DC Current-Regulated 300B filaments?

[rongon]

Hello there...

I'm thinking of building a new power amp. Will most likely use PP parallel 2A3's or PP 300B's. DHT's, of course. I would like to keep the hum down this time. I was reading some posts, found that voltage-regulated DC for DHT filaments doesn't sound good to most. Most prefer current-regulated DC for this.

Can anyone recommend a simple-but-decent DC current regulator circuit for something like 5VDC @ 2.5A (pair of 300B)? Or 2.5VDC @ 10A (quad of 2A3)?

Many thanks...

 

[tomchr]

I'm using an LM317 to power the filament of one JJ 300B. Set R1=120 Ohm; R2=360 Ohm and you're set (see fig. 1 of the data sheet). The LM317 will provide 1.5 A with proper heatsinking.

For higher output currents, look into the LM338. The LM338 will deliver up to 5 A with proper heatsinking. Again R1=120 Ohm; R2 = 360 Ohm for 5.0 V out.

~Tom

 

[Arnulf]

He's looking for current regulator. LM317 can be used in this role as well, put 1R (yes, one ohm) between output and adjust pin and connect load to the junction of resistor and adjust pin for 1.25A (a single tube). This will allow you to distribute the heat around by using multiple regulators.

A higher current version (LM338/LM350) should be substituted in the case of higher output current and current sense resistor should be appropriately adjusted.

 

[tomchr]

Arnulf: It isn't clear to me if he wants a voltage regulator or a current regulator. I assumed a voltage regulator as the specs are given in volts.

But for a constant current source, the circuits you mention will work fine. One catch, though: The resistor from the output to the adjust pin (the one that carries the load current) will need to be a power resistor. I'd calculate Presistor = 2*1.25*Iout and round off to the nearest higher power rating. I use a factor of 2 to give a little margin on the resistor power. Expect the resistor to be at least 80~100 deg C even with this margin.

~Tom

 

[Eli Duttman]

AAMOF, DC heating may be unnecessary, as PP topology is being employed. Cancellation of filament hum, in the O/P trafo, will occur. This is particularly true of the 2A3, which is known to work well SE, with AC heating.

 

[kevinkr]

I made the mistake of using AC heating in a line of commercial 300B PP stereo amps I manufactured about a decade ago, shortly after their release into the wild I had to recall and start retrofitting the units with dc filament heating in the output stages. The level of output "buzz" was a function of how closely the tubes stayed matched over time and the owner's willingness to adjust the filament hum null pots periodically. It proved entirely unsatisfactory..

I have found however that 2.5V dhts like the 45 and 2A3 work fine in both SE and PP amplifiers I have designed with AC heating. Higher filament voltages seem to be at the root of the problem, differing filament construction (symmettry?) may play a limited role as well.

 

[rongon]

Thanks for the replies.

A couple of people have mentioned that voltage regulated DC filament supplies can sound 'cold' or 'harsh.' I saw something KYW wrote, summing up DC vs. AC filaments for DHTs -- to paraphrase: softer, more distorted sound of AC fils can be compensated for by use of more aggressive sounding passive parts, while DC fils with more aggressive sound can be compensated for by using softer sounding passive parts.

However, I did build a PP 2A3 amp (two stages DC coupled, 6 watts per ch, but needs some gain from the preamp). I had some hum problems but a hum balancing pot in the filament of each pair of 2A3's fixed that well enough for my purposes. But there's still a tiny bit of hum down in there. I suppose it would be the same with 300B's, but I wanted to try to do better this time.

So let's say I use an LM338 as a 5A current regulator.

I have a 6.3V 5A winding on the power transformer. That's just enough current for four 300B filaments (1.25A * 4). Rectification minus voltage drop across the diode bridge would yield about 7.5 to 7.8V, correct?

So, how much voltage drop would I expect across the LM338 when used as a current regulator? About 2.5V? That right there would get me to my 5V for the 300B's.

 

[rongon]

shortly after their release into the wild I had to recall and start retrofitting the units with dc filament heating in the output stages.

Kevin, I hope this is OK -- can I ask what type of DC supply you used for the 300B filaments? Anything special? Passive w/ choke? Some kind of SS voltage regulator?

 

[Arnulf]

A couple of people have mentioned that voltage regulated DC filament supplies can sound 'cold' or 'harsh.' I saw something KYW wrote, summing up DC vs. AC filaments for DHTs -- to paraphrase: softer, more distorted sound of AC fils can be compensated for by use of more aggressive sounding passive parts, while DC fils with more aggressive sound can be compensated for by using softer sounding passive parts.

Yes ... and Homer Simpson said: d'oh !

I have a 6.3V 5A winding on the power transformer. That's just enough current for four 300B filaments (1.25A * 4). Rectification minus voltage drop across the diode bridge would yield about 7.5 to 7.8V, correct?

So, how much voltage drop would I expect across the LM338 when used as a current regulator? About 2.5V? That right there would get me to my 5V for the 300B's.

"Under 3V", according to datasheet. Pretty realistic figure, but then again you're forgetting the 1.25V reference drop that must also be subtracted from total. As in: not going to happen in your case, you're approximately one volt short even if you used schotty rectifiers

Hooking up multiple filaments in parallel to a single CCS seems like an extremely bad idea anyway since they weren't manufactured for series operation and might therefore differ significantly in their parameters and therefore also in current consumption (and consequently emission). Guess what happens when the most thermally stressed filament (one of them will always take the brunt at the pwoer-up) gives up with such arrangement ? Why yes, the 5A current is shared by three remaining filaments, provided that CCS's supply rail is high enough (*). When next filament moves south there remaining two share 5A. Then another one goes farming and it's just last one dealing with 5A until it goes up in smoke as well

*: CCS only has to supply something in excess of N*If current, where N = number of filaments still alive and If = current each filament requires to keep the cascade going, it needn't (and with your supply arrangement indeed couldn't) actually supply full 5A until the bitter end to take out all four filaments.

 

[tomchr]

Using a 6.3 V winding may be a bit optimistic. I'd use the following worst-case estimate for the drop-out voltage:

Vrectified = 0.9*Vnominal*1.3-2*Vd = 0.9*6.3*1.3-2*0.7 = 5.97 V.
--> Vdropout = 5.97-5 = 0.97 V.

The mains voltage may vary +/-10 % in many places, hence the factor of 0.9 for worst case. I use a factor of 1.3 rather than the theoretical sqrt(2) to calculate the rectified voltage. After all the conduction angle is non-zero. A diode drop of 0.7 V is assuming a fairly beefy power diode. Note that many diodes approach 1.0 V drop at high currents due to their internal resistance.

With 970 mV available for the drop-out, you're looking at something other than an LM317/338 as these require about 2.5~3 V drop if used as voltage regulators, 3.75~4.25 V if used as current sources. Linear Technology has some regulators that can work with the low 970 mV drop-out, but they're quite pricey ($7~8/each). I think Sharp has one too.

Personally, I'd use one LM317 per 300B and run it as a voltage regulator. The 1.5 A current limit of the LM317 will provide a controlled in-rush current. I'd power four 300B's with a 9 V, 50~60 VA transformer. The transformers from Antek are quite good and very inexpensive. If you insist on using a current source to power the filaments, I'd probably go with LM317's and a 10 V, 50~60 VA transformer.

Note that 3~4 W will be dissipated in each LM317. You'll need to provide adequate heatsinking to get rid of the heat.

~Tom

 

[jrenkin]

Slight thread diversion, sorry.
I have a 9V 100VA toroid I want to use for the filaments of 4 EL34 in PP.
I figure I have a few options.
1. Parallel secondaries and plop in an ~ 1 ohm 20 + W resistor (ow....)
2. One secondary for each pair of EL34 and plop in an ~ 1 ohm 12W in each side (less ow)
3. try a voltage regulator idea like is being discussed in this thread (hence the post here).
I probably don't need voltage (or current) regulation, but I want to do whatever will works best. Any other options or critiques?

 

[Arnulf]

Slight thread diversion, sorry.
I have a 9V 100VA toroid I want to use for the filaments of 4 EL34 in PP.
I figure I have a few options.
1. Parallel secondaries and plop in an ~ 1 ohm 20 + W resistor (ow....)
2. One secondary for each pair of EL34 and plop in an ~ 1 ohm 12W in each side (less ow)
3. try a voltage regulator idea like is being discussed in this thread (hence the post here).
I probably don't need voltage (or current) regulation, but I want to do whatever will works best. Any other options or critiques?

A combination of 2 and 3 would be my choice if you insist on using that transformer. Use resistor (in RC) filter before regulator to put less strain on the rectifier and capacitor. One regulator with large heatsink per pair would allow you to distribute the heat more evenly around the enclosure.

OTOH, my primary choice would be a dedicated heater supply transformer and AC heating. These can be ordered custom made for the price of what you're likely to sink into your DC regulation scheme (~20 EUR across the big pond from you). I'm sure you can find already made pieces cheaper in the US - all things electrical tend to be more expensive here, transformers must usually be custom ordered, etc.

 

[jrenkin]

A combination of 2 and 3 would be my choice if you insist on using that transformer. Use resistor (in RC) filter before regulator to put less strain on the rectifier and capacitor. One regulator with large heatsink per pair would allow you to distribute the heat more evenly around the enclosure.

OTOH, my primary choice would be a dedicated heater supply transformer and AC heating. These can be ordered custom made for the price of what you're likely to sink into your DC regulation scheme (~20 EUR across the big pond from you). I'm sure you can find already made pieces cheaper in the US - all things electrical tend to be more expensive here, transformers must usually be custom ordered, etc.

Thanks! Based on your comments I think I will choose option #4 (I forgot to add it). This is an Antek Transformer and I should be able to unwind it enough to get what I need. Or wind a little bucking winding.
Just out of curiosity, is there anything intrinsically wrong about option #2, aside from heat?

 

[rongon]

Yes ... and Homer Simpson said: d'oh !

<snip>

Hooking up multiple filaments in parallel to a single CCS seems like an extremely bad idea anyway since they weren't manufactured for series operation and might therefore differ significantly in their parameters and therefore also in current consumption (and consequently emission). Guess what happens when the most thermally stressed filament (one of them will always take the brunt at the pwoer-up) gives up with such arrangement ? Why yes, the 5A current is shared by three remaining filaments, provided that CCS's supply rail is high enough (*). When next filament moves south there remaining two share 5A. Then another one goes farming and it's just last one dealing with 5A until it goes up in smoke as well

Does it matter at all if the fils are strung up in parallel? Or am I just showing my ignorance (again)?

*: CCS only has to supply something in excess of N*If current, where N = number of filaments still alive and If = current each filament requires to keep the cascade going, it needn't (and with your supply arrangement indeed couldn't) actually supply full 5A until the bitter end to take out all four filaments.

OK, I get the picture. 6.3VCT to 5V regulator won't work... and 300B filaments would need to be regulated individually to avoid catastrophic burnouts due to unequal sharing of current between tube filaments.

So what is a correct way of current regulating two 300B filaments? Individual LM317 on each 300B?

Would I be safe to assume a drop across the LM317 or LM338 of approx. 2.75V + 1.25V = 4V ?

Thanks to tomchr for the info.

 

[Eli Duttman]

Dude,

The "right" way to supply current regulated DC to DHT filaments is an individual 2 stage setup for each tube.

The winding you rectify has to be AC RMS rated for approx. 2X the total DC draw. You need 4.8 A. of 5 VDC. A "10" A. rated winding is needed. An AnTek model An-2212 looks like it's just up to this job. Bridge rectify the paralleled secondaries with 4X Schottky diodes into a BIG 25 WVDC 'lytic.

Each 300B is fed by a cascade of regulator ICs. Voltage regulate the filtered DC down to a point it's enough to prevent dropouts of the 2nd regulator IC and no higher. The 2nd IC is setup to deliver 1.2 A. of current.

 

[Arnulf]

Does it matter at all if the fils are strung up in parallel? Or am I just showing my ignorance (again)?

I'm not sure I understand this question Uneven current sharing only happens when filaments are in parallel, regardless of whether filament supply is unregulated, voltage regulated or current regulated. However, when the filaments have their supply voltage regulated, each filament can only ever take itself out and the voltage regulator will keep the voltage constant (more or less ...) even with reduced load, which in turn limits the current flowing through each remaining filament depending on that particular filament's resistance. Current regulator will not limit the voltage, it will simply try to stuff same current down the increased resistance - which means voltage rise - until it hits the vicinity of its supply rail. By that time each filament will have more than initial If running through it, again mostlikely unevenly distributed ...

If however you use individual CCSs (one per filament) you (A) get to spread the heat around much better and (B) cannot kill anything else if a filament goes open. The added bonus is controlled start-up which should actually prolong the life of the filaments (no overshoot and no huge temperature gradient on start-up).

 

[Arnulf]

Just out of curiosity, is there anything intrinsically wrong about option #2, aside from heat?

Nothing at all, aside from the heat - 16+W is quite alot of heat once it builds up in a cramped unventilated space.

It is my opinion that a good design will always call for minimum wasted power; an additional winding sounds great in this regard.

 

[JimW]

I have dht pp amps using both 2a3 and 300b. I use ac on both. With the 2a3, hum is less than .5 mv, using separate transformer and hum pot on each filament. One humpot can be locked in center, and adjustment made to the other. On the 300b, I use a separate transformer for each tube, the center tap is connected to the shared cathode resistor, no humpots. The hum from this arrangement is less than 2mv, and has been stable for about three years with the same tubes. It is important to get the phasing of the ac correct on the filaments. To dial in the filament voltage, I tried using a bucking transformer ahead of the two filament transformers, noise went up by a large factor, so I opted to use a resistance in series with the primary of the two transformers. I also tried running the filaments in parallel using a single transformer, was noisier than using a separate transformer for each tube.

 

[jrenkin]

I have dht pp amps using both 2a3 and 300b. I use ac on both. With the 2a3, hum is less than .5 mv, using separate transformer and hum pot on each filament. One humpot can be locked in center, and adjustment made to the other. On the 300b, I use a separate transformer for each tube, the center tap is connected to the shared cathode resistor, no humpots. The hum from this arrangement is less than 2mv, and has been stable for about three years with the same tubes. It is important to get the phasing of the ac correct on the filaments. To dial in the filament voltage, I tried using a bucking transformer ahead of the two filament transformers, noise went up by a large factor, so I opted to use a resistance in series with the primary of the two transformers. I also tried running the filaments in parallel using a single transformer, was noisier than using a separate transformer for each tube.

Very helpful and interesting. What wattage did you use for the resistor on the primary side? In my case, I would be reducing 120V to ~90V to get 6.3 V out of the secondary. This should dissipate about the same as if the resistor was in series with the secondary, about 10W.

 

[kevinkr]

Kevin, I hope this is OK -- can I ask what type of DC supply you used for the 300B filaments? Anything special? Passive w/ choke? Some kind of SS voltage regulator?

Hi rongon,
In this case it was a very simple retrofit with bridge rectifiers, a resistor and a cap. Space and budget constraints permitted nothing else. It did fix the problem.

The transformer you plan to use is too small as pointed out in several other posts, figure on at least 50VA rating for heating a quartet of 300B. I use a single CCS for each 300B in my current amplifiers, and recommend that to those who build my designs. An LM317 or LT1086 will prove adequate, but must be properly heatsinked.

You haven't mentioned whether or not you are using fixed bias or cathode bias in your output stage. Cathode bias would require 4 separate secondaries, bridges, filter caps, and regulator assemblies. Antec transformers have two secondaries that can be used independently to power 2 300B which is how I do it even though I use fixed bias as this allows me to measure the cathode current independently for each output tube. Fixed bias would allow for the use of a single raw supply and separate CCS for each tube - however I really recommend a dedicated supply for each tube.

An Antec 50VA transformer will power a pair very comfortably and all you need is a 9V secondary. Antec doesn't make an 8V secondary which would be even better, however Amveco does - it however costs more. (8V is high enough with schottkies and an LT1086 including drop out for 10% low line.)

CCS heating IMHO sounds as good as AC in my experience and provides a nice controlled warm up extending the filament life of your 300Bs.