When can you share cathodes on preamp triodes?

[Baguette]

Hello,

I'm building something where real estate does seriously matter and I'm wondering when it's possible to tie together 2 triodes' cathodes to the same resistor / cap.
Like, Fender often have the first gain stages of 2 channels tied together. How about tying together two serial gain stages (like tying the two gain stages of a Champ / Princeton to the same 820 ohm / 25uF cap)???
Is it risky?

Thanks for helping,

[RicharD]

You cannot share cathodes of series gain stages w/o creating a feedback loop.

[jeff]

 I've often wondered about this myself. If the bypass cap is large enough isn't it keeping the bias voltage constant?

Also didn't fender share cathode resistors in the 2 series reverb return stages? ([E] V4A & V4B )
Was this done specifically to create feedback?

[RicharD]

Maybe I'm completely wrong.   

[jeff]

I'm not saying that you're wrong but maybe there is something to this feedback thing. It'd be cheaper to share cathodes but it's not done so there must be a reason.

The second stages of both channels are share a cathode but not the first. There's got to be a reason.

What about this, the two channels are out of phase. Since the cathodes are shared anything on the cathode of one is on the cathode of the other. The cap blocks the signal but anything the cap can't smooth will apear as a change in the grid to cathode voltage of the channel not being used. What happens if that cap is removed? Now the cathode resistor is unbypassed. The grid of the other channel is 0V but now the cathode is swinging.

Just thinking out loud.

[quayhog]

The very modable Fender PA-100 has the cathodes tied together for all four channel's first stages. (V1a,v1b,v2a,v2b). They all reference ground through a single unbypassed 270 ohm resistor.
http://www.webphix.com/schematic%20heaven/www.schematicheaven.com/fenderamps/pa100.pdf

When modding these the first thing you do is seperate those out and voice each channel differently.

[jeff]

The very modable Fender PA-100 has the cathodes tied together for all four channel's first stages. (V1a,v1b,v2a,v2b). They all reference ground through a single unbypassed 270 ohm resistor.

They are bypassed by a 250uF cap. That's what I'm wondering what happens if it wasn't bypassed?

Playing through CH1 would change the current through an unbypassed cathode resistor which would change the bias voltage of the other channels. So the tube sees a signal between the 0V grid and the now swinging cathode. If it were unbypassed would CH2-4 knobs do anything while playing through CH1?

[kagliostro]

what happens if it was unbypassed?

Less gain and restricted frequency response

Kagliostro

[jeff]

Right but what happens to the other channels if they share an unbypassed cathode resistor?

[kagliostro]

The second stages of both channels are share a cathode but not the first. There's got to be a reason.

Thinking to output tubes that shares a common resistor and cap it seems to me that a single resistor arrangement will give interaction between tubes and the current that flows

without the bypass cap the result, seems to me, is limited to have the effect of a lower output, for that reason I told less gain and frequency response, but as the tubes are in some way paralleled some interaction between it is implicit, what is to be verified is the amount of this interaction

Kagliostro

[jeff]

 this is what's going on in a long tailed phase inverter?

[chuggy]

I once built a brown Deluxe-ish amp with a shared bypassed cathode for serial first and second gain stages.  It thumped, and the thump stopped when I separated the cathodes.

Jim R

[tubeswell]

The oscillation happens because you have two inverting stages one after the other sharing the same triode.

In the fender scheme posted, the stages with the shared cathodes at 'A' are not in series with each other, so there is no problem.

In regard to the shared cathodes at 'E', the following stage is not only dealing with inverting the reverb recovery signal that is in opposite phase to the reverb recovery stage, but it is also dealing with the signal from the vibrato channel TS recovery stage that is (also) connected to the grid of the 2nd stage, and does not share the same cathode resistor with that subsequent stage, which mitigates the tendency for oscillation.  The phase inversion of the reverb recovery stage is independent of the phase inversion of the TS recovery stage (the reverb recovery stage picks up its own completely unique signal phase from the output of the reverb tank).

But generally, if you use the same cathode resistor for a subsequent stage where the signal going into the 2nd grid is inverted 180 degrees from the phase of the signal going into the first grid, you will get oscillation.

(So you can relax again buttery   You were right after all, even tho' you probably couldn't remember why)

(Edited and re-edited whilst tying myself in verbal knots)

[Blind Lemon]

It happened to me, I did a gain stage/Tone stack/gain and the amp didn't like it a bit. I separated the triodes on either side of the tones stack and the oscillation went away.

[PRR]

> You cannot share cathodes of series gain stages w/o creating a feedback loop.

Correct.

> If the bypass cap is large enough isn't it keeping the bias voltage constant?

Not "forever". Now large is "large"?

> didn't fender share cathode resistors in the 2 series reverb return stages? ([E] V4A & V4B )

If the bass-loss from 1st plate to 2nd grid is greater than the loss through the common cathode node, it is (marginally) stable.

Reverb wants bass-cut. The 0.003 against 100K pot is 500Hz. There is another 1:0.3 of broadband loss in the 470K:220K divider. The cathode common cap is 20uFd? That's nominally good to below 50Hz.

So grid cut-off is 10 times higher than cathode cutoff, we shed 10:1 of gain before we approach trouble. There's another 3:1 loss in divider, we are 30:1 away from trouble. The nominal gain of V4B (first rev recovery stage) is 50. We are very close to, maybe in, trouble. Now comes the sharper pencil -OR- separate cathode networks.

There's really 2:1 loss from one cathode to the other, plus the 20uFd is good far below 50Hz, more than 10:1 lower than the 500Hz grid coupling. This plan IS stable as-is. But it took more than ten cents of thinking to be sure. In a one-off, don't think, use separate cathode networks. When you build a thousand, it may be worth the math/experiment to save the pennies.

> The oscillation happens because you have two inverting stages one after the other sharing the same triode.

Correct. Even when you don't get outright bass howl, you can have subsonic instability, "thumps" on transients or all-the-time. (Even some well-regarded amps had this problem: one Dynaco phono preamp was tolerably right 20Hz-20KHz but BIG bump near 0.5Hz.... the warp-rate of 33RPM playback!)

> cathodes tied together for all four channel's first stages.

The signals bleed to each other; also flip-phase and cancel at low channel gain.

But think real-world. Four instruments or mikes on the same stage, too cheap to use four amps or a proper mixer. Signals are gonna bleed anyway. Gain/balance is all by ear. Mostly all pots are full-up, no sneakage. And this cheap trick put a mixer-amp 50 cents closer to musician affordability.

> real estate does seriously matter

Mistake. Pain in the butt. If you try, and it isn't right, you can't get back in to try somthin else. Build bigger. (Learn from my mistakes.)

You can use incredibly small resistor and cap. As small as you can hold. 0.05W, 3V, even surface-mount if you can wire it.

[RicharD]

I feel a lot better now.  That reverb recovery threw me for a loop.  Thanks for explaining it away.

I agree with PRR about the real estate.  Hi gain tube circuits packed in too tight tend to squeal.  More than once we've gone from bread-board to chassis and had circuits start to oscillate.

[38Super]

this is what's going on in a long tailed phase inverter?

The long tailed PI is a different case.  They do share a cathode resistor, but the circuit is configured as a differential pair, similar to that which you would find on the (+) and (-) inputs on an opamp.

It is a little easier to look at the large common resistor in PI cathode as a constant current source: the grid voltage on each input determine how the current is shared between the two triodes.

When using a l/t PI w/o feedback, one input is driven from preamp, other is at AC ground.  When OT secondary f/b is applied, many times it is applied to the other input of l/t PI, very similar to an opamp front end.  Here's a link to a paper describing diff pairs.

http://www.ece.mtu.edu/faculty/goel/EE-4232/Diff-Amps.pdf


cheers,

rob

[Baguette]

Thanks everybody,
PRR, that's an awesome explanation.
Hats off!

[PRR]

Here's the part you must avoid.

Two amplifier stages, one feeding the other, standard multi-stage audio amp, but with common cathode resistor.

You don't need an input signal.

When you turn it on, cap C1 forces tube 2 "on", raises cathode node. C1 eventually charges-up enough that tube 2 becomes less-on, cathode node drops. This turns-on tube 1, which SLAMS C1 and tube 2 to "off". So far off, that it takes a while to recover. Then it happens again. Forever.



Yes, if you put a huge cap on the cathode, so it moves VERY much slower than the grid, it may be "stable". However there is a large rage of cathode caps which may not oscillate, but still get wobbly when signal hits it. The large cap will probably be electrolytic; recall that before electros outright fail, they often lose capacitance.

[38Super]

Yes, if you put a huge cap on the cathode, so it moves VERY much slower than the grid, it may be "stable"

In addition to PRR's good points:  when we bypass cathode resistors, we are using non ideal capacitors.  Caps have ESR, which from a modeling perspective is resistance in series with a perfect capacitor.  When an RC is placed in the cathode ckt, it has a gain boost effect on the triode gain stage.  Traditional analysis would suggest that as frequency continues to increase, the bypass cap impedance decreases linearly: this this is true for an ideal capacitance, but the effective series resistance does not have the same frequency response.  To boil it down, at some higher frequency, the bypass cap behaves like a shorted capacitor in series with a resistor (ESR).  In this application, the cap ESR creates a zero in the frequency response. Gain rolls up due to a zero and rolls off from a pole or corner.  The effect is that the cathode parallel RC never actually looks like a short circuit due to ESR and finite cap reactance.  For example at 500 Hz, the impedance [1/(2*pi*f*C)] of 22uF cap is about 14 Ohms + ESR.  Similar effect occurs with filter caps - many designers place a high quality 0.1u or 0.01u across filter caps due the effect of ESR and ESL.  The conclusion is that in using real world components in shared cathode configuration, the designer can never totally decouple the AC signal between stages - that means feedback - sometimes good, sometimes bad  :shocked:

One last comment: many here are sinking lots of hours into making/improving quality, "boutique on a budget" amps.  I can't think of any sonic improvements to be had with shared cathode circuits.  It seems like cost is the primary driver for designing preamps stages with common cathodes. With this approach, you lose the ability to independently voice each preamp stage and preceding posts show you open yourself up to inadvertent feedback and possibly circuit instability.

cheers,

rob

[PRR]

> Caps have ESR

Yes; and the classic limit on electrolytics is that ESR will not spoil ripple-rejection. That ESR is much-less than capacitive reactance at 120Hz. This virtually assures that when you pick a cap for ample (full bass) audio bypassing, you will get "negligible" ESR.

No, the problem with common cap in sequential stages is that the cap must be larger than usual in proportion to the gain. For full-bass, a single cap could be 10uFd. If stage gain is 25-50, then a sequentially shared cap must be 250-500uFd, a LARGE cap.

It works in the reverb plan because reverb "bass" is limited to 500Hz. We could use a 1uFd or less single cap. Also stage gain is clobbered to less than 15 by the reverb return network. 15uFd may be marginal; Fender used 20uFd. Works, but cutting it close.

It works in the 4-input PA100 because the four cathodes are NOT sequential, no positive feedback path. "Some" bass leaks from say V1A cathode across the 750uFd to V1B V2A V2B cathodes and into the other channel mix-controls. But 750uFd against four cathodes is below 0.1Hz for full cross-talk. At 80Hz the crosstalk is 800 times lower, or utterly inaudible for the purposes the mixer was designed for: four players on the SAME stage playing the SAME song.