Help with dropping B+ with Zeners

[Dr Tony Balls]

Alright, it appears that i'm limited in the performance on the two multimeters I have. Only one measures AC current, and in ranges of up to 200mA and up to 10A. So i'm losing significant digits on the 10A setting but it's coming up with .24A (240mA). This is more than I was calculating (~220mA) but not wildly so. 240mA and a 12V drop comes out 2.88W per zener which I would think would be fine with a 5W rating. Am I crazy or does that sound reasonable?

I'm stumped by this. A lot of folks seem to use this option for dropping B+ with success and while I know that it's not the only option I cant understand why its having problems in this case. I suppose I could try the balancing resistors that RG hinted at? Or a new brand of diode, possibly with a 10W power rating. Or the power mosfet/zener option.

[bepone]

2.9W is too much for those small zeners.. this is a lot of heat!
also 1W causing a lot of heat! not even mention 3W,
usually the best is to take 3-4x times more power on passive element which you are not cooling down , than you are consuming (resistors etc)

[bepone]

kill 20V with 50w resistor in Ub+ line, make cathode bias which will eat another 10V, and you will be fine with the amp

[bepone]

critical is voltage on g2, so Vg2 must be around 300V.. so dropping resistor to Vg2 is more important than voltage on anode (EL84 are working also on 400V on anodes with low Vg2 voltage), this 360 now - 12 V on cathode is Vgk= less than 350.. acceptable even without touching Ub line, but Vg2 must be corrected

[R.G.]

I may be confused. You mention AC current scale on the meter. The zeners are getting DC only, I think. Maybe I read that wrong.

On some items bepone brought up. You have a surplus of voltage. You can also put a temporary resistor in series with the zeners. A 1 ohm resistor would only drop 0.24V, and a 10 ohm would drop 2.4V. The voltage across these resistors would show the current, perhaps on a better scale for your meter.

If your amp runs class A and the current does not vary much, then a dropping resistor would work, excepting when the tubes are warming up and current is low.

Yes, devices getting hot is in general worse than devices staying touchable. However, a 50% power derating is generally OK to start with. I looked at the datasheets for some 5W rated zeners in this package. Through hole zeners actually get rid of most of their internal heat by conduction out through the leads. The datasheets show that the devices are rated for 100% of the rated 5W up to a lead temperature of 75C / 167F and about 3.3W at lead temps of 100C/212F. So yes, they're getting hot, but not catastrophically so. 2.5W is OK, in my internal checklist, they're not running too hot. It still really worries me that two of them get hot, the others don't. Something is wrong there. Maybe even two of them being marked the wrong way around for anode/cathode, or internally shorted; something.

Zeners in series with the rectifiers have higher heating than zeners in series with DC after a filter cap. The RMS current in rectifiers is bigger than the average current after the first filter cap But still, that's not likely to be what it happening. Maybe, not likely.

You could solder "wings" of copper to the joints where the zeners are soldered to the tag strip. This would suck heat out of the leads very effectively for cooling them, but of course the heat sinks are at high voltage and need to be protected from shorts and accidental touching.

The resistor paralleling the zeners is the passive version of the active zener with MOSFET. They act the same, with the paralleled resistor or MOSFET conducting current to keep it out of the zeners. Just sayin'
It suffers from needing a heat sink too. But a 600V / 5A 49W MOSFET (IPA60R1K5CE) is in stock at Mouser for US$0.92. That particular one is a full pack, with epoxy insulation so the heat sink is insulated from the device. There are probably other, better fits, but this came up in my "sort by cheapest" on Mouser. Yeah, the MOSFET amplified zener sounds complicated, but it's not much more complicated. To the the four zeners, you add the MOSFET, two resistors and a gate protection baby zener if the MOSFET doesn't already have one built in.

But I get off into the tall grass easily. There is still something really wrong with the basic zener string you have. The basic zener string should be working.

[Dr Tony Balls]

Cutting to the chase on this...

There is still something really wrong with the basic zener string you have. The basic zener string should be working.

Thanks, that's what it seems to me as well. The method seems popular and trusted and i'm just not sure why its not working in this case. ESPECIALLY with the irregularity between diodes in the string.

2.9W is too much for those small zeners.. this is a lot of heat!
also 1W causing a lot of heat! not even mention 3W,
usually the best is to take 3-4x times more power on passive element which you are not cooling down , than you are consuming (resistors etc)

Thanks for the reply. I guess I didnt think that 2.5W or so would be too much heat for a 5W diode. Especially given the test condition in an open chassis with plenty of place for any heat to rise off.

kill 20V with 50w resistor in Ub+ line, make cathode bias which will eat another 10V, and you will be fine with the amp

Right i'm aware of the alternatives. Not necessarily looking for how to shed B+ but the reason why my implementation isnt working.

I may be confused. You mention AC current scale on the meter. The zeners are getting DC only, I think. Maybe I read that wrong.

The diode string is between the PT center tap and ground, so that should be AC, no?


I think at the end of the day this is going to require a Mouser order. 1) for some other similar Zeners of a different make just to sanity check this excercise. And two a mosfet/zener combo to try another implementation. Something like this:

[sluckey]

The diode string is between the PT center tap and ground, so that should be AC, no?

No, that's dc. Current may be pulsing but it's only flowing in one direction, ie, direct current.

AC current would be flowing in two directions, alternating between the two directions, ie, alternating current.

[R.G.]

I guess I didnt think that 2.5W or so would be too much heat for a 5W diode. Especially given the test condition in an open chassis with plenty of place for any heat to rise off.

IMHO, you're right. I would not hesitate to do that. I might sub in some bigger dissipation types later after the thing worked and I decided to cool things off for longer life.
Funny thing is, the diodes that are acting weird, with little or no voltage drop, are the ones that are not hot. The hot ones are the ones acting more normal. So something is weird.

Sluckey's right. The diodes on the ends of the two B+ half-secondaries force the return currents going back to the CT to only go one way, so it's pulsed DC.

[Dr Tony Balls]

The diode string is between the PT center tap and ground, so that should be AC, no?

No, that's dc. Current may be pulsing but it's only flowing in one direction, ie, direct current.

Well in that case i'm measuring 160mA, which explains things even LESS. That's measuring between the CT and the top of the diode string. If I bypass the diode string and measure from CT to ground i'm measuring 185mA.

Using the PT current calc here (https://thesubjectmatter.com/calcptcurrent.html) this is what i'd expect to see for 4 x EL84 and 3 x 12AX7:

Selected transformer voltage: 290-0-290

Calculated voltage at first capacitor (B+): 397.3V. Subtract 6V, if using choke.

Calculated filament current (typically the 6.3v secondary): 4.03A

Calculated current: 214.78mA at 4000R load (10% plus factored in). Preamp valves current draw is estimated at typical 12AX7 max dissipation of 1.2W at 330V, i.e. 3.6mA per triode

[R.G.]

Well in that case i'm measuring 160mA, which explains things even LESS. That's measuring between the CT and the top of the diode string. If I bypass the diode string and measure from CT to ground i'm measuring 185mA.

It gets complicated. Most meters that measure "AC" don't really measure AC, they use a capacitor to block real DC, then a diode to rectify the peak of the "AC" and get average rectified DC and report that voltage, but scaled down to correct the peak voltage down to something like RMS for a sine wave. It works fine for measuring mostly sine-wave things, like 50-60 Hz sines. It works poorly on other waveforms, like especially high peak waveforms and ones mixed with DC. Some fancy "true rms" meters work great with pulses, AC, DC, whatever, but most of lesser meters use the rectify-and-guess technique.
The current waveform flowing through the zeners is high-pulse, low duty cycle DC, which throws off the assumptions on rectify-and-guess. Even on DC scales, there are oddities about peak to average pulse trains. I guess the advice is to not sweat the exact value until you figure out why two diodes do and two diodes don't. You're close to the theoretical value, figure out later why the meter doesn't tell you the expected value.

If I was at home now with my "math and engineering" computer, I'd zap up a simulation of the circuit and run some waveforms and readings. But I only have my laptop with me now, so I had to break the glass cover over my brain for emergency use.

Overall, if it were me, I would put a 1 ohm or 10 ohm resistor in series with the zeners to measure current, then fire it up and measure (1) the voltage drop across the resistor and (2) the voltage drop across each diode individually.

Well, actually, I would disconnect those zeners from the circuit, and use a power supply and series resistor on clip leads on each diode to see the voltage drop on each one both forward and backward. Something like a 15V-24V DC power supply, or even three 9V batteries in series, series-ed with a 1K resistor. You ought to see 12V one way, 0.7V the other way on each one. This tells you open, shorted, cathode marking backwards, mislabeling, etc. It's quicker than the amount of typing already expended on this. Sorry - I have developed low patience with simple devices that act funny. I also like Bob Widlar's solution to new-but-defective parts which was to Widlar-ize it: put it on an anvil and pound it into dust with a ball-peen hammer.

[Roe]

If the power amp works roughly like an ac30, I'd try the following CRC filter:

rectifiter to mains cap 1: 32-47uf

big power resistor: 50-100R 10w

main caps 2: 20-33uf.

A 50R resistor will drop 10v at idle and 12,5 at full distortion with a typical ac30 circuit with a shared 50R cathode resistor.
A 100R will drop twice of that.

If this is insufficient, you can add an extra power resistor between the ground of cap2 and ground of cap1 (disconnecting cap1 from ground)

[Dr Tony Balls]

Update on this following a part order...

First, I did put a 1ohm resistor between the center tap and the start of the Zener string and measure 160mV across it, so to answer the question we're drawing 160mA of current.

So initial attack was to try redoing the Zener string with new diodes. I know there are other, potentially better solutions but I have to satisfy my curiosity. Made a new string with new 1N5349B, 12V, 5W diodes and got largely the same results. The overall drop was more like 9 volts per diode, instead of 12, but that's within reason. But sure enough after about a minute they start to smoke. And, wildly, the temperature measurements are the same as before! Going CT > D1 > D2 > D3 > D4 > Gnd i'm still seeing D1 and D3 at ~180-200F but D2 and D4 are at room temperature. How they could possibly be acting that differently and in the same spot blows my mind.

I suppose I could try higher powered zeners but they start getting expensive and big. I could try a string of eight 6V/5W zeners to achieve the same result but that also starts taking up space and may be prone to heat issues nonetheless, even if they are less egregious.

I then went with what I expected I was going to wind up on anyway...the MOSFET and zener solution. I went with an IRF840 MOSFET, 47V zener, and a 220R and 2.2K resistor. Seems so far to work like a champ. I'm measuring ~37V drop between the CT and ground and initially 308V on the plates. Temperature wise its just bolted to the chassis, but i've had it on for ~10 minutes and no smoke. I hit it with the infrared thermometer and it reads at room temp. Touching the chassis underneath the MOSFET its hot, but not burn-a-finger hot.

That had me at like 86% dissipation, so I rebiased it. Now getting 301V on the plates, 293 on the screens with bias at 95.8%. I still have zero clue why those zeners behaved the way they did, but I think i've got everything in order.

THANKS EVERYONE FOR YOUR HELP AND INPUT!

[Dr Tony Balls]

Also for anyone curious here's my implementation:

[bepone]

nice..but whats about 95.8% bias??

[Dr Tony Balls]

nice..but whats about 95.8% bias??

not sure the question. For a cathode biased amp I usually shoot for 95% dissipation. Do you or others disagree?