Idea: parallel input tridoes with different cathode bias/bypass. Good? Bad?

[mooreamps]

I haven't posted to this forum in weeks and the first thread I click on has you being a jerk to another forum member. Seriously dude, you've gotta knock it off. I'm sure you have good thoughts to contribute. Please try to do so without insulting other users.

Seriouslly Dude ; if he's going to go ahead and "pimp" my designs, he could at least get it right...

-g

[Merlinb]

So if that's true then I guess you could lower the cathode resistor further to 470 ohms and get another 3db of noise reduction. And on down to 220 ohms and get another 3db, and so on and so on?

Russ

Don't fall for it Russ, the cathode resistor has almost zero effect on the noise. (Otherwise, if you kept on halving its resistance, you would get to zero noise, wouldn't you...) It's a bizarre mistake than Gary has made; he enjoys posting it here now and again.

Just to make things clear, it is perfectly acceptable to operate two triodes with a shared anode resistor but different bias circuitry. Noise is reduced by 3dB since the signals add arithmetically, but the noise adds by RMS addition. However, in practice this is a moot point, because far more noise comes from the source resistance of the guitar+grid stopper, which hardly changes when using parallel valves.

[R.G.]

Now this has me curious. I thought that using diodes on the cathode, be they zenners, switchers, rectifiers or light emitters makes for a constant voltage source. Isn't that how the p-n junction in the diode works? Once the threshhold forward or reverse voltage is reached, it stays fairly constant regardless of current through the junction?

That's pretty much right, subject to the little nitpicks about non-perfect devices. Semiconductor diodes do conduct very little until they hit an exponential "knee", where the current increases, well, exponentially with not a whole lot of change of the forward voltage. The voltage change amounts to a sudden change to a low resistance on top of a DC voltage. So yes, using a diode forward biased, or a zener reverse biased in breakover amounts to putting a constant-ish voltage there in series with resistor of a few ohms.

I was more referring to a matter of intent. You can bias a tube (or bipolar, or MOSFET, or JFET, or tunnel diode, or...) with either constant voltage, constant current or resistively. And this can be fixed or sliding depending on operation. The subset Gary meant was that he wants his cathodes held to a fixed voltage and the grid connected to a fixed voltage under that by some resistor network. I think, at least. Language is a slippery thing.

Lurking at the bottom of this bag is that DC biasing is somewhat related to AC gain. The voltage gain of a triode depends on the tube's internal characteristics and on the external resistances. And the internal characteristics change depending on both the tube, type, and operating point.

The tube amplifies a voltage which is the difference between the grid and cathode. If the cathode has a resistor in series with it, there is a cathode follower action, and the tube makes a large fraction of the incoming signal appear across the cathode resistor. This is in series with and opposing the input signal, so it's negative feedback, and tends to stabilize the tube in both AC and DC sense, as well as lowering the AC gain. It was a happy accident for the early tube designers who had to use battery power supplies that varied all over the map that you only needed one resistor to stabilize a triode operation while battery voltage varied.

Today we have relatively GREAT power supplies compared to then, so we have the luxury of not needing as much DC stabilization. A triode needs to have its cathode a bit above its grid. You can do that with a resistor and let the triode stabilize itself. You can do that with a constant voltage source, or you can do that with a constant current source, which is something the Golden Age people could never have afforded.

The resistor method means that whatever tube you pop in will seek its own bias level moderately well, subject to the usual nitpicking about tubes and parts varying. The constant voltage method presumes that you're always using the same tube parameters, as it forces every tube in the socket to be at the same grid/cathode voltage. This may not be a happy choice for swapping between 12AX7, 12AY7, 12AU7, etc. but maybe you could get away with it. The constant current bias setup is one which is almost unknown for tubes, but well loved in semiconductor use. It simply forces the tube to run at a fixed current. This can be done most easily with a bipolar or FET device between the cathode and ground and a bias network that makes the semiconductor run at a fixed current. That does a number of things, including nailing the plate to a fixed voltage down from the power supply level.

All of these can get to the exact same voltages and currents on the grid, plate and cathode. What changes is AC gain, and that leads us back to the matter of design intent. The AC gain of the tube depends on the AC impedance at the cathode to ground and the plate to power supply (which ought to be AC ground if you did it right). Low impedance at the cathode means all the gain the tube can otherwise do, high impedance at the cathode runs the gain to the plate all the way down to nearly zero.

A resistor bias at the cathode has a gain determined by the tube parameters, the plate resistor and the cathode resistor. If you put a capacitor across the cathode resistor, then at frequencies where the capacitor is relatively a short circuit compared to the resistor, the tube acts like it has its cathode tied to ground and has all the gain it can do. The value of the resistor in concert with the tube parameters set the DC operating point and the frequency where the gain goes from low to high is determined by the RC time constant of the cathode resistor and capacitor.

A fixed DC voltage on the source is almost the same as the bypassed resistor except that there is some small resistance at all frequencies, so the ultimate gain is fractionally lower unless you also bypass the DC voltage source with a capacitor; and that there is no falloff of gain at low frequencies as there would be with a resistor/cap. This increases bass response below the RC rolloff point, and forces you to decide (or ignore, I guess) whether that's something you wanted to do.

As an example, a Fender-ish triode circuit with a 1K and a 25uF cap has a rolloff of 6.3Hz if I punched the right numbers in the calculator. Low E on a guitar is 81Hz. So whether you'd hear a difference in a constant voltage bias and a resistor/cap bias is something that at least needs some investigation. To a first order, the frequency response of both would be the same for guitar. Purists will argue about the capacitor contaminating the sound, maybe. The residual resistance of the "constant voltage" diode source will reduce AC gain a tiny amount unless bypassed.

A constant current bias source is a different animal. It lets a more-or-less fixed current through and lets the voltage across it vary to make that come true. A bipolar transistor, JFET, MOSFET, or pentode above the knee of their startup region all act like moderately good constant current sources if their bias is held fixed. A constant current source, if unbypassed, makes a plate gain of **zero** and a cathode gain of as near unity as the device can do. If bypassed, it has the same gain as the bypassed resistor, but can be bypassed by a much smaller capacitor. In this case the cap is acting as the AC "Cathode resistor" and the current source as the DC.

So it's a matter of (a) knowing the methods that are available (b) knowing the first and second order effects they give (c ) selecting the right one for your purposes. It helps if you know what matters and what doesn't in achieving whatever you're trying to do. Even if that's just trying to be different, like all the other boutique amp builders. LOL.

I really need to read more about this stuff.

I believe in that as a way of life.

[octal]

Even if mooreamps assertion about the cathode resistor being a significant contributor to the noise of a triode gain stage was true (and I don't believe that to be the case) if the cathode is bypassed, there's essentially an AC short for all audio band frequencies across that noise generating resistor. Hence, no noise contribution in the audible range.

I love how Gary seems to think that sticking a couple of diodes between the cathode and ground is some sort of revolutionary innovation and everybody is lining up to "steal his work." Two diodes to ground is one of many ways to bias a tube and it will work just about as well as a bypassed resistor. End of.

Nathan

[soundmasterg]

It does clamp the cathode to a constant level, and there is very little you can read about it, since I have not published any of my research on the subject.

-g

Boy aren't you a rather self-important git?

Greg

[Russ]

True, but you get less gain. I just pop the resistor completely out of the circuit, and just be done with it.

-g

Firstly, by your reasoning, with no cathode resistor, your noise level would be zero! (as Merlin also pointed out)

Secondly, if just reducing the cathode resistor, Rk, from 1k5 to 820 ohm with parallel triodes reduces the noise by 3db, then you could just do the same with a single triode and also get 3db noise reduction. And if Rk is fully bypassed, there's no gain change, just a shift in bias point. So why bother with parallel triodes to get some noise reduction, huh Gary?
Just use a single one, lower Rk, and save yourselve a triode.

Russ

[mooreamps]

Firstly, by your reasoning, with no cathode resistor, your noise level would be zero! (as Merlin also pointed out)

Secondly, if just reducing the cathode resistor, Rk, from 1k5 to 820 ohm with parallel triodes reduces the noise by 3db, then you could just do the same with a single triode and also get 3db noise reduction. And if Rk is fully bypassed, there's no gain change, just a shift in bias point. So why bother with parallel triodes to get some noise reduction, huh Gary?
Just use a single one, lower Rk, and save yourselve a triode.
Russ

1. Parallel gain stages does add some smoothness ; IE : Matchless amps.

2. It can also be a question of bias. If using resistor bias of 820 ohms, one gain stage will give you an input bias of approx 0.8 volts. If using two gain stages, gives you an input bias of approx 1.6 volts. An input bias of 1.6 volts maybe more pedal friendly than 0.8.

3. What do you mean as Merlin pointed out ? I am writing the paper on fixed bias, not him.

-g

[tmod]

Ya, somebody tell "merlin" you don't need bypass caps when using fixed bias... :|

-g

Somebody tell Gary (again) to READ the book before critisizing what he BELIEVE is omitted... :|

From the valvewizard site on common gain stage:

Page 34:
"This ensures that the cathode voltage will remain constant regardless of anode current, or even what valve is plugged in, so there is no need for a bypass capacitor."

Page 35:
"However, if the stage is likely to be overdriven, a ‘snubbing’ capacitor of around 100nF may be added in parallel with the diodes to reduce any noise created as they switch on and off with the valve."

[tmod]

A new Randall Smith?

Seriouslly Dude ; if he's going to go ahead and "pimp" my designs, he could at least get it right...

-g

I'm relatively new in this game, but I first read about diode bias in 2006 with the Melissa amp using constant a current diode (1N5294). Diode bias in guitar pre amps has been discussed over at ax84 at least since 2001; I suspect it's been mentioned at the old ampage waaaaaay earlier. In HiFi it's certainly been around for a long time.

I'm awaiting the anouncement of a new patent on revolutionary idea of using a steel (or perhaps, wait for it: aluminium!) chassis for guitar amps...

[tmod]

It does clamp the cathode to a constant level, and there is very little you can read about it, since I have not published any of my research on the subject.

-g

...or you can read about the experiences of countless other who have tried it. The ax84 Blues preamp use LED biasing with two different LEDs for a thicker sound, and there are many discussions about them. I'm sure a search here would also yield a lot of info!

[R.G.]

I once had a brother in law who thought he was out standing in his own field. He became a legend in his own mind.

[Russ]

1. Parallel gain stages does add some smoothness ; IE : Matchless amps.

Maybe, but they also reduce noise by 3db. BTW, all of the Matchless parallel triode amps use 1.5k Rk, except the Clubman which uses an unbypassed 2.2k

2. It can also be a question of bias. If using resistor bias of 820 ohms, one gain stage will give you an input bias of approx 0.8 volts. If using two gain stages, gives you an input bias of approx 1.6 volts. An input bias of 1.6 volts maybe more pedal friendly than 0.8.

Once again, maybe, but bias point, as Rk, doesn't have anything to do with noise reduction in parallel triodes.

3. What do you mean as Merlin pointed out ? I am writing the paper on fixed bias, not him.

So why is it that using a lower Rk in a single triode doesn't get you any noise reduction, but (according to you) it does with parallel triodes?

Russ

[mooreamps]

Somebody tell Gary (again) to READ the book before critisizing what he BELIEVE is omitted... :|

From the valvewizard site on common gain stage:

Page 35:
"However, if the stage is likely to be overdriven, a ‘snubbing’ capacitor of around 100nF may be added in parallel with the diodes to reduce any noise created as they switch on and off with the valve."

not enough gain to over drive them, not in the conventional way like when using resistors. I know, I've had a few set-up that way.
don't believe ? You build a few and find out for yourself.

-g

[tmod]

not enough gain to over drive them, not in the conventional way like when using resistors. I know, I've had a few set-up that way.
don't believe ? You build a few and find out for yourself.

-g

Quite on the contrary! In theory a diode will yield the exact same gain as a bypassed resistor. A diode will yield considerably more gain than an unbypassed resistor giving the same bias voltage.

I tend to use diode biasing on the first stage to maximize signal to noise ratio at the input. Because there is effectively no feedback path I get wide frequency response, high gain and low noise. In this capacity there is no risk of overdriving them, but that's only because the signal in is very small. I dare say very few tube gain stages will be overdriven by a signal in the mV range. But hit it with a bigger signal, and it will be overdriven!

The sensitivity of a tube gain stage (how large signal it takes to drive it to maximum unclipped output) is not determined by the bias method, but by the load line/tube characteristics. If a stage biased by a bypassed resistor will overdrive given a certain voltage, it will overdrive at the same input signal bias by the same voltage with diodes.

Surely, you must agree that if we slam a diode biased stage with a 50 volts signal it will be overdriven?

[tmod]

I once had a brother in law who thought he was out standing in his own field. He became a legend in his own mind.

Ha ha, priceless!