Hi all, I'm looking for guidelines on B+ ripple magnitude in a DHT output stage that will have about 220V B+. I have read previous threads and posts such as this one, and found them somewhat useful. I'd like feedback on the following design and suggestions.
With the below inductor input design, I see about 90 mV ripple peak to peak over about a 5 second window. I'm aiming for about 200V on B+ and, for now, the diode bridge is just a stand-in for what will eventually be 6AX4. The initial current surge through the first inductor is below the max rating of the 6AX4, and the circuit seems reasonably well damped with no major overshoot and some initial oscillation that settles in under 1s. The Hammond chokes are highish in DCR.
Feedback much appreciated.
With the below inductor input design, I see about 90 mV ripple peak to peak over about a 5 second window. I'm aiming for about 200V on B+ and, for now, the diode bridge is just a stand-in for what will eventually be 6AX4. The initial current surge through the first inductor is below the max rating of the 6AX4, and the circuit seems reasonably well damped with no major overshoot and some initial oscillation that settles in under 1s. The Hammond chokes are highish in DCR.
Feedback much appreciated.
@mickeymoose good point! OK, with the ~120R plate impedance, I've got better startup damping. Ripple is roughly the same.
So what are general recommendations for B+ ripple? What do folks normally target? This also happens to be a headphone amplifier, so hum is a concern.
So what are general recommendations for B+ ripple? What do folks normally target? This also happens to be a headphone amplifier, so hum is a concern.
Last edited:
The first inductor can be "tuned" to 120 Hz with a parallel capacitor (0.5-0.6 uF). This will reduce attenuation of higher harmonics of 120 Hz, but the two stages that follow have attenuation increasing with frequency, so that's not likely to be a problem.
The mosfet regulator used on the RJM phono stage is nice and simple. https://phonoclone.com/diy-pho3.html
Output stage? As in a power amplifier? In that case, the max ripple is the max ripple on the output of the amp divided by the turns ratio of the output transformer. You can figure that out from the impedance ratio.
Turns ratio, N = sqrt(Zpri/Zsec), so N = sqrt(5000/8) = 25 for an OPT with 5 kΩ primary and 8 Ω secondary.
If the max hum voltage you're willing to tolerate on the output of the amp is, say, 0.5 mV, you'll need ripple better than 0.5*25 = 12.5 mV on the B+.
If you're working on a line stage (as your low B+ could indicate), you'll need to figure out what the transfer function from the B+ to the output is and go from there. If your tube models are good you should be able to simulate that.
Tom
Turns ratio, N = sqrt(Zpri/Zsec), so N = sqrt(5000/8) = 25 for an OPT with 5 kΩ primary and 8 Ω secondary.
If the max hum voltage you're willing to tolerate on the output of the amp is, say, 0.5 mV, you'll need ripple better than 0.5*25 = 12.5 mV on the B+.
If you're working on a line stage (as your low B+ could indicate), you'll need to figure out what the transfer function from the B+ to the output is and go from there. If your tube models are good you should be able to simulate that.
Tom
Hi all, thank you for the feedback. @tomchr @v4lve lover the output stage will be a 71a. This will be a headphone amplifier. I'm not sure the max hum voltage I should target to tolerate, thus this thread. How does one determine this value?
In single ended, the ripple voltage on the supply rail will be reduced by the transformer winding ratio. Now how much ripple is audible depends on the headphone sensitivity. the problem with your approach is that depending on headphone impedance, the transformer ratio can be quite low, as opposed to a loudspeaker. This means you need a quiet supply.
Also heating a DHT triode can induce additional hum into the output.
Also heating a DHT triode can induce additional hum into the output.
I'll probably go with a dual tap OPT 20:1 and 10:1. For filaments, I'm using the @Rod Coleman v9 regulator. I like the simple MOSFET linked by @astouffer above. Simple and effective is good.
Thanks!
Thanks!
You'll definitely want DC on the filament. I'd also regulate the B+. The MOSFET capacitance multiplier linked to above doesn't work so well in practice, though. First off, you'll want a zener clamp on the gate so you don't violate the Vgs(max) voltage of the MOSFET. Without that clamp you risk blowing the MOSFET. With that clamp in place, the capacitor in the cap multiplier will be discharged through the zener clamp if the input voltage drops. This means that the regulator will have zero regulation after brief brown-outs and such. Such voltage drops are actually very common. They happen every time your fridge, furnace, or any other heavy load turns on. The end result is an amp that works well most of the time but every now and then the music will be interrupted by hum that lasts until the capacitor in the cap multiplier has recharged. That's pretty annoying. I know because I used a capacitor multiplier exactly like the one linked to in Post #6 in my first Spud amp.
Tom
Tom
Thanks, Tom. Though it's a bit hard to envision needing the clamp. With Vgs(max) of 30V on the MOSFET, and Vgs in my circuit nominally being about 5V, you'd be looking at big droop on the mains for this to come into play. Is that realistic?
Thanks again.
The droop only has to be for a fraction of a microsecond for the gate oxide in that MOSFET to blow, so, yes, I would take that spec seriously. Also consider what happens during start-up and shut-down. Can you guarantee that the gate voltage remains within ±30 V of the source at all times?
If the spec on the MOSFET is ±30 V, I'd use two 27 V Zeners in series. For a MOS with ±20 V max I'd use two 15 V zener diodes. You'll want to connect them anode to anode and connect the two cathodes to gate and source, respectively. That'll give you a clamping voltage around 28 V in either direction. Place these close to the MOSFET to minimize the series inductance.
Tom
If the spec on the MOSFET is ±30 V, I'd use two 27 V Zeners in series. For a MOS with ±20 V max I'd use two 15 V zener diodes. You'll want to connect them anode to anode and connect the two cathodes to gate and source, respectively. That'll give you a clamping voltage around 28 V in either direction. Place these close to the MOSFET to minimize the series inductance.
Tom
Tom, I completely respect your adherence to a defensive design philosophy. And furthermore, it's a great discussion and I appreciate your input. So, I'm curious, did you actually observe 30V drops on your mains service in Canada? Or, was the Vgs max of the circuit in your spud amp substantially lower than 30V? In the US, the NEC specifies a voltage drop of no more than 5% for branch circuits and no more than 3% for feeders. These are max values and that's 6 volts and 3.6 volts, respectively. A 25% or 30V drop, even instantaneously for a microsecond or less, during heavy load conditions on a 120V residential service is just so far out of bounds. I would also guess that 30V drops would risk damaging many other types of household appliances as well. Now, during power-on, that's an interesting thought experiment... And I have seen the back to back zeners that you describe in many Ale Moglia schematics as well. I suppose if I had an audio company or was publishing schematics for other folks to use, I'd do the same. 🙂
I did a quick PSUD simulation, and I got 62mV pk-pk at the second capacitor and 77mV at the 3rd one. That's obviously wrong; I think you are into the numerical noise of the calculation and those values are not real. Look at the ripple waveform!
As for the target ripple, my experience is with loudspeakers. An output ripple of 5mV rms (typical of AC heated 300Bs) is usually good enough for speakers of 92dB or lower sensitivity. I believe another 20dB of reduction, (to 0.5mV rms, or about 2mV pk-pk) would be acceptable for most headphones but not for all. However, I'm not a headphone user - hopefully someone with direct experience will comment.
As for the target ripple, my experience is with loudspeakers. An output ripple of 5mV rms (typical of AC heated 300Bs) is usually good enough for speakers of 92dB or lower sensitivity. I believe another 20dB of reduction, (to 0.5mV rms, or about 2mV pk-pk) would be acceptable for most headphones but not for all. However, I'm not a headphone user - hopefully someone with direct experience will comment.
And this appears to have zero observable 120 Hz ripple at V(OUT) in LTSPICE. L3 seems to eliminate what little there is without that last stage inductor.
Why do you want to use a vacuum tube rectifier , and then a transistor regulator/cap multiplier ?
With solid state diodes and C1 big enough , let's say 1000uF , you don't need even the second choke L2 ...
Or with tube rectifier more LC filters ... if needed . Second cap C2 can be much bigger without loading the tube too much .
With solid state diodes and C1 big enough , let's say 1000uF , you don't need even the second choke L2 ...
Or with tube rectifier more LC filters ... if needed . Second cap C2 can be much bigger without loading the tube too much .
Last edited: