I am sure some of you may have considered using DC for heater filaments to reduce noise, so I am looking to confirm some results of napkin calculations and spice sim.
Rectifying a typical heater winding with 6.3VAC, filtering with a 100mf cap then loading it with a load resistor representative of the current of multiple 12ax7s in parallel seems to reveal a high 250mV ripple voltage on the filaments. I have a design using seven (7) 12ax7s which are drawing 2.1A of heater current. This appears to have an approximate impedance on the circuit of ~ 3ohms. There is basically no amount of practical filtering I could provide to this circuit to put clean DC on every preamp tube. I am pretty sure I am wrong in stating this, but the quick "math" and the sims are in line.
When I apply the Vripple calc, Vrip=(Iload/FC), then the necessary filtering comes out to 1.75 farads (!) for 20mV of ripple. To be honest, I don't know how much ripple is too much, but I don't even think a factor of 10 is going to make a difference.
Is this in-line with anyone else's observation? I may just elevate the filaments 75vdc instead.
Rectifying a typical heater winding with 6.3VAC, filtering with a 100mf cap then loading it with a load resistor representative of the current of multiple 12ax7s in parallel seems to reveal a high 250mV ripple voltage on the filaments. I have a design using seven (7) 12ax7s which are drawing 2.1A of heater current. This appears to have an approximate impedance on the circuit of ~ 3ohms. There is basically no amount of practical filtering I could provide to this circuit to put clean DC on every preamp tube. I am pretty sure I am wrong in stating this, but the quick "math" and the sims are in line.
When I apply the Vripple calc, Vrip=(Iload/FC), then the necessary filtering comes out to 1.75 farads (!) for 20mV of ripple. To be honest, I don't know how much ripple is too much, but I don't even think a factor of 10 is going to make a difference.
Is this in-line with anyone else's observation? I may just elevate the filaments 75vdc instead.
Last edited:
Your Math is good but your evaluation of the problem is incomplete.
You do not need ZERO ripple to get an improvement, by any means.
* Plain 6.3VAC has 17.8Vpp "ripple" .
* the raw DC you got has 250mVpp ripple.
Improvement is 0.25/17.8=37dB ... a HUGE improvement.
Somewhat less in practicl terms because AC is 50/60Hz (less audible) and ripple is 100/120Hz and to boot waveform is wonkier and has more harmonics, but anyway very worthwhile.
You do not need ZERO ripple to get an improvement, by any means.
* Plain 6.3VAC has 17.8Vpp "ripple" .
* the raw DC you got has 250mVpp ripple.
Improvement is 0.25/17.8=37dB ... a HUGE improvement.
Somewhat less in practicl terms because AC is 50/60Hz (less audible) and ripple is 100/120Hz and to boot waveform is wonkier and has more harmonics, but anyway very worthwhile.
Ah fair enough. I was thinking there was some sort of cancelling out in a parallel AC filament circuit but I don't know why I would assume that. There's clearly no way for that to happen. I'll commonly assume Va+(-Vb) which just cancels out and is not what's happening with outputs from transformers. It's more like Va - (-Vb) if that makes any sense. I forget the phase is inverted AND the voltage is negative, respectively.
Last edited:
try raising the heater voltage to +15v relative to the ground. By doing this, I have reduced the AC hum on 6922 to below the hiss levels produced by the tube gain. btw my tubes are AC heated.
Biasing the heaters off B+ should be adequate, with the exception of a phono preamp. Pick up even more noise control by using a 7025 equivalent, like the Sovtek 12AX7LPS, which contains a spiral wound, hum bucking, heater.
If a DC heater supply is insisted on, select a 6.3 VAC filament trafo rated for 4X the total heater draw at "12" V. "Full wave" voltage double the 6.3 VAC into a huge valued cap. stack, using 2X SB520 Schottky diodes. Follow with a very well heat sinked 1.5 A. rated 7812 regulator.
If a DC heater supply is insisted on, select a 6.3 VAC filament trafo rated for 4X the total heater draw at "12" V. "Full wave" voltage double the 6.3 VAC into a huge valued cap. stack, using 2X SB520 Schottky diodes. Follow with a very well heat sinked 1.5 A. rated 7812 regulator.
Attachments
There is only an improvement when the valve, 12AX7 in this case, is sub standard. AC heaters, until people who know more than the designers of valves, decided that DC was best.
I know a few manufacturers tried DC heaters with poor results. Poor reliability and extra cost.
Why not purchase a valve, that is not heater-cathode leaking and use the prescribed method of heating.
I know a few manufacturers tried DC heaters with poor results. Poor reliability and extra cost.
Why not purchase a valve, that is not heater-cathode leaking and use the prescribed method of heating.
I have some 125 watt ballast chokes with an inductance of 0.4 H, 11 ohms DCR and a 1 amp capability. I've always thought they would make a great choke input supply for a series heater chain.
Apply the critical current (in mA.) approximation of V/L. If V = 6.3 and L =0.4, the critical current is 15.75 mA. Given the 1 A. rating, the highest voltage allowed is 400 V., provided you draw all of that 1 A.
JMFahey is on the money..... You have made a huge improvement in the reduction of ripple... Consider that you are only using a 1st order filter.....you could better with a 2nd order filter or even better with a "tuned" second order filter...ie. a Band-Stop filter..however the weight and size of the filter will grow... Some may have forgotten that tubes were first intended for DC operation from batteries.. Later tubes are designed for either AC or DC for mobile applications as well as MIL equipment still operated from batteries in the field.. If you still need low ripple...I suggest + - Dual-Rail Linear reg...or a small SMPS ... I have designed small SMPS at 5 AMPS 3.3V @ 2.2 Mhz switching frequency..so it should not bother the audio...this can be simply be designed for 6.3V ...it can be placed in 1.5" square area... you should not need ISO-lated supply since heaters are "technically" floating..
Last edited:
It really depends on the application. My Dynaco PAS-3 has been working fine for nearly half a century after I replaced the original selenium rectifiers with silicon. I also cured a slight hum issue in my Fender Champ and Pignose G60 guitar amps with DC heater: a simple bridge, a resistor and a cap was all it took, plus it let me run repurposed computer fans to cool down the innards.
My current project is based on a ready made board; it has on board heater rectifers and filtering, but it's reported it has hum issues due to poor layout. Just to play it safe, I'm adding an off board regulator: it's only 10 bucks!
Merlin, I love your book and I'd highly recommend it!
Well I was thinking of elevating the heater taps to combat the typical high voltage cathode follower issues instead of running DC. I've got some amps that burn through russian tubes if used in cathode follower positions. However, from what I can gather, the actual benefit of DC comes from eliminated the excess AC wiring in the chassis. It doesn't appear that the small heater filaments are AC coupled in any significant sense to the cathode, plate, or grid. I don't think anyone is denying the expertise of the designers, rather thinking "how can I simplify implementation?"
The ol' money budget doesn't really allot for an external heater transformer so I am bound by the 3.15-0-3.15 taps of the power tx. I won't be able to run a series scheme with that.
Last edited:
Cheap and simple, but maybe not as elegant... 12V 5A 60W Power Supply AC to DC Adapter for 5050 3528 Flexible LED Strip Light | eBay
Begs the question of, why use a transformer at all? Just use a transformerless scheme for the pre-amp heater supply...
I am mostly concerned with safety when it comes to these cheap circuits.
I am mostly concerned with safety when it comes to these cheap circuits.
I use transformers (or SMPS) for isolation.
FWIW I've been using power supplies like this for years without any issues. I mean, don't beat the snot out of it (why I suggested a 5A rating for 1A loading) and it'll last for years.
I use a 12V 10A version for a phono amp based off of the Aikido, complete with a double Janus regulator. The high voltage before the regulator is 400V@50ma from a 12V DC-DC boost converter, and the heaters need 12V @ 1.8A.
FWIW I've been using power supplies like this for years without any issues. I mean, don't beat the snot out of it (why I suggested a 5A rating for 1A loading) and it'll last for years.
I use a 12V 10A version for a phono amp based off of the Aikido, complete with a double Janus regulator. The high voltage before the regulator is 400V@50ma from a 12V DC-DC boost converter, and the heaters need 12V @ 1.8A.
Last edited:
Okay you've convinced me, but I'd rather implement my own design of that style of circuit, so now I have to figure out how to do that. Thanks.
You know, the combination of Schottky diodes (minimizing the VF voltage drop) reasonably modest first-stage capacitive filtering, then a resistive filtering network (which admittedly only works for a constant current draw) with a follow-on capacitor really seems to optimize things.
6.3 VAC × 1.414 = 8.91 VPEAK
8.91 - 0.55 × 2 = 7.81 VDC peak
We want to drop to 6.3, with a 2.1 amp load, right?
E = IR…
(7.8 - 6.3) = 2.1 A • R
R = 0.72 Ω (at 3.2 watts)
OK, that handles the 12AX7 string in parallel. Indeed: if you don't mind more, smaller components, it'd also be fine to tap the 7.8 V DC (first capacitor) with individual filtering / ballast resistors, and individual post-drop filtering capacitors, for each valve. At least that way, you could pull a tube, and not affect the rest. And smaller capacitors and resistors are definitely cheaper apiece.
(7.8 - 6.3) = 0.3 A • R
R = 5 Ω (4.7Ω closest), P = ½ watt
See? Using a 47,000 µF first capacitor, and “only” 10,000 µF capacitors for each tube is quite cost-effective.
Do your SPICE modeling, and see.
It also gives opportunity to put a much larger cap on the first tube ... where hum amplification is worst.
Anyway, late to the party.
Sorry if I'm disrupting the thread.
GoatGuy
6.3 VAC × 1.414 = 8.91 VPEAK
8.91 - 0.55 × 2 = 7.81 VDC peak
We want to drop to 6.3, with a 2.1 amp load, right?
E = IR…
(7.8 - 6.3) = 2.1 A • R
R = 0.72 Ω (at 3.2 watts)
OK, that handles the 12AX7 string in parallel. Indeed: if you don't mind more, smaller components, it'd also be fine to tap the 7.8 V DC (first capacitor) with individual filtering / ballast resistors, and individual post-drop filtering capacitors, for each valve. At least that way, you could pull a tube, and not affect the rest. And smaller capacitors and resistors are definitely cheaper apiece.
(7.8 - 6.3) = 0.3 A • R
R = 5 Ω (4.7Ω closest), P = ½ watt
See? Using a 47,000 µF first capacitor, and “only” 10,000 µF capacitors for each tube is quite cost-effective.
Do your SPICE modeling, and see.
It also gives opportunity to put a much larger cap on the first tube ... where hum amplification is worst.
Anyway, late to the party.
Sorry if I'm disrupting the thread.
GoatGuy
Last edited:
If I model a 3.15VRMS source in LTSpice it gives terrible performance for voltage, under 3VDC. I can assemble the circuit on the bench and see that it puts out over 7 without a regulator. I've attached the circuit.
Attachments
Why the 100 ohm resistors?
They do nothing good and short 2 diodes.
Either AC and artificial centertap OR DC and the bridge has one end grounded.
But not both.
They do nothing good and short 2 diodes.
Either AC and artificial centertap OR DC and the bridge has one end grounded.
But not both.
- Status
- Not open for further replies.