# 6.3VDC Choke-Input Power Supply, exceeding inverse voltage?

#### Nicholas K. Heinrich

Joined Feb 25, 2012
102
I am designing a filament supply of 15 Amps capability (~12A actual load) at 6.3 VDC. Basic circuit is 20V C.T. transformer (10V per leg) full wave rectification via 2 NL623 Tungar rectifiers, smoothed by 2 LC filters in series. I was using PSU Designer II, though it doesn't list the nl623, so I used a suitable(ish) solid state equivalent. Inductor values are 2mH 1 Ohm, caps are 1mF. When I simulated it, it said I exceeded 1320V inverse voltage on the diodes? Surely it must be a fluke in the program? I know inductors and capacitors don't always play nice and can create voltage spikes, but still. Usually increasing inductance, or reducing capacitance reduced peak inverse voltage, I tried both and nothing helped until I hit 5mH on the inductors. The calculated inductance required for 12A at 6.3V is around 630uH, from an old guide to tube rectifier and filter circuits. Surely it shouldn't be this hard to smooth 6.3V... Input is much appreciated, thanks.

EDIT: So I did some more tweaking on PSU Designer II, and I'm pretty sure it is the culprit. I changed the rectifier to a vacuum tube type, then I tweaked inductance to 50mH, 100mH, and 30mH. I settled on 30mH,because it is $300, as compared to$600 and $1200 for 50 and 100mH respectively. Unfortunately this leaves 2v of ripple, but it is only filament voltage after all, less hum is better than all of the hum I suppose. Either way, the program is saying there is 6.5V output into a 3.3 ohm load, at a current of 2 amp... Obviously that math doesn't add up, so I'm wondering if the peak inverse voltage is also incorrectly calculated, or if any of it is right at all. Anyone know where to find the equations to calculate ripple based on an LC filter? #### Attachments • 370.6 KB Views: 18 Last edited: #### MrAl Joined Jun 17, 2014 7,806 Hello, That giant picture and we cant even see the waveform. The red pointer points to nothing. Run the simulation and show the waveforms on both sides of the top diode. In this kind of circuit the inductances usually keeps the diodes conducting longer than they would without them and that would prevent a back emf problem, so seeing the waveforms would help figure out what is happening here. Thread Starter #### Nicholas K. Heinrich Joined Feb 25, 2012 102 The red pointer is the start of the simulation, and shows the voltage rise, the giant red blob is the waveform. I'll run it again when I'm home, but if I zoom way in it shows the waveform. It is jumping between about 2.7 and 7.2V. It basically looks like massive spikes very close together, a lot like a high amplitude high frequency sine wave. Also, the program can only show the waveform of the output of the supply. It's not particularly advanced. It is however, free. Last edited: #### ian field Joined Oct 27, 2012 6,539 I am designing a filament supply of 15 Amps capability (~12A actual load) at 6.3 VDC. Basic circuit is 20V C.T. transformer (10V per leg) full wave rectification via 2 NL623 Tungar rectifiers, smoothed by 2 LC filters in series. I was using PSU Designer II, though it doesn't list the nl623, so I used a suitable(ish) solid state equivalent. Inductor values are 2mH 1 Ohm, caps are 1mF. When I simulated it, it said I exceeded 1320V inverse voltage on the diodes? Surely it must be a fluke in the program? I know inductors and capacitors don't always play nice and can create voltage spikes, but still. Usually increasing inductance, or reducing capacitance reduced peak inverse voltage, I tried both and nothing helped until I hit 5mH on the inductors. The calculated inductance required for 12A at 6.3V is around 630uH, from an old guide to tube rectifier and filter circuits. Surely it shouldn't be this hard to smooth 6.3V... Input is much appreciated, thanks. EDIT: So I did some more tweaking on PSU Designer II, and I'm pretty sure it is the culprit. I changed the rectifier to a vacuum tube type, then I tweaked inductance to 50mH, 100mH, and 30mH. I settled on 30mH,because it is$300, as compared to $600 and$1200 for 50 and 100mH respectively. Unfortunately this leaves 2v of ripple, but it is only filament voltage after all, less hum is better than all of the hum I suppose. Either way, the program is saying there is 6.5V output into a 3.3 ohm load, at a current of 2 amp... Obviously that math doesn't add up, so I'm wondering if the peak inverse voltage is also incorrectly calculated, or if any of it is right at all. Anyone know where to find the equations to calculate ripple based on an LC filter?
Making the distinction between filaments and heaters - that's probably the good reason for wanting a smoothed DC supply. Anything with cathodes rather than directly heated is better off with AC. Its a good practice thing that doesn't have a big impact on life expectancy, but some people are particular about that sort of thing.

3-terminal regulators are not unheard of in heater circuits and regulation can help supress ripple - but at 15A, you'll need the application notes for current boosting techniques. A variation of the discrete regulator emitter follower can be configured into a capacitance multiplier for enhanced smoothing.

For that much current, I'd look for an off the shelf 5V switcher with enough adjustment range - some use 6.3V filter electrolytics, so there could be a small risk involved.

#### Nicholas K. Heinrich

Joined Feb 25, 2012
102
I actually should have said heaters. As far as I know 5751's, 6sn7's, and KT88's all have indirectly heated cathodes. I didn't know AC increases the life of the heater. Even though they're not directly connected, wouldn't the proximity of the AC to the cathode induce hum? It's also an excuse to have two giant glowing bulbs if I use DC lol. Glowing tubes and bulbs is kind of the theme of this build, from the rectifiers to the voltage regulator tubes. The thing with switchers is that I don't trust them to have a potential to ground, which this circuit may. Also I'm trying to eschew all things solid state on this build, just for fun. Though if AC is better, then each plate transformer already has a 6.3v 12 amp winding, and there are two of them.

#### ian field

Joined Oct 27, 2012
6,539
I actually should have said heaters. As far as I know 5751's, 6sn7's, and KT88's all have indirectly heated cathodes. I didn't know AC increases the life of the heater. Even though they're not directly connected, wouldn't the proximity of the AC to the cathode induce hum? It's also an excuse to have two giant glowing bulbs if I use DC lol. Glowing tubes and bulbs is kind of the theme of this build, from the rectifiers to the voltage regulator tubes. The thing with switchers is that I don't trust them to have a potential to ground, which this circuit may. Also I'm trying to eschew all things solid state on this build, just for fun. Though if AC is better, then each plate transformer already has a 6.3v 12 amp winding, and there are two of them.
The heater secondary is usually floating and in most cases the coupling is too small to affect the low impedance cathode circuits. Distribution around the chassis is usually by twisted pair to minimise coupling to high impedance circuitry.

Some signal triode pairs have series heaters with the tap brought out to a pin - you can run them series or parallel. There are "tricks of the trade" to arrange successive stages so cathode hum tends to cancel rather than augment.

If all else fails - put a linear pot across the heater secondary and ground the wiper. This practice is frowned upon, but sometimes its the only way out.

#### Nicholas K. Heinrich

Joined Feb 25, 2012
102
The 5751 has a center tapped heater, there will be two per channel for the preamp, then one 5751 for the phase inverter for both channels (using half of the dual triode for each), two 6sn7 per channel for the driver, then two KT88 output per channel. So 5 tubes have a center tapped heater, while the other 8 do not. How should I arrange them? Does it depend on physical layout? Also, how does a pot across the heater secondary help, and why is it frowned upon?

Last edited:

#### recklessrog

Joined May 23, 2013
985
KT88's are generally considered the "Holy Grail" of output valves (tubes) for Hi Fi audio Valve amplifiers. Use heavy gauge tightly twisted wires to the heaters, and run them away from high gain areas in the amplifier but keep them as short as possible. Feed the KT88 heaters first in the line, better still, use a centre tapped separate heater transformer or winding to supply them.
One guitar amplifier manufacturer uses a DC supply to the heaters only because their original designer was not brought up on thermionic valve technology and used theory only in designing a Hybrid amplifier. Only after many owners complained about the residual hum, did they address the problem, and so as not to completely re-design it as they had hundreds of ready made pcb's and chassis in stock, they opted to use DC for the heaters which worked well. All they had to do was use a modified mains transformer, a few diodes and capacitors and a re-run of the heater wiring.
Had it originally been designed by an engineer experienced in Valve amps, this would never have happened.

#### Nicholas K. Heinrich

Joined Feb 25, 2012
102
A separate center tapped heater winding is no problem, since many of the other transformers for bias voltage and the regulator circuitry have a filament winding that would go unused otherwise. Switching to AC actually eliminates the separate heater transformer that was to be for the dc filament supply. So that being said, what about the high gain stages? How should I arrange the heater supply to the driver, a separate center tapped winding with one side going to the "push" tube, and the other to the "pull" tube? And what of the 5751s with their center tapped heater?

#### recklessrog

Joined May 23, 2013
985
For 6.3 Volts, supply to pin 9 and Join pins 4&5 together . . Or for 12.6 Volts to pin 4&5 and leave pin 9 unconnected.
See the spec sheet below. you can connect all the heaters in parallel but feed the KT88's first, drivers second, phase splitter next, then pre-amps. avoid ground loops at all costs. I would use 12 or 14 swg single strand tightly twisted wire to the heaters.

#### Attachments

• 631.8 KB Views: 1

#### ian field

Joined Oct 27, 2012
6,539
The 5751 has a center tapped heater, there will be two per channel for the preamp, then one 5751 for the phase inverter for both channels (using half of the dual triode for each), two 6sn7 per channel for the driver, then two KT88 output per channel. So 5 tubes have a center tapped heater, while the other 8 do not. How should I arrange them? Does it depend on physical layout? Also, how does a pot across the heater secondary help, and why is it frowned upon?
Capacitance between the heater windings and others can cause assymetric floating supply. In rare cases it can cause cathode hum. The pot is used to balance the AC waveform. Not all heater/cathode capacitances are created equal - artificial balancing can produce unpredictable results.

Some high end stuff has separate heater circuits for signal and power stages - some people use sweep tubes in the power output, which are often series heater chain type. You end up with large AC voltages on some of the heaters, but its not as big a problem as in signal stages.

A lot of UK TV sets had half wave rectified heater chains to reduce dropper dissipation - but I wouldn't recommend it on audio gear.

#### MrAl

Joined Jun 17, 2014
7,806
The red pointer is the start of the simulation, and shows the voltage rise, the giant red blob is the waveform. I'll run it again when I'm home, but if I zoom way in it shows the waveform. It is jumping between about 2.7 and 7.2V. It basically looks like massive spikes very close together, a lot like a high amplitude high frequency sine wave. Also, the program can only show the waveform of the output of the supply. It's not particularly advanced. It is however, free.
Hi,

It shows that on your software perhaps, but not on the picture you posted. It shows nothing on the picture you posted. Take another snap shot and make sure the waves can be seen in the picture, not in the software.

#### Nicholas K. Heinrich

Joined Feb 25, 2012
102
Hi,

It shows that on your software perhaps, but not on the picture you posted. It shows nothing on the picture you posted. Take another snap shot and make sure the waves can be seen in the picture, not in the software.
I know the pic doesn't show that, I was just trying to describe it. It'll be a while before I can run that simulation again , since my pc turned into a large paperweight today. Though since it's AC on the filaments now, it's also not much of a worry anymore.
For 6.3 Volts, supply to pin 9 and Join pins 4&5 together . . Or for 12.6 Volts to pin 4&5 and leave pin 9 unconnected.
See the spec sheet below. you can connect all the heaters in parallel but feed the KT88's first, drivers second, phase splitter next, then pre-amps. avoid ground loops at all costs. I would use 12 or 14 swg single strand tightly twisted wire to the heaters.
Capacitance between the heater windings and others can cause assymetric floating supply. In rare cases it can cause cathode hum. The pot is used to balance the AC waveform. Not all heater/cathode capacitances are created equal - artificial balancing can produce unpredictable results.

Some high end stuff has separate heater circuits for signal and power stages - some people use sweep tubes in the power output, which are often series heater chain type. You end up with large AC voltages on some of the heaters, but its not as big a problem as in signal stages.

A lot of UK TV sets had half wave rectified heater chains to reduce dropper dissipation - but I wouldn't recommend it on audio gear.
I'll keep this all in mind, and come up with a full schematic and post it, though as said above my pc is out of commission for a while, so it'll be a week or two. Thanks for letting me in on some "tricks of the trade"!