LA-2A, THD rise with compression action

Hello

Excuse me my wrong English, Is the first time that I Have wrote in this forum and, of course, the first time that I write in English language forum.
I have finished a LA-2A replica with amazing and warm sound. They have some of the following characteristics:

Cell:
LDR : NSL5910 (two for audio path and one for VU-meter)
Panel, Farnell, colour aqua.

Input transformer: SOWTER 4383 (1:4)
Output transformer: Home made audio transformer (2,5:1)
R7 = 1K (more resistance reveal “negative compression” in Limiter position.
6AQ5 is substitute with 6V6.
Is there another change without importance now.

Now I am proceeding at the specific and precise audio measurements.
I measure the output THD, first without compression for different output levels, (with a low distortion audio generator and ARTA computer Software).

Output    THD (%)
+10dB    0,12%
+15%    0,23%
+20dB    0,42%
23,2dB    1%
24dB      Clip

I thing are good characteristics, but the problem is when the compression is present.

THE PROBLEM

When the compression is present, the THD is rising, until very high values, more than 3 or 4%...... I have seeing the THD is rising until the compression arrive at -3dB (of compression), with more compression the THD is come down until 1% more or less. I DON’T UNDERSTANT ANYTHING, because the LDR action is pure resistive….

Any body have an idea or can help me?

Thanks

Miquel

This is present in all dynamics processors.

Even a perfect gain element will cause distortion to the audio path because the clean dry signal is modulated by the side chain control signal that isn't a steady DC.

The faster/shorter the time constants in the control path, the higher the distortion. You can visualize what is going on for a very low frequency sinewave. While the sine waveform edge is rising, it will reduce gain, and during the falling part of the waveform it will restore gain. This changing gain across a single cycle of the waveform causes the distortion. 

You should measure lower distortion with higher frequency sine waves for a given time constant speed.

JR

Thanks for your answer

I did not say the analysis was at 1.000 Hz.

I know this problem. I have worked with mu variable compressors and this problem is clear with very fast attack time and low frequencies, for example drums and bass. But I has not thought never that occour in a no fast optocompressors and at frequencies than 1.000 Hz. (Or perhaps the optocompressors are more fast than I think...?)

I has not measure the attack time but i think is slow than a mu variable compressors (in them it is a simple mathematical calculation RC...). Tomorrow will continue experimenting by varying the frequency, but I think there are another mystery.

Miquel

If you're talking about mainly 3rd harmonic, it's something which happens in the CdS photocells. Bear in mind that the photocells are nonlinear in their 'release' characteristics.

I DON’T UNDERSTANT ANYTHING, because the LDR action is pure resistive….

Click to expand...

Firstly, the opto cells are NOT purely resistive, they have a residual nonlinearity which sometimes accounts for some small (0.1% or so) distortion even with NO GR...

More importantly, however, they are a VARIABLE resistance, and they modulate their resistance according to stimulus. The pulsed stimulus from the sidechain, -coupled with the nonlinear resistance over time- produces distortion.

Keith

Also... the attack time of an LA-2a is NOT slow... -Not a bit of it.

It's lightning fast.

Keith

Very thanks for the answer (and excuse my cryptically English)

I suppose, in my last post there is a mistake. I think the “modulation” of the sidechain it is due to the excessive fast release, not for the fast attack, I suppose, if the release time constant is long, the low frequencies (for example 100 Hz) can’t “modulate” the gain. But I am thinking about it now……

I think the LDR release is complex (“program dependent” in mu variable compressors language), and I don’t know if there is an initial fast release time, and it can “modulate” the sidechain in low frequencies, but I think 1.000 Hz is very high frequency. (I use 1.000Hz for measurements because I have a very low distortion generator in that only frequency. My variable frequency generator its cheap and show excessive THD for accurate measurements, but I can utilize it, of course).

I had not thought about the non linearity of the LDR, but I am surprise because I have read that the Optocompresors were born to avoid the THD generate in mu variable tubes (in the mu variable compressors….)

RESUME

But, …nobody has observed this problem?, (THD rise at rise the compression). Is it only my problem?.

I am sure the circuit work well and sounds amazing. Only the NSL5910 is not the Teletronix original, but I think is there a lot of people that use it.

I read the manuals of different compressors and any says nothing of that problem. Only the mu variable Faichild offers a graphic of THD evolution at rise the compression.

This week I do more measures and I communicate it.

Miquel

miquel said:

Very thanks for the answer (and excuse my cryptically English)

I suppose, in my last post there is a mistake. I think the “modulation” of the sidechain it is due to the excessive fast release, not for the fast attack,

Click to expand...

The modulation of the waveform comes from both fast attack AND fast release.  Either will permit distortion. It doesn't matter which direction you look at the waveform along the time axis, a sine wave is symmetrical, and bending the top part is a distortion whether it happens on the leading or trailing edge.

I have observed this many times. All LA-2a's do it. So does the LA-3. It's part of the nature of the unit.

Remove the T4 and it doesn't happen... by all means plug in another T4 using differently-sourced components, and you'll still observe the same characteristic peak in distortion at a certain level of GR.

Keith

Keith

Thanks for the answer

I am testing with 1.1 input transformer.

I made some measurements, first with a SOWTER 1:4 input transformer, and second with a 1:1 Lundahl 1545A transformer in high impedance connection. I measure the THD with different compressions levels and for different Input levels, (every time at 1.000Hz).

1:4 transformer

- Input +15dB
Atten 0dB  THD 0,54%
Atten 5dB  THD 4,20%
Atten 10dB  THD 2,10%
Atten 15dB  THD 1,20%
Maximum THD = 4,40% at 2,5 dB attenuation

- Input +20dB
Intolerable THD around 5dB compression.


1:1 transformer

- Input +15dB
Atten 0dB  THD 0,26%
Atten 5dB  THD 0,97%
Atten 10dB  THD 1,37%
Atten 15dB  THD 1,33%
Maximum distortion for 3rt. Harmonic al 1,5dB attenuation

- Input +20dB
Atten 0dB  THD 0,47%
Atten 5dB  THD 1,74%
Atten 10dB  THD 1,19%
Atten 15dB  THD 1,34%
Maximum THD = 1,92% at 1,8 dB attenuation

CONCLUSSIONS

For high level line signals, the numerical results are much much better for the 1:1 transformer than 1:4 transformer.

I don’t understand the possible behabiour with the UTC original 1:10 transformer..., my God!....

In the operative and real studio work, the difference is that with 1:4 transformer the normal gain knob position is 20/30 (more o less) for example, and with 1:1 transformer you need work at 50/60 number (only magnitude algorithms…)

I am very perplexed. I think the people don't measure the THD in compression, or no divulgate the results. The 1:1 transformer is, in tests, better, but the knob position is very different to LA-2A typical.

Is there anybody with experiece in input transformers for the LA-2A?

Miquel

I don't get what you're trying to achieve there.

the 1:4 sowter input is rated to handle +16 dBu max and even then it's going to produce some harmonics. Obviously it's going to clip like mad when driven +20 dBu.

Also,
the THD measurements with attenuation/compression tell us only that it's compressing. Audio is successfully going through the sidechain amp. You can not judge anything else based on that number. If you want to know what's really happening you should be running test audio with squarewaves and sines and looking at the shapes and speeds of the compression action in oscilloscope or even better, some wave editor. Compression in a way is harmonic distortion. We are changing the shape of the waveform. Just often with program dependency, whether frequency or level dependent (or both like in LA2A), and slower rate of change than with plain clipping.

Look at this site for some pictures of LA2A compression action: http://www.igsaudio.pl/index.php?option=com_content&task=view&id=51&lang=english

The Lundahl you have at 1:1 is clearly rated to handle + 22 dBu and your test results reflect that perfectly. Again the THD measurements when the unit is compressing are quite useless.

miquel said:

The 1:1 transformer is, in tests, better, but the knob position is very different to LA-2A typical.

Click to expand...

That because you are starving the sidechain amp. The make up amp has plenty of clean gain available (something like 45dB), but you still need to input healthy levels to the sidechain, which gets it's input before the make up gain amp attenuator, in fact directly from the input transformer.

I don’t understand the possible behabiour with the UTC original 1:10 transformer..., my God!....

Click to expand...

It is 20 dB of gain unto itself, so +20 in equals +40 at the T4 and volume control.  That's 10 watts of power. 

Expected input levels are not equivalent to modern full scale DA converter output, it expects about 20 dB less. 

This looks like the typical misunderstanding people have about tube equipment and high ratio (gain or loss) transformers in fixed gain systems.  Understand that the input transformer is a +20 dB fixed gain system, followed by the T4 and a passive attenuator, then another +20 dB fixed gain system.  It's up to you to adjust to it's parameters. 

Even with a 1:1 input, feeding it +15-20 is excessive, and outside of the intended usage.  Attenuator fully open would add another 20 dB to that; we're back at 10 watts. 

The side chain has more gain than the signal amp, so it's abuse is even greater. 

I am not familiar with voltage coefficients associated with CDS cells. There can be a distortion mechanism at higher power and very low frequency, if the resistance changes with heating within a single cycle, but there will always be distortion caused by any gain change occurring within a single cycle of the waveform.

This distortion is not caused by fast release or fast attack alone, but the product of both. If you have a very slow attack, or a very slow release the amount of gain change within one cycle will be small. A not slow attack combined with a not slow release will both try to attack and release within the same cycle of the waveform and generate distortion.

This can also impacted by the detection scheme used in the side chain. Peak detection will be more waveshape sensitive than average or RMS detection.  Advanced side chain techniques add hold circuitry and adaptive rate circuitry that slows down the side chain and control voltage generated distortion for small level changes.

JR

Very thanks for all the answers, you are the best.

Now I have a lot of thinks in my brain. I saw, in the first communication, one of the possible problem is my bad English explanations… agggh!. Now I need arrange my ideas and explain point by point.

Topic: Dynamics.

When you speak of dynamics and show dynamics graphs, but I don’t have dynamics problems, I made dynamics EXCEL graphics for all peak reduction knob numbers (x dB input, y dB output). With that analysis I have observed that I need adjust R3 in a precise point for a precise, progressive and good compression. Remember that I have two NSL5910 in parallel in audio path. Another cells has another light and black resistance, and, of course, another adjustments.

In the future I am going to measure the time attack and time recovery with oscilloscope dataloger. But there will be another possible problem… or no…

My “problem” is with continuous wave, not impulsions waves. I saw: when rise the compression, for high levels and CONTINUOUS WAVE, rise the output Total Harmonic Distortion. Without compression the general THD comportment is good..


Topic: saturation transformer.

OK and thanks for the information. I give you the THD for both transformers at high levels, of course, without compression.

+15dB input, system gain 0dB, (+15 dB output)
Sowter = 0,54%
Lundahl = 0,26%

+20dB input, id.
Sowter = 0,70%
Lundahl = 0,47%

The transformer effect is clear but not dramatically, (remember, the problem is THD rise when rise the compression and is more important than that numbers.


Topic Cells CdSe and AC non linearity.

I am perplexed, because the LA-2A and similar compressors (remember I have done a reproduction with a little changes in the cells) are considered a very good devices. I think my problem is to do too much analysis….


I think, there are two problems.

1.- Little problem. The Sowter 1:4 transformer is little, but for my, is not problem because de THD at +20dB (for all the system) is 0,70%... not bad, (and 0,05% for 0dB)

MAIN PROBLEM HIPOTESIS

2.- I think, with 1:4 transformer, and with the modern studio line levels, there is too much audio voltage in the LDR circuitry and input general circuitry. But I don’t understand that effect in de deformation of sine wave waveform. I suppose this theory because when compression is present, (and with 1:4 transformer), the notable THD appears with +10dB input, at +15 input is very big, and at +20dB is intolerable, and with a 1:1 transformer (12dB less voltage AC in the circuitry) the problem appears, but only in the proximity of +20dB input.

Very thanks for all, you are very big.

I try to show a photo of the prototype (is not, of course the definitive). It’s only an analysis board.

http://s749.photobucket.com/albums/xx138/miquel_bucket/?action=view&current=Prototip0.jpg&newest=1

(I don't know if I have been good the opperation)

Miquel

I don't think you would ever put +20dB level into an LA-2.

It has 1:50(?) or 34dB gain. The output really isn't good for over +10dBm or so. The level at the first grid should not be over -24dBu or 0.05V. If gain is full-up, then 0.05V on the photo resistor. At 20dB limiting, 0.5V into the fixed resistor which feed the photo resistor. Closer to zero dB levels, not +20!

Of course everything before the gain pot will strain if you turn the gain pot way down. Leave it open. Reduce whatever is going in. There's no call for +20dB on a regular basis.

As others have said, your input transformer may be the main THD. You may check this by bypassing it and going in with an unbalanced signal.

> LA-2A .... are considered a very good devices.

The maximum "clean" level, and thus the signal to hiss ratio, is better than in several other techniques (diode-bridges, JFETs). The control path is utterly isolated from the audio, avoiding the complications (and expense) of working both signal and control in the same device path or on common pins.

The photo-resistor DOES distort. It is NOT a "perfect resistor". It is magic crystals which are sensitive to energy. They are used for their sensitivity to light energy. But electric field is another kind of energy, and it does affect them. The resistance changes. At much higher values (over 100V), even in full dark, the crystals will turn-on and short out. At our voltages the change is "small", but THD is sensitive to small changes.

I don't have numbers handy, but IIRC the distortion is "very low" for under 0.1V and "significant" at several volts. I know I did not see (on 'scope) any bent waves with a 200mV photo-limiter, but when I worked one at 3V limiting level the sound was muddled (but not harsh like power amps clipping).

Since LA-2 has a gain pot after the photo resistor, you "may" work it at full-gain 0.05V or choke-down so it has 0.5V or 5V at limiting level. If you want "clean", you must work it full-gain and get some control of what is going in.

BTW, under 1% THD below limiting and "several" % THD in limiting would have been "good" for 1963, especially at the LA-2's price (MUCH less than the G.E. many-tube limiter). AM radio or phono-disc at high modulation has THD rising from 2% to 20% very quickly. The limiter would hold this down to mostly 2%. If not constantly limiting, the sound would not be ugly.

Very thanks for the LDR non linearity explanations, Is a big surprise for my, I thought they was a perfect resistance… my God…!.

In really, I read (I don’t remember where), the optocompressors born to avoid the mu variable tubes THD….!). I have built mu variable stereo compressors (with ECC189 and semiconductors) with very good THD measurements, better than expected.

I am not a studio operator, I am a technician, but I have seen an operator compressing a bas line, in high level, with a LA-2A reissue (remember, that unity is possible has original 1:10 input transformer, but I think no). The gain knob was a very low possition (evidently...), number 15 for example.

But…..if is really that the compression action generate THD. ¿Is that possible that some people use the LA-2a as just a processor, not strictly a compressor?, and… that THD is good for that user….?.

¿Is a too revolutionary theory?.

Ah, excuse me my wrong English, I am Shamed.

Miquel