Marshall JVM205 Cathode Follower Death, Sprial Filaments and Internet Mendacity

[nickb]

Ugh. A JVM205 just came back to me. Again. The first time I saw it the CF tube was bad so it was replaced.

Now in case you are not familiar with these the standby switch does kill not the HT. This means that when turn on from cold, the full 450V HT gets applied between the gird and cathode of V8b. So, I did what I do and added a neon between the grid and cathode to limit the grid current, replaced the tube with a 12AX7LPS spiral filament type and after other checking sent it home.

And then it came back with the same tube dead. So, thinking maybe the spiral filament lore is true I fitted a non-spiral type and packed it off. After all I'd already done the neon grid protection trick so no need to worry about that, right? You guessed it, that one failed too.

Hmm what about the heater cathode voltage? With a lack of actual data we turn to the Internet, naturally. Many forums will teach you that the heater cathode insulation fails when running since the cathode is about 165V and equally common is the statement to avoid spiral filament types as the insulation will fail due to not to be trusted new-fangled spiral technology. On the other hand I read somewhere that the Electro-Harmonix version was especially suitable for this task. The funny thing is that the data-sheets don't support this view. Here are the max heater to cathode limits from some datasheets. The LPS spiral filament is equal best at 200V. EH and Sovtek A/B/C all a dismal 100V.

SOVTEK 12AX7LPS	200V
SOVTEK 12AX7A/B/C	100V
ELECRO-HARMONIX	100V
GENELEX 12AX7	200V
GENERAL ELECTRIC 12AX7	180V
SVETLANA	200V
TUNGSOL RI	180V

We know to take contemporary data sheets not too literally so, suspecting that the insulation might be failing I set up a little test jig:



The heater had volts applied and I monitored the half sine on the scope while raising the variac. Any failure would show as a flattening on the top of the waveform I tested the two 12AX7LPS and the 12AX7WC that had failed. I went all the way to 500V with no sign of failure. Same results with the heaters cold. Can you say surprised? Well that ruled out that failure mode.

Time to look at these failures a different way so popped them into the curve tracer. By the way the "A" section is the DC coupled CF so it's easy to compare the two sections. The solid lines are section A and dashed are the B section.



Also of note is then when running in the DC CF position you get a 40 to 50V drop from grid to cathode. I conclude from all this that the failure is to do with the grid. Presumably the incredibly fine wire is being destroyed during warm up by the current of <2mA, even with the neon from grid to cathode. By the way, ALL the datasheets specify a max grid to cathode voltage of zero. I think a current limit might have been more helpful.

So, I've gone to a new strategy. The neon has now been replaced with a yellow LED with a forward voltage of 1.8V (measured, and because I have a pile left over from a different project) in series with a diode to protect the LED form reverse voltage. I did try just a diode but it caused too much flattening of the waveform. Another technique is so use a diode and a resistor in series. I wanted to have as low as an impedance as I could so that I why I went for the LED. Also changed the plate resistor from 220k to 330k to bring the cathode voltage to under 100V to be on the safe side. Marshall really wanted this to distort. The lower plate current available is somewhat mitigated by the increased plate to cathode voltage on the CF as that reduces the grid to cathode operating voltage. Interestingly the load line suggests this should be typically around zero. I'm seeing -0.5V, which is good:



Here's a scope shot of the warm-up with a neon and with the LED and 330K plate resistor.



Finally, when operating full tilt:



So, I'm hoping this will be a permanent solution. Unless anyone has a better idea...

PS: Maybe I wasn't clear so these are the methods I tried:

[Gregg]

On the other hand I read somewhere that the Electro-Harmonix version was especially suitable for this task.

My first problems with 12AX7 failing as CF started 10 years ago. The funny thing is they started exactly with EH. Some of them were dying after the first power up. Some of them later. Interestingly pieces from the previous batch were OK. Go figure. So after some digging in the net I learned that the problem was not specific to the EHs only and other brands would also fail as CF. After that I started raising the heaters to 70-80V just to be on the safe side and that cured the problem permanently.
Then I asked myself the question how come Mesa DR is running the CF tube crazy voltages where the cathode to heater difference is more than 200V and I found out that they were using predominantly chinese tubes. I noticed the same thing with other companies as well. So if you need a tube that can withstand the abuse it looks like the chinese 12AX7 can handle it well.
Concerning the protection I'm using the technique described here:

The Valve Wizard

[Pedro Vecino]

I have never accessed tests of that type (thanks!) but my idea is that Sovtek LP / LPS and Tung Sol 12AX7 are not the ones indicated for that function. I discovered it the day that in a classic Marshall two consecutive LPS that had worked perfectly in tests on other functions failed suddenly in ten minutes. I speak about 14 years ago or more.
With AC30 Vox CC series I consider relevant that they come equipped with two ECC83 Tung Sol in the first and third socket but in the second always includes a different one, usually chinese avoiding the Tung Sol in the CF.
Mesa uses the same chinese 12Ax7 in CFs most of the time, but I have also seen units with CF equipped entirely with the model they use more frequently (Russian II) without problems. This tube is of Russian origin for sure but does not seem to correspond to any Russian model marketed by other distributors. It is not exactly WA, WB, WC, WXT, LP, LPS, Electro Harmonix or Tung Sol. The most similar is Electro Harmonix, but not identical.

[nickb]

In order to try to understand better what is going on, I've done more testing. Some pretty interesting results.

First I tested the 12AX7LPS for grid current. This first plot used a signal generator. The grid/cathode behaves like a diode with zero forward voltage and 720 ohms in series. No sign of anything bad.



Next, I used a power supply feeding the grid via a 1k resistor and noted the power supply voltage and the drop across the resistor. The objective was to see how much current I could feed into the grid before it failed. I ran increasing currents for about 10mins at each level and after each test ran it through the curve tracer at Vg=0 and Vg=-0.5 to see how it was doing. The reject tubes had only one bad section so I tested the good section. I was happy to sacrifice since they were bad already.

This is at the start of testing. Solid line is the bad section and the dashed line is the section being tested.



and this is at 8mA. Still OK. Still maintaining the 720 ohm slope. Note the bad section which just has the heater running seems to have fixed itself. More on that later.



After 8mA, things started to go very non-linear, the grid current would fall and the grid to cathode voltage would increase i.e. the effective grid resistance was increasing. Eventually I left it with about 28V on the grid and just 2mA flowing, an effective resistance of 14K and left for 10mins. This is what we got. Notice the remarkable likeness to the bad section especially from the first curve trace in post #1. You can see that the section being tested has failed. This is somewhat supportive of grid current being the failure cause. It also meas that the LED + diode limit the grid to cathode voltage is a better strategy, not withstanding the other resistive current limiting in the circuit.



Now, noting that what was the original bad section seemed to have miraculously fixed itself in the course of the tests, I ran the newly damaged section with 0.75mA with the heaters at 14.2 (as opposed to 12.6v) for 20mins and re-tested:



We have recovered the newly made bad section. This tube now has both sections working. Why, I have no idea but it's very much like a build up of cathode interface resistance.

Now taking Gregg and Pedro's anecdotal input on the Chinese 12AX7, I grabbed one and did the same to it. It would happily take 23.1 mA with the internal resistance staying steady at about 1.2K ohms.



Final test on grid to cathode arcing. I applied 360V via a 4.7Meg to the grid, no heater of course. No sign of any problems. I'll redo at a higher voltage but I'll have to re-jig to get more volts.

I think I might re-do the heater cathode insulation test to use DC rather than the half wave rectified voltage just in case it shows a different behavior due to a time effect.

So, I was able to reproduce the failure but I had to use much higher currents (8mA) than can happen in the JVM205 where it's limited to about 1.8mA, Perhaps it's a time & repetition effect. Certainly not conclusive, but supportive data. So far, I have no evidence to support the heater-cathode insulation theory, other than anecdotal. Also it look like Pedro is on to something, the Chinese tubes are better in this regard.

Coda:
I re-did the heater cathode and grid to cathode tests at 600V Dc for > 1min. No problems at all. I'm not advocating that anyone use these as design guidelines, it was just to rule these out as failure mechanisms in this case. Clearly these tubes a a lot tougher than I thought.

Lastly ,I did the just discovered heater tube recovery method on the last remaining damaged tube a 12AX7WC. That shows a 100% recovery too.

If anyone knows what is really going on here, I sure would like to know.

[nickb]

I woke up this morning and it hit me. The problem is not grid current, but simply current. The design of the cathode determines the maximum current that can flow. This is thermal saturation. Once that is reached the cathode can be damaged due to the excessive current density(*) . I'll have to research the literature to find the exact details.


Update:
I tested the saturation current of the Sovteks by connecting the grid to the anode and sweeping up to 50V. They came out in a 25-35mA range. The Chinese one went all the way to the tester current limit of 400mA at 75V. So the problem is not thermal saturation since the current magnitudes are different but it's closely related. My guess is that as the grid is a spiral of fine wire close the the cathode, the current density on the cathode is excessive directly opposite the wires and so the cathode is being damaged.

(*)Physical Electronics: Handbook of Vacuum Physics
edited by A. H. Beck pg 233.

[nickb]

I guess not very many are interested in this. I spoke with Merlin about this. We agree that damage to the cathode during warmup is the probable cause. The cathode can only supply so much current as it warms, demand exceeds supply and the cathode is damaged. It's clear this damage can happen very quickly. Here is a suggestion of a mod to prevent this from happening.

R1 senses the current through the first stage and when high enough turns FET M1 on. C1 prevents it turning on to quickly to avoid a thump and also give a some delay which helps mitigates an early turn on due a FET that happens to have a gate threshold at the the low end of the range.

[Enzo]

Why doesn't this happen to every JVM205 then?

[alexradium]

Why doesn't this happen to every JVM205 then?

this happens mostly to russian tubes,if you have a JJ or good chinese tube it won't happen so easily,best thing would be to elevate the voltage reference of filaments to at least 80V.

[nickb]

Why doesn't this happen to every JVM205 then?

It depends on the tube. The Electro-Harmonix types that I tested have about 1/4 or less the emission at new than the Chinese. The warm up characteristics will have an effect too. I started looking into this after seeing many failures in Marshall amps in which the B+ is always on.

[nickb]

this happens mostly to russian tubes,if you have a JJ or good chinese tube it won't happen so easily,best thing would be to elevate the voltage reference of filaments to at least 80V.

Agree up to a point. In my tests I proved that heater to cathode insulation is not the the issue as I put 600V across them without it breaking down. It was clear form the curve tracing that the emission had collapsed.

[Gregg]

Here is a suggestion of a mod to prevent this from happening.

IMO this is not a realistic mod. It needs to be quick and dirty like the one with the 47k resistor and 1N4007. Next quick (if the scheamtic allows) elevate heaters (or just the CF one) up to 70-80V, even 90V.

[nickb]

IMO this is not a realistic mod. It needs to be quick and dirty like the one with the 47k resistor and 1N4007. Next quick (if the scheamtic allows) elevate heaters (or just the CF one) up to 70-80V, even 90V.

I agree Greg. This would not be practical as a mod and I didn't intend it as such. It's just a suggestion for one way of dealing with the issue.

PS: I have no evidence whatsoever that elevating the heaters does any good - see earlier tests.

[strato56]

Hi,
Looking at schematic : R81 and R83 form a voltage divider that elevate the heaters to 67V DC.
Do measurements confirm this ?

[bordonbert]

Hi guys. I found this thread with a Google search and decided to become a member on the strength of it. I'm a JVM205H user with a real interest in this very subject after I came across Merlin's Valve Wizard site and his description of it there a good while back. I have implemented the resistor+diode fix but I am now intrigued by your LED+diode version. Also by the fact that you don't think the evidence shows that the mechanism is a feature of the grid/cathode overvoltage at switch on.

I can confirm one fact straight away. In the JVM series Marshall have definitely already implemented lifting of the heater voltage by what must be around 60V. This is created direct from the initial Preamp HT voltage across a divider made with resistors of 82k and 470k, (400V -> 59.5V, 450V -> 67V).

Seeing what you have tested and found here is really interesting and I take my hat off to your methodology. Sadly NickB seems to have been worried that no one is interested, well I for one am intrigued. Has there been any progress on this subject since the post before this?

Edit: It does occur to me that the exploratory tests you did on grid current with your PSU and 1k resistor could be set up to simulate the real world conditions of the DCCF fed from the 220k anode resistor of the previous gain stage with the 47k cathode resistor of the CF itself, as that would perhaps relate to the real current limits within the grid circuitry. Or is that not relevant?

[nickb]

Hi guys. I found this thread with a Google search and decided to become a member on the strength of it. I'm a JVM205H user with a real interest in this very subject after I came across Merlin's Valve Wizard site and his description of it there a good while back. I have implemented the resistor+diode fix but I am now intrigued by your LED+diode version. Also by the fact that you don't think the evidence shows that the mechanism is a feature of the grid/cathode overvoltage at switch on.

I can confirm one fact straight away. In the JVM series Marshall have definitely already implemented lifting of the heater voltage by what must be around 60V. This is created direct from the initial Preamp HT voltage across a divider made with resistors of 82k and 470k, (400V -> 59.5V, 450V -> 67V).

Seeing what you have tested and found here is really interesting and I take my hat off to your methodology. Sadly NickB seems to have been worried that no one is interested, well I for one am intrigued. Has there been any progress on this subject since the post before this?

Edit: It does occur to me that the exploratory tests you did on grid current with your PSU and 1k resistor could be set up to simulate the real world conditions of the DCCF fed from the 220k anode resistor of the previous gain stage with the 47k cathode resistor of the CF itself, as that would perhaps relate to the real current limits within the grid circuitry. Or is that not relevant?

Welcome and thanks for the interest. I could modify the test set up to mimic the actual map but I prefer have the controlled conditions. I remain convinced that it is the excessive cathode current that damages the tube. it is (was) easily demonstrated that heater to cathode insulation breakdown is not the problem.

The easy fix seems to be to use Chinese tubes. The Sovtek ones that caused me so much pain are to be avoided.