I built a parafeed SET headphone amp and have been generally pleased with it, except some mechanical buzz which seems to be coming from the rectifier diode. I'm hoping to get some help to tame the buzz/rattle.
First, here's a schematic and PSUD sim:
The tube is a damper diode. I've tested a few different 6BY5GAs. Here's a link to the datasheet:
https://frank.pocnet.net/sheets/049/6/6BY5GA.pdf
The power transformer is a custom SumR toroid made to the specs below. Secondary 1 is for B+ whereas secondary 2 is used only to heat the rectifier. The center tap of the heater supply is tied to B+ (after the choke) to ensure that the Vhk(max) rating is respected.
Sec # 1: CT (400-0-400) @ 100mA. Off-load = 425.4Vper side.
Winding DCR = 56 ohm per phase (measured)
Sec # 2: 6.3V CT (3.15 - 0 - 3.15) @ 3A. Off-load = 3.4V per side.
Primary: 117V.
Core Bands, Static Shields.
All the caps are film. The 220nF is a recent addition to snub the choke, and doesn't seem to impact mechanical noise levels. (Or anything else as far as I can tell, really.)
The choke is a Hammond 193H. It is supposed to be 5H, but I took the below impedance measurement and estimate its actual inductance under no DC bias to be closer to 3.8H. (The x-axis is frequency in Hz, y-axis is reactance in Ohms.) I know, inductance depends on frequency, level, bias, weather, etc.-- but long story short, it seems sensible to derate it. The inductor DCR is 60 ohms, and I added an additional 150 ohm resistor such that the LC pole defined by it and the 418uF of capacitance is <0.5.
The load is a shunt regulator, so it is guaranteed to draw a constant 86mA.
My first intuition was to replace the rectifier. I was using a GE tube, which wasn't silent, but not outrageous, either. I switched in a Selectron tube, and the noise out of that one was outrageous. They are physically identical as far as I can tell, so I assume Selectron were seconds from GE.
My sense is that I'm riding a bit too close to the 6BY5's PIV value of 1400V, particularly as the primary was designed for 117V when my mains fluctuates between 120V and 125V. I'm considering adding a resistor in series with one of the windings to drop the voltage, but I'm not too keen on the heat. Does anything stand out as not quite right in this design? Short of changing out the transformer, are there any other tricks worth testing to get the diodes quiet?
First, here's a schematic and PSUD sim:
The tube is a damper diode. I've tested a few different 6BY5GAs. Here's a link to the datasheet:
https://frank.pocnet.net/sheets/049/6/6BY5GA.pdf
The power transformer is a custom SumR toroid made to the specs below. Secondary 1 is for B+ whereas secondary 2 is used only to heat the rectifier. The center tap of the heater supply is tied to B+ (after the choke) to ensure that the Vhk(max) rating is respected.
Sec # 1: CT (400-0-400) @ 100mA. Off-load = 425.4Vper side.
Winding DCR = 56 ohm per phase (measured)
Sec # 2: 6.3V CT (3.15 - 0 - 3.15) @ 3A. Off-load = 3.4V per side.
Primary: 117V.
Core Bands, Static Shields.
All the caps are film. The 220nF is a recent addition to snub the choke, and doesn't seem to impact mechanical noise levels. (Or anything else as far as I can tell, really.)
The choke is a Hammond 193H. It is supposed to be 5H, but I took the below impedance measurement and estimate its actual inductance under no DC bias to be closer to 3.8H. (The x-axis is frequency in Hz, y-axis is reactance in Ohms.) I know, inductance depends on frequency, level, bias, weather, etc.-- but long story short, it seems sensible to derate it. The inductor DCR is 60 ohms, and I added an additional 150 ohm resistor such that the LC pole defined by it and the 418uF of capacitance is <0.5.
The load is a shunt regulator, so it is guaranteed to draw a constant 86mA.
My first intuition was to replace the rectifier. I was using a GE tube, which wasn't silent, but not outrageous, either. I switched in a Selectron tube, and the noise out of that one was outrageous. They are physically identical as far as I can tell, so I assume Selectron were seconds from GE.
My sense is that I'm riding a bit too close to the 6BY5's PIV value of 1400V, particularly as the primary was designed for 117V when my mains fluctuates between 120V and 125V. I'm considering adding a resistor in series with one of the windings to drop the voltage, but I'm not too keen on the heat. Does anything stand out as not quite right in this design? Short of changing out the transformer, are there any other tricks worth testing to get the diodes quiet?
Attachments
Last edited:
Shoot, I made an error in the original sim and posted it with a 120mA load. I reran the simulation with the correct 86mA load. Thank you for catching this! I updated the OP with the corrected sim.
Also worth mentioning that the PT is fully encapsulated. I measured the noise with my phone and found it was loudest next to the rectifier tube itself, and not next to the transformes nor the PS choke (which was problematic in the past).
I have heard that sort of noise with an Altec 1570B amplifier which uses four 5U4 rectifiers in a bridge.
Replacing the rectifier tubes with a string of four 1n4007 per tube did not have the same noise.
I really have to agree that it can be some noise made by the tube and not choke or transformer noise. More like a glassy tinkling noise and not humming from metal.
Replacing the rectifier tubes with a string of four 1n4007 per tube did not have the same noise.
I really have to agree that it can be some noise made by the tube and not choke or transformer noise. More like a glassy tinkling noise and not humming from metal.
I would call it a capacitor input supply instead of choke input. You would have to remove C1 to call it a choke input supply.
Also lowering the value of C1 may reduce the noise. 220 gives you a bit of hot switching current which tends to make noise.
Also lowering the value of C1 may reduce the noise. 220 gives you a bit of hot switching current which tends to make noise.
I think your choke might be at the critical inductance point with 86mA load. Can you add a resistor load to the output to get back to 120 mA and see if the noise changes? Also for testing remove C1 even though it probably makes no difference. If the noise continues and you are sure it is the rectifier tubes, put one or two 1n4007 diodes in series with each tube to guarantee there is no reverse breakdown. Test for noise with the 1n4007's in place and 120 mA load total.
Little bit overstressed rectifier.
At power up the oversized C2 charging takes half a second.
If it's 86mA steady current SET PSU, at the full power the peek load is 172mA (in this test at 1s, takes 100ms). See the current consumption.
At power up the oversized C2 charging takes half a second.
If it's 86mA steady current SET PSU, at the full power the peek load is 172mA (in this test at 1s, takes 100ms). See the current consumption.
This sounds familiar. Twenty years or so ago I had problems with mechanical noises from TV dampers that I used as rectifiers in a project. The tubes where in question EY88 or PY88 if my memory serves, european noval tubes.
I've also noticed more lately when experimenting with filament supplies that choke input filters can cause a lot of strange noises, not only from the chokes but also from the power transformers. Small Cs after the rectifiers and snubbers across the secondary windings can be useful.
I've also noticed more lately when experimenting with filament supplies that choke input filters can cause a lot of strange noises, not only from the chokes but also from the power transformers. Small Cs after the rectifiers and snubbers across the secondary windings can be useful.
Your supply is not choke input, it's cap input. If it is buzzing maybe your power transformer is over-stressed. If the secondary's are rated for 100mA, then you should not try to draw more than 100mA (1/1.4) = 70mA since you are doing cap input.
But I see you are drawing 86mA...
But I see you are drawing 86mA...
RMS sure. But I have never seen a rattling rectifier before euro21... Rattling mains, rattling choke most certainly.
Everybody,
The input capacitor is Not 220uF (it is Not 220 x 10-6 Farads).
If the installed input capacitor was 220uF by mistake . . . remove it; replace with 220nF.
The schematic says:
The input capacitor is 220nf (200 x 10-9 Farads).
Xc @ 100Hz = 7.23k Ohms
I would use a 1000V rated 220nF cap.
3.8H
XL @ 100Hz = 2.39k Ohms
The inductance of 3.8H is just a little bit low for 86mA.
Critical inductance is about 4.9 Henry.
I use the Hammond 5H 200mA choke on all my B+ choke input filters. But I have 120Hz full wave rectifier frequency (2 x 60Hz)
(There might be not quite enough laminations for 100 Hz full wave).
Have Fun!
The input capacitor is Not 220uF (it is Not 220 x 10-6 Farads).
If the installed input capacitor was 220uF by mistake . . . remove it; replace with 220nF.
The schematic says:
The input capacitor is 220nf (200 x 10-9 Farads).
Xc @ 100Hz = 7.23k Ohms
I would use a 1000V rated 220nF cap.
3.8H
XL @ 100Hz = 2.39k Ohms
The inductance of 3.8H is just a little bit low for 86mA.
Critical inductance is about 4.9 Henry.
I use the Hammond 5H 200mA choke on all my B+ choke input filters. But I have 120Hz full wave rectifier frequency (2 x 60Hz)
(There might be not quite enough laminations for 100 Hz full wave).
Have Fun!
Last edited:
Thanks for all the feedback!
https://www.diyaudio.com/community/threads/clarification-on-morgan-jones-choke-input-snubber.405826/
Either way, the cap is out now, and the noise is unchanged. It really was just 220nF.
My math says that the critical current is about 89mA after plugging 3.8H, 419VRMS (per the above sim), and 60Hz into the below formula. 120mA doesn't sound so extravagant after all?
Your analysis is bang on, I'll try to test the critical inductance theory by boosting the current further.
I would tend to agree, but Morgan Jones says we're still allowed to call it a choke input supply. 🙂 A few posts recomend using his modified snubber, which effectively converts it to a cap-input supply that somehow still performs like a choke-input supply. I am not smart enough to explain it, so here's a thread that coveres some of the details:
https://www.diyaudio.com/community/threads/clarification-on-morgan-jones-choke-input-snubber.405826/
Either way, the cap is out now, and the noise is unchanged. It really was just 220nF.
Good thinking. The best I could do with parts I had on hand was to push it to 100mA, and this didn't seem to change much. I'll try to push it further, maybe by altering the shunt reg's CCS.
My math says that the critical current is about 89mA after plugging 3.8H, 419VRMS (per the above sim), and 60Hz into the below formula. 120mA doesn't sound so extravagant after all?
I'm not too worried about this. The sticker on the transformer says the core is rated for 96VA, and I'm drawing 420V * 92mA + 6.3V * 1.6A = 48VA. That's only 50% of the core's rating. The wire might heat up, but the core is nowhere near saturation. My understanding of the spec is that I can get 200mA at 400V or 100mA at 800V (windings in series).
Since the raw B+ supply feeds a shunt regulator, the rectifier really does see a constant 86mA and not 172mA peaks. Peaks are sourced from the shunt leg of the regulator.
I can be a clutz, but I would be hard pressed to mix up a 220nF 1000V cap with a 220uF 1000V cap. 🙂
Your analysis is bang on, I'll try to test the critical inductance theory by boosting the current further.
Is this actually an issue? The 418uF is actually made up of three DC-link caps. I had set it up as an LCRCRC filter, but found the bass significantly improved by tying all the Cs in parallel, and removing the Rs. I was very skeptical as I assumed the shunt reg kept the raw B+ supply out of the picture, but a couple of blind, electronically illiterate listeners found that the difference was not subtle.
I use the following critical inductance formula (a simple, approximating formula):
Critical Inductance = 350/mA DC load current
120 Hz full wave; 350/mA load.
350/86mA = 4.07 Henry
Remember this is an Approximation of Critical Inductance.
So, in this case I would use a 5H or 7H choke rated at least at 2 x 86mA DC.
For 50Hz mains (100Hz full wave), multiply 350 x (120/100) = 420
The simple formula becomes: Critical Inductance = 420/mA load
420/86mA = 4.88 Henry.
To be conservative use the next higher inductance. 5H is close; use a 7H.
86 x 2 = 172mA. Use a 200mA or 300mA rated choke.
A lot is dependent on the quality of the choke, and how they measured the choke ratings.
For example, is 5H and 200mA actually 5H @ 200mA, or is the inductance -20% at 200mA.
Then, there is the issue of the test frequency, 100Hz, 120Hz, or what?
There is another formula for calculating Critical Inductance, it give slightly different answers.
In general, use parts that are rated that have some margin of error versus what you calculate, and it will be more likely to work as planned.
Critical Inductance = 350/mA DC load current
120 Hz full wave; 350/mA load.
350/86mA = 4.07 Henry
Remember this is an Approximation of Critical Inductance.
So, in this case I would use a 5H or 7H choke rated at least at 2 x 86mA DC.
For 50Hz mains (100Hz full wave), multiply 350 x (120/100) = 420
The simple formula becomes: Critical Inductance = 420/mA load
420/86mA = 4.88 Henry.
To be conservative use the next higher inductance. 5H is close; use a 7H.
86 x 2 = 172mA. Use a 200mA or 300mA rated choke.
A lot is dependent on the quality of the choke, and how they measured the choke ratings.
For example, is 5H and 200mA actually 5H @ 200mA, or is the inductance -20% at 200mA.
Then, there is the issue of the test frequency, 100Hz, 120Hz, or what?
There is another formula for calculating Critical Inductance, it give slightly different answers.
In general, use parts that are rated that have some margin of error versus what you calculate, and it will be more likely to work as planned.
An advantage of choke input filtering:
Given a specific DC load current (86mA for example) . . .
The power transformer runs Much Cooler.
Versus . . .
Given a specific DC load current (86mA for example) . . .
The power transformer runs Much Hotter.
Ignoring all the voltage losses in a power supply . . .
Choke input:
Vrms x 0.9 = VDC
If there is a pure sine wave from the power transformer.
The 0.9 factor is the maximum Vrms to VDC, if there were not other loss factors.
Versus . . .
Cap input:
Vrms x 1.414 = VDC
If there is a pure sine wave from the power transformer.
The 1.414 factor is the maximum Vrms to VDC, if there were not other loss factors.
The power mains waveform is often clipped, so there are less Peak Volts versus Vrms.
The cap input factor becomes less than 1.414 x Vrms.
Other voltage losses: Power transformer primary DCR and secondary DCR, rectifier voltage loss, choke DCR loss, Series resistors from Cap to Cap voltage loss, etc.
Given a specific DC load current (86mA for example) . . .
The power transformer runs Much Cooler.
Versus . . .
Given a specific DC load current (86mA for example) . . .
The power transformer runs Much Hotter.
Ignoring all the voltage losses in a power supply . . .
Choke input:
Vrms x 0.9 = VDC
If there is a pure sine wave from the power transformer.
The 0.9 factor is the maximum Vrms to VDC, if there were not other loss factors.
Versus . . .
Cap input:
Vrms x 1.414 = VDC
If there is a pure sine wave from the power transformer.
The 1.414 factor is the maximum Vrms to VDC, if there were not other loss factors.
The power mains waveform is often clipped, so there are less Peak Volts versus Vrms.
The cap input factor becomes less than 1.414 x Vrms.
Other voltage losses: Power transformer primary DCR and secondary DCR, rectifier voltage loss, choke DCR loss, Series resistors from Cap to Cap voltage loss, etc.
If you want to eliminate the rectifier tube as the culprit, could you use a diode instead, along with a variac so the rectified output voltage is the same as with the tube? Then if it rattles at least you can say it is not the damper.
"Since the raw B+ supply feeds a shunt regulator, the rectifier really does see a constant 86mA and not 172mA peaks. Peaks are sourced from the shunt leg of the regulator."
You didn't write anything about that the load is steady shunt.
Is this what it is about?
Are you sure that these conditions are met?
You wrote that filament tied to B+ (BTW why not to cathode?).
Simmed cathode and B+ voltages.
