Will This Work? 26 Linestage

[Rod Coleman]

i am concerned about the filament hum issues described and the relative complexity/cost involved in attempting to alleviate them. Especially in terms of my own abilities, budget and patience. This is a project I would actually like to finish Two things come to mind batteries and smps's. Batteries would seem to be the simplest, dumbest, brute force method and I would guess be very quiet. I will start looking but if anyone can point me to particularly good thread or web page on using them for dht filaments I would appreciate it.

Most chemistries of battery are noisy, sad to say. Lead-Acid gel-cells are the worst of all, and conduct a wide spectrum of load-dependent artifacts into the filament. In terms of energy density, these are the only reasonably costed choice, but they are not hifi!

SMPS can be used, but they usually show a leakage-current (to ground) that is a serious fraction of the anode current, and will inject noise. If your 26 cathode is routed to round, a current will flow. The SMPS involve an electrolytic cap across the filament, and this certainly degrades the sound. Some builders like these for output DHTs, but they are unlikely to be quiet enough in this application.

You're welcome to copies of my PDF manuals for the regulator kits to see how they work in an amp - the cost of these for a pair of 26s is comparable to a pair of 4Ah gel cells.

Thomas gives good advice: unless the power supply design is first rate, it is better to work on a 6J5, and invest what you can on good basics. DHTs operating in a noisy environment do not sound well.

 

[andyjevans]

I'm surprised the 126C is $74. Earlier this year I was buying them in the UK from Bluebell Audio for £26 each. That's really good value.

Andy

 

[funkytek]

funkytek: I just assumed AC filaments would be to noisy. Will have to do some more research.

Yes, you should try it. As others have mentioned, AC on a 26 can be a hum problem. But you should try it first, with a high quality filament transformer and careful wiring. Then go to a DC supply if the hum is too obnoxious.

What I found was that the AC sounded better. But, for me, the hum was not workable, especially with my old globe tubes. I ended up with an LCLC DC supply but Rod's boards are an elegant alternative that would be much more compact. Actually mine was an LCLCL where the last choke was a common mode choke. It was custom wound but the Bottlehead FC1 looks like a good commercial choice.
Bottlehead Filament Choke FC-1
My supply was in a separate chassis, with the plate supply, but I put the choke in the signal chassis right at the tube sockets.

 

[tomchr]

I had a lot of problems with LM1084 switching off.

Nine times out of ten -- if not 99 out of 100 -- the LM1084 switching off is caused by overheating the LM1084. It has built-in thermal protection and will shut off once the junction temperature hits 150 deg C (approx).

Many people don't realize how much power they're dissipating in the filament regulators and, thus, don't bother with the thermal design.

~Tom

 

[andyjevans]

Hmmm - I have the LM1084s on big heat sinks with good thermal pads. I'm surprised, but maybe the heat isn't dissipating fast enough. I suspect it's more that the supply voltage is over 22v

andy

 

[tomchr]

I use a switchmode regulator for the filament in my 300B amp. Works and sounds great. I can't tell a difference in measured performance or sound quality between that and a linear supply (constant voltage or constant current).

The LMZ12002 from National looks like it'll fit the bill just fine. It's a bit spendy, but the layout is extremely easy and it has a nice soft-start feature. On my 300B filament, I use a 5-second start-up.

~Tom

 

[tomchr]

Hmmm - I have the LM1084s on big heat sinks with good thermal pads. I'm surprised, but maybe the heat isn't dissipating fast enough. I suspect it's more that the supply voltage is over 22v

The two go hand in hand. Here's the thermal math:

Pdissipated = (Vin-Vout)*Iout
Trise = Pdissipated*ThetaSA
Theatsink = Tambient + Trise

Use SI units all the way.

I don't know what "big" means to you. You need to know the thermal resistance (ThetaSA). Is this a "big" heat sink? It probably would be to some... Its thermal resistance is 3 degrees C per Watt (C/W). If you were to use this heatsink, the thermal math would look like this:

Pdissipated = (22 - 1.5) * 1.05 = 21.53 W
Trise = 21.53 * 3 = 64.6 degrees C
So in a warm living room, assuming free air flow across the heat sink and fins oriented vertically, the temperature of the heat sink would be:
Theatsink = 30 + 64.6 = 94.6 degrees C.

The thermal washer probably adds another 0.2 C/W. In addition, there's a thermal resistance from the case to the junction, ThetaJC. In the LM317 datasheet from National Semiconductor, this is specified to be 2 C/W.

So....

Tcase = Theatsink + Pdissipated*ThetaCS = 94.6+(21.53*0.2) = 98.9 deg C
Tjunction = Tcase + Pdissipated*ThetaJC = 98.9+(21.53*2) = 142 deg C

That's beyond the recommended operating point of +125 C and definitely within the range where I'd expect the thermal shutdown to kick in.

The easiest way to fix this is to lower the input voltage (different trafo). Alternatively, you can use a much, much bigger heatsink. I have some that measure 30x15x5 cm, weigh about 2.5 kg each, and have a thermal resistance of 0.4 C/W. They would allow your regulators to operate at 86 deg C junction temperature. That's a big heatsink...

Or you could use a switchmode supply. Those tend to run at about 90 % efficiency for those kinds of output voltages, currents. So Pdissipated = 0.1*Pload. Or Pdissipated = 0.1*1.5*1.05 = 52.5 mW. Heck, you can dissipate that in the part alone. No need for a heatsink...

In above, I assumed you were powering a 26 filament (1.5 V, 1.05 A).

~Tom

 

[andyjevans]

Hi Tom

I'm using a 10Y - 1.25A filament. Right now I'm using the LM1084 in current mode. 21v in, 18.7v out. Heat sink is 4" by 2" by 1.5" and is cool to the touch. That works. It didn't work in voltage mode - kept switching off on fire up. Even with the LM1084 in current mode it's not ideal - with 12v at the cathode the filament is at 6.7v not the specified 7.5v. That would require the cathode to be at 12.5v for a 10R resistor (filament 1.25A) and thus the input to be at 20v - 12.5v bias plus the 7.5v filament. So the supply should be 23v or more. I prefer the sound of the voltage reg mode, and would also like to end with a CMC which would absorb a little more voltage - around 1v. Does this help?

 

[Rod Coleman]

Hi Tom, I believe Andy's amp is in filament bias mode, as in my diagram in post above:

http://www.diyaudio.com/forums/tubes-valves/179197-will-work-26-linestage.html#post2400596

So if the chip delivers 18.7V out, and a 10R bias resistor is used, then for current-connected mode [1.2A]: bias resistor drops 12V, filament drops 6.7V = 18.7V - OK.

so the thermal load for the LT1084 is 2.3V x 1.2A = 2.76W
The heatsink Andy describes will work fine.

At switch-ON the story is changed by the filament resistance dropping to ~1R. So transient thermal pulse is 1.2 x 7.5 = 9W, which still should be OK, providing the heatsink was below 30 deg C at switch-ON.

For voltage mode, at switch-ON you get a current ~18v/(10+1) = 1.6A, and a drop of 10V =16W

The LT1084 in TO220 pkg, which shows 2.7K/W to the heatsink (more like 3.2K/W with a mica insulator). To keep the chip below 125 deg C, you'll need to keep the rise below 100 deg C. At 16W, that's 6.25K/W - or 3K/W from the heatsink, bare minimum! THat's a bigger sink than Andy describes, so the cutout is not surprising.

 

[Rod Coleman]

Andy, I'll get your kit made up for the filament bias operation. Would you like it for #26, or for 10Y? actually I can send both resistors so that you can adapt it.

Take care that you have enough voltage headroom in each case - 4.5V minimum up to about 8V. For 10Y and 10R, your proper output voltage is 20V, so 25.5V to 26.5V is the target for the raw supply.

With (say) 6V headroom, you get about 6W burned in 2 TO220s, and the heatsink should be 5 or 6K/W.

 

[simon2a3]

The two go hand in hand. Here's the thermal math:

...

That's a great explanation Tom, I've only ever half understood it before but you made the maths easy.