What happens when you starve a heater?

[leadbelly]

So, for my EL84M amp, the power xfrmr has a 6.3V winding. I was planning to use this for the DC heater supply, using LM1086 low dropout regulators for current limiting to 760 mA.

So, 6.3VAC * 1.414 bridge - 1.414VDC rect losses - 1.5VDC reg dropout = 6VDC, just a tad low.

Anything bad happen if you starve a heater slightly?

 

[cerrem]

Yes and NO.....
If your just building a quick and dirty circuit...then 6V is fine and will be fine....
If your building a Hi-End very "anal" phono-stage..ect..
The you will get reduced emmision.... assuming yo are using hand selected valves..
If you were to replot the plate curves at 6V and then design to that then that would be OK.... But one would statistically argue that the emmision variation is no worse than standard process variation..and 6V is within the spec sheet working voltage ..
For example:
The U47 mic uses a VF-14 tube which was intended for 60V filament.....In the mic circuit it runs at about 32V ....But this was done on purpose to reduce the space charge while being wired for TRIODE...This was to reduce the grid current since there is a 60MEG resistor on the grid....
The plate curves in the TELEFUNKEN data book do not apply anymore...I did a trace of the plate curves with 32V on the filament...wow, what a totally different tube characteristics..Plate resistance obviously was increased ....
Getting back to the question...
I would go about the DC conversion differently....
But I remember doing a trick by adding a diode and resistor to a LDO to get the 6.3V ...I will go home and check my notes and get back ...

Chris

 

[wa2ise]

Starved heaters on small stuff like 12AX7s can actually improve linearity. But you can't expect to draw the published max rated cathode current. This would be an issue with a power output tube. But DC on power output tubes won't really make it less hummy. If current limiting is what you're after, a simple series resistor will get you close to the benefit you are looking for. Rectify and filter the 6.3VAC to about 9VDC. Then each tube gets a 3.5 ohm power resistor from that in series with its heater. This will make for a more gentle inrush current on the heater vs. direct connection to the heater winding.

BTW tube heaters are designed to take a fair amount of high inrush currents. When tubes are manufactured they run the heaters at about twice rated voltage as they suck the vacuum and flash the getters (to bake out the trapped gasses inside). If it can take that, normal life is not a big deal to the heater.

 

[Brion55]

I wouldn't think .3 volts on the heater would make that much of a difference. Variations in line voltage, transformer windings, etc. probably account for wider variations in AC heater supplies than that.

Have you actually tried to breadboard the circuit to see what the actual voltage loss is?

 

[rdf]

Here's one opinion on the matter, less voltage = longer life, from Bill Pearl's site if I recall. Somewhere I have another graph from a different source demonstrating any deviation from 6.3 volt high or low results in rapidly decreased tube life. Take your pick.

Edit: Incidentally, the latest audioXpress has an interesting article full of measurements showing a massive (tenfold) reduction in Intermod Distortion is possible by tuning filament voltages in the case of DHTs.

 

[ray_moth]

I reckon you'll be OK. I live in Indonesia where the official mains voltage is 220v @ 50Hz. At my address it actually measures between 190v and 200v. My amp uses two transformers, both made by Amplimo and rated for 230v on the primary, with no taps for lower input voltages.

The main tranny, has, in addition to its HT and bias windings, a 6.3v winding rated @ 6.7A. I use this for 4xEL34s and am getting, at best, 6.3x200/230v = 5.4v. The other tranny has two 6v windings rated @ 6.8A each, which I use for 1x6D22s, 2x6SL7s, 2x6SN7s and 2x6AU6s. From this tranny, I'm getting only 6*200/230v = 5.21v.

The amp sounds great but the power must be limited and I daresay it would sound better with 230v mains. I'll have to rig an autotransformer to get better performance, I guess.

 

[Colt45]

thats only 5% out, no big deal

 

[ilimzn]

There is a GREAT difference between tungsten, thoriated tungsten filament direct heated cathodes and oxide coated cathodes in this respect. You cannot use the bright tungsten graph posted above to conclude much about indirectly heated oxide coated cathodes.

 

[leadbelly]

Thanks for all the replies. I'm sure I will sort something out, even if it means adding an extra xfrmr.

But now I'm more intrigued, because unless I'm mistaken nobody really answered the theoretical question, just the practical. My example is of when you are literally starving the tube, i.e. both current draw and voltage are fixed. The examples of undervoltage alone don't completely apply to this case because if you just fix the voltage at some value then the current draw will settle to a corresponding point on the heater curve, right?

 

[PRR]

> There is a GREAT difference between tungsten, thoriated tungsten filament direct heated cathodes and oxide coated cathodes in this respect. You cannot use the bright tungsten graph posted above to conclude much about indirectly heated oxide coated cathodes.

Very correct.

Note that this Bright Tungsten graph shows temperatures like 2,500 degrees. You have to run tungsten that hot to get any emission out of it.

Now find data for COATED cathodes (such as EL84 etc etc). Temperature is more like 1,000 degrees. Looking again at the bare tungsten graph, at 1/2.5 times the temperature, life is up to about 150,000%! Or about 1,500 times longer than a bare tungsten filament.

Oxide-coated heater life is NOT an issue.

Under-heating Thorium impregnated Tungsten while working at high currents is a problem. But oxide cathodes generally have WAY ample emission, so they never "strip" and there is no problem running them +/-10%, even +/-20% away from nominal voltage.

> 6.3VAC * 1.414 bridge - 1.414VDC rect losses - 1.5VDC reg dropout = 6VDC, just a tad low

The "low" is a total non-problem. But you have neglected to account for ripple. If a DC meter reads 7.5V into the regulator, and you have 1V p-p ripple (not an unlikely value), then you will have -0.5V pips on your "6VDC" output, which totally defeats the purpose of a DC heater supply (those pips will contaminate all the audio).

Tube heaters do not need regulation. Ray's Indonesian power is marginal (-15%) and he's happy. Take your AC, rectify, put in a BIG cap, a low-ohm high-watt resistor, and another BIG cap. 10,000uFd at 10V. A big C-R-C filter delivers smooooth power. Fiddle the resistor to get 6.0 to 6.5V at the heaters.

 

[ilimzn]

6.3VAC * 1.414 bridge - 1.414VDC rect losses - 1.5VDC reg dropout = 6VDC, just a tad low

The "low" is a total non-problem. But you have neglected to account for ripple. If a DC meter reads 7.5V into the regulator, and you have 1V p-p ripple (not an unlikely value), then you will have -0.5V pips on your "6VDC" output, which totally defeats the purpose of a DC heater supply (those pips will contaminate all the audio).

Agreed, and this is interestingly, a rather common mistake.
If your heater supply is overdimensioned and you really MUST regulate it (see below) you may want to use a voltage doubler to get around the dropout problem, but it's still a crutch.

Tube heaters do not need regulation... Take your AC, rectify, put in a BIG cap, a low-ohm high-watt resistor, and another BIG cap. 10,000uFd at 10V. A big C-R-C filter delivers smooooth power. Fiddle the resistor to get 6.0 to 6.5V at the heaters.

If I may add something:

People tend to forget that tube heaters are not resistors. They are nonlinear and to an extent are self-compensating: Lower voltage means lower temperature means lower resistance means higher current means more power means higher temperature.
If this were not so, series heating would never be possible.

Regarding the CRC filter mentioned above, I would actually make it a RCRC filter. The problem with rectifying low voltage and high current transformer windings is the need for very large capacitors, which in turn decreases the rectifier conduction angle, and increases the peak rectifier current. This in turn does two things which you REALLY do not want:

1) The reverse recovery time lengthens because of the higher forward current, resulting in 'inexplicable' overheating of the rectifier and filter caps. This is often 'solved' by fast soft recovery or schottky diodes.
2) The high current pulses generate substantial magnetic fields from the wiring and especially from the transformer as well. They can momentairly saturate the transformer at which point they become a stray field. They can and do also bleed through other windings to other power supplies, especially if a toroidal power transofrmer is used (remember that other rectifiers conduct at the same tim, probably with a larger conduction angle). Since the pulses are very short, their high order harmonic content is very high. Fast recovery or schottky diodes will actually make this worse, often producing HF hash into the MHz range. Self resonant transforemr problems are also possible, as damping is relatively low, since the caps, rectifiers, and winding all have low resistance.

A simple resistor in series, before the first smoothing cap (and you will be dropin voltage on a resistor anyway so no big deal) will widen the conduction angle, reduce peak current (often dramatically) and very cheaply solve all of the above problems.

 

[anatech]

Yes, absolutely. That's the way I do it.

-Chris