I’m working on a one-tube differential cathode follower to drive loads from 600R to 10K, with cable capacitance from 330pF to 4.7nF. Safe to assume it will always follow a 12AT7 phase inverter or a post phase inverter differential volume control.
My actual application for the trial run is in a Fender Pro Reverb that I will be modding in a way that frees up one whole noval socket, but I’m posting here instead of in Instruments & Amps because the design priorities are totally different. I need lowest possible output impedance and low(ish) distortion without going nuts on parts count. Keeping capacitor size small is also a priority, so here I have selected the 100u/100V Nichicon UEP bipolar, and that has provided some interesting design limitations. I’m also keen to keep Rk and Rp at 2W or less. This will probably get built on a tiny turret board sized to fit in the gap between V3 and V4, where the reverb driver transformer underhangs from the chassis, and I can take advantage of existing eyelets on the stock board for plate connections.
In this amp, the tap off the choke is 440V, and I can drop down to ~230V with a 10W wire wound in the 5-7K range with 30mA-40mA pull, filtered by 22uF/350V. Seems like the hardest-driving lowest-impedance dual triode available is the 6N6P, so that’s my first port of call. I will cherry-pick a balanced one from a handful. I’m running 20mA Iq each triode in the sketch below, though I think I can get away with less, and I may be approaching idle current totally wrong to begin with.
I don’t know how to calc cathode source impedance. Actual output goal is +24dBm into 600R with 100R per side source Z maximum, i.e. I need to put out +26.5dBu across the two cathodes, and 75% of that (600R/800R) is +24dBu. I calc’d 634mW output per cathode but my math could be off. Obviously 10K loads will be a walk in the park.
I’m mixing my Broskie and my Blencowe here (👋🏼 @Merlinb ), so any suggestions for optimization are welcome. I would love to minimize strain on the power transformer.
My actual application for the trial run is in a Fender Pro Reverb that I will be modding in a way that frees up one whole noval socket, but I’m posting here instead of in Instruments & Amps because the design priorities are totally different. I need lowest possible output impedance and low(ish) distortion without going nuts on parts count. Keeping capacitor size small is also a priority, so here I have selected the 100u/100V Nichicon UEP bipolar, and that has provided some interesting design limitations. I’m also keen to keep Rk and Rp at 2W or less. This will probably get built on a tiny turret board sized to fit in the gap between V3 and V4, where the reverb driver transformer underhangs from the chassis, and I can take advantage of existing eyelets on the stock board for plate connections.
In this amp, the tap off the choke is 440V, and I can drop down to ~230V with a 10W wire wound in the 5-7K range with 30mA-40mA pull, filtered by 22uF/350V. Seems like the hardest-driving lowest-impedance dual triode available is the 6N6P, so that’s my first port of call. I will cherry-pick a balanced one from a handful. I’m running 20mA Iq each triode in the sketch below, though I think I can get away with less, and I may be approaching idle current totally wrong to begin with.
I don’t know how to calc cathode source impedance. Actual output goal is +24dBm into 600R with 100R per side source Z maximum, i.e. I need to put out +26.5dBu across the two cathodes, and 75% of that (600R/800R) is +24dBu. I calc’d 634mW output per cathode but my math could be off. Obviously 10K loads will be a walk in the park.
I’m mixing my Broskie and my Blencowe here (👋🏼 @Merlinb ), so any suggestions for optimization are welcome. I would love to minimize strain on the power transformer.
Last edited:
100uF 100V caps makes me nervous.
Two 230V B+ voltage sources, slow starting of cold output tubes, and whatever voltages appear across the 100V caps during warmup might be too much, especially if the B+ warms up instantly versus the output tubes, there will be more than +230V B+.
The voltage on the top plate is going to be more than 158V before the tube warms up.
You better have a voltage limiter on the +230V B+, or before the tube warms up, you will have 440V on the tube plates.
Cold tubes do not draw 30ma through a 7k voltage dropping resistor.
100uF to drive a 600 Ohm impedance load . . .
100uF is 600 Ohms capacitive reactance at 2.7Hz!
Unless you are trying to reproduce the 6Hz Canon on Telarc's recording of the 1812 Overture, what is the point of a capacitor that is only -3dB at 2.7Hz; and -1dB at 5.4Hz?
A 30uF cap, will be -3dB at 10Hz, and -1dB at 20Hz.
Well, since the two 30uF are in series with the 600 Ohm load, it will be -3dB at 20Hz, and -1dB at 40Hz.
Limit the un-loaded +230V supply, or plan on using some 450V or 600V caps.
Two 230V B+ voltage sources, slow starting of cold output tubes, and whatever voltages appear across the 100V caps during warmup might be too much, especially if the B+ warms up instantly versus the output tubes, there will be more than +230V B+.
The voltage on the top plate is going to be more than 158V before the tube warms up.
You better have a voltage limiter on the +230V B+, or before the tube warms up, you will have 440V on the tube plates.
Cold tubes do not draw 30ma through a 7k voltage dropping resistor.
100uF to drive a 600 Ohm impedance load . . .
100uF is 600 Ohms capacitive reactance at 2.7Hz!
Unless you are trying to reproduce the 6Hz Canon on Telarc's recording of the 1812 Overture, what is the point of a capacitor that is only -3dB at 2.7Hz; and -1dB at 5.4Hz?
A 30uF cap, will be -3dB at 10Hz, and -1dB at 20Hz.
Well, since the two 30uF are in series with the 600 Ohm load, it will be -3dB at 20Hz, and -1dB at 40Hz.
Limit the un-loaded +230V supply, or plan on using some 450V or 600V caps.
Last edited:
@6A3sUMMER the HPF is designed for minimal group delay and phase shift as we approach 100Hz; not subsonic accuracy. In single supply op amp output designs, we often use 220u or even 470u when a 600R load is expected. For my guitar amp application you’re 100% right, I could use smaller caps. For other applications, there are some arguments to be made for 100u or greater. And by a limiter, do you mean a big honkin’ zener after the dropping resistor? Example please.
Guitar amps get away with lots, that is OK, different appliction.
Hi Fi and Stereo, indeed.
- 1dB at 20Hz has a 26 degrees phase shift.
1080 feet / second sound velocity
26 degrees/360 degrees = 0.722 feet
The phase shift at 20Hz is the same as a woofer that is Only 0.722 feet away from the tweeter.
12 x 0.722 = 0.866 Inches
In what world is your woofer only 0.866 inches away from the tweeter.
Put them on a flat plane, and adjust their time alignment, OK for a listener at a distance.
Even if they are time aligned, that is only true for a flat woofer cone.
Most woofer cones are Conical, and deeper than 0.866 inches. Right?
Do we merely take the integrated average distance of the woofer cone. (Conical, Elliptical, Hyperbolic?).
Start improving the system by peeling off the outer layers of the onion.
The inside layer can only be touched, After you peel off the outer layers.
Well at least, everyone can sleep at night if they know you use at least 100uF (-1dB at 5.4Hz).
That is 0.234 inches of distance time shift.
Just my opinions of how to make something more practical; and not to be used as a sleep medication.
If the 100V caps short, at least the phase shift at DC will be 0 degrees.
Voltage limiters.
How complex do you want? Zeners; voltage divider driving a comparator versus a reference voltage, with the comparator driving a bipolar or MOSFET to shunt the voltage?
I hope that helps.
Hi Fi and Stereo, indeed.
- 1dB at 20Hz has a 26 degrees phase shift.
1080 feet / second sound velocity
26 degrees/360 degrees = 0.722 feet
The phase shift at 20Hz is the same as a woofer that is Only 0.722 feet away from the tweeter.
12 x 0.722 = 0.866 Inches
In what world is your woofer only 0.866 inches away from the tweeter.
Put them on a flat plane, and adjust their time alignment, OK for a listener at a distance.
Even if they are time aligned, that is only true for a flat woofer cone.
Most woofer cones are Conical, and deeper than 0.866 inches. Right?
Do we merely take the integrated average distance of the woofer cone. (Conical, Elliptical, Hyperbolic?).
Start improving the system by peeling off the outer layers of the onion.
The inside layer can only be touched, After you peel off the outer layers.
Well at least, everyone can sleep at night if they know you use at least 100uF (-1dB at 5.4Hz).
That is 0.234 inches of distance time shift.
Just my opinions of how to make something more practical; and not to be used as a sleep medication.
If the 100V caps short, at least the phase shift at DC will be 0 degrees.
Voltage limiters.
How complex do you want? Zeners; voltage divider driving a comparator versus a reference voltage, with the comparator driving a bipolar or MOSFET to shunt the voltage?
I hope that helps.
Last edited:
@6A3sUMMER I can look at cap values. The output caps are easier to sub in than the CMR caps — those need to be bipolar. And re: limiting, least parts count = zener. I’ve never looked up HV zeners but there’s a first time for everything I suppose.
My bias was screwy, I was looking at the wrong line on the data sheet. And then I had to adjust some values to keep resistors around 75% wattage. And separating the buildout resistance from the bias network nets a wee bit more level.
Anyways, here’s a very worked-over V2:
There’s a little PSU reference in the top left, with the regular Fender RCRC omitted after the stock CLC. I’m just adding the 4K7 10W and the 22u 350V for the 6N6P.
I don’t see how cathode voltage might spike on a cold start (tell me I’m wrong?), so I left the output caps alone for now. I marked the PSR caps TBD for now, because there is no good 450V bipolar solution. I could take two 220u/250V caps and put ‘em back to back. Higher parts count, more board space. Anybody know the most miniaturized long life cap series in that WV, offhand?
None of this is as important as the basic questions of A) output impedance, and B) what the lowest quiescent current is that I can run here to get 24dBu into 600R at reasonably low THD. I would really appreciate someone looking at that math.
Anyways, here’s a very worked-over V2:
There’s a little PSU reference in the top left, with the regular Fender RCRC omitted after the stock CLC. I’m just adding the 4K7 10W and the 22u 350V for the 6N6P.
I don’t see how cathode voltage might spike on a cold start (tell me I’m wrong?), so I left the output caps alone for now. I marked the PSR caps TBD for now, because there is no good 450V bipolar solution. I could take two 220u/250V caps and put ‘em back to back. Higher parts count, more board space. Anybody know the most miniaturized long life cap series in that WV, offhand?
None of this is as important as the basic questions of A) output impedance, and B) what the lowest quiescent current is that I can run here to get 24dBu into 600R at reasonably low THD. I would really appreciate someone looking at that math.
Last edited:
Seems like you may be inspired by this Broskie circuit below in his blog. He claims the output impedance is 50 ohms with 6DJ8 which should be able to drive 600 ohm headphone load. I prefer his arrangement of the 4 output caps.
The output stage of course can be turned into a differential buffer in the second circuit that allows you drive it with your preferred balanced gain stage.
@directdriver yes, essentially the split load balanced buffer — but with grid bias drawn from the cathode divider, and a little bit of buildout resistance. I had seen that post and then lost it!
The best one-tube alternative is a BCF or WCF with an impedance balanced cold. WCF will have the most robust output current out of all of these circuits I believe; I am not sure how it fares in terms of PSR.
I am not enamored of these big caps in the split load balanced buffer TBH. And IIRC a WCF has lower THD than this (and certainly will at way higher voltage into 600R):
The best one-tube alternative is a BCF or WCF with an impedance balanced cold. WCF will have the most robust output current out of all of these circuits I believe; I am not sure how it fares in terms of PSR.
I am not enamored of these big caps in the split load balanced buffer TBH. And IIRC a WCF has lower THD than this (and certainly will at way higher voltage into 600R):
Last edited:
Well, technically, shifting to an impedance-balanced WCF still comports with the thread title.
Thoughts? I’m not sure if I’m supposed to calc my dropping resistor(s) based on 16.7mA or 33.3mA. Also unsure about the WCF output Z.
Goal is 40Vpp into Riso + 600R, which should result in 24dBu at the + output.
forgot to label the bottom triode coupling cap, it’s the same 100n.
Thoughts? I’m not sure if I’m supposed to calc my dropping resistor(s) based on 16.7mA or 33.3mA. Also unsure about the WCF output Z.
Goal is 40Vpp into Riso + 600R, which should result in 24dBu at the + output.
forgot to label the bottom triode coupling cap, it’s the same 100n.
Anybody muck about with this form? Retains the push-pull aspect of the White (unlike the Aikido) but adds a degree of PSR. I’d have to recalc my cathode and plate resistors. More learning curve, but it would get me closer to the cross-coupled rejection I came into this thread with. C1:C2 is 10:1 or more in one Broskie example from another blog post. He didn’t seem to spend too much time on this form.
PSR?
"What Me Worry?"
("Mad" a magazine for those of you who have never seen a copy).
I generally use Brute Force B+ circuits.
Low DCR choke input, lots of capacitance after the choke, lots of capacitance in the following capacitors, low Ohm resistors from cap to cap, etc.
The ripple is very low at 120Hz full wave rectification frequency; But because of the very large capacitances . . .
The output stage current demands at 20Hz are also taken care of (no 20Hz B+ "ripple").
Therefore, I can use simpler circuits in the amplifier, because I do not have to worry about PSR (Power Supply Rejection).
Just my design criteria.
"What Me Worry?"
("Mad" a magazine for those of you who have never seen a copy).
I generally use Brute Force B+ circuits.
Low DCR choke input, lots of capacitance after the choke, lots of capacitance in the following capacitors, low Ohm resistors from cap to cap, etc.
The ripple is very low at 120Hz full wave rectification frequency; But because of the very large capacitances . . .
The output stage current demands at 20Hz are also taken care of (no 20Hz B+ "ripple").
Therefore, I can use simpler circuits in the amplifier, because I do not have to worry about PSR (Power Supply Rejection).
Just my design criteria.
Last edited:
@6A3sUMMER That makes sense when one has control over the power supply design, and what else hangs off the B+, and what else dumps to ground. In this case, I do not.
https://el34world.com/charts/Schematics/files/Fender/Fender_pro_reverb_aa165_schem.pdf
https://el34world.com/charts/Schematics/files/Fender/Fender_pro_reverb_aa165_schem.pdf
Sounds like you're spending a lot of money on the power supply to avoid one extra cheap capacitor in the amplifier circuit
Merlinb,
I adhere to the rule that says something like . . . start your amplifier design with a good power supply.
Good, not just for hum, but good when the amplifier load versus signal varies widely, such as at an organ's 32 foot pipes low frequency.
Others choose to follow other rules.
Many rules can be viewed as if they are bandaids.
My choice of fundamental rules works for me (less than 100uV of hum into 8 Ohms, is OK to run low to medium efficiency loudspeakers, and for efficient loudspeakers that are at least 12 feet away.
Headphones . . . too much hum.
I adhere to the rule that says something like . . . start your amplifier design with a good power supply.
Good, not just for hum, but good when the amplifier load versus signal varies widely, such as at an organ's 32 foot pipes low frequency.
Others choose to follow other rules.
Many rules can be viewed as if they are bandaids.
My choice of fundamental rules works for me (less than 100uV of hum into 8 Ohms, is OK to run low to medium efficiency loudspeakers, and for efficient loudspeakers that are at least 12 feet away.
Headphones . . . too much hum.
Last edited:
Okay so anyways, I need to get my dropping string right, and the Class AB function is confusing me.
1) Does a Totem pole (WCF, BCF, ACF, whatever) draw a single triode’s Iq, or the total Iq of both triodes? 16.7mA vs 33.3mA, respectively, in this case. My string is set up for 33.3 mA above. And
2) if I need to deliver 57.8mA / 40.5Vpp at max load, am I correct that setting the pair for 33.3mA will (more than) do it, or do I need to set EACH triode for about 30mA quiescent? Re: #1, is that 30mA or 60mA total?
^ Big difference on this 200mA power trafo, which is already supplying two 6L6GC, two 12AT7, et al.
Surely the experts have some insight on these noob Qs.
1) Does a Totem pole (WCF, BCF, ACF, whatever) draw a single triode’s Iq, or the total Iq of both triodes? 16.7mA vs 33.3mA, respectively, in this case. My string is set up for 33.3 mA above. And
2) if I need to deliver 57.8mA / 40.5Vpp at max load, am I correct that setting the pair for 33.3mA will (more than) do it, or do I need to set EACH triode for about 30mA quiescent? Re: #1, is that 30mA or 60mA total?
^ Big difference on this 200mA power trafo, which is already supplying two 6L6GC, two 12AT7, et al.
Surely the experts have some insight on these noob Qs.
Last edited:
If you have no control over the B+ circuit, then you have to be willing to accept the results that the complete amplifier gives . . .
(often less than optimal).
A totem pole is almost always a Serial Connection.
Series circuits draw the same quiescent current, all the way through, from end to end.
Take a mono totem pole that draws 16.7mA; Then two of them will draw 33.3mA for a stereo pair of totem poles.
You do have to consider the top tube in the totem pole, how much Peak current does it need to supply to the load?
Any power supply has to deal with both the output stage quiescent current, and the output stage peak current.
For ease of discussion, use a Class A single ended 300B output stage.
With a quiescent plate current of 60mA, if you drive the 300B to severe clipping, you get 0mA at cutoff in one signal direction (signal polarity), and about 132mA peak current at the other signal current direction (other signal polarity). In Class A operation, the power supply has to provide the average current, in this case about 66mA. A large uF value B+ capacitor that connects to the output transformer's B+ connection, not only provides that 66mA average current, it also provides the 132mA peak current.
The totem pole is not a lot different than that, with one end at ground, and the other end at B+. The B+ filter cap has to have enough capacitance to provide the quiescent current, it also has to provide the average current, and it has to provide the peak current, providing all those current levels, without the B+ voltage dropping significantly.
(often less than optimal).
A totem pole is almost always a Serial Connection.
Series circuits draw the same quiescent current, all the way through, from end to end.
Take a mono totem pole that draws 16.7mA; Then two of them will draw 33.3mA for a stereo pair of totem poles.
You do have to consider the top tube in the totem pole, how much Peak current does it need to supply to the load?
Any power supply has to deal with both the output stage quiescent current, and the output stage peak current.
For ease of discussion, use a Class A single ended 300B output stage.
With a quiescent plate current of 60mA, if you drive the 300B to severe clipping, you get 0mA at cutoff in one signal direction (signal polarity), and about 132mA peak current at the other signal current direction (other signal polarity). In Class A operation, the power supply has to provide the average current, in this case about 66mA. A large uF value B+ capacitor that connects to the output transformer's B+ connection, not only provides that 66mA average current, it also provides the 132mA peak current.
The totem pole is not a lot different than that, with one end at ground, and the other end at B+. The B+ filter cap has to have enough capacitance to provide the quiescent current, it also has to provide the average current, and it has to provide the peak current, providing all those current levels, without the B+ voltage dropping significantly.
Last edited:
Atavacron, without any Ub stabilisation and all that voltage dropping serial resistance in your power supply only class A is an option.
40Vpk into 600ohm is doable, but your response to one of the most helpful people on this site, it wont be me to figure out the details for you
40Vpk into 600ohm is doable, but your response to one of the most helpful people on this site, it wont be me to figure out the details for you
Last edited:
I guess I could do something right after the choke. The next dual triode I’d look at would be the 12BH7A; greater heater-cathode range. But at 440V and a minimal value at the top triode’s plate, I’d have to float the heater supply regardless. 6N6P has curves up to 220V.