Elevating Heaters with No Center Tap?

[wizard333]

With all the talk of elevated heaters here lately, I haven't seen anything about how one goes about doing it without a CT on the PT for the heaters. How would you do that?

[R.G.]

With all the talk of elevated heaters here lately, I haven't seen anything about how one goes about doing it without a CT on the PT for the heaters. How would you do that?

I would construct an artificial center tap with two 100R/1% 1/5W resistors. Connect them in series across the heater supply, connect the center "tap" to your elevated voltage.

In many instances of lesser-quality power transformers, this can be a more accurate centertap than a winding one. Sometimes it's a lower hum option to use the resistors as a CT even if there's a winding CT.

[tubeswell]

Or in place of the 2 x 100Rs, use a 250R-500R pot with the input and output respectively connected one to each side of the heater winding, and the wiper becomes the CT, which gets connected to the elevated DC supply = an elevated humdinger. Should be really quiet when you dial the pot in (by ear)

[wizard333]

So on say a vintage Fender, where they create the virtual center tap over at the pilot light with the two 100ohm resistors, you'd lift those off ground, then that becomes what you'd connect your dc bias to?

Any reason for those to be 100ohms vs 1K, etc?

I like the idea of a pot to, brings the whole "hum balance" control into it along with the biased heaters.

Any suggestions for pots in that value range that can stand up to the constant DC? Can't really toss a 10k bias pot in there.

On the schematic Tubeswell posted, the resistors are labeled as precentages. Does the value matter so long as that ratio is maintained? Assuming you arent subbing for one of the 220k resistors across the resevior caps.

Also, if I wasn't subbing for one of the 220k resistors across the resevior caps, should I draw my voltage from further up the power rail, like the screen supply, PI supply, or pre-amp supply?

Thanks!

[defaced]

So on say a vintage Fender, where they create the virtual center tap over at the pilot light with the two 100ohm resistors, you'd lift those off ground, then that becomes what you'd connect your dc bias to?

Yes, exactly

Any reason for those to be 100ohms vs 1K, etc?

I think I know the answer, but I'll let someone else answer.

Any suggestions for pots in that value range that can stand up to the constant DC? Can't really toss a 10k bias pot in there.

I would wager that a cermet pot, the sames ones recommended for bias supplies, would be ok.

On the schematic Tubeswell posted, the resistors are labeled as precentages. Does the value matter so long as that ratio is maintained? Assuming you arent subbing for one of the 220k resistors across the resevior caps.

Sorta. For the voltage part of it, the ratio is all that matters. But if the resistors are too small, you'll tax the power supply. If they're too big, well, I don't know. 1M/100K is a pretty popular combination. It gets me about 50 volts from a 525v screen supply.

Also, if I wasn't subbing for one of the 220k resistors across the resevior caps, should I draw my voltage from further up the power rail, like the screen supply, PI supply, or pre-amp supply?

I don't totally follower the first part of your question about the 220k and the resivor cap. It sounds like you're talking about the balancing resistors usually used for stacked capacitors. But to answer the basic questions, I don't think I've ever seen elevated heaters pulled from anywhere but the screen supply node.

[pdf64]

What's the rationale for creating an artificial centre tap with elevated heaters?
My understanding was that the heaters should be elevated such that the (pre-amp) heaters are always above the cathodes.
That can be achieved by referencing 1 of the heater legs to say 20V (and leave the other leg floating) - why mess around with 100 ohm resistors / pot?

[tubeswell]

Well my take on the pot is that it works like two 100R resistors, only you can finely tune the balance between the two AC sides of the heater winding, which is theoretically better (but it is also just more knobs to fiddle with). Any 1/2Watt linear pot would be fine.

Whereas the elevation lifts the entire VAC wobble above the ground potential reference point of the cathodes, so that this doesn't interfere with the signal. It certainly makes it quieter.

I'd never heard of just ground referencing one side of the heater winding and leaving the other one floating before. I thought the whole point was to have both sides of the winding referenced evenly, in order to get the same amount of opposing VAC in each side, which cancels out better when the AC pairs are close/twisted together.(But my ejamacatian stands to be improved as usual)

As for the voltage divider, if it is coming from the HT the 220k total is a good standard load (doubles as bleeder circuit), and 90/10 could be achieved with two 390k 1W in parallel (giving 195K 2W) for the "90" in series with a 22k 1W for the "10", and put the 10uF cap in parallel with the 22k to get about 40V from a 400V B+.

But you don't need 90/10, you could go for 85/15, or 80/20 etc. You probably don't want to go much above 90V h-k elevation with modern 12AX7s to be on the safe side.

[otto pärt]

As for the voltage divider, if it is coming from the HT the 220k total is a good standard load (doubles as bleeder circuit), and 90/10 could be achieved with two 390k 1W in parallel (giving 195K 2W) for the "90" in series with a 22k 1W for the "10", and put the 10uF cap in parallel with the 22k to get about 40V from a 400V B+.

If my cypherin' is correct, you could use 1/2W resistors if you went with 470k + 56k or something like that. Is there downside to using even greater resistances? Or do you just lose the bleeder benefit?

[tubeswell]

If my cypherin' is correct, you could use 1/2W resistors if you went with 470k + 56k or something like that. Is there downside to using even greater resistances? Or do you just lose the bleeder benefit?

Nah 500k odd would still work, but the bleedin' would be slow (bleedin' slow - heh heh), and the dissipation with be double AFAICT.

[otto pärt]

Nah 500k odd would still work, but the bleedin' would be slow (bleedin' slow - heh heh), and the dissipation with be double AFAICT.

You're confusing me. The formula is V^2 divided by R, right?

[wizard333]

I'm clear on all of it except why the 2x100ohm resistors need to be 100ohm?

Fender used a 100ohm pot for hum balance which would be 2x50 ohm if it were centered for instance.

Any reason they need to be 100ohm? Or 50 ohm? What range of values works?

[R.G.]

Or in place of the 2 x 100Rs, use a 250R-500R pot with the input and output respectively connected one to each side of the heater winding, and the wiper becomes the CT, which gets connected to the elevated DC supply = an elevated humdinger. Should be really quiet when you dial the pot in (by ear)

This is a good candidate for being titled "Hum reduction by the method of offsetting errors". It's handy if you have other hum sources that you can't get rid of. So you introduce a slight cancelling hum from the heaters. It works great if:
- the other sources of hum are power line frequency, not power supply ripple frequency
- the other hum sources are either in phase with or out of phase with the AC line. If they're, say, 60 degrees out of phase with the line, no amount of pot diddling will get perfect cancellation. This can happen with magnetically induced hum sometimes.
- the summation of the other hum sources which you're cancelling do not change. If they change, the cancellation is wrong.

But you can often get usefully less hum this way.

[wizard333]

Is there a better way?

[tubeswell]

You're confusing me. The formula is V^2 divided by R, right?

Yep (V^2)/R = (V/R) x V = W (Remember that V/R = A, and V x A = W just a different way of arriving at the same answer)

So (450V/500k) x 450V = 0.0009A x 450V = .405W

Whereas (450/250k) x 450V = 0.0018A x 450V = .81W

[tubeswell]

100R is good because they don't get too hot and 1/2W is okay. If they were 50R they would heat up more and 1/2W wouldn't be enough. But it'd still probably work, although quite a bit less current would be available for the filaments (and don't forget the 6.3V lamp), so you have to be careful about the load on the winding.

On the other hand 200R-250R per side should work okay as well AFAICT. But if the resistance got too high, (say several hundred kohms) then you lose the ground reference I guess (but I never tried going really high).