300B SwitchMode filament supply

[tomchr]

Forgive me Father for I have sinned... I have rigged my 300B to ... <sob> ... a SWITCHMODE SUPPLY! On the filament no less...

So I've been designing this 300B amp. I like the anode follower. I'm using only two 6J5's total - not two in parallel as shown in the schematic in above thread. I really like it. But so far I've been running the filaments on lab supplies.

Then I started designing the power supply. I've settled on using regulated supplies for the HV and the filaments. The HV supply is easy. The Maida regulator (LM317 with a BJT cascode to handle the high input voltage) is a proven concept. The power dissipation (worst case) for the Maida regulator in my setup is about 10 W (5.6 W typ). But the filament supplies are another story. At 5.0 V, 1.4 A for the 300B, 6.3 V, 600 mA for the two 6J5's, the filament supplies would dissipate 16 W (worst case -- 11 W typ). So the heat sink for the power supply needs to dissipate 52 W for two FIVE WATT amplifier channels. I'm sorry. That's ridiculous. My worst case isn't even that conservative - basically, I'm assuming that the mains voltage will be 10 % high and I disregard that the power dissipated in the regulators will be slightly less due to ripple on the supply.

So I started looking at switchmode options. I generally default to National Semiconductors for stuff as I like their data sheets and their sample program. The parts are good too... I chose the LMZ12002 for the 5.0 V supply as it only requires a few passives to get to run. Unfortunately, its max output voltage is 6 V, so I can't use it for the 6.3 V supply. For this, I used an LM2734. I threw together a small prototype. After running for an hour, the LM2734 is up to about 60 deg C, and the LMZ is barely lukewarm (40 C). My need for a heat sink in the filament supply just evaporated.

Sound... It sounds like the midrange has cleaned up and become even more transparant. The bass seems tighter. It may be a product of my imagination, but at least I don't experience any significant drawbacks.

According to my measurements, the noise floor is 15 dB (!) lower with the switchmode supply. And there is a bit of clean-up happening around 7~8 kHz. I think the 'garbage' at 7~8 kHz with the linear supply is due to a stability problem in my ancient Sorensen 40-2 supply that I use for the 300B filament.

Measurement results, schematics and such attached. 0W_8R_Overlay.gif shows the noise floor (no input signal); 1W_8R_Overlay.gif shows the output spectrum for 1 kHz at 1 W into 8 ohm. Enjoy.

[Lingwendil]

Looks like a good design choice to go switchmode supply for the filaments, heresy aside. You may be able to get away with AC for your preamp tubes though, but if not, switchmode seems a good choice.

[CLS]

Well done, congratulations!

Would it be the switching noise of the rectifiers in your linear supply? Those 7~8kHz 'garbage' you said? And what are the upstream circuits for the SMPS chips (J1 and J3) see? They supposed to be DC sources, right? What's your solution for those?

[analog_sa]

Originally Posted by tomchr

I think the 'garbage' at 7~8 kHz with the linear supply is due to a stability problem

Very cool thread. However, it will be more beneficial to compare noise and sound to a decent heater supply and not to 'garbage'. I have never seen any decent linear regulator generate such noise.

Have you tried a CCS for heaters? Or straight AC? Either will provide some reasonable sound reference.

[Serge66]

Son, your sins are forgiven.

As it is exactly how I intend to power up the heaters of my next project: PP with 6N1P - 6SN7 - KT88. However, I will use industrial SMPS from Traco Power.

Cheers,

Serge

[Sch3mat1c]

Why not the HV? Haven't gotten that far yet?

Now can you do it with a non-isolated winding, instead of wasting two floating windings on them?

Lastly, maybe you'll realize the 300B is a crummy old dinosaur and there are better ones out there... maybe a class D EL519 >

Tim

[tomchr]

Originally Posted by analog_sa

Very cool thread. However, it will be more beneficial to compare noise and sound to a decent heater supply and not to 'garbage'. I have never seen any decent linear regulator generate such noise.

I should clarify. I do realize that the buzz my Sorensen supply puts out is not ideal. But I take the output from the Sorensen into a 15000 uF cap and from there into an LM317 that sits right next to the 300B and provides the 5 V. So there is a fair amount of regulation between the Sorensen and the 300B filament. The buzz at 7~8 kHz is from the Sorensen. The noise floor, I bet is from the LM317.

Originally Posted by analog_sa

Have you tried a CCS for heaters? Or straight AC? Either will provide some reasonable sound reference.

I have tried an LM317-based CCS. I was not able to distinguish between CCS and CVS. I have not tried AC, as I don't really care for 60 Hz IM products.

~Tom

[tomchr]

Originally Posted by CLS

Would it be the switching noise of the rectifiers in your linear supply? Those 7~8kHz 'garbage' you said?

The buzz at 7~8 kHz is a stability problem with the linear supply. I'm sure it's caused by an electrolytic cap that's dried out. I can easily see the buzz on the output voltage from the Sorensen supply when using an oscilloscope. That said, I do follow up with a hefty cap and an LM317, so the Sorensen really only serves as a pre-regulator.

Originally Posted by CLS

And what are the upstream circuits for the SMPS chips (J1 and J3) see? They supposed to be DC sources, right? What's your solution for those?

Currently, they are HP lab supplies. HP 6236B. In the final implementation, they'll be fed from a 12 V, 50 VA transformer through the usual arrangement of diode bridges and caps.

Originally Posted by Sch3mat1c

Why not the HV? Haven't gotten that far yet?

Not a bad idea. But it's a project for another day. I can deal with the power dissipated in the Maida regulators pretty easily...

Discrete switchmode supplies tend to open up a Costco-sized can of worms. I did once design a 2.5 kV SMPS supply for a research satellite, but that was 10 years ago. Given the low amount of power needed for this project ... hmmm... NO! Must resist!

~Tom

[Sch3mat1c]

Why resist? Give in.. give iiiinnnn....



Beef this up... ooh, it's even automotive



Versions for 90-250VAC input, 100W abound...

Tim

[tomchr]

Originally Posted by Sch3mat1c

Why resist? Give in.. give iiiinnnn....

Versions for 90-250VAC input, 100W abound...

Damn you! Ahhh... The good ol' Unitrode controllers. I remember those.

Zo... I'll need roughly 160 mA at 400 V. I'd probably design for 200 mA (80 W). That's starting to look a lot like a forward converter, though, a flyback like you're using might work. If I power it from a 12 V trafo, I technically won't need the galvanic isolation. But if I power it off the mains, I would. No biggie, though, it makes the feedback (or switch drive) a bit more tricky. Last time I used two opto couplers for the feedback and some tricks to linearize them. Worked well.

If you're doing the 85~264 VAC thing, you might as well throw in the power factor correction...

I may have asked you before, but which sources do you recommend for magnetics (cores, bobbins, transformer tape)?

~Tom