• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

'nipple' in waveform - What is it?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've been working on a push-pull amp in LTspice. I was checking out a design, looking at clipping behavior. Using a 5965 or 12AT7 as a cathodyne phase splitter, I'm seeing a weird waveform coming from the inverting side (plate) of the cathodyne stage (green trace). I've attached a screenshot.

I changed the cathodyne tube to a 12AU7 with 10mA plate current, and that cures the weird 'nipple' in the inverting side waveform. Putting a 6SN7 or 5687 there also cures the problem.

What is it about the higher mu 12AT7 or 5965 that could be causing this behavior when the output stage goes into clipping?
--
 

Attachments

  • 'nipple' waveform.png
    'nipple' waveform.png
    53.6 KB · Views: 478
I think it has something to do with the standing current in the cathodyne phase splitter, because if I change things around so that the 6SN7 cathodyne has 4.5mA and 22k plate and cathode load resistors, it too shows that 'nipple' in the waveform from its plate (inverting) output.

Is this a higher frequency caused by slew limiting?

--
 
EDIT: The following does not make sense.

I don't know a whole lot about tube design, but I can tell you that your phase splitter is sending different signal levels to each of the power tubes. I know you can't get rid of R1 for biasing, but it is the culprit. You need to connect C1 directly to the cathode of U2 and raise the resistance of R21 to match the bottom. You need to add 1.2k so you might just add one in series. (23.2k ohms would be hard to find.)

Basically what you see is one power tube clipping before the other, if I'm not mistaken.
 
Last edited:
There was a wave of threads arguing over the characteristics of the split load phase inverter (aka 'cathodyne' or 'concertina'), not too long ago. In the end, it was pretty well proven that the cathodyne phase splitter's output impedance is matched side to side as long as it's working into a high enough impedance load. So that's not the origin of the problem unless the output stage starts drawing grid current.

If I use a 6SN7 with 8mA or more going through it as the cathodyne stage, the nipple in the waveform doesn't appear on its inverting output.

Too see if the unbypassed cathode bias resistor could be adding to the load on the cathode (non-inverting) side of the phase splitter, I got rid of it by DC-coupling the first and second stages using a voltage divider (470k series resistor with a 0.22uF cap in parallel, and a 560k to ground did the trick). No change, the nipple's still there on the plate feed from the cathodyne. I think that exonerates the cathode resistor from imposing imbalance.

I also tried adding a 1.2k resistor in series with R21 (plate load resistor on the cathodyne). That didn't do anything either.

Maybe grid current drawn by one of the output tubes is causing the behavior? OK, so I'll reduce the value of R6 and R7, down from 330k to 100k. Nope, that makes it worse.

Next, I went the other way. I put 1M resistors in for R6 and R7, which actually helps a little. Unfortunately, that wouldn't work in the real world, as 1M is double the permissible grid-to-ground resistance for a cathode biased 6L6 (0.5M ohm is the max for cathode bias).

All through this, the 12AX7 output waveform is looking good.

All I can figure is that it's something to do with whether the cathodyne stage is able to sink enough current into the grid of the output tubes during overload. A cathodyne with 4mA standing current has the problem, somewhat independent of tube type (although the higher mu triodes are more prone to the problem). 6SN7 with 8mA doesn't show the problem at all.

--
 
On my guitar amp I got rid of C2, R2 and R22.
This gives balanced 22k on upper and lower legs of phase splitter.
Just make previous stage anode voltage B+/4.


In other words, you DC-coupled the plate of U1 to the grid of U2, correct? I tried that in simulation, and the problem still happened.


The lower part is probably too but I haven't seen clipping like that before.

Bear in mind that the waveforms are from the plate and cathode of the phase splitter, not from the output. I've seen this kind of thing before on a scope when a differential driver stage clips before the push-pull output stage it's driving. This is covered in the RDH4 somewhere, but I can't seem to find it...

--
 
Last edited:
I'd look at the various grid currents. They may provide clues to what's going on.

Thanks, that's good thinking! OK, so nothing is drawing grid current except the two 6L6s. There's 22.5pV on U4's grid (, while there's 6pV on U3's grid. The currents across the 6L6 grid stopper resistors are in the fractions of fV (femto-volts), so I don't know if they're useful.


Increase R10 to 300-350 ohms and see what happens.

That does something in a positive direction, but only because it takes more input signal voltage to drive the output stage into clipping.

LTspice is showing that biasing 6L6-triodes at something like 350V plate voltage with 60mA plate current reduces their usually too-high odd order distortions. You don't get much power out of them this way, but I'd rather have the lowest distortion I can get from 1 to 5 watts output, and whatever I have above that is gravy. That does mean it's more likely I'll drive the output stage into clipping from time to time, so I check clipping behavior.

The problem here is that the 5965 cathodyne version is showing really nice low distortion, but with this weird clipping behavior at output stage overload. If I put 6SN7 in instead, distortion doubles, but clipping behavior has good symmetry and no weird bumps.

--
 
Valve Wizard,yes!

CathodyneShots.jpg


That's exactly it in the top picture. It looks like the cathodyne's cathode (non-inverting) output has more drive capability than its plate (inverting) output. Increasing the value of the 6L6 grid stoppers to 47k did flatten out the nipple on the waveform, just as predicted.

What is the downside to using grid stoppers of this large a value? I checked the simulated frequency response and that hardly changed at all. Still -0.8dB at 20kHz. If there's no downside, why don't more amplifiers use them?

--
 
It looks like the cathodyne's cathode (non-inverting) output has more drive capability than its plate (inverting) output.
Of course, and that's the point, the cathode side has way lower impedance than the plate side, not a big deal while amp is clean but all the difference in the World when power tube grids start pulling current.
The cathode driven side will go "a little further" while tha plate side will be hard clipped.
And that's what you see in tye waveform.

Another contributor to the "nipple" waveform is feedback, when power tube clips the PI will try to send more drive to compensate, of course only one of both halves has "muscle" to do so.

Adding large grid stoppers way over what's normal is only a Band Aid, you are solving nothing, just making the cathode side weaker and the anode side separated fron the load, which is the power tube grid.
 
This has been argued about over and over again, and by people who know a lot more than I do, but I'll go with the consensus that the output resistances of the anode and cathode outputs are the same in a cathodyne phase splitter when driving a Class A output stage. Morgan Jones' explanation in "Valve Amplifiers" basically states that the plate and cathode have the same output resistance as long as the output stage doesn't start drawing grid current or go into cut-off. That stands to reason, since the balance is so good when used that way (gain is exactly the same from both anode and cathode).

"...triode phase splitters should only ever be loaded by a stage that can be guaranteed to be Class A1 at all times if the balance of the phase splitter is not to be upset. (Any transition into Class B temporarily eliminates Miller capacitance at the input of the switched-off valve, whilst Class A2 dramatically reduces input resistance of the switched-on valve once grid current flows.)"

I think this tidbit from The Valve Wizard explains what's happening when the 'nipple' in the waveform appears:

"When the cathode output begins to overdrive its power valve, the input impedance of the power valve suddenly drops, preventing the cathode voltage from rising further. This clamping of the cathode voltage makes the cathodyne look briefly like its cathode is bypassed, so its gain to the anode suddenly increases."

I think the problem is grid current. When the cathode output overdrives its 6L6, the input impedance of the driven 6L6 takes a dive, and the cathode output is working into relatively nothing. This sudden 'shorting out' of the cathode looks to the tube like its cathode resistor has been bypassed, and the gain of the anode spikes up, causing the nipple in the waveform.

Morgan Jones again,

"If there is a possibility of Class B loading, the concertina should use a low mu valve because this reduces (the output resistance from the anode) and consequent imbalance (during overload). ECC88 and E182CC are good, but 6AH4 and 12B4-A are even better." I'm finding that a 6SN7 or 12AU7 with 10mA is very adequate, but have also been looking at a 5687 cathodyne with 14mA going through it. Perhaps a 6V6 in triode?

At any rate, the balance from a concertina is nearly perfect under normal operation in Class A. I'm not trying to use it as a Class AB driver, but I would like the overload to not be too nasty...

--
 
Last edited:
I think I'm abandoning the 5965 or 12AT7 for the cathodyne. It looks like it will be either a 5687 if I have to use a twin-triode, or a triode-wired EL84. Both of those look really good. The higher current and lower plate resistance yield very good clipping behavior. With the 5687, THD at lower output levels is higher than with the 5965 cathodyne. But using a trioded EL84 yields the lowest distortion levels overall -- at least within LTspice. 6V6-triode doesn't work out as well, for some reason.

--
 
Disabled Account
Joined 2013
If output is clipping both sides, the nipples should appear at both sides too, the squaring of the top appears only at the last drivers. Since the squaring happened before last driver, the problem could be in splitter operating point (clipped before output clipped). Where the exact point snapshots are taken? See the sim of clipping for my amp.
 

Attachments

  • clipping -1.png
    clipping -1.png
    336.9 KB · Views: 147
Last edited:
Disabled Account
Joined 2013
In the first post, both curves are taken from the 6L6 grids, just after C1 and C2.
--

Ok, the nipple look similar to my amp, but for yours the top and bottom are not quite the same level. If you increase the level of blue one to same as green one, you should also get nipple on that. The open loop gain (different mu tube) and amount of nfb will affect the driver levels as well, since when the damping factor is increased, it will try to pull up more level (hence the nipples) during clipping.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.