shutting a tube off with a relay

[Gainzilla]

I'm working on a project that uses an MCU to automatically set the bias of a pair of tubes (individually). It measures plate voltage, cathode voltage drop across a 1Ω resistor, and adjusts to the desired setting in realtime. It has an app you can use to change the bias, tube type, look at overall tube hours, see historical data, etc. If it detects a fault, it can trigger an AC relay to lift the hot leg of the AC inlet to protect the amp from damage.

What I'm curious about is the possibility of shutting down the actual failing tube. I'd thought initially it would be easy to sever the cathode connection to ground for the bad tube, but I've read that could put voltage on the chassis? I could add a relay to switch off the grid voltage like a half power switch, but the concern there is that the relays would likely be very large and expensive. I guess the question is, is it safe to "turn off" a tube by disconnecting its cathode path to ground?

The initial design is for a pair of tubes in push-pull, which I realize would prevent the amp from making sound. Alternately, maybe it lifts all tubes, and the app lets you know which tube is bad (along with a status LED on the actual amp)?

Thanks in advance.

Bryce

1_GridControl.png

[xtian]

Sure, I've seen designs that disconnect cathode from ground to mute the power section.


Push-pull amps still make sound with only one functioning tube!

[Gainzilla]

Nice, thanks. I'll play around with it.

[nuke]

Maybe a low-side mosfet switch instead of a relay?

Just need to be sure the Vds is rated for full supply voltage.

[pdf64]

A problem with lifting the cathode etc is that a pretty common output valve failure mode has the HT somehow shorting to the heater.
So cutting the HT feed to the affected circuit might be better.

[Gainzilla]

Maybe a low-side mosfet switch instead of a relay?

Just need to be sure the Vds is rated for full supply voltage.

Do you mean on the cathode or on the HT instead of a relay, like PDF mentioned?

[Gainzilla]

A problem with lifting the cathode etc is that a pretty common output valve failure mode has the HT somehow shorting to the heater.
So cutting the HT feed to the affected circuit might be better.

That’s a good call. This is why I wound up just cutting the main power instead. But now I’m wondering if a high voltage mosfet WOULD be ok for HT…

[R.G.]

I did a variant of sense-the-current-and-shut-down, including a bad-tube indicator in about 2003, and a processor controlled auto-bias in about 2005. The auto-bias didn't go into production because I didn't have the computing power available on a uC to separate the zero signal bias current from the shifted average currents when audio was going through the output stage. By the way, look into that and be sure you're picking out the right bias points with various signal levels.

One approach I got running was a bias-me-now button. You pressed the button, held it down until the green I'm-all-biased-now light came on. The button muted the audio path and told the auto-bias to go do its work. Took just a few seconds and didn't have to mess with confusing audio levels.

We actually went with an indicated bias; each tube had a red/green/blue LED that was tied to current sensing. Blue is too cold, red is too hot, and green is a tiny, tiny band around the desired bias current. In practice, people biased amp in about five seconds per tube, staring into the hole in the back of the amp with the bias indicator. You could re-bias between songs if you were so inclined. Turned out that the people twisting a screwdriver and watching LEDs were about as fast as the controller.

No phone app for it, of course.

A high voltage MOSFET will work fine for turning off the HT in an amp. I did it with the overcurrent sense and hold app. Worked great. I used N-channel, but with proper gate switching design, it would work fine with a P-channel (or, well, N-channel for that matter) high-side MOSFET. You just have to get the gate switching right. I never ran into the short-to-heater, so N-channels in the cathodes worked fine. But high side is probably smarter.

I designed a number of power supply controller things with MOSFETs that might fit into a similar scheme. One was a current clamp for the HT. (http://geofex.com/Article_Folders/mosfe ... 0Clamp.pdf) that prevents overcurrent surges into filter capacitors and hit tubes coming out of standby. A single optoisolator can turn that that thing completely off from an isolated logic signal. The circuit could be simplified and enhanced to be one-per-tube. An LED-to-photovoltaic isolator instead of a resistor self bias would make it much lower loss and might let you get away without a heat sink.

[Gainzilla]

Wow, that's amazing. I kind of like the idea of the bias now button. Very clever!

My setup will keep each tube individually biased to whatever range you want. The default value is "Normal" which is like 60%. It will work out of the box without the GUI. There is a 3 color LED that provides visual feedback upon startup, during normal usage, and if there is a fault. Non-nerdy users will just be able to plug a set of tubes in and play.

The cool thing is, though, that with the app, you can choose any type of tube your amp can support, and even mix and match tube types. Matched or not matched, shouldn't matter. Their amp will keep the tubes perfectly biased, regardless of wall voltage. Variac? cool. Still biased. The app just gives them greater control and visibility into the behavior of their tubes over time, plus the ability to change bias settings on the fly. There's even a fan header and temp header.

ANYWAY, I'll definitely have a look at high voltage N-channel Mosfets to cut power to tubes. That would take things back into he realm of feasible.

Oh, yeah, the MCU has a lookup table of tube data. Just selecting a different tube type should load the appropriate tube data, and therefore, be able to calculate the appropriate bias points.

Cheers!

Bryce

[R.G.]

This thread got me started thinking about the circuits again. The high-side MOSFET may be trickier to do because of the MOSFET body diode.

This may need more thinking or very high voltage MOSFETs. In a P-P stage, the turned-off tube plate swings to nearly two times the HT voltage by transformer action. A similar thing happens in an SE stage because the OT has stored up enough energy to supply the signal in the turning-off direction. A high-side MOSFET's body diode will be forward biased by the higher voltage and clamp the off-side plate in P-P to the HT supply. In SE, the output signal going high on the tube plate will similarly be clamped to the HT supply. Several things could happen here, but none of them are likely to be good.

The MOSFET switch would probably need to be a two-MOSFET setup in series, like the common MOSFET output AC relays. The LED-photovoltaic setup could still run this; but the cost for MOSFET doubles. You can get an LED-PV for about $3, and about the same for a 1kV/3A N-channel MOSFET. So if you did one switch per tube, you'd need about $9+ of semiconductors for the switch.

[nuke]

In addition to the body diode/flyback-effect, the high-side, n-channel switch also needs a gate drive that can supply voltage higher than the source pin to keep it turned on, yet not exceed the g-s voltage maximum.

I've been down the high-side switch rabbit hole a bit on a non-audio application at lower voltages.

[R.G.]

In addition to the body diode/flyback-effect, the high-side, n-channel switch also needs a gate drive that can supply voltage higher than the source pin to keep it turned on, yet not exceed the g-s voltage maximum.

Yep. That was one reason I mentioned the LED-PV drivers. They're too slow to run fast-switched power supplies and PWM stuff, but they're fine for driving on-off switches. The LED light makes a truly floating 7-10V source for driving floating gates. The Vishay VOM1271 ( https://www.vishay.com/docs/83469/vom1271.pdf ) was the one that caught my eye this time around. It's in stock at $1.87 at Mouser. The datasheet shows the two-MOSFET series connection that would be needed.

[nuke]

Good tip on the VOM1271. That looks like a good choice for a number of applications. I guess it generates the g-s voltage from pure magic, but as long as it does what it says, would do the trick.

[R.G.]

Per Arthur C. Clarke's Third Law: "Any sufficiently advanced technology is indistinguishable from magic".

It's a small solar cell ("photovoltaic") illuminated by an LED plus a little integrated junk to make it turn off faster.

[lonote]

I am following this out of interest, but I have no real knowledge of SS & while I understand the concept, I am at a loss to imagine the implementation (mostly).

Could you guys maybe rough-out a sketch of how something like this would work?