KT90 Arcing Internally

[tweedeluxe]

I just built a single-ended KT90 amp. It sounds great, but the EHKT90 arcs internally when playing big low chords with the volume dimed. The symptom comes on more quickly with humbuckers/boosted input. I only let it happen twice, and I didn't notice which pins arced - though I could see that it occurred near the tube base.

The arc is only occurring inside the tube itself.

Specs:
-4k primary Z (custom 25W Heyboer)
-430V on the plate
-Cathode biased to 40W (max disss is 46W) at 90mA.
-I'm using a resistor in series with a rheostat for Rk, bypassed by a 100uf cap.
-Screens are filtered through a 15H choke/40uf (as is the preamp), but I added a 2k/22uF leg off that node for the screens, then finally a 1k8 screen grid R.

I can draw up a schem if my description is confusing.

I am looking for some advice. Does this just sound like a weak tube, or is there something obviously wrong with my design? I am assuming that the tube, for whatever reason, can't handle the high voltage spikes when pushed far into distortion.

If that's the case, would reducing the bias have any effect?

Thanks for any help, Lowell.

[Structo]

Did you figure out what the tube is biased at dissipation wise?

Most likely just bad tube(s).

They don't make them like they used to!

[Merlinb]

Over 400V is a LOT of voltage to put on the screen with a 4k load, it sounds like it's suffering (common with KT designs). You should really be looking for about 300V on the screen, or use MUCH bigger screen stoppers.

[Andy Le Blanc]

I have a KT90 PP at higher plate and screen voltage that seems very
comfortable with 2.2k screen grid resistors. Id suspect the tube first.

[tweedeluxe]

I have one spare EH KT90. I'll try that first before modifying the screen supply.

I chose a 4K load based on the New Sensor datasheet. It is my first time designing an amp directly from the curve traces/mutual characteristics graphs, so I am a little wary of my calculations.

However, the tube seems to be operating within about 15% of where I'd calculated it would - which I figured that was within the normal tolerance range. To be honest, I was surprised that the amp even worked!

[Firestorm]

What value grid resistor are you using? All KT tubes seem to be very fussy about this.

[tweedeluxe]

10K grid stopper, 100k grid reference.

I know the KT66 datasheet recommends 10k-50k for a grid stopper. The KT90 datasheet doesn't mention anything, but I assumed the same recommendation applied.

[Firestorm]

The newer datasheets don't tell you this, but the older ones for KT tubes always say it is "essential" to have separate cathode resistors for each tube (or separate bias sources in fixed bias amps). This sort of defeats the feedback effect of a common cathode resistor, but since you're using a bypass cap, you've sort of defeated the feedback anyway.

The KT90 EH was designed from some other family of tubes (I forget which), so it's not an upgrade of the KT88, but something entirely different. It is a "high emission" tube, meaning the cathode is especially effective at spitting out an electron cloud. Supposedly, tubes like this will spark internally if you exceed maximum plate current (and thus, I suppose, maximum cathode current) even momentarily. The audiophile sites say this is always a circuit problem; not a tube problem (it masquerades as one because it sometimes happens to only one tube). Also most common in new tubes (they still have the highest emissions).

So, my guess, you're probably biased too hot and hitting big chords with the controls dimed you're exceeding the current limits.

[tweedeluxe]

Firestorm - thank you for that information! I will try cooling the bias down and see if that helps.

This is a single-ended amp using just one KT90. I did build a PP amp with KT66s using separate cathode bias resistors and it sounded great!

Thanks for all the advice guys, I'll update the thread when I get it all figured out.

[Firestorm]

Oh, single-ended. That's a little different since it is running in Class A. You should be able to bias it hot. I don't have a spec sheet for the tube, but is your voltage proper for Class A operation? Is the load right for that voltage? The only other thought I have (besides the possibility of a bad tube) is that when dimed, the control grid is going positive WRT the cathode. I think the other rule with these tubes is the grid should never be positive or the electrons get unruly.

[tweedeluxe]

Well, I THINK the load is OK, but I am still new to this (designing directly from the datasheet), so I could very well be off. I certainly don't feel very confident now that the tube is sparking under heavy signal .

Thankfully I ordered an OT with 3 primary taps: 2.5k, 4k, and 6k. I figured that this would give me more than enough options, especially with a mismatched speaker load, in case I experienced problems like an arcing tube!

This thread has the curves that New Sensor sent me:
https://tubeamparchive.com/viewtopic.ph ... light=kt90

I had calculated that 4k should be just right with a plate voltage of B+ of 460V. My actual B+ ended up about 30V below that, which I suppose would make the load line move down below the knee of the screen grid curves.

I just double checked my bias and the tube is running at about 43W idle (EH KT90 rated at 46W). Though I am running in Class A, is there still a chance that the tube is biased too hot, causing the arc?

I'm a little out of my league here. Thanks for all your help!

[tweedeluxe]

Is there a way to prevent the grid from going positive (if that's indeed what's occurring)? Decreasing the grid reference R? Decreasing preamp gain maybe?

-Lowell

[Merlinb]

I assumed push-pull too! With 4k single-ended then you're definitely toasting that screen! As Andy said, 2k2 as a bare minimum on the screens, and personally I would increase the screen dropping resistor too. Or lower the transformer impedance (but then you'll hit cut-off sooner, although most SE guitar amps run closer to cut-off, and nobody seems to mind.)

[tweedeluxe]

I guess that's why they call you the Wizard! Advice taken, thank you!

[Firestorm]

Is there a way to prevent the grid from going positive (if that's indeed what's occurring)? Decreasing the grid reference R? Decreasing preamp gain maybe?

-Lowell

Since the problem happens with the controls dimed and is worse with humbuckers, you might try trimming the gain a bit in the preamp. Change a 100K plate load to 82K, or if you have an unbypassed cathode resistor somewhere. increase its value.