Current regulation for filament of say a 300B

[Bas Horneman]

Apparently one of the shortcomings of say a LM317 or even a LT1086 as current or voltage regulator is that they essentially do little attenuation above 1k.

So 50/60Hz is dealt with effectively but that is about it....in terms of a filament supply.

Matthijs de Vries's solution (MachMat.com) is to use a fet. This got me thinking.

Could I simply make a current source with say an IRFP240 and have something with much higher impedance so that a lot more rubbish above 1kHz is attenuated?

Or is the problem with such a device that it needs a higher voltage across it to work? A typical application I would want to use it for is say the filemant of a 300B.

 

[Rod Coleman]

I have bored everyone to death about how much you can improve over LT1086 or any 3-leg regulator.

current source featuring short feedback loop is needed together with a gyrator on the other filament terminal.

Here is the original design, still going strong. It has been built by other folks on the forum, and found to be a real improvement -

http://www.diyaudio.com/forums/showthread.php?s=&threadid=38248&perpage=30&highlight=&pagenumber=2

You can experiment with a FET in either the gyrator ( may work well) or the current source (you'll need a high performance scope to check you're not oscillating at 1 to 50MHz) but I prefer BJT here.

You can easily check when your 300B filament design is working well - just do a quick comparison to ac-heating. If it is not better in every way you can imagine - you need to keep working on it!

 

[Bas Horneman]

Thanks Rod,

I remember that one. I was just too afraid to build something that complex. But I think if I concentrate enough. I might just pull it off!

Regards,
Bas

 

[Sheldon]

I remember that one. I was just too afraid to build something that complex. But I think if I concentrate enough. I might just pull it off!

Rod's version has three elements; DC supply, regulator, and filament balancing circuit. If you look at the lower left quadrant of the schematic, you'll see that the actual regulator part is not really complicated. Adding the dual transformers and spiffed up cathode resistor set up, makes it look more complicated. But, and Rod will correct me if I'm wrong, you would gain the greater part of the benefits by using the regulator even with a more conventional DC supply and cathode resistor.

I ended up buying Guido's kit, because I didn't have room for more transformers and because it has a low enough drop out to use the regular AC filament winding. If you are starting a new amp, these issues may not be present.

Sheldon

 

[Bas Horneman]

I ended up buying Guido's kit, because

I thought about that too. And diyhifisupply's and the one by Matthijs. But due to my parts buying sprees in the years gone past. I thought to myself...why should I get myself into even more debt if I can build something half decent with cheap parts...after all I'm a diy' er. Obviously not the type to come up with solid state designs..but one that could build something simple that someone else designed.

I'm still thinking about my original question. After all a DN2540 n-channel mosfet makes a pretty good CCS. Take something similar but one that can handle the current...and voila? A filament CCS with a very high impedance?

you would gain the greater part of the benefits by using the regulator even with a more conventional DC supply and cathode resistor.

That is what I thought as well.

So I was thinking that when I've got some time to re-draw the thing so that it would become evident( to me) how to wire it up.

 

[Rod Coleman]

correct me if I'm wrong, you would gain the greater part of the benefits by using the regulator even with a more conventional DC supply and cathode resistor.

I think that is fair. If the build looks daunting, construct a regular transformer-bridge power supply to try it.

But some things really should be observed:

- don't share the trafo winding with the other channel or connect anything to ground - or the 300B cathode circuit will be shorted;
- use schottky diodes for the bridge/FW rectifier;
- good elko caps in the supply please! Panasonic FC or Nichicon low impedance - they are cheap, after all.

You can compare the sound with ac-heat to see how well you are doing.

 

[Rod Coleman]

I'm still thinking about my original question. After all a DN2540 n-channel mosfet makes a pretty good CCS. Take something similar but one that can handle the current...and voila? A filament CCS with a very high impededance?

The DN2540 would probably be well worth trying. just be sure to put the circuit in the negative side of the circuit (connect the DRAIN to the filament, and the source resistor/gate to power supply negative).

Using the Drain as a current source is superior to the source circuit, since the source transitions into a LOW impedance node for frequencies outside the control loop's response. The drain is high impedance for all frequencies, other than spoiling effects of parasitic capacitance.

Remember to allow enough volts across the FET to prevent unwanted capacitance effects (2-3V should do.. I couldn't find a data sheet for it tho)

FETs make good gyrators too, but N-channel enhancement type are best (IRF740s are in many a DIYer's toy box). Try 1M gate resistor and 470nF film cap to V-).

Remember to use an insulating pad/shouldered washer to attach each FET to a big piece of chassis or heatsink.

 

[Sheldon]

I'm still thinking about my original question. After all a DN2540 n-channel mosfet makes a pretty good CCS. Take something similar but one that can handle the current...and voila? A filament CCS with a very high impedance?

My knowledge is not exhaustive by any stretch of imagination, but devices meeting that description are not common (the 2540 only handles about 150mA). The only one I know of, is the Lovoltech LU1014D. You might be able to make a single device version or even a cascoded CCS. But it's a TO251 package, which is much smaller than a TO220. I made an amp based on Susan Parker's Zeus (you can find it here). I used some low temperature solder to mount them directly to copper heat sinks.

Sheldon

 

[Bas Horneman]

The DN2540 would probably be well worth trying. just be sure to put the circuit in the negative side of the circuit (connect the DRAIN to the filament, and the source resistor/gate to power supply negative). Using the Drain as a current source is superior to the source circuit, since the source transitions into a LOW impedance node for frequencies outside the control loop's response. The drain is high impedance for all frequencies, other than spoiling effects of parasitic capacitance. Remember to allow enough volts across the FET to prevent unwanted capacitance effects (2-3V should do.. I couldn't find a data sheet for it tho)

I'll draw up something next week or so. I used the DN2540 as an example...since it is often used in tube amp. But it can only pass 150mA hence the question about the IRFP 240 that I have a few extra of. (Still planning on one day using them for a Zen V9)

 

[Rod Coleman]

The IRFP240 can be used, but unlike the DN2450, it is an enhancement mode type - so you can't just use it as a current source by adding a resistor in the source circuit (like you can with Depletion-mode FETs).

you could build it into my circuit using the extra transistor to control the current - but it will be inferior to the ordinary transistor in that position, mainly due to much higher capacitance. FETs are often harder to tame for oscillation, too.

Look out for some nice 2SC4381 NPNs or equivalent sometime. These can heat a 300B with 3V across them, using only 100x100mm aluminium 2mm thick for a heatsink. Best of all they have tiny parasitic capacitance - about 1/10 of the FET.

http://www.cricklewoodelectronics.c...p?cat=178&sort=title&sort_direction=0&page=16

 

[Bas Horneman]

but it will be inferior to the ordinary transistor in that position

Ok..I'll build it as you've done it then.