Univox Tremolo, hows it work.

[adamasd]

http://www.univox.org/pics/schematics/1221.gif

The grids of the 12AU7 are connected together, just not drawn that way.

I thought the 12AU7 was working as a voltage controlled resistor but then I noticed that the second half of the tube is connected upside down. Wouldn't that mean the as one tube goes up in resistance the other would go down and they would cancel each other out? I am stumped, I can't figure it out.

adam

 

[solder_city]

a pair of voltage controlled diodes is more like it. imagine those as two diodes with a 'control' terminal and youll start to uderstand how it works.

 

[Guest]

The "triode square law" means that Ri depends on a grid bias, so one triode works for a positive half-wave, another triode works for a negative half-wave.

 

[adamasd]

ahh! That is pretty nifty. Current only flows on way in a tube, so you need a backwards one to do fullwave. I think I gave up to soon on this one, I should have been able to sort that out, or maybe not, my brain was pretty stuck on it acting like a voltage controlled resistor. Thanks for explaining it.

 

[PRR]

> my brain was pretty stuck on it acting like a voltage controlled resistor.

A triode IS a voltage controlled resistor.

But controlled by two voltages: grid-cathode, and plate-cathode.

When plate-cathode is negative, the resistance is always infinity.

Two back-to-back ensure that when one is negative, the other is positive.

Makes a full-wave resistor instead of a half-wave resistor.

But it is much more subtle than that.

What is the plate resistance at ZERO volts plate-cathode?

Hmmmmm.....

According to your 3rd-grade tube textbook, it's probably zero current, infinite resistance.

And Shockley's Law (which does pertain to hollow-state, though usually masked by other effects) predicts zero transconductance (thus infinite resistance) at zero current.

That leaves the middle of the wave, the zero-crossings, un-tremmed.

Sorta reverse crossover distortion. Which is probably annoying.

BTW: the signal level here is around 150mV for full output.

So we are never talking "real plate voltage", not 100V or even 10V. We are way down the crack at the origin of the plate curves.

Do what I did (when this amp was new). Wire it up on the bench and 'scope it.

What I recall (and confirmed again much later) is that MOST tubes won't work. Indeed the plate current and resistance around a few mV is so very high that it does "nothing" against the ~200K source and load impedance.

BUT. Many 12AX7s do have significant plate current at zero plate and grid voltage. Many microAmps. Plate resistance under 100K for AC signals under a few milliVolts.

How can that be? Why would electrons flow to a plate at the same potential as the cathode they just left?

Electrons leave the cathode at varying energies. Some drift away, some fall back to cathode, and some have several volts of random energy rushing them along even in the face of equipotential indifference.

The 12AX7 has a relatively energetic cathode (about the same as 12AU7 which can handle 10 times the current) and a fairly small plate. So the number of hi-energy electrons is more than zero, and the chance of them running into the plate is high.

And tho the plan says 12AU7, my UniVOX had 12AX7 there. I do remember it being critical: the tube type was maybe not as important as finding one with this undocumented zero-voltage current effect.

Compare to a JFET. We use it the same, working shunt against a series resistor. But the thermionic cathode is strongly unidirectional, so we need two devices. However we do not need the Vc/2 correction for lowest THD. But not all tubes have a useful zero-volt redistance; JFETs always do. The tubes' resistance is an order of magnitude higher, the maximum "linear" voltage is (suspiciously) similar to a JFET, so the signal-noise-level tradeoff is about 10dB in favor of the JFET. However, the year this was built, a case of possibly "under-spec" tubes was still cheaper than JFETs (we had just gone to silicon power diodes, which are far more compelling than tremolo techniques). If you were to do it today, the JFET is the "logical choice".

 

[adamasd]

Ok, so grid voltage controls current which means resistance varies. Current only flows one way in a tube so we need a pair to do the fullwave. Makes sense. Pulls it all together nicely.

-Sorta reverse crossover distortion. Which is probably annoying.

That would explain why the LFO sounds almost like a square wave at higher speeds. At least if I understand what you mean by reverse crossover distortion.

-Do what I did (when this amp was new). Wire it up on the bench and 'scope it.

thats the plan for when I have time to sit down and finish replacing all those leaking Oil caps. For now it will stay closed up so I can play it.

What I recall (and confirmed again much later) is that MOST tubes won't work...tube type was maybe not as important as finding one with this undocumented zero-voltage current effect.

Hmm, I found that every 12AU7 that I tried worked, all with a different end result, but worked, must have been lucky. I think I will have to sit down with a pile of 12A_7 type tubes and try them out, maybe can get it abit better sounding. Maybe one of the old 12AT7s that had the small round plate would work even better, more plate surface area up close to the cathode and all.

Will have to think on all this abit more, but it is starting to make real sense to me.

thanks

 

[Guest]

[quote author="adamasd"]Ok, so grid voltage controls current which means resistance varies. [/quote] No. Grid voltage controls resistance. As soon as resistance varies plate voltage varies as well that changes resistance also, so the "variable resistor" is not linear. Paul suggested to use one FET with added 1/2 of a signal to the controlling signal on a gate for less distortions.

 

[adamasd]

Almost, had it backwards.

I have been playing with the idea of making a small jfet board that can plug into the socket. But that can wait, I do not have time to worry about improving something that is already pretty great sounding.

 

[PRR]

> so grid voltage controls current which means resistance varies.

It's a tricky concept, since it is so far from "normal" tube use.

I'd say the grid-cathode voltage influences the ratio of voltage/current in the plate-cathode path.

And of course the ratio of voltage/current is resistance.

And as the shunt element of an L-pad, it changes attenuation.

The grid voltage probably influences the plate voltage needed to shunt a specified audio current.

A tube of very different Mu may need different bias and LFO levels, may even have a different optimum max signal level. Or maybe not: at these very low currents (~1uA) the Mu is far off from book value, largely determined by end leakage, and may be similar in tubes with different grid spacing (different book Mus) but similar overall electrode dimensions.


me> we do not need the Vc/2 correction for lowest THD

I may be wrong here. The key parameter is probably the grid-cathode voltage. But one cathode is bopping up and down with the audio. So the control voltage should ideally be referenced to the audio. But only for one half-cycle. The other flows in the side with grounded cathode. And there is an overlap. So there may not be an elegant THD correction.

Doesn't matter. The audio swing is ~150mV. The grid swing is more like 3V. That's a 5% error. We may have another 5% error in mis-match between two units in the same bottle. And it's a dang guitar amp, and a strong effect. And indeed some of the favored tremolo schemes actually force tubes in and out of serious distortion. It appears to increase/decrease the (fuzzed-up) highs, slightly like the variable low-pass of the flap-doors on a pipe organ swell box.