LA-2A Theory of Operation - EL Panel Characteristics?

[Joe Tritschler]

Hey folks, sorry to start a new LA-2A topic when there are so many very useful ones here, mostly involving cloning...but this strictly involves theory.  I can't seem to find any voltage vs. current characteristic curves for EL panels and was wondering if they conduct in both directions, or only one.  It seems the electroluminescent panel in an LA-2A is driven with the audio signal directly, and I gather the panel has a threshold voltage above which it conducts and emits light.  But does it start to conduct beyond this threshold in both positive and negative directions?  If not, then does the LA-2A only respond to half the audio peaks?  If this has been discussed before, pointing me in the right direction would be most helpful.  Seems if the EL panel only conducts one way, we should be using precision rectifiers in our LA-2A's.  We didn't study the physical mechanism of electroluminescence here at Wright State University, at least not in any of the EE programs.  You usually see 'em driven with AC, right?

"Crazy" Joe Tritschler

 

[gary o]

Hi the EL panel responds to the AC from side chain amp so it responds to both peaks theres lots of discussion here , Im always interested in the LA2A myself

 

[Joe Tritschler]

Hi Gary,

Do you know how these things are modeled?  Like are they resistive over some dynamic range or what?

Joe

 

[abbey road d enfer]

These EL panels are very hard to model because they are highly non-linear with serious hysteresis effect. Basically, when submitted to a high voltage through a resistor, nothing happens until reaching about a breakdown voltage of about 100V (dependant on gas composition, electrodes, geometry,...) then it suddenly goes into conduction, dropping the voltage abruptly to 70V, where the ionization ceases; the voltage ramps up until reaching the breakdown voltage and the cycle starts again. This is a relaxation oscillator, the frequency of which depends mainly on the capacitance of the EL element and the series resistor. With an AC signal, this happens at every half-wave. When you see a neon bulb, the light it emits is a 60 (or 50) Hz modulated HF.
Modelling an EL element involves a negative-resistance.

 

[Joe Tritschler]

Well, funny you should mention negative resistance because the only thing I could find in text was that EL panels don't have a negative-resistance characteristic like neon or fluorescent lamps and therefore can be used in nightlights and so forth without any external resistance or ballasting.  The breakdown mechanism you just described sounds just like a gas-discharge or glow tube but I'm not sure it applies to EL panels.

Incidentally, I just realized this post should probably be in the "Drawing Board" forum so please feel free to move it at any time.

Joe

 

[gary o]

Im always interested in La2a related stuff sadly I dont understand most of the theory kinda stuff so i have done lots of crude experiments , Ive tried different El panels Light bulbs neons leds at one stage i made a box that had several different light sources & several different cells vactrols but to my ears at least as long as the lights attack is quicker the the cells attack & the EL and LEDs seem to be then it sounds the same to me .....Iv only tested vocals by the way....I find LEDs more sensitive in general ie switch on with less volts I suppose, I tried single LED and 2 LEDs in antiparalel ...i found much wider range of sounds as it were with different cells.

 

[Kit]

Basically, when submitted to a high voltage through a resistor, nothing happens until reaching about a breakdown voltage of about 100V (dependant on gas composition, electrodes, geometry,...) then it suddenly goes into conduction,

Maybe thats true for the "old style" EL panel. These modern japanese panels have a much lower threshold, about 20-30V. I used them in a opto comp I recently built.

Also, an interesting find was that these panels does not have a timeconstant or "afterglow" that some people claim that the T4B has. They are faster than a speeding bullet.

 

[abbey road d enfer]

Well, funny you should mention negative resistance because the only thing I could find in text was that EL panels don't have a negative-resistance characteristic like neon or fluorescent lamps and therefore can be used in nightlights and so forth without any external resistance or ballasting.  The breakdown mechanism you just described sounds just like a gas-discharge or glow tube but I'm not sure it applies to EL panels.

You're absolutely right; I was thinking of the gas-discharge mechanism. Yes, these "new" EL panels seem to have a much gentler I/V curve. Googling the subject does not give much. Researching semi-conductor for EL panel drivers gives some results, they all seem to indicate 80-90V ac as a requirement.

 

[abbey road d enfer]

Basically, when submitted to a high voltage through a resistor, nothing happens until reaching about a breakdown voltage of about 100V (dependant on gas composition, electrodes, geometry,...) then it suddenly goes into conduction,

Maybe thats true for the "old style" EL panel. These modern japanese panels have a much lower threshold, about 20-30V. I used them in a opto comp I recently built.

Also, an interesting find was that these panels does not have a timeconstant or "afterglow" that some people claim that the T4B has. They are faster than a speeding bullet.

As I said in the previous post, I've googled for info. All i got researching EL panel drivers seem to indicate 80-90V ac as a requirement. So it looks like yours are a different animal. What is the brand and model? Since you have them, could you make some basic measurement (I/V curve, DC behaviour)?

 

[Joe Tritschler]

Hey Abbey  -

I'm seeing a few folks calling these Light Emitting Capacitors - first I've heard of this.

http://www.e-lite.com/about-el/
http://en.wikipedia.org/wiki/Light_emitting_capacitor

Crazy Joe

 

[abbey road d enfer]

Joe, thanks for the links. Since the load is largely capacitive, it seems that the detector would have a natural pre-emphasis, making it compressing high frequencies more than lows. I've also seen a mention that the color depends on the frequency, so due to the varying sensitivity of the photoresistor with wavelength, there is an added frequency-dependant aspect. It would be more or less tempered by the feedback configuration.

 

[Jonte Knif]

Yes, the luminance is strongly dependent on frequency. I remember having the peak at several kHz, perhaps even 10k.

Therefore LA2A needs some compensation network in SC. But that is not the whole story. It actually needs less pre-emphasis in treble because the freq dependency is not as strong as simple capacitor model would give. This empahis is done with a simple low value cathode bypass cap in the original in one of the triodes in SC. Basic pre-emphasis simply comes from the pentode driver.

The illumination vs. freq curve is different at different voltages. You tell me why, I have no idea. The element gets progressively more efficient at ca. 7 kHz when level is increased. This is easily measured from the compressor so the feedback really doesn't tamper all these effects. And yes, the color changes _a lot_ with both freq and level.

About modeling I know nothing particular. I designed my side chain assuming that the element is simply capacitive. Worked just fine and gave predictable results. Perhaps it could be used as a crude model at low levels?

-Jonte

 

[Joe Tritschler]

Abbey - when you postulate that there is a "natural pre-emphasis," making it compress high frequencies more, are you suggesting that the capacitive load draws more AC current at high frequencies and therefore gives higher luminous intensity, resulting in more gain reduction?  Because it seems Jonte is suggesting "pre-emphasis simply comes from the pentode driver," which I would think would tend to make the HF current constant down to the frequency where the 6AQ5 driver's 10k output resistance is comparable to the capacitive reactance of the panel - in  other words, a falling voltage response above this frequency but flat current response.  So I guess this begs the question - Jonte, if you ignore the non-linear frequency dependence with level (which is VERY interesting), did you find the gain reduction being constant with frequency with constant current applied to the panel or constant voltage?  Or neither?  Is the HF lift in the cathode of the SC 12AX7 to compensate for falling voltage response due to the capacitive load, or panel/detector non-linearities when driven with fairly constant current?  I guess if I knew what the low-level capacitance was, it would be easy to at least determine what that 10k driver output impedance is doing...

Crazy Joe

 

[abbey road d enfer]

Abbey - when you postulate that there is a "natural pre-emphasis," making it compress high frequencies more, are you suggesting that the capacitive load draws more AC current at high frequencies and therefore gives higher luminous intensity, resulting in more gain reduction? 

That would be probably the case if using voltage-drive. Current-drive should make the process more or less constant. In-between, i.e. voltage-drive with a non-zero resistance, there's likely to be some kind of intermediate pre-emphasis. The $64 question: is the light flux proportional to current or apparent power (V.I)?

Because it seems Jonte is suggesting "pre-emphasis simply comes from the pentode driver," which I would think would tend to make the HF current constant down to the frequency where the 6AQ5 driver's 10k output resistance is comparable to the capacitive reactance of the panel - in  other words, a falling voltage response above this frequency but flat current response. 

Yes, pentode-drive certainly adheres to the definition of voltage-drive with a non-zero resistance (or current-drive with a non-infinite resistance, according to Thevenin).

So I guess this begs the question - Jonte, if you ignore the non-linear frequency dependence with level (which is VERY interesting), did you find the gain reduction being constant with frequency with constant current applied to the panel or constant voltage?   Or neither?  Is the HF lift in the cathode of the SC 12AX7 to compensate for falling voltage response due to the capacitive load, or panel/detector non-linearities when driven with fairly constant current?   I guess if I knew what the low-level capacitance was, it would be easy to at least determine what that 10k driver output impedance is doing...

Crazy Joe

yes; experiments!

 

[Jonte Knif]

Ooops,

forget about half of what I wrote. I just found some of my measurements and SC-simulations. It was years ago when I run the experiments. Sorry.

Measured "cold" C was about 1,8 nF in my own element. That is not very far from T4B if I recall. Perhaps more. That leads to a corner frequency of about 8 kKz from 10k driver. I compensated this with a couple of dB:s rise in SC response. Pretty much unnecessary because compression curves seem to indicate clear rise in luminosity at high freq.

I remember driving the element from 10k impedance and checking voltage response and seeing corner freq at around the right value, so the "hot" C is pretty much the "cold" C too.

So, basically the capacitance is so low that 10k anode resistor produces pretty much constant voltage drive to the element. I have a family of compression curves in front of me and the non linear luminosity vs voltage vs freq seems to produce pretty simple slope, i.e.. if I have pretty straight line at -3dB compression, at -12 dB it is sloping towards treble at about 2dB per decade.

Jonte, if you ignore the non-linear frequency dependence with level (which is VERY interesting), did you find the gain reduction being constant with frequency with constant current applied to the panel or constant voltage?  Or neither?

Well, it this question just can not be answered without determining certain level because it is level dependent. It seems that at low levels it is more voltage dependent and at high levels more current dependent. My head spins.

Is the HF lift in the cathode of the SC 12AX7 to compensate for falling voltage response due to the capacitive load, or panel/detector non-linearities when driven with fairly constant current?

The first one. But it seems to overcompensate a bit after all.

I kind of got interested in the topic again. Perhaps I could measure something tomorrow. Skip the compressor and measure just the element. I obviously have some gaps in my memory and understanding. The work I did way back then was based on the assumption that the panel _basically_ is to be driven with constant voltage, but that is not quite the case. (And add that level dependence to the mess...) Obviously if the EL material is basically a cap, then the amount of electrons bombarding the luminescent material will be current and not voltage dependent. How on earth did I get such weird results. I'll try again tomorrow.

 

[Joe Tritschler]

Please report, this is fantastic stuff guys!  I'm waiting on two sets of T4B guts from Anthony DeMaria, so you can bet I'll be conducting my own experiments, too.  Honestly, I have great respect for the old guys and the LA-2A was undoubtedly very carefully and thoughtfully designed around components available at the time.  I always enjoy tinkering, though.

Crazy Joe 

 

[Jonte Knif]

Here you are. Constant voltage curves.  Figures on the right are EL drive RMS voltages. Values are 2 dB apart.
It might not be obvious from the first look, but the slope really changes a lot with different drive levels. Look at the difference between 100Hz and 10kHz.
10kHz-40kHz behaves quite differently at different levels, interesting but not important in real life.

Capacitance is fairly linear. There is some increase at higher levels. My element has 1,6 nF at very low levels and 1,9 nF at the hottest one ever needs. Measurement freq was around 8kHz. Again, interesting but not important. Capacitance also increases a bit towards higher freqs, but not a lot. There are no visible anomalies anywhere, just a bit tilted I/V curve (line in Log/Log graph)

Constant current drive gives extreme drop in luminance towards high freqs. Not usable. It didn't even make sense to draw a graph of it.

So the answer to the question:

if you ignore the non-linear frequency dependence with level (which is VERY interesting), did you find the gain reduction being constant with frequency with constant current applied to the panel or constant voltage?

obviously is "neither" but constant voltage makes some sense (and works in a feed back design) but constant current absolutely not.

Well, that's todays results. I hope I didn't do any mistakes. Measuring the resistance curves is a bit frustrating because of long memory effects. I hope someone could explain the physics behind the results. Also remember that these results were measured from my own "T4B" which uses modern materials. I don't have an original or a copy available now.

 

[Joe Tritschler]

Jonte -

All I can say is WOW.  So, if for a given AC drive level, high-frequency signals tend to pull the resistance down more than low-frequency signals, even though it's only a couple dB at low levels, why do you suppose they didn't boost the low-end a little bit in the side-chain?  In fact, the "LIM RESP" control would do the opposite.  More musical un-equalized?  Loud bass notes don't pull the whole mix down, but sharp transients get limited?  Your graph really explains a lot about the very complex way these things process audio.  I realize that in the context of a feedback limiter, the net result may not be so dramatic.  But especially useful to me is knowing the impedance of the EL panel and how it likes to be driven. Thanks for sharing your hard work with us.  I would kind of like to see what a series of output voltage curves looks like vs. frequency with a set of constant input voltages separated by a couple dB for the actual attenuator output in a feedback arrangement and with a true constant-voltage driver.   Sounds like a good experiment for when I get my optos from ADL...

Many thanks Jonte,

Joe

 

[PRR]

Yes, WOW. I've done this (and with a pencil), many-many years ago; this is tedious. Thanks for the update.

> for the actual attenuator output in a feedback arrangement

With few ideal assumptions, you can read it right off Jonte's plot. I just tore out a kitchen so I was a bit sloppy but here's my hasty read and you can refine my calculations.

I took just three frequencies. While the curves look steep, they don't have any wild bends or kinks, so the response at other frequencies will be on the same trend as 100Hz, 1KHz, 10KHz.

Attenuation is a simple function of resistor and LDR. Again the curves are smooth enough we don't have to plot a lot. I forget what the LA-2a's attenuator is. There's a 68K and some other stuff? I took 30K as the equivalent fixed resistance. Then thinking very fast, LDR>30K does little, LDR=30K gives significant cut, LDR~~3K gives ~~20dB cut which is heavy limiting. I plotted this, then realized somebody would claim limiting starts at less than 6dB, added 120K as a ~~2dB cut.

Assume a "flat" amplifier in feedback topology. Assume same near-zero output impedance as Jonte's test rig.

Then the EL voltage is the output voltage. We see that 2dB limiting happens with 18V to 14V output, which is pretty darn smooth. If we wanted dead-flat, it would be easy to compensate. 20dB limiting needs 47V to 23V, a larger range. We could compensate that easily. I took -6dB at 1KHz as "zero dB" and noted the relative output levels for other points.

We can't easily get "perfect flatness" for both -2dB and -20dB limiting: one needs 2dB correction and the other needs 6dB correction. I don't think speech/music limiters really need to be same-threshold at all limiting depths.

The rise at 1KHz is 4dB into 1.5dB (2.7:1) and 18dB into 5.5dB (3.3:1). The ratio is looser for 100Hz: 18:8 and tighter for 10KHz: 18:4.

We must find out how the EL and LDR "average" complex audio signals; but life may be too short to know this. "Sounds good" is the best test.

 

[CJ]

you can measure the EL panel to death, but as stated above, the compressing is in the LDR.

so any quirks in the panel get averaged out by that variable resistor.

the panel is very fast, but the LDR is very Slooooooooooooowwwwwwwww,

on the way down, at least.  up time is fast with the first hit, then it slows a bit.