B+ voltage drop via Zener/MOSFET question

[casey73]

I've rebuilt an old Kay 720 for a friend. It had been modified far from the original and it sounded thin with no bottom end. I told him I'd set it up kinda like a Princeton with 6L6 outputs. The original PT had been replaced with a Hammond 278X, and it had 545V on the plates when I first turned it on. I put in five 15 volt, 5 watt zeners and a 20 volt 5 watt zener in series on the PT center tap to ground, and got the plate volts down to a more reasonable 455 volts. I had the zeners around so I used them. This is my first zener/center tap B+ voltage drop attempt.

I then set the up the pair of 6L6s as cathode bias with two 470 ohm 5 watt resistors in parallel giving me at cathode resistance of 234 ohms. I measure 31 volts at the cathode and 413 on the plates now. If my math is correct I'm drawing 132 ma current, which comes out to about 25 watts per tube.

It sounds good, but with the chassis open still open I smelled something getting hot, saw one of the zeners smoking and shut down. I figured 18 watts of dissipation across six 5 watt zeners would be 3 watts per diode. Either my math is off or I'm pushing the zeners to hard.

I found RG Keen's MOSFET B+ drop circuit (MOSFET Follies) so thought I'd try that. I happen to have a couple IRF130 MOSFETS around. I will use a 100V, 5 watt zener in conjunction with the IRF130 (max 75W dissipation). The MOSFET will be running less than 30% of max dissipation, so I don't think it will get real hot? I have just enough room for a small TO-3 heat sink in the chassis. According to the schemo on RG's site, the drain goes to ground, so I shouldn't have to insulate it from the chassis. Am I missing anything here?

[R.G.]

... I smelled something getting hot, saw one of the zeners smoking and shut down. I figured 18 watts of dissipation across six 5 watt zeners would be 3 watts per diode. Either my math is off or I'm pushing the zeners to hard.

I'm pretty sure it was the 20V zener that was smoking. All the zeners had the same current, so each had a power being generated in it of 15V times the current ... except the 20V one. That one is dissipating 33% more than the 15V ones. It's probably the one that fried.

I found RG Keen's MOSFET B+ drop circuit (MOSFET Follies) so thought I'd try that. I happen to have a couple IRF130 MOSFETS around. I will use a 100V, 5 watt zener in conjunction with the IRF130 (max 75W dissipation). The MOSFET will be running less than 30% of max dissipation, so I don't think it will get real hot? I have just enough room for a small TO-3 heat sink in the chassis. According to the schemo on RG's site, the drain goes to ground, so I shouldn't have to insulate it from the chassis. Am I missing anything here?

There are some details here that the devil can hide in. the 75W max dissipation is probably stated on the data sheet at 25C *case temperature*, which is almost impossible to do unless it's bolted to a solid copper block which has ice water flowing through holes in it. I took a quick look - yep, that's specified at 25C case temp, which you can't get to.

Something that's not widely appreciated is that there is NO correlation between the amount of heat generated inside something and the temperature it reaches without also stating how hard it is for that heat to get out. The best illustration is a grain of wheat incandescent light bulb. This may dissipate a watt or less, but the filament gets hot enough to glow white. When something generates heat, its temperature rises until the heat leaves at the same rate it's being generated.

For most silicon semiconductors having the junction above about 150C is the dead line (literally). You can dissipate more power in it if you bolt it to a heat sink that gets the heat out at a lower temperature drop. The critical point will be whether the MOSFET/chassis combination can get the heat out at a low temperature rise.

You're planning on dropping 100V, and 132ma. Actually, I'm guessing that's idle and it may rise a bit at full output. Call it 150ma. So the power to be gotten rid of is 100V*0.15A = 15W. Not too bad.

Oops. I looked at the maximum voltage on the IRF130. It's specified for a breakdown of 100V. It will almost certainly die in this application. I'd advise going for a power device that can take at least 600V, ideally 800-1kV. This makes it much more robust if there's a nanosecond of a transient. These are easily available and not that expensive, seeing that you only need to get one rated for over a couple of amperes Id and 50W or more power. You'll need to use a gate protection zener if there's not already one in there.

If you use the chassis for a heat sink, be careful to bolt the device to a flat spot and use thermal compound between the device and the chassis. It might be even better to form a "U" from some aluminum sheet for the device to rest in and bolt that to the chassis, again with heat sink goo between the device, the aluminum and the chassis.

I'm frequently amazed at how much of power electronics depends primarily on how you channel the heat away, no matter how clever the circuit is.

[trobbins]

As RG indicates - solid state junctions are much more friable (fryable ) than tube plates. Most designers don't go above 110C junction - and you really need to know the ambient temp of the heatsink and the thermal resistance of the heatsink and the design equation that links those values to the junction temperature.

You have a good example of this with the zeners - they are only 5W rated with excellent heatsinking, and soldering one to another to another doesn't allow the heat to escape via the leads, so you'd be lucky to dissipate 1W safely from each zener. Plus the position of the zener is very onerous, because the current has a high crest factor. So although you'd expect 2.6W dissipation in the 20V zener, which will fry without a good heatsink, the junction temp will be even worse with those peak currents.

[casey73]

I'm pretty sure it was the 20V zener that was smoking. All the zeners had the same current, so each had a power being generated in it of 15V times the current ... except the 20V one. That one is dissipating 33% more than the 15V ones. It's probably the one that fried.


There are some details here that the devil can hide in. the 75W max dissipation is probably stated on the data sheet at 25C *case temperature*, which is almost impossible to do unless it's bolted to a solid copper block which has ice water flowing through holes in it. I took a quick look - yep, that's specified at 25C case temp, which you can't get to.

Something that's not widely appreciated is that there is NO correlation between the amount of heat generated inside something and the temperature it reaches without also stating how hard it is for that heat to get out. The best illustration is a grain of wheat incandescent light bulb. This may dissipate a watt or less, but the filament gets hot enough to glow white. When something generates heat, its temperature rises until the heat leaves at the same rate it's being generated.

For most silicon semiconductors having the junction above about 150C is the dead line (literally). You can dissipate more power in it if you bolt it to a heat sink that gets the heat out at a lower temperature drop. The critical point will be whether the MOSFET/chassis combination can get the heat out at a low temperature rise.

You're planning on dropping 100V, and 132ma. Actually, I'm guessing that's idle and it may rise a bit at full output. Call it 150ma. So the power to be gotten rid of is 100V*0.15A = 15W. Not too bad.

Oops. I looked at the maximum voltage on the IRF130. It's specified for a breakdown of 100V. It will almost certainly die in this application. I'd advise going for a power device that can take at least 600V, ideally 800-1kV. This makes it much more robust if there's a nanosecond of a transient. These are easily available and not that expensive, seeing that you only need to get one rated for over a couple of amperes Id and 50W or more power. You'll need to use a gate protection zener if there's not already one in there.

If you use the chassis for a heat sink, be careful to bolt the device to a flat spot and use thermal compound between the device and the chassis. It might be even better to form a "U" from some aluminum sheet for the device to rest in and bolt that to the chassis, again with heat sink goo between the device, the aluminum and the chassis.

I'm frequently amazed at how much of power electronics depends primarily on how you channel the heat away, no matter how clever the circuit is.

Thanks RG! I was hoping you'd chime in. Not just an answer but a very practical lesson in thermodynamics. I had a feeling my idea was too good to be true. The IRF130 was from a solid state amp I was attempting build around 1990 when I knew even less than I know now about electronics, and just before I discovered tube amps and abandoned solid state, as I found tubes to be much more forgiving.

I hadn't paid attention to those specs outside of dissipation, which you pointed out can be misleading, using your "solid copper block which has ice water flowing through holes in it" metaphor My sum total of MOSFET experience are the BS170 and 2N7000 in stompbox applications. So I either have to opt for less voltage drop or find a better MOSFET. I found one locally, an NTE2387, TO-220 (which I have ready made aluminum heat sinks for) and the specs say it's pretty rugged with a breakdown voltage of 800V. It's about $8. I've heard NTE often sells components that fall way short of specs, so I'm a little reluctant, but am willing to take a chance on $8. Are there any other specs I need to pay attention to? Also, the spec sheet I saw does not show an internal gate protection diode, so what type would you recommend for gate protection?

[R.G.]

... So I either have to opt for less voltage drop or find a better MOSFET. I found one locally, an NTE2387, TO-220 (which I have ready made aluminum heat sinks for) and the specs say it's pretty rugged with a breakdown voltage of 800V. It's about $8. I've heard NTE often sells components that fall way short of specs, so I'm a little reluctant, but am willing to take a chance on $8.

Mouser sells several 800-1kV parts in the TO-220 package for under $2 each. I can suggest some part numbers if you want.

Are there any other specs I need to pay attention to?

One good one is the derating factor for power dissipation with the case over 25C. What that number really is is the inverse of the thermal resistance from chip to case. Sometimes they'll specify the thermal resistance separately too.

Thermal resistance is the analogy to Ohm's law for heat. If you consider temperature to be a "voltage" and heat flow to be a "current", then Rt = T/Q where T is the temperature difference and Q is the heat flow rate. It's usually quoted in degrees C per watt. The nice thing about doing it this way is that thermal resistances add like electrical ones, and so if you have 1.47 C/W from chip to case, 0.5C/W from case to heat sink, and then 2C/W from sink to air, then the junction temp goes up 1.47+0.5+2 = 3.97 degrees C per watt dissipated. The only real fly in this ointment is that the resistance from sink to air is nonlinear, and critically dependent on orientation and air flow rates. The numbers I quoted are reasonable quesses for the kind of setup you're describing, so if you dissipate 15W as you're envisioning, the junction will heat about 60C above the air temp around the sink. So if you're under 150C-60 = 90C, the junction is OK. Most equipment never gets above 40-50C, so there's some margin.

The 1.47 C/W I used is from one of the TO-220 devices I looked at. It's representative.

Also, the spec sheet I saw does not show an internal gate protection diode, so what type would you recommend for gate protection?

The gates on most MOSFETs are specified at +/-20 to +/-30V with respect to the source. A good way to protect them is to put a 12V zener with anode to source and cathode to gate. This keeps the gate from ever going more than 12V above the source or 0.7V below, which is fine in this application. A 100R resistor right at the gate is handy for preventing it from becoming an RF oscillator at frequencies too high to see on mid-range scopes 8-|

Most TO-220 and bigger MOSFETS have transconductance (Gm) ratings of 1-3Siemen; a Siemen is the inverse of an ohm, or amps per volt of change on the gate. So if you change the gate by a volt above the threshold voltage, the current in the channel increases by Gm, 1 to 3 A depending on the device. You're looking at fractional amps, so the gate voltage will always be only negligibly above Vt.

There is a lot more you can do with this kind of setup. You can build in current limiting, for example, although the circuitry gets a little tricky.

[casey73]

[QUOTE=R.G.;341376]Mouser sells several 800-1kV parts in the TO-220 package for under $2 each. I can suggest some part numbers if you want.

At the risk of sounding lazy, suggested part numbers would be great. Mouser's on-line catalog filters seem rather Byzantine at times unless you have a specific part number on hand. Waiting a few more days won't matter since my friend isn't in a big hurry.

[casey73]

And it is pretty difficult to "heat sink" a diode the size of a half watt resistor. Point well taken, lesson learned. Take dissipation specs with a large grain of salt.

[R.G.]

No problem. I usually avoid saying this in public, but I kind of like the Mouser catalog filters. It's a pain to get used to, but if you once start thinking of it as sequential database queries to narrow search results, it gets handy.

OK, so I'm a little strange.

Here's one I would consider: Mouser stock number 511-STP3NK90Z, $1.79 each. It's nominally rated for 90W, 900V, 3A, but more importantly with a Tr of 1.38 C/W to help get the heat out. They have over 500 in stock. It's really intended for switching applications, but like most of them, it will work in linear mode if you get it right. This one has internal gate zeners so you don't need to put an external gate protector zener on it.

[casey73]

Thanks R.G.. I'll let you know how it turns out.

[casey73]

Thanks R.G.. I'll let you know how it turns out.

I got the zener/mosfet circuit together using a 100 volt zener and got the voltage in reasonable range, but the heat was pretty extreme, and it's not practical to add more heat sinking because lack of space. So I went with a 43 volt zener and it was barely warm, but the voltage was still too high. This morning I used a 56 volt zener but the voltage is still to high.

Can I run two zeners in series in the MOSFET/Zener combo?

[R.G.]

Hmmmm... where was the heat extreme? The zener or the MOSFET? In general, the MOSFET is the one that should be hot, and the zener will be nearly room temperature. The current in the zener is just the current across the 2K resistor, which is only the MOSFET threshold voltage of maybe 4-5V. That puts 1/4W in a 100V zener.

And yes, you can stick two zeners in series from the drain to the gate of the MOSFET.


If you meant "Can I use two MOSFETs, each set up as an amplified zener, in series?" then the answer is still yes. In fact, that's a good way to spread out the heat dissipation.

The physical setup is the challenge, it seems.

[casey73]

Sorry R.G. I should clarify. The heat from the MOSFET mounted on the heat sink, which is mounted to the chassis (with heat sink silicone on all surfaces) got real hot real fast. After two minutes, it was too hot to hold my finger on the heat sink. The diode and resistors mounted to a small perfboard were fine. This is mounted inside the chassis. There is no room for much additional sinking outside the chassis.

With the 56 volt zener, it is barely warm, but the output is idling at 32 watts per tube. I did fire it up, turned out the lights in my garage and played it dimed for a few minutes, and got no red plates or orange glow swell that might indicate the screen were getting to hot (1K resistors on the screens). But I'm still not real comfortable with it running this hot. I think an additional 15 volt zener should get me there.

[R.G.]

OK, go for version 2.

Temperature is the result of how fast heat can be removed from a heat source. Heat sinking keeps temperatures lower by reducing the thermal resistance (degrees C per watt of heat) from the heat source to ambient (air, in this case). You can also cut temperature for the same amount of heat being generated by dividing the heat source into several pieces, each with their own path to ambient air.

So you could use a second MOSFET/zener/resistors/etc. in series with the first one, each running at half the voltage you want to drop, and the same current, as they're in series. However, you really need to make these two heat dissipators have separate thermal paths to air. And one of them needs an insulator from chassis ground, as it's floating on the source of the first one.

Another issue may be the spreading resistance from the heat sink to the chassis, or the chassis thermal resistance to the outside air. The MOSFET can be fine, the heat sink fine, but the chassis simply not able to get the heat into the surrounding air very well. When the heat sink is too hot to keep your finger on, how hot is the chassis just away from the mounting bolts, and then 1" and 2" away? If it's not nearly as hot, or is only hot at the mounting bolts, the chassis isn't carrying its part of the bargain in transferring the heat to the air. If that's true, the chassis is simply not a good heat sink, and the prospects get dim without modifying the chassis.

[casey73]

Hmmmm... where was the heat extreme? The zener or the MOSFET? In general, the MOSFET is the one that should be hot, and the zener will be nearly room temperature. The current in the zener is just the current across the 2K resistor, which is only the MOSFET threshold voltage of maybe 4-5V. That puts 1/4W in a 100V zener.

And yes, you can stick two zeners in series from the drain to the gate of the MOSFET.

If you meant "Can I use two MOSFETs, each set up as an amplified zener, in series?" then the answer is still yes. In fact, that's a good way to spread out the heat dissipation.

The physical setup is the challenge, it seems.

Okay. Short story, I got it working and working correctly. When I reattached the MOSFET/perfboard module with the new Zeners I failed to connect the ground, which is why it didn't get warm and why my voltage didn't drop as anticipated. I'd claim rookie mistake but I've been doing this stuff too long, so I have to claim senior moment. Once I got it right I still had considerable heat to deal with. I had some aluminum angle stock laying around so I cut two pieces about 6" long and shaped them to fit against the back of the chassis, overlapping them so the form a U shape. I used the screw/nut used to secure the MOSFET to the inside of the chassis to hold them on, with an additional small sheet metal screw to secure it. It gets warm but not hot so it's doing the job.

Mucho thanks R.G. for helping me out on this one. I returned the amp to my friend and he was quite happy with it.

[R.G.]

Hey, congratulations! Good work!