I would agree where some power can be traded off the 5K OPT will provide slightly lower output impedance and significantly lower distortion, from experience 5 - 6dB improvement can be achieved. I use 5K transformers with 300B, 400V and fixed bias, plate current 75mA. At clip about 6.5Wrms, about 2% THD at 6W.
In the 3K case under the same conditions approximately 10W @ clip and 8W or so at 4% THD.
Your mileage will vary and there can be significant variation depending on the tubes used - these numbers applied to 2000s vintage WE 300B and JJ 300B over the past 20 years I have used them.
I recommend Rod Coleman's filament regulators - I've used a lot of them over the past 10 years, they're reliable, limit inrush current, and are inexpensive. I think they are a subjective improvement over the Ronan style regulators I used in the past. (Just my opinion!)
Oh . . . did I forget to mention that all the brute force DC filament supplies I use . . . Are intrinsically Soft Start supplies.
A 300B filament, 5V/1.25A = 4 Ohms when hot. But the Cold 300B filament is much lower resistance than that. The 2 Ohm series resistor in the brute force CRC filter will take much more than half of the voltage, until the 300B filament warms up. I measured four 300B filaments, two from one manufacturer, and two from another manufacturer. Two were 0.5 Ohms cold, and two were 1.2 Ohms.
I also measured two 6A3 from one manufacturer, 1.25 Ohms cold. Of course when the filaments are hot, by definition they are 6.3 Ohms (1A @ 6.3V).
Just be sure to use a Bleeder resistor across that last 22,000uF cap. Otherwise if you tube-roll some 300B tubes (or have a 300B filament failure), the filament supply caps will be charged to more than 5V. When you turn the amp on without a working 300B filament, you will have about 8.6V for center tapped full wave supply, or 8.3V for a bridge supply. That is Not a soft start under those conditions. And the charge will remain until you subject the "unsuspecting" poor 300B. So you can see, a filament supply Bleeder is required.
A 300B filament, 5V/1.25A = 4 Ohms when hot. But the Cold 300B filament is much lower resistance than that. The 2 Ohm series resistor in the brute force CRC filter will take much more than half of the voltage, until the 300B filament warms up. I measured four 300B filaments, two from one manufacturer, and two from another manufacturer. Two were 0.5 Ohms cold, and two were 1.2 Ohms.
I also measured two 6A3 from one manufacturer, 1.25 Ohms cold. Of course when the filaments are hot, by definition they are 6.3 Ohms (1A @ 6.3V).
Just be sure to use a Bleeder resistor across that last 22,000uF cap. Otherwise if you tube-roll some 300B tubes (or have a 300B filament failure), the filament supply caps will be charged to more than 5V. When you turn the amp on without a working 300B filament, you will have about 8.6V for center tapped full wave supply, or 8.3V for a bridge supply. That is Not a soft start under those conditions. And the charge will remain until you subject the "unsuspecting" poor 300B. So you can see, a filament supply Bleeder is required.
Last edited:
300b's with ac filament heating is hard to tame humwise, unlike the 2a3's where i used only ac heating and yet the amp was very quiet.....
here is an anecdote told to me by a client, he says' "a little bit of hum actually made the music more dynamic to my ears", i had previously set the humdinger pot to null the hum...he adjusted the pot to give a little bit more hum....though not really audible at his listening position....how do i explain that? who am i to tell him he is wrong?
here is an anecdote told to me by a client, he says' "a little bit of hum actually made the music more dynamic to my ears", i had previously set the humdinger pot to null the hum...he adjusted the pot to give a little bit more hum....though not really audible at his listening position....how do i explain that? who am i to tell him he is wrong?
i have a pair of EH300b's to play with...
i purchased a pair of 5v/2A offline smps power bricks to try, i would have wanted it rated for more than 2A but we'll see how it fares...
my power traffo has 100 volt ac extra windings to power up the power bricks..
imho, SMPS is the way to go, i am asking a friend to design smps for b+ and heaters...
With AC filaments on a DHT 2A3, you Can cancel the hum for all intents and purposes, correct. Actually, your mileage (hum) may vary, some tubes are better than others. No test tones, no music, so no hum.
What you can not cancel are the 2X line frequency sidebands. There are no sidebands there until you apply a test tone, or a music tone (flute, etc.). At that point, the grid is now moving with the test tone, or the music tone. Remember, as the grid voltage changes more and more, it gets closer or further away from the filament. The transconductance varies with signal.
When the AC filament voltage is at the crest of one alternation direction . . . The transconductance of the center of the filament is one value, the transconductance of the filament at one end is different than that, and the transconductance at the other end of the filament is a third value (the opposite direction of alternation).
But the parallel total transconductance of those 3 filament areas is not quite the same as the total transconductance of the same filament when the AC filament voltage is at 0 (Zero crossing of the AC).
Wow, a total transconductance that changes some as the AC filament voltage varies. And that effect changes even more as the grid moves with signal.
Remember how I said Alternations. The Alternations are 2X line frequency. And that is the nature of those sidebands that appear on each test tone, or each music note.
Example: 60Hz mains in the US. 2X mains = 120Hz Hmm, a Cello playing a 440 A note, will have 440 +/- 120 = 320, 440, and 560. That is just the fundamental, not to mention the harmonics of 440 that will have +/-120Hz sidebands on each of them.
A flute that is playing a 1320Hz note, will have 1320Hz +/-120Hz sidebands on it. 1200Hz, 1320Hz, and 1440 Hz will be the result. Again, the harmonics of the 1320 note will also have +/- 120Hz sidebands on them.
Can you hear them? I leave that up to you, your particular amp, your particular 2A3, and your particular music.
Can you measure them? Only if you have a spectrum analyzer or other FFT that has some dynamic range.
Then, when you are measuring them, be sure to turn the test tone up more (far before clipping, but to a fairly large signal). You may be surprised how quickly those 2X sidebands start growing relative to the increase of test tone amplitude.
For proof of the different transconductance along the length of the DHT filament: Connect a 2.5VDC supply to the 2A3 filament. Connect a series of two 20 Ohm resistors across the filament. Connect the center of the two 20 Ohm resistors to a 750 Ohm resistor to ground. Connect the plate to +295V (250V + 45V). Measure from one end of the filament to the top of the 750 Ohm resistor. It will not be 1.25V. Measure the other filament to the top of the 750 Ohm resistor. It will not be 1.25V. The Sum total will be 2.5V, but uneven from one end to the other.
Ever wonder why the 2A3 data sheet lists the operating condition as -45V grid, 250V plate to filament, and 60mA, but with a Caveat? It says the condition is the midpoint of the AC powered filament. You can use either a center tapped filament secondary, or those two 20 Ohm resistors I mentioned. Both methods give the correct results.
When you use DC power on the filament, the midpoint of the two 20 Ohm resistors versus the ends will not be +/- 1.25VDC. Non linear transconductance anyone.
What you can not cancel are the 2X line frequency sidebands. There are no sidebands there until you apply a test tone, or a music tone (flute, etc.). At that point, the grid is now moving with the test tone, or the music tone. Remember, as the grid voltage changes more and more, it gets closer or further away from the filament. The transconductance varies with signal.
When the AC filament voltage is at the crest of one alternation direction . . . The transconductance of the center of the filament is one value, the transconductance of the filament at one end is different than that, and the transconductance at the other end of the filament is a third value (the opposite direction of alternation).
But the parallel total transconductance of those 3 filament areas is not quite the same as the total transconductance of the same filament when the AC filament voltage is at 0 (Zero crossing of the AC).
Wow, a total transconductance that changes some as the AC filament voltage varies. And that effect changes even more as the grid moves with signal.
Remember how I said Alternations. The Alternations are 2X line frequency. And that is the nature of those sidebands that appear on each test tone, or each music note.
Example: 60Hz mains in the US. 2X mains = 120Hz Hmm, a Cello playing a 440 A note, will have 440 +/- 120 = 320, 440, and 560. That is just the fundamental, not to mention the harmonics of 440 that will have +/-120Hz sidebands on each of them.
A flute that is playing a 1320Hz note, will have 1320Hz +/-120Hz sidebands on it. 1200Hz, 1320Hz, and 1440 Hz will be the result. Again, the harmonics of the 1320 note will also have +/- 120Hz sidebands on them.
Can you hear them? I leave that up to you, your particular amp, your particular 2A3, and your particular music.
Can you measure them? Only if you have a spectrum analyzer or other FFT that has some dynamic range.
Then, when you are measuring them, be sure to turn the test tone up more (far before clipping, but to a fairly large signal). You may be surprised how quickly those 2X sidebands start growing relative to the increase of test tone amplitude.
For proof of the different transconductance along the length of the DHT filament: Connect a 2.5VDC supply to the 2A3 filament. Connect a series of two 20 Ohm resistors across the filament. Connect the center of the two 20 Ohm resistors to a 750 Ohm resistor to ground. Connect the plate to +295V (250V + 45V). Measure from one end of the filament to the top of the 750 Ohm resistor. It will not be 1.25V. Measure the other filament to the top of the 750 Ohm resistor. It will not be 1.25V. The Sum total will be 2.5V, but uneven from one end to the other.
Ever wonder why the 2A3 data sheet lists the operating condition as -45V grid, 250V plate to filament, and 60mA, but with a Caveat? It says the condition is the midpoint of the AC powered filament. You can use either a center tapped filament secondary, or those two 20 Ohm resistors I mentioned. Both methods give the correct results.
When you use DC power on the filament, the midpoint of the two 20 Ohm resistors versus the ends will not be +/- 1.25VDC. Non linear transconductance anyone.
Last edited:
I run mine at 450v with a 5k transformer
Its well worth the trade off in power as long as your good with about 4 watts
Its well worth the trade off in power as long as your good with about 4 watts
Hi - with the power supply if I was to build two channels on one supply, how and where would I split the supply? Would it be best to split after the first cap and use two chokes, or is it ok to just double the current rating of the choke and connect both after the 2nd CLC cap? Thanks
I had a play with linux sox program to generate the sidebands for 1KHz.
So for 1KHz + 1100KHz I can hear sideband down to -30dB. For 1KHz + 900Hz I can hear down to -36dB. This was in headphones. I think this means you should aim for better then -40dB max sideband levels. If you want to experiment:
play -n synth 2.25 sine 900 gain -36 synth 2.25 sine mix 1000
So for 1KHz + 1100KHz I can hear sideband down to -30dB. For 1KHz + 900Hz I can hear down to -36dB. This was in headphones. I think this means you should aim for better then -40dB max sideband levels. If you want to experiment:
play -n synth 2.25 sine 900 gain -36 synth 2.25 sine mix 1000
baudouin0,
Thanks for doing that test.
However, the contribution of Both Upper and Lower sidebands on the 1kHz tone is like AM modulation.
Those 2 sidebands are Additive (different than just creating one sideband at a time in the test).
Try building an AM modulated 1kHz tone.
When the desired signal (1kHz) is small, the sidebands are relatively much smaller. It is OK at that amplitude.
When the signal is very large, the sidebands grow to be a larger portion of the total signal.
A 300B with AC filaments generally has a 'virtual center' at the middle of the filaments (no actual connection there).
So, Each end of the filament is varying by +3.54V peak to -3.54V peak relative to the grid.
As the grid voltage comes very near to the voltage at one end of the filament, you will find bigger sidebands riding on that large signal.
I can not remember the word that describes the grid action, but a grid starts conducting, or changing its characteristic, or whatever you want to call it, Before the grid gets to 0V relative to a filament end, or relative to a cathode.
Oh, I remember, that is called Contact Potential, Right?
AC direct heated filaments will modulate at 2X Mains frequency, according to Contact Potential effect; and DC direct heated filaments will not.
Just Sayin'
Thanks for doing that test.
However, the contribution of Both Upper and Lower sidebands on the 1kHz tone is like AM modulation.
Those 2 sidebands are Additive (different than just creating one sideband at a time in the test).
Try building an AM modulated 1kHz tone.
When the desired signal (1kHz) is small, the sidebands are relatively much smaller. It is OK at that amplitude.
When the signal is very large, the sidebands grow to be a larger portion of the total signal.
A 300B with AC filaments generally has a 'virtual center' at the middle of the filaments (no actual connection there).
So, Each end of the filament is varying by +3.54V peak to -3.54V peak relative to the grid.
As the grid voltage comes very near to the voltage at one end of the filament, you will find bigger sidebands riding on that large signal.
I can not remember the word that describes the grid action, but a grid starts conducting, or changing its characteristic, or whatever you want to call it, Before the grid gets to 0V relative to a filament end, or relative to a cathode.
Oh, I remember, that is called Contact Potential, Right?
AC direct heated filaments will modulate at 2X Mains frequency, according to Contact Potential effect; and DC direct heated filaments will not.
Just Sayin'
Last edited:
Does the filament go back on itself and does this have make a difference to the effect or does the filament behave as two separate cathodes.
I put this on another thread - I idea is to model the distributed cathode/heater with 5x12au7 in parallel. It compliments Mr Summers' fine explanation rather well.
I put this on another thread - I idea is to model the distributed cathode/heater with 5x12au7 in parallel. It compliments Mr Summers' fine explanation rather well.
Last edited:
baudouin0,
Perfect!
I never could find the new thread, so I came up with this:
That is a great idea, and a good application for simulation software.
(and I do not even like simulation software, or any software)
I am not certain, but it seems you have 200V on the plate load resistor, and -5V grid bias.
I suggest putting a "desired" 0.707Vrms 1kHz sine wave signal (1V peak) in series with the -5V bias.
I do not know the meaning of: sine 0 9 50. 50Hz?
A 300B filament is 1.25A at 5V, so it is 4 Ohms resistance.
I suggest using a floating 5Vrms 50Hz sine wave across the end-to-end 12AU7 cathodes.
But change the resistors from cathode to cathode to 1 Ohm each resistor (0.1 Ohm needs too much current at 5V).
The 5V floating sine wave will never be more than 2.5Vrms (3.54V peak) above 0V (at either 'end' cathode).
So, -5V bias +3.54V peak = -1.46V on the grid with respect to the cathodes that are on the end.
And that leaves the 1kHz 1V peak sine wave on the grid to bring the grid to within 0.16V of the cathode.
Now the question is, what is the FFT result of simulating it that way?
I bet there will be 900Hz lower sideband, 1,000Hz desired signal, and 1,100Hz upper sideband.
That will simulate a response of a 300B with AC filaments.
I believe you asked if the filament wrapped back on itself.
The 300B filament is a W or an M shape. But the center is the same as the center of the W or M, and the voltage increases from the center to the ends. Of course in all cases, AC or DC the voltage polarity from the center to one end is opposite to the polarity from the center to the other end.
Perfect!
I never could find the new thread, so I came up with this:
That is a great idea, and a good application for simulation software.
(and I do not even like simulation software, or any software)
I am not certain, but it seems you have 200V on the plate load resistor, and -5V grid bias.
I suggest putting a "desired" 0.707Vrms 1kHz sine wave signal (1V peak) in series with the -5V bias.
I do not know the meaning of: sine 0 9 50. 50Hz?
A 300B filament is 1.25A at 5V, so it is 4 Ohms resistance.
I suggest using a floating 5Vrms 50Hz sine wave across the end-to-end 12AU7 cathodes.
But change the resistors from cathode to cathode to 1 Ohm each resistor (0.1 Ohm needs too much current at 5V).
The 5V floating sine wave will never be more than 2.5Vrms (3.54V peak) above 0V (at either 'end' cathode).
So, -5V bias +3.54V peak = -1.46V on the grid with respect to the cathodes that are on the end.
And that leaves the 1kHz 1V peak sine wave on the grid to bring the grid to within 0.16V of the cathode.
Now the question is, what is the FFT result of simulating it that way?
I bet there will be 900Hz lower sideband, 1,000Hz desired signal, and 1,100Hz upper sideband.
That will simulate a response of a 300B with AC filaments.
I believe you asked if the filament wrapped back on itself.
The 300B filament is a W or an M shape. But the center is the same as the center of the W or M, and the voltage increases from the center to the ends. Of course in all cases, AC or DC the voltage polarity from the center to one end is opposite to the polarity from the center to the other end.
Last edited:
Sorry sin 0 9 50 means sinewave 50Hz with pk of 9V (approx 6.3V rms). I used .1 R resistors. It does not actually matter the idea is that the cathodes get the correct ratios of the hum with the centre valve getting none. Its like breaking the heater/cathode into 5 sections. The 5x12au7 is not the same as a 300b its just to illustrate.
Here's a 2v peak sinewave added to the grid.
This is a PP version. The 100Hz cancels but the 1KHz side bands do not.
Here's a 2v peak sinewave added to the grid.
This is a PP version. The 100Hz cancels but the 1KHz side bands do not.
Last edited: