Why bother?

[Fancy Lime]

..I believe it uses a GE miniature lamp #19.

I remember that wobbling preamp. If that is indeed the bulb, 100mA is a lot for a small pot to control. (LEDs were not even dreamed of yet.)

My entry for nitpick of the month: ACTUALLY, the LED predates the BJT by over four decades in terms of prototypes. Dreams of LEDs are as old as the triode tube. But yeah, practically useful devices... I still remember when digital alarm clocks with LED displays were the hot $#!+ and I only had one of them flippy panel things, which I now wish I had kept because how cool were those!

Total range in resistance from tens of Megaohms to a few kiloohms may indeed be the thinking behind the LDR. Although to me, the problem of not getting to near-zero resistance would by far outweigh the advantage of an unnecessarily large Rmax. With an R(c) of 22k on Q2, I doubt that an Rmax of 100k vs 10M would make an audible difference, given that this whole thing is probably not the most hifi amp ever. A few Ohm vs a few kOhm on the other hand may make a real difference here. Of course, the "hard clipping end" of the spectrum may not have been desired, so an Rmin of a few k mat have been just what they wanted. I think we'll need to find out who designed this and track down their descendants to sift through old notes and diary entries to really get into the mindset of the designer. If we feed this extensive research into an advanced AI, it may be able to tell us what the hell they were thinking when they designed this amp.

Andy

[Rob Strand]

Mark's beef is why not just use a pot *in series with the diodes*.  I don't know, it's pretty weird.   

The whole set-up is asking for tolerance related variations from unit to unit.  Even if they wanted a funky taper you would think the part tolerances would screw that up a lot more than just a pot with perhaps a parallel resistor added.

[digi2t]

I have one. 2x12 version. I've had it since I was 16. I got my first guitar, a Goya strat at 14, had to work for over a year to afford an amp. Anyway, I can confirm that yes, the Harmonic Multiplier part of the amp is truly useless. Soft clipping is a misnomer. It sounds more like there's an opamp in there that's about to poop the bed. It's always stayed at zero. For a while there I thought that maybe the lamp or LDR had gotten weak over the years, and maybe the clipping supposed to be... better? Anyway, I just left it alone. The clean however, is wonderful. Maybe not as rich (and loud) as my Ampeg G212, or as sparkly as a Jazz Chorus, but very respectful. I still have the original speakers, but I replaced them with a pair of Texas Heats.

[Rob Strand]

Soft clipping is a misnomer. It sounds more like there's an opamp in there that's about to poop the bed. It's always stayed at zero.

Thinking out allowed: I guess what happens is when the clipper is set to hard (ie. LDR low impedance) the diodes clip the output of the Q2 transistor stage but at the same time limit the swing on the collector of Q2 is resitricted.   When you set it to soft the output of Q2 can swing more and crash harder against the rails.  It could end up sounding harsher and that would undermine the softness, well unless your input is low enough to prevent Q2 clipping.   (I don't agree series resistors soften diodes anyway.)

A more obvious omission is any low-pass filtering after the clipper.

[soggybag]

The drawing has "NOTE [1]" is that annotated somewhere what does it say?

[Rob Strand]

The drawing has "NOTE [1]" is that annotated somewhere what does it say?

It's just below and to the right of that annotation:

"[1] To be covered while being serviced."

All that are saying is if the ambient light gets on the LDR you will get inconsistent results.

I'm not sure how the lamp is physically coupled to the LDR.  Since they are using a lamp (which gets warm) they might not seal-up the lamp and the LDR to stop the ambient light getting in.  You might see coupling tubes, covers, or even see a neon lamp (or LED) and LDR taped together.

[PRR]

Mark's beef is why not just use a pot *in series with the diodes*. ....

What rating of pot would you specify to dim a 1.4 WATT incandescent?

Remember the worst-case is not half-voltage as it would be for resistor load; the lamp current falls off slower than voltage.

[Rob Strand]

Mark's beef is why not just use a pot *in series with the diodes*. ....

What rating of pot would you specify to dim a 1.4 WATT incandescent?

Remember the worst-case is not half-voltage as it would be for resistor load; the lamp current falls off slower than voltage.

It is possible to work out but the solution isn't convenient for hand calculations.   More of a problem for pots is estimating the power dissipation when the wiper *isn't on full*.   If you set the pot to 1 (on a 0 to 10 scale) the
the power has to be dissipated in 1/10 the physical region of the pot.    For common pots such info is rarely given.

However, Mark's beef is why even use a lamp.  Just put a low power pot in series with the diodes.
no lamp, no power considerations, does the same job.

Just as an example we can use the crappiest lamp model of VL = k * IL^2 and workout the worst case pot dissipation.

Lamp:  1.4W @ 12V we have IL_nom = 117mA and the lamp resistance is RL_nom = VL/IL = 103 ohm, also k = 855.

Suppose now we put a resistance Rs in series with the lamp.
We know that VDC = I * Rs + VL = I Rs + k I^2,
which we can write as k I^2 + I Rs - VDC = 0  ; which is solvable by hand as a quadratic.
For each Rs setting we can calculate I.  Once that is done we
can calculate P_lamp,  P_Rs, Vlamp etc.

Going further we can find which Rs gives the maximum dissipation in the resistor P_Rs.

Going through the motions the maximum power dissipation in Rs occurs when Rs = 2*RL.
So 2/3rd of VDC is across Rs (pot) and 1/3rd of VDC is across RL (the lamp).
When I calculated it out I got:

Rs = 118 ohm,    RL = 59 ohm
V_Rs = 8V,          VL=4V
I = 0.0673 A
P_Rs = 540mW (=2*PL)    PL = 270mW

The current is a bit over half the nominal lamp rating, and the resistance Rs is a little over the nominal lamp resistance

If we use more accurate lamp models, we use the same approach but the calculations need to be done
numerically on a computer.

I may have stuffed up but it looks ball-park.

[Mark Hammer]

I didn't have a "beef", as much as a bewilderment at why such a complicated approach was being used when, as far as I could tell, there was a much simpler way of doing it.  However, as Gus's post clearly implied, finding a pot that would accomplish,on its own, what the optical arrangement does, would be tough slogging.  So, I'm much less bewildered than before.

[puretube]

The drawing has "NOTE [1]" is that annotated somewhere what does it say?

Go to +400% view here:
https://el34world.com/charts/Schematics/files/Gibson/Gibson_g_105.pdf

Maybe they just took over the LDR-lighting from an older Tube-Amp Tremolo circuit (GA-79RVT),
and replaced the tubes with transistors ... ?
https://www.schematicsunlimited.com/g/gibson/gibson-ga77rvt-crestline-schematic

[Rob Strand]

However, as Gus's post clearly implied, finding a pot that would accomplish,on its own, what the optical arrangement does, would be tough slogging.  So, I'm much less bewildered than before.

I can see his point.  A wide range of reisistance would allow the clipper to be switched out. 

However, I'm not 100% convinced it can't be done with a pot.
Unfortunately it's one of those devils in the details problems.
It doesn't help we have no info on the lamp or LDR.  (is the lamp even 12V?)

If you look at the original circuit the lamp driver (Q3 etc) is essentially applies a variable
*voltage* to the lamp.
https://postlmg.cc/R3bNWSBB

Here's where I'm not 100% convinced of the extreme wide range of resistance:  The lamp is restricted to about 1.1V to 4.3V. The resistor R7 sets the lower voltage limit and the resistor R8 sets the upper voltage limit.    The circuit deliberately prevents the lamp voltage being set to zero.   It's not a small voltage range but it doesn't go to zero to switch it out.

Here's where the devils in the details comes in:   Where I see the possibility of LDR resistance range being high is LDRs respond to light level (lux) and we can see the lumen output from an incandescent bulb is a strong function of the voltage.  So maybe we don't need to go to zero volts.



It depends on the range of voltage vs the rated lamp voltage.
Nonetheless you can see the possibility of getting wide variations.

[Vivek]

Study of Total Harmonic Distortion versus Compliance Resistor in a Hard clipper

Spice Schematic :
Assumptions:
6Vp signal source with 2K internal resistance


Transient Analysis with Compliance resistor stepped 0.00001, 10, 50, 100, 200, 500, 1K,  2K,  4K, 10k, 50K, 100k, 200k, 400K,2Megs, 5Megs ohms


Total Harmonic Distortion Table:

Code Select Expand


Rc      THD
0.00001 35.15%
10 33.88%
50 29.46%
100 25.37%
200 19.83%
500 11.91%
1000 7.08%
2000 3.86%
4000 1.98%
10000 0.77%
50000 0.14%
100000 0.07%
200000 0.03%
400000 0.02%
2000000 0.01%
5000000 0.01%


Total Harmonic Distortion graph:


Observation: This is plotted for 6Vp inputs, and that was meant to represent a guitar signal of max 1Vp going through a gain stage of 6x. Suppose guitar signal is normally 300mvp after the initial attack, the THD for most of the guitar signal will be much lower than the THD at the attack.

[Vivek]

Does the human ear respond linearly or logarithmically to THD ?

Or is it a very complex function similar to the well known Fletcher- Munson and Dunning–Kruger effects ?

[amptramp]

When you are mentioning light output for control of CdS or CdSe photoresistive cells, you should be aware that CdS has a wavelength peak sensitivity of 640 nm and CdSe has a wavelength peak of 790 nm so what we see as visible light does not necessarily correspond to the cells are most sensitive to.  There are some mixed CdS - CdSe cells that have peak response between these wavelengths.  The time response of an incandescent lamp is faster the greater the filament temperature, so the added efficiency of a CdSe photocell at low temperature levels may be overcome by the difficulty in getting good response times.

[iainpunk]

Does the human ear respond linearly of logarithmically to THD ?

Or is it a very complex function similar to the well known Fletcher- Munson and Dunning–Kruger effects ?

iirc i read somewhere that its a reverse log function when it comes to pure sines, and a linear function for complex musical signals. 3% thd on a sine wave is extremely obvious, 3% thd on a guitar is what we consider adding coloration, not even distortion or overdrive.

thd measurements can be thrown off a lot by tone controls as well, the perfect square wave has 48-ish% THD, but a Big muff on its shrillest and its gain cranked can trick the test in giving percentages over 100%, because the tone stack takes out the fundamental sine (100Hz) enough.

than there is intermodulation distortion, which complicates the whole story even more, as it adds non-harmonic distortion, which are quite noticeable, but arent measured with THD.
big parts of corssover distortion and wave folding also add non-harmonic overtones.

THD is a very narrow scope to look at circuits with.

cheers

[Fancy Lime]

Does the human ear respond linearly or logarithmically to THD ?

Or is it a very complex function similar to the well known Fletcher- Munson and Dunning–Kruger effects ?

The equivalent for distortion is called the Herman-Munster or Freddy-Krueger effect.

Sorry, I'll see myself out... Worth it, though 

[Mark Hammer]

Humans likely perceive added harmonic content in logarithmic fashion.  That is, it takes very little increment to harmonic content to be able to hear that increment, when starting out from near zero, and takes much larger increments to be able to detect/perceive that increment when the level is already pretty high.  No different than easily tasting the difference between a 1/4 and 3/8 teaspoon of salt in a litre of water, but having a difficult time tasting the difference between 2 vs 3 tablespoons of salt in a litre of water.

But that's a somewhat instantaneous comparison, similar to when the eye doctor asks you to squint and look through the eyepiece, and then asks "Which of these is clearer to you, number 1 or number 2?".  Under what circumstances would we make such near-instantaneous audio comparisons?

That said, most improvements in audio technology have emerged because listening to a recording of a natural sound, vs the original source, live, resulted in people identifying that there was a little less of X,and maybe a little more Y, in the real thing, when they had the chance to compare.  They might not have been able to tell the difference between Ella Fitzgerald and Memorex in 1972, but probably could in 1952, when tape recording was poorer quality, and again in 1992, when our expectations and standards for sound had increased beyond those in 1972.

EDIT: I realized this may be too obscure a reference for some people, so here is the ad.

So, while it is difficult to identify ANY dimension of perceived intensity or other change (e.g., rate, pitch) that does not correspond to the "power law" in psychophysics, real world instances where we notice it are not always easily identified or replicated.

(And I'm unsure if your mention of Dunning-Krueger was meant as a joke for our benefit, or was a simple misattribution)

[Phend]

And don't forget the newear vs oldear factor which can be as high as 10.

[anotherjim]

THD is a very narrow scope to look at circuits with.

Agree. IMD is the test I look for, but I'm too lazy to do it much myself even though I could. There was a time when some equipment reviewers used test software like RMAA but it all seems to have gone back to cork sniffing.

[Rob Strand]

When you are mentioning light output for control of CdS or CdSe photoresistive cells, you should be aware that CdS has a wavelength peak sensitivity of 640 nm and CdSe has a wavelength peak of 790 nm so what we see as visible light does not necessarily correspond to the cells are most sensitive to.  There are some mixed CdS - CdSe cells that have peak response between these wavelengths.  The time response of an incandescent lamp is faster the greater the filament temperature, so the added efficiency of a CdSe photocell at low temperature levels may be overcome by the difficulty in getting good response times.

It's a valid point.  When we use lux and lumens units those effects are already taken into account (at least to first order).

Does the human ear respond linearly or logarithmically to THD ?

Or is it a very complex function similar to the well known Fletcher- Munson and Dunning–Kruger effects ?

You need to use psycho-acoustics.  If you have a certain amount of fizz from distortion it stands out when the spectrum of the fizz is more distant (in frequency) from the wanted signal.   That's why nasty fuzzy distortion
stands out on bass - it clearly sounds like the crap is added to a good signal, not part of it.  Also why hiss stands out while music is playing, but hiss is made worse by the ear's response (Fletcher-Munson) and the fact your brain can separate the hiss out as not being part of the music.