jonte - you hinted that you ended up driving the panels with constant voltage - was that in actual use? I have some ideas on how to do that with adequate bandwidth and slew rate into the capacitive load to do some pretty heavy peak limiting fairly cleanly...whether or not it's an "improvement" over the ol' 6AQ5 driver might be debatable. Regarding the matching of highly inconsistent cells...isn't that basically what you're paying for when you spend the big bucks on T4B replacements? That and the octal can?
> "Sounds good" is the best test.
And that they do. Many of 'em.
Jonte - you hinted that you ended up driving the panels with constant voltage - was that in actual use? I have some ideas on how to do that with adequate bandwidth and slew rate into the capacitive load to do some pretty heavy peak limiting fairly cleanly...whether or not it's an "improvement" over the ol' 6AQ5 driver might be debatable. Regarding the matching of highly inconsistent cells...isn't that basically what you're paying for when you spend the big bucks on T4B replacements? That and the octal can?
Jonte Knif
Well-known member
I use mosfet driver and a little bit of treble boost to compensate the capacitive load. Essentially equal to the original. I don't think you have to worry about slew rate at all. A simple tube or mosfet without feedback does not have a slew rate.
The panel resistance measurements were done with my AP and a step up transformer with one input of AP sensing the voltage applied to the element. Pretty simple and accurate.
I have no idea what kind of matching commercial elements have in them. In a mono unit it is not too essential.
I start with measuring distortion of LDR:s. There are _big_ differences. I use the higher dist ones for metering and sort the rest in a couple of "grades".
Then I measure resistances at couple of points. I have noticed that the R/luminosity curves don't seem to have anything particularly strange in them, so 2 points is enough. Ready compressors track very well.
I haven't measured time constants. Hmmm...perhaps I should, but I check my stereo compressor release characteristics and so far they have tracked perfectly. Perhaps by grading the other two parameters I end up having well behaved pairs regarding the time constants too. It seems so.
AnalogPackrat: Thanks for the links! Something to study today.
The panel resistance measurements were done with my AP and a step up transformer with one input of AP sensing the voltage applied to the element. Pretty simple and accurate.
I have no idea what kind of matching commercial elements have in them. In a mono unit it is not too essential.
I start with measuring distortion of LDR:s. There are _big_ differences. I use the higher dist ones for metering and sort the rest in a couple of "grades".
Then I measure resistances at couple of points. I have noticed that the R/luminosity curves don't seem to have anything particularly strange in them, so 2 points is enough. Ready compressors track very well.
I haven't measured time constants. Hmmm...perhaps I should, but I check my stereo compressor release characteristics and so far they have tracked perfectly. Perhaps by grading the other two parameters I end up having well behaved pairs regarding the time constants too. It seems so.
AnalogPackrat: Thanks for the links! Something to study today.
Joe Tritschler
Member
Jonte -
Maybe I'm wrong, but as I understand it, any time you drive a capacitive load with a finite amount of available peak current you limit the frequency at which you can get full voltage swing. I've had to deal with this driving electrostatic speaker panels. I did a rough calculation just to see if the original was slew limiting and, assuming you can get the full 14 mA standing current in the 6AQ5 into the (very approx.) 1.8 nF load, I come up with about 8V/us. Assuming that a huge peak might slam the panel with 50Vrms which translates to about 70V pk, I get a full-voltage bandwidth of 18 kHz which is probably fine. I bet the push-pull transistor stage driving the LA-3A's autoformer drives the crap out of the EL panel - wonder if that explains the purported difference in response (besides the different solid-state amplifier sound).
Joe
Maybe I'm wrong, but as I understand it, any time you drive a capacitive load with a finite amount of available peak current you limit the frequency at which you can get full voltage swing. I've had to deal with this driving electrostatic speaker panels. I did a rough calculation just to see if the original was slew limiting and, assuming you can get the full 14 mA standing current in the 6AQ5 into the (very approx.) 1.8 nF load, I come up with about 8V/us. Assuming that a huge peak might slam the panel with 50Vrms which translates to about 70V pk, I get a full-voltage bandwidth of 18 kHz which is probably fine. I bet the push-pull transistor stage driving the LA-3A's autoformer drives the crap out of the EL panel - wonder if that explains the purported difference in response (besides the different solid-state amplifier sound).
Joe
Jonte Knif
Well-known member
Ok, I see. Yes, of course there is slew. My mistake.
If we only had measurements of an original now...
But anyhow, I once measured a deep limiting attack time around 2ms for a T4B copy So you might wonder if lightning fast EL action has any merit.
And if the "real thing" has about same luminance properties as the new stuff (a BIG if) you never need the full voltage swing at high freqs. If your audio signal peak is very fast it obviously has high energy content at high freqs.
So I guess we might really be on the safe side with that 14mA maximum.
If we only had measurements of an original now...
But anyhow, I once measured a deep limiting attack time around 2ms for a T4B copy So you might wonder if lightning fast EL action has any merit.
And if the "real thing" has about same luminance properties as the new stuff (a BIG if) you never need the full voltage swing at high freqs. If your audio signal peak is very fast it obviously has high energy content at high freqs.
So I guess we might really be on the safe side with that 14mA maximum.
Joe Tritschler
Member
You know, I just realized an unexpected parallel between this and the guys designing vari-mu limiters. The Fairchild being a good example of this, isn't the objective with vari-mu side chains to have plenty of peak current capability so you can charge the capacitor with as quick an attack time as possible (at least for disc mastering, not "sound-effect" limiting)? Because that beastly push-pull power amp in the Fairchild limiter is strictly to drive the rectifier and capacitor. Now we're talking about quickly charging an EL panel at high peak voltage, but in this case it seems a moot point if, like you say, other mechanisms (particularly the photocell) have much longer time constants.
I'm thinking WAY too much about all this, aren't I...haha.
I'm thinking WAY too much about all this, aren't I...haha.
AnalogPackrat
Well-known member
Joe--the only time you can "think too much" is when you're being paid to do something for someone else. Even then it's often hard to define "too much" thinking. This is DIY. Think all you want!
The moot point may be the actual necessity or benefit in increasing the attack speed, but it is definitely possible to do it by using a servo-loop (basically a second LDR - either submitted to same light excitation in a dual element or a separate optoisolator receiving the same drive). Obviously, in order to work properly, the drive circuitry must be an order of magnitude stronger. That means driving LED's with 200mAmp peaks, or EL panels with serious high-voltage. I believe Ted Fletcher had a compressor with selectable drive circuits for the opto isolator and he claimed being capable of emulating almost any existing VCA response.Joe Tritschler said:
Joe Tritschler
Member
Wow, now that's interesting. EL panels already have a tendency to wear out over a few thousand hours; wonder how fast they'd wear out with the kind of current pulses it would take to make them much faster. Very clever idea.
So I've been screwing around with SPICE simulations and found some very interesting characteristics of the LA-2A side-chain circuit (man, this thing is a LOT more complicated than I thought).
1) The combination of the HF lift in the cathode of V3 (C8-1 and C8-2), the HF roll-offs of the driver/EL combination and the variable capacitor in the shunt network at the input of V4 (not designated, 50-380 pF variable), and the HF lift caused by C13 which shunts screen-current feedback (amounts to 2-3 dB at 20 kHz; surprised the hell out of me) DOES result in pretty flat HF voltage drive at the EL panel.
2) The "LIM RESP" control (R37), associated capacitor (C12), output blocking cap (C9), and LF shunt network in the plate circuit of V3 (C6 & R30) create a very interesting variable LF characteristic that gives a healthy bass boost when the control is maxed out but cuts QUITE a bit of lower midrange and upper bass frequencies as the control is rotated...check out the response curves for each 10% wiper position. I isolated the HF stuff from the LF stuff...hence the circuit simplification.
The bass boost makes sense with the extra LF drive required for a given gain reduction; but what the hell is that big notch for? The manual for the Requisite Audio opto limiter implies using the control to shape the threshold of gain reduction...I believe it's even on the front panel.
So I've been screwing around with SPICE simulations and found some very interesting characteristics of the LA-2A side-chain circuit (man, this thing is a LOT more complicated than I thought).
1) The combination of the HF lift in the cathode of V3 (C8-1 and C8-2), the HF roll-offs of the driver/EL combination and the variable capacitor in the shunt network at the input of V4 (not designated, 50-380 pF variable), and the HF lift caused by C13 which shunts screen-current feedback (amounts to 2-3 dB at 20 kHz; surprised the hell out of me) DOES result in pretty flat HF voltage drive at the EL panel.
2) The "LIM RESP" control (R37), associated capacitor (C12), output blocking cap (C9), and LF shunt network in the plate circuit of V3 (C6 & R30) create a very interesting variable LF characteristic that gives a healthy bass boost when the control is maxed out but cuts QUITE a bit of lower midrange and upper bass frequencies as the control is rotated...check out the response curves for each 10% wiper position. I isolated the HF stuff from the LF stuff...hence the circuit simplification.
The bass boost makes sense with the extra LF drive required for a given gain reduction; but what the hell is that big notch for? The manual for the Requisite Audio opto limiter implies using the control to shape the threshold of gain reduction...I believe it's even on the front panel.
