Hum in directly-heated single-ended output valves

[GrimJosef]

It's a fact that audio amps using the simplest possible output stage are still pretty popular. This isn't a thread about why they are popular or whether they should be - posts about those subjects belong over in the audiophoolery (or otherwise) thread please. What I'm curious about here is one of the basic problems of amps like this - hum. I'm actually going to narrow it down further and ask just about hum in the output valve, which is almost always a directly-heated triode (DHT).

The attached pic shows such an output stage circuit, which could hardly be simpler. The anode load is a transformer capable of passing the quiescent DC. The 100k grid leak resistor holds the grid at zero volts DC and the resistors between the filament and ground provide the (auto) bias. The component values and voltages I've shown correspond to an actual amp I have on the bench at the moment.

I can think of three sources of hum in this circuit (please let me know if you can think of more)

i) leakage into the circuitry from stray electromagnetic fields - everything from magnetic flux leaking out of the mains transformer and directly into the output transformer, to the electric field from the mains live wiring inducing voltages in the stage's input wiring. With the exception of flux from the rectifier/reservoir capacitor loops in power supplies, leakage hum is almost always at 50Hz and the odd harmonics. I'm not going to say any more about this because it's nothing to do with the valve really. It can usually be controlled by good 'hum hygiene',

ii) ripple on the HT supply. The amp I'm working on is a stereo one, so the total HT draw is about 130mA at 1125V (mostly for the 211's, with an additional 15mA or so voltage-dropped to supply the small-signal stages). Smoothing that isn't trivial. The amp was supplied with a 50uF reservoir feeding a 110uF smoother via a 1k resistor (both capacitors were actually three 500V electrolytics in series). I added a 5H choke in series with the resistor and dropped the latter's value a little to keep the HT voltage the same. Without the choke there was about 250mV pk-pk of ripple on the HT rail, mostly at 100Hz but with significant contributions from higher-frequency harmonics. With the choke the ripple fell to 70mV pk-pk, still at 100Hz obviously but now somewhat more sinusoidal i.e. with lower amplitude harmonics,

iii) ripple on the filament supply. Again generating and smoothing this - 3.25A at 10V - isn't quite trivial. The brave might have built a solid-state regulator and floated it up to the 60V or so filament potential. But sooner or later all valves are inclined to hiccup and, when they do, anything but the simplest solid-state circuitry attached to them can expire in clouds of smoke. So the original filament supply was a brute-force one - a chunky bridge rectifier with a 10,000uF reservoir capacitor. The impedance of the mains transformer winding and the rectifier meant that the waveform at the reservoir capacitor had a bit less than 4ms trough-to-peak risetime and roughly 2V pk-pk amplitude. This is pretty poor DC, but it's still much better than the raw 50Hz AC which is what would have been used (assuming mains power) in the days before cheap high-current low-forward-voltage rectifiers. Luckily for me the output voltage from this supply was too high - nearly 11.5V - so I was able to add 0.4ohms of series resistance followed by another 10,000uF smoother which gave me much closer to 10V DC, now with about 0.55V pk-pk ripple.

So much for the long preamble (apologies for that). Getting to the point, the pot across the filament can be adjusted to minimise any effect on the valve's anode current of the filament supply's voltage ripple. This needs to be done because any current modulation generates a voltage across the anode load and thus across the speaker - the dreaded 100Hz hum.

In practice I've found that I can indeed minimise the hum by adjusting the pot. But the reason I've started a thread about it is that the minimum I can reach differs much more than I would have expected between the two 211 valves that I have. I'm curious to know whether this is often true and also for any enlightenment about what might be causing it. To give the numbers, with one valve I can drop the 100Hz component of the hum spectrum down to a little below -70dBV (less than 0.3mV RMS) measured across the 6ohm load on the output transformer secondary. With conventional cone speakers this is to all intents and purposes inaudible. With the very sensitive 'horn' speaker systems popular with the folks who also like single-ended DHT amps this level of hum is audible, but acceptable. (For the record I should also add that the same folks tend to have a religious objection to the use of negative feedback. So this amp doesn't have any, even though it would reduce the hum as well as doing lots of other good things.)

When I put the other 211 into the same socket that the first valve had been tested in I found I couldn't get the 100Hz hum component below -59dBV or so (more than 1.1mV RMS). This would be audible, if quiet, through typical cone speakers and would be irritatingly loud through horn ones.

My question is "What difference(s) is (are) there between the two valves that might explain the different levels, -59dBV and -70dBV, of their 100Hz hum minima ?". It is true of course that the total 100Hz hum at the output transformer primary has contributions from the HT rail contribution, the 100Hz variation in average voltage of the filament (and therefore the average grid-cathode voltage) weighted by any emission variations along the filament and also, as the filament supply voltage varies, with any resulting filament temperature variations (will these be negligible at 100Hz ?). The pot adjustment is an attempt to minimise the sum of all of these by varying just one, or maybe two, of them. Are there any other contributions that I've missed ?

The anode load impedance is about 11k (6ohms transformed at 43:1 turns ratio) which is a good deal larger than the valve's anode impedance. So most of the 70mV pk-pk HT ripple appears across the output transformer primary and ought therefore to generate more than 1mV pk-pk - perhaps 0.4mV RMS -at the secondary. I find, though, that the ripple at the secondary due to the filament supply can be much larger than this - many mV RMS when the filament pot is moved away from the null position. So the reason why I can't completely null the 100Hz hum is not that the fixed HT contribution is larger than the variable filament supply one. Furthermore, although the pot is wirewound, and therefore has a 'stepped' response, it doesn't feel as though the null position is so tight that my inability to find a position 'between' adjacent wires explains the failure to null completely. Complete cancellation could be impossible if the different contributions to the 100Hz total were phase-shifted relative to one another. Then just varying the amplitude of one of them with the pot wouldn't serve to cancel the others. I can imagine that the ripple from the HT's strong CLRC filter and the filament's weaker CRC one might be relatively phase-shifted. But this wouldn't be different for the two valves, since they were tested with the same supplies in the same socket.

So for now I've hit a bit of a wall in terms of explaining the hum difference. I'd be grateful for any ideas !

If it helps I should add that the 211 datasheet is here https://frank.pocnet.net/sheets/049/2/211.pdf.

Cheers,

GJ

[Radio Wrangler]

Bridge rectifying the heater supply into an immensely large capacitor filter will give low-ish droop between charging cycles, and so the charging cycles will be short in time and especially large in amplitude. Care is needed to get much attenuation between these massive pulse currents and the signal paths and their returns.

David

[pmmunro]

Could the filaments of the two valves have different thermal inertias due to some difference in their construction?

PMM

[GrimJosef]

Yes, that is always an issue that has to be balanced. It's why I mentioned the nearly 4ms charge time, confirming that the 10,000uF used here isn't really immensely large. When I added the second 10,000uF I also put 0.4ohms in series with it, again to limit the charge time a bit. At the moment the last capacitor (originally the reservoir, now the smoother) is on the amp pcb. I plan to move it away from the audio components and closer to the power supply ones at the back of the chassis.

Important though it is, it's the same for both valves. So for now I've parked it to see if I can work out what's 'wrong' with the 'bad' valve and perhaps work out a circuit way of fixing its problem.

Cheers,

GJ

[GrimJosef]

Originally Posted by pmmunro

Could the filaments of the two valves have different thermal inertias due to some difference in their construction?

I wondered about that. The valves are modern ones from the same manufacturer (PSVane, altough I don't know where they were sourced from - here perhaps https://audionote.shop/valves/rohren...11-vt4c-psvane ?). They were bought together and look as alike as two peas in a pod. I'm not sure how to estimate the thermal inertia of a valve filament so not really clear whether I'd expect significant temperature change, and therefore emission change, at 100Hz. Is it known whether or not this is generally an issue ? Maybe it would be easiest actually to measure it rather than to try to predict it ?

Cheers,

GJ

[turretslug]

Could susceptibility to anode current modulation by filament hum be dependent on the degree of 3D symmetry of the electrode structure, particularly the cathode-grid space where the directly-heated emissive filament zig-zags up-and-down inside the grid structure? Perhaps slight displacement to one side, or bowing/sag of the filament (manufacturing spread/rough handling/hours running?) results in inconsistency/asymmetry (relative to an ideally central null point) of emission current distribution along the filament and consequent possible depth of hum nulling?

I don't think I'm explaining that very well- I think I know what I mean, but maybe not getting it across very well! I might work on the wording and express things better in the morning....

I first came across the distinctive-looking "spud-masher" high voltage version of the 5R4 rectifier with its matching deeply-recessed IO socket in a 211 amp that had a bi-phase full-wave rectified HT supply. More than a little frightening!

[Herald1360]

Does the hum disappear if the filament is fed from a 12V accumulator via a suitable resistor? (Or even five 2V cells in series?)


Is there any mileage in moving the grid ground point around physically with respect to the valve socket?

[joebog1]

I have "fiddled" with 211's. I had origional GE made in USA military ones and I had 5 to choose from. My circuit was ideally as yours using my hand wound OPT's. The heater supply is also similar to yours EXCEPT, I used a round ferrite toroid obtained from a dead UPS. Just black ferrite!!!, so don't ask for data. I wound some hand made Litz wire
(10 strands of 1mm enamel copper) onto this toroid and that drastically reduced the hum. I seem to remember that the winding on the toroid "fitted" the window space available without overwinding, around ten turns. I had originally used big 10,000 uF Rodestein caps, but after reading a bit on the net I ended up with a single Rodestein after the rectifier and 5 X 2200 uF graphite caps by Nichicon after the toroid choke.

Exact details I have forgotten and I didnt keep notes unfortunately.

These are just dim thoughts remembered,

Joe
Said toroid was about 1/4" square cross section, 1" centre hole and obviously about 1 1/2" diameter. In the end I gave up on the CRO and used my ears to detect minimum hum using the hum twiddler.

[kalee20]

I can't see any mechanism in the sketched circuit for producing hum. The filament supply is DC; the HT rail is low ripple; the grid is shorted to deck.

The 200Ω potentiometer - what is that for? To equalise DC current throughout the filament? It ought not make any difference to hum UNLESS the DC heater supply has a common-mode hum voltage to chassis. But Herald1360's idea of using a battery to power the filament would prove that.

Generally, I have considered it bad practice to power directly-heated valves with DC (unless specifically designed for it). There will be a voltage gradient along the filament, meaning the cathode-grid voltage is different at different points along it - so emission will be non-uniform and anode dissipation likewise. Better to power from AC. I have read (but can't quote references!) that filament voltages and currents are carefully chosen such that asymmetry caused by the voltage gradient, and electron flow perturbation caused by the magnetic field close to the filament, can be arranged to partially cancel.

Using AC on the filament will at least, over time, give uniform emission over the filament, and you won't 'wear out' one part of the filament prematurely. Failing this, I would want a changeover switch to reverse polarity from time to time, as was done I believe with fluorescent tubes operating directly from DC.

You could consider driving the heaters from an ultrasonic 25kHz AC source - a simple power oscillator will do that for you, and a ferrite toroid will make an excellent transformer to allow floating of the heater supply to the filament.

As to thermal inertia - my approach to get a feel for it would be to drive the heater with a pulsed supply (relay to interrupt it), nearly 100% duty cycle, and 'scope the anode current. Immediately on breaking the filament current, there will be a change in anode current due to a change in the voltage distribution along the filament; this will be followed by a gentle ramp as the filament cools. You can compare this ramp between samples of valves.

[turretslug]

Yes, Herald's battery (unambiguously ripple-free DC!) sounds (!) like a necessary next-step in the chasing-down of the culprit. The filament is being fed DC at the moment, but it is ripply DC even with GJ's extra filtering and it's in close proximity to a sensitive anode-current controlling element, viz. the grid. Difficulty in achieving an exact and symmetrical electronic centre-tap and precise balancing of the anode current-influencing hum component along the filament length as a result of mechanical asymmetry and emission unevenness could maybe limit the depth of null achievable with the external pot.

[GrimJosef]

Originally Posted by turretslug

Could susceptibility to anode current modulation by filament hum be dependent on the degree of 3D symmetry of the electrode structure ...

I suspect it could, however a) I think the details might be tough to work out (I was rather hoping this would have been done already back when directly-heated valves were more common, and someone here would either know the answers or could point me towards a book ) and b) somehow it feels as though this might account for a shift in the pot setting for a complete null but have much less of an effect on the depth of the null. If we imagine rotating the pot to the ends of its travel I think it's clear that the resulting 100Hz hum at the speaker will go from one very large value to roughly the same very large value but with the opposite polarity (180 degree phase change). Somewhere in between, actually very close to the centre of the pot's rotation, the hum from this source must either go through zero amplitude or through a 90 degree phase shift. There are no other ways from plus to minus.

Originally Posted by Herald1360

Does the hum disappear if the filament is fed from a 12V accumulator via a suitable resistor? (Or even five 2V cells in series?)

Is there any mileage in moving the grid ground point around physically with respect to the valve socket?

I haven't tried using a true DC source on the filament. I actually need some (adjustable) 100Hz hum there though, so I can use it to null the (fixed and unavoidable) 100Hz hum from the HT rail.

There might well be mileage in tidying up the grounding in this amp. I'm looking at it separately because as well as this 100Hz hum there is also a good deal of 50Hz (and odd harmonics) coming from other sources. As it happens, though, the 211 grid ground is very close to the bottom of the 1k filament resistor which should be the best place for it. I keep being forced back to the point that while this is interesting it's just the same for both the 211's I tried. Yet they have different null depths.

Originally Posted by joebog1

... The heater supply is also similar to yours EXCEPT, I used a round ferrite toroid obtained from a dead UPS ... and that drastically reduced the hum ...

Interesting. If it were my amp I might well have designed a different filament supply. And I might still see if I can lower the hum on this one (easiest would be by uprating the 10,000uF smoother to 47,000uF - it shouldn't cause serious charging spikes since it's not the reservoir cap). But I do still need to leave some hum at the filament so I can adjust it to null out the HT hum.

Originally Posted by kalee20

I can't see any mechanism in the sketched circuit for producing hum. The filament supply is DC; the HT rail is low ripple; the grid is shorted to deck.

Sadly the filament supply is far from DC (10V average with 2V pk-pk ripple) and the 70mV ripple on the HT mostly appears across the output transformer, generating hum on the secondary (some sums in my first post)

The 200Ω potentiometer - what is that for? To equalise DC current throughout the filament? It ought not make any difference to hum ...

Imagine one end of the filament - say the negative end - is connected to the 1k bias resistor. When the filament supply is at minimum voltage (9V) the average filament voltage will be 4.5V above 57V. When the filament supply is at maximum, 11V, the average filament voltage will be 5.5V above 57V. Simplistically that's a 1V pk-pk signal fed into the grid-cathode circuit (in practice the finite cathode impedance of the valve will reduce this, but not by much). 1V g-k will generate a lot of hum at the load ! The job of the 200ohm pot is to 'centre' the 1k resistor on the filament and thereby null out this signal. It's a very commonly used arrangement with DHT's.

Generally, I have considered it bad practice to power directly-heated valves with DC (unless specifically designed for it) ... Better to power from AC. I have read (but can't quote references!) that filament voltages and currents are carefully chosen such that asymmetry caused by the voltage gradient, and electron flow perturbation caused by the magnetic field close to the filament, can be arranged to partially cancel ...

There might well be an electrical case for an AC filament supply. But in an audio circuit the sensitivity required of the pot balancing to eliminate the resulting hum (which would now be at 50Hz unless the AC supply was exotic) would be extreme. With 2V pk-pk on a 10V average my hum voltage can easily double as I move a few degrees away from the null setting of the pot. This is a bad enough problem already. If I had 28V pk-pk (10V RMS AC) the sensitivity would be 14 times worse.

Cheers,

GJ

[turretslug]

It's difficult to visualise (and I accept that I might be barking up the wrong tree anyway!) but I was wondering if the mechanical or electronic asymmetry I've been thinking of results in a transfer characteristic that has a slightly different law either side of the null point (e.g. uneven edge effects that wouldn't be capable of complete annulment)- it might only be tiny, but "tiny" might still result in the observed 60-70dB variability of of achievable minimum as opposed to perfection. Does the harmonic characteristic of the hum change from maximum to minimum amplitude? We need someone thoroughly steeped in late '20s/early '30s AF power amps, I'm sure there would have been thorough in-depth research into the problems and their amelioration, though it is likely to be obscure nearly a century on!

[GMB]

Well I have tried doing the maths, and it looks to me like any non-linear response will generate a voltage that cannot be cancelled out by the pot, so the observed difference between the valves is a reflection of their slightly different linearity. I am amazed it is as good as observed.

To think of it graphically, imagine the gradient of emission along the filament that is constantly varying with the instantaneous filament voltage and view it relative to the central point. With linear response the high voltage section is up by the loss in the low voltage section. Your signal must be applied relative to the exact zero point but as this is not accessible the pot recreates it externally.

But any non-linear response to the filament voltage results in an imbalance. You might think that we can move the centre point with our pot to cancel it, but the snag is that the imbalance is a moving target. So if you move the centre point you no longer have the low side exactly mirroring the high side to cancel at all times. Another way to look at it is that the exact centre point at any instant is moving back and forwards during the cycle and so applies a signal relative to the centre point to be amplified by the valve.

The best residual hum is going to be harmonics of the original supply which unfortunately will make it more noticable that the original I suspect.

[trobbins]

Is the voltage gain of the output stage the same for each valve?

I would also recommend using a 12V VRLA and rheostat as a way to eliminate any parasitic voltage and current influences, even due to mains earth loops and heater charging current transients and stray fields from heater winding, etc.

If you had a spectrum analyser you could better confirm the harmonic levels (than just eyeballing a scope waveform), and use that to assist comparisons.

You could vary the decoupling on the cathode, and certainly make sure the loop area of the grid and cathode circuitry is minimal with a star 0V, as any induced voltage on the grid due to loop area and parasitic voltage on the cathode due to non-ideal decoupling could come in. There is also 14pF from anode to grid, which could creep in if star ground and loop area were a bit loose.

Is your PT suitably removed from the OT, and is the OT screened from any other stray wiring/equipment fields.

You could further attenuate the 100Hz ripple in B+ by tuning the choke.

[GrimJosef]

Originally Posted by GMB

Well I have tried doing the maths, and it looks to me like any non-linear response will generate a voltage that cannot be cancelled out by the pot ... To think of it graphically, imagine the gradient of emission along the filament that is constantly varying with the instantaneous filament voltage and view it relative to the central point ... any non-linear response to the filament voltage results in an imbalance. You might think that we can move the centre point with our pot to cancel it, but the snag is that the imbalance is a moving target. So if you move the centre point you no longer have the low side exactly mirroring the high side to cancel at all times. Another way to look at it is that the exact centre point at any instant is moving back and forwards during the cycle and so applies a signal relative to the centre point to be amplified by the valve ...

Thanks for the clarity there. I fear you're right. The fact that I am actually aiming not just to null the filament supply ripple effects but also the HT supply ones too may make the problem even worse. The null offset on the filament supply may increase the effects of the nonlinearities in that.

In this case perhaps the best solution will be to try to reduce the fundamental amplitudes of the HT and LT ripples as far as possible, since I think the nonlinearity will generate a residual whose own amplitude will be proportional (roughly) to those ripples.

In practice I don't have much more room for manoeuvre with the HT ripple. My current 5H 250mA choke is dangling outboard (with 1125V on it I have to remember to give it a wide berth on the bench !) but I think the largest commercial unit I can squeeze into the amp itself might be the 7H 150mA Hammond 159Q. I could raise the smoothing capacitor from 110uF (3 x 330uF in series) to 157uF (3 x 470uF) but whether the cost would be justified for 3dB or so reduction is open to question.

Reducing the LT ripple, which has a larger amplitude, by increasing the smoothing capacitors (one for each valve) might well be worthwhile.

Cheers,

GJ

[GrimJosef]

In reply to trobbins:

The output stage gain is indeed very similar for the two valves. The 12V battery would eliminate filament supply ripple and thus allow me to confirm the effects of the HT ripple. But to be honest I think those are already well understood so I don't think I'd learn very much. A battery wouldn't be a long-term solution because I need some filament supply ripple to null out the HT ripple (this is working very effectively in the 'good' valve). I think the chap who owns the amp (or more accurately, let's be honest, his wife) would also fail to see the funny side.

I do have a spectrum analyser - an HP3561A in fact, which I find invaluable for amp work. I mentioned above that I'm also dealing with some 50Hz noise in this amp. Without the analyser I wouldn't have been able to watch the 100Hz peak go up and down in the fixed comb of larger (when the 100Hz is minimised) 50Hz and odd-harmonic neighbours. It's also how I know the discrepant -59dBV and -70dBV numbers so very precisely.

The external layout points are well made and come under the category of 'hum hygiene' I mentioned in my first post. I'm doing what I can, but the manufacturer laid out most of the amp's circuit on a pcb, so flexibility is limited. He also used a steel chassis which distributes the magnetic noise pretty widely. But all of this stuff doesn't help me with the root problem - that one valve is performing very well while, in the same socket, the other one is more than 10dB worse.

Cheers,

GJ

[G6Tanuki]

Thinking of my time playing with RF balanced mixers (as used in SSB generation and direct-conversion receivers) which used something very similar to the "pot to produce a centre-tap" you have to set the balance, one issue there is unequal capacitance-to-ground from either side of the pot, which can make it difficult to get proper balance.

The answer in the balanced-mixer case is to add a small variable capacitance from one end of the pot to its wiper.

*which* end of the pot needs the capacitance added is determined by trial and error, same goes for the value of the capacitance. For balanced mixers a 3-30pF trimmer is usually used as the variable capacitance - at audio frequencies you'll need something a bit larger. If you put the capacitor on the wrong side it makes the null achievable by tweaking the pot worse.

I'm wondering if something like this - trying by substitution a few fixed-value capacitors [starting at maybe 0.05uF and going up to 1uF or more] might provide enlightenment?

[GMB]

Originally Posted by GrimJosef

The 12V battery would eliminate filament supply ripple and thus allow me to confirm the effects of the HT ripple. But to be honest I think those are already well understood so I don't think I'd learn very much.

I think assuming things is not a good idea. Try it.

With battery heater you can see what else you are getting from HT and stray pickup. You might want to also try the same but with the LT run into a resistor to check for magnetic pickup.

Then the extra stuff you get with the heater on a/c is from that and I suggest there will be two sources. Adjusting the pot should remove the 100Hz+harmonics from the direct effect, but leaving the non-linear effect as unremovable extras. Any extra 50Hz is likely to be capacitive coupling from heater winding to the mains or other windings - and unless the transformer is well screened then I would expect to see that too.

[GrimJosef]

I didn't say I'd assumed it. I said I thought I understood it.

What I'm doing is putting two resistors in series - an 11k one (representing the 6ohm load resistor reflected into the output transformer primary) and a 3k8 one (this is the anode resistance of the 211 with ~1000V a-k and 50-odd milliamps Ia - the resistance is printed in the valve's datasheet). The 70mV pk-pk is a modeled figure from the well-known psud2 simulation package. But this is very widely used and genuinely reliable and I made significant effort to set the component values correctly. It predicted the DC voltages on the reservoir and smoothing capacitors accurately so I've got no reason to believe it's got the ripple wrong.

What this boils down to is whether it's 'not a good idea' to believe that the behaviour of a resistive divider with a small AC voltage across it is easily understood. To be honest I think it is.

The experiment, however, is rather a long way from easy. It wouldn't just be a matter of 'trying it'. First I'd have to find a 12V car battery (I do actually have a couple in a synthetic mains supply, but getting one out would be a nuisance and I don't have any spare terminal connectors). Then I'd need a resistor to drop 2.5V or so at 3.25A, so a little over 0.7ohms at, say 10W. I don't have one of those either. As you say I should really load the disconnected filament supply with another resistor - 10V at 3.25A is about 3 ohms but now that needs to shift 32.5W. That's going to need a bit of cooling ! All of this stuff has to be connected (battery) and disconnected (filament supply) from the pcb, which has the humbucking pot on it. 211s have to be operated either vertically or (riskier) horizontally with their filament plane vertical. The latter would involve propping the heavy, somewhat delicate (projecting valves) amp at an awkward angle. The former would mean the new high current leads have to come out of the bottom, also a bit of a fiddle.

What really puts me off getting distracted by this though is that whatever the filament and HT supplies' properties are, they'll be the same for both valves. My problem is not a fundamental one of nulling the 100Hz hum. Having made the supply changes that I already have, I can now null that hum perfectly well in one valve.

My problem is why I can't null it so well in the other valve, when everything in the external circuitry is unchanged.

As far as 50Hz on the filament supply goes, there might be some. But I suspect there's a lot more 50Hz leaking into the (extremely sensitive) small signal stages. With these connected I can look at the spectrum analyser and use the pot to null the 100Hz peak from being much larger than all the others down to being more than 10dB below the 50Hz (and in one channel also the 150Hz) peaks. Turning the pot doesn't change the heights of these peaks at all.

Cheers,

GJ

[Radio Wrangler]

Could there be a mismatch in the relative positions of the centroid of the emissive surface and the centroid of the heater voltage?

David