Hello,
As per title, what kind of PSRR value is to be expected from a typical push-pull power output stage, say a 6L6 in regular transformer-coupled circuit without negative feedback, perhaps in a realistic situation where tubes age and the stage might not be perfectly balanced ?
I just need a rough ballpark figure to estimate how crazy I have to get on power supply ripple filtering; the input stage will be regulated.
Thanks in advance,
Joris
As per title, what kind of PSRR value is to be expected from a typical push-pull power output stage, say a 6L6 in regular transformer-coupled circuit without negative feedback, perhaps in a realistic situation where tubes age and the stage might not be perfectly balanced ?
I just need a rough ballpark figure to estimate how crazy I have to get on power supply ripple filtering; the input stage will be regulated.
Thanks in advance,
Joris
Last edited:
When it comes to power supplies, I am a believer in Brute Force.
I not only consider rectifier ripple (2X line frequency), I consider the load signal current ripple.
Consider the capacitive reactance at 120Hz (full wave center tapped, or bridge rectification).
I use 200uF or 500uF to "power" the center tap of a push pull output transformer.
200uF @ 120Hz, Xc = 6.63 Ohms. (13.3 Ohms at 60Hz)
13.3 Ohms is probably good enough when the upright bass plays a 60Hz note
200uF @ 20Hz, Xc = 39.8 Ohms.
Suppose the 20Hz signal current is 70mA peak, that is 0.070A x 39.8 Ohms = 2.79 Volts peak 20Hz signal ripple.
Is that good enough?
Suppose the plate swing at 20Hz is 200 volts, and the 20Hz ripple is 2.79V.
20 Log (2.79V/200V) = -37.1 dB.
Want to reduce it by another 7.96 dB, then use a 500uF cap instead of a 200uF cap.
-37.1 dB or -45dB; 200uF versus 500uF respectively.
Consider 6L6 at 50mA each, quiescent current. During Class AB1 operation, one tube cuts off, one tube goes to 50mA plus about 70mA 120mA total, (tubes go to more than 2x the quiescent current, when the control grid = cathode voltage (2nd harmonic non- linearity; cancelled by push pull, but the current character is the same of each tube for single ended or for push pull).
If a well matched push pull tube output has 20dB PSRR, the 500uF cap at 20 Hz will give -65dB, the total rejection.
Both 200uF or a 500uF are probably good enough, a 20 second long very loud 30 foot organ pipe will reduce your B+ anyway, but that is one extreme example for the average listener.
Do not even worry about the definitive Telarc recording of the 6Hz Canon signal in the 1812 Overture.
Your loudspeakers will fold up, distort like crazy, and you just have to enjoy whatever sound you get for that impossible to reproduce signal.
Fix that . . . Just get a real canon, and set it off at the appropriate times on the Score. Realism at its best.
I have been on a USS Destroyer, when the 5 inch 38 caliber rounds were "pushed out" of the end of the barrel, so to speak.
LOUD and CONCUSSIVE.
5 inch diameter round, and a barell length that is 5 inch x 38 = 190 inch barell.
The biggest 50 Caliber in the world were on USS Battleships. 16 Inch diameter rounds. 16 x 50 = 800 inch, 66.7 foot barell.
The USS New Jersey with its new barells could "Hurl" the equivalent of the original small Volkswagen Bug 63 miles. (fastest car trip ever).
Seriously, Have Fun!
Happy power supply designing.
I not only consider rectifier ripple (2X line frequency), I consider the load signal current ripple.
Consider the capacitive reactance at 120Hz (full wave center tapped, or bridge rectification).
I use 200uF or 500uF to "power" the center tap of a push pull output transformer.
200uF @ 120Hz, Xc = 6.63 Ohms. (13.3 Ohms at 60Hz)
13.3 Ohms is probably good enough when the upright bass plays a 60Hz note
200uF @ 20Hz, Xc = 39.8 Ohms.
Suppose the 20Hz signal current is 70mA peak, that is 0.070A x 39.8 Ohms = 2.79 Volts peak 20Hz signal ripple.
Is that good enough?
Suppose the plate swing at 20Hz is 200 volts, and the 20Hz ripple is 2.79V.
20 Log (2.79V/200V) = -37.1 dB.
Want to reduce it by another 7.96 dB, then use a 500uF cap instead of a 200uF cap.
-37.1 dB or -45dB; 200uF versus 500uF respectively.
Consider 6L6 at 50mA each, quiescent current. During Class AB1 operation, one tube cuts off, one tube goes to 50mA plus about 70mA 120mA total, (tubes go to more than 2x the quiescent current, when the control grid = cathode voltage (2nd harmonic non- linearity; cancelled by push pull, but the current character is the same of each tube for single ended or for push pull).
If a well matched push pull tube output has 20dB PSRR, the 500uF cap at 20 Hz will give -65dB, the total rejection.
Both 200uF or a 500uF are probably good enough, a 20 second long very loud 30 foot organ pipe will reduce your B+ anyway, but that is one extreme example for the average listener.
Do not even worry about the definitive Telarc recording of the 6Hz Canon signal in the 1812 Overture.
Your loudspeakers will fold up, distort like crazy, and you just have to enjoy whatever sound you get for that impossible to reproduce signal.
Fix that . . . Just get a real canon, and set it off at the appropriate times on the Score. Realism at its best.
I have been on a USS Destroyer, when the 5 inch 38 caliber rounds were "pushed out" of the end of the barrel, so to speak.
LOUD and CONCUSSIVE.
5 inch diameter round, and a barell length that is 5 inch x 38 = 190 inch barell.
The biggest 50 Caliber in the world were on USS Battleships. 16 Inch diameter rounds. 16 x 50 = 800 inch, 66.7 foot barell.
The USS New Jersey with its new barells could "Hurl" the equivalent of the original small Volkswagen Bug 63 miles. (fastest car trip ever).
Seriously, Have Fun!
Happy power supply designing.
Last edited:
You laugh but for the 400th of Quebec city's founding they performed the 1812 Overture outside in a park where the Brits defeated the French and conquered Canada. They actually had the army dust off a couple Howitzers to perform the canon parts... To this day I still regret I didn't attend the event.
One summer concert in downtown Portland, OR . . .
I was there for the 1812 Overture, and the Oregon Guard's Howitzer's Debut.
I was there for the 1812 Overture, and the Oregon Guard's Howitzer's Debut.
I'm more into lower capacitances figures as I use paper in oil caps (motor runs). My current design has only 45uF finals. If bass isn't satifactory I have the options to go to maybe 80-100uF or crank up the input cap to get more voltage from the PT and stick a regulator on the B+. BUT I want to try it without regulation first.
That being said I've pretty much given up hopes of having bass on par with solid-state from a tube amp. The figures you mention would required a hefty power supply... Still I try to get better chances with a PT having 3.7x the required VA rating and a low impedance filter chain (36 ohms series DC resistance).
Last edited:
Another way to look at it . . .
Try 45uF!
A 6000 Ohm plate to plate output transformer; when one tube is cut off, the other tube is driving 1500 Ohms.
Consider a low frequency note of 30Hz.
45uF Xc @ 30Hz = 117.9 Ohms
117.9 Ohms is in series with 1500 Ohms.
That is 117.9 + 1500 Ohms = 1617.9 Ohms the increased impedance that the driving 6L6 plate "sees"
As a voltage divider at 30Hz, 1500 / 1617.9 = 0.927.
0.927 = -0.657 dB.
Of course, the output transformer's primary inductance may be attenuating the signal even further.
At 100H plate to plate, that would be only 25H plate to center tap, the 6L6 has to drive this too.
25H @ 30 Hz = 4712 Ohms inductive reactance.
Compare that to the 1500 Ohms, it increased the load by about 30% at 30Hz.
An average electric bass guitar plays down to just over 40Hz.
OK. I want to hear your amplifier. When it is up and running, turn the volume up!
Try 45uF!
A 6000 Ohm plate to plate output transformer; when one tube is cut off, the other tube is driving 1500 Ohms.
Consider a low frequency note of 30Hz.
45uF Xc @ 30Hz = 117.9 Ohms
117.9 Ohms is in series with 1500 Ohms.
That is 117.9 + 1500 Ohms = 1617.9 Ohms the increased impedance that the driving 6L6 plate "sees"
As a voltage divider at 30Hz, 1500 / 1617.9 = 0.927.
0.927 = -0.657 dB.
Of course, the output transformer's primary inductance may be attenuating the signal even further.
At 100H plate to plate, that would be only 25H plate to center tap, the 6L6 has to drive this too.
25H @ 30 Hz = 4712 Ohms inductive reactance.
Compare that to the 1500 Ohms, it increased the load by about 30% at 30Hz.
An average electric bass guitar plays down to just over 40Hz.
OK. I want to hear your amplifier. When it is up and running, turn the volume up!
Last edited:
Oh I forgot to mention the output stage is class A. I'm not advanced enough to design a proper AB stage...
This eases the power supply requirements a lot since the load is more constant and almost perfectly balanced hence better PSRR through common-mode rejection.
Last edited:
Actually, the j117.9 ohms adds at right angles to the 1500 ohms, so the attenuation is even less.
The way to look at it is the pole frequency between the load impedance and the cap. 45 uF and 1500 ohms is a 2.35 Hz cutoff frequency. You want this frequency to be at least 5 times lower than the lowest frequency you want to reproduce. Rule of thumb says this is half a dB down. For 30 Hz operation, you are doing quite well. Even if you consider the loading for two channels plus some for previous stages. Doesn’t take near as much cap with loads of a k or more, compared to 4 ohms for two channels of SS.
The way to look at it is the pole frequency between the load impedance and the cap. 45 uF and 1500 ohms is a 2.35 Hz cutoff frequency. You want this frequency to be at least 5 times lower than the lowest frequency you want to reproduce. Rule of thumb says this is half a dB down. For 30 Hz operation, you are doing quite well. Even if you consider the loading for two channels plus some for previous stages. Doesn’t take near as much cap with loads of a k or more, compared to 4 ohms for two channels of SS.
SomeJoe,
wg_ski,
Class A.
Suppose one 6L6 current ranges from 10mA (near cutoff, but not at cutoff), to quiescent of 50mA, to peak of 110mA.
(40 mA less than quiescent, and 60 mA more than quiescent) . . .
While the other tube is doing the opposite . . . going from 60mA more than quiescent, to quiescent of 50mA, to 40mA less than quiescent.
Those are merely approximations, but as you can see, the Class A load on the B+ is fairly constant.
Yes, wg_ski was right, it is a complex load.
The error in this case is just like using an inexpensive old 5 inch slide rule, and not interpolating the readings between the very few marks.
Using my excellent old bamboo Post Versalog, I would have got a bad grade for not doing the trig, and getting a more accurate number.
I would rather eat a piece of Pie, than use Pi in a calculation.
wg_ski,
Class A.
Suppose one 6L6 current ranges from 10mA (near cutoff, but not at cutoff), to quiescent of 50mA, to peak of 110mA.
(40 mA less than quiescent, and 60 mA more than quiescent) . . .
While the other tube is doing the opposite . . . going from 60mA more than quiescent, to quiescent of 50mA, to 40mA less than quiescent.
Those are merely approximations, but as you can see, the Class A load on the B+ is fairly constant.
Yes, wg_ski was right, it is a complex load.
The error in this case is just like using an inexpensive old 5 inch slide rule, and not interpolating the readings between the very few marks.
Using my excellent old bamboo Post Versalog, I would have got a bad grade for not doing the trig, and getting a more accurate number.
I would rather eat a piece of Pie, than use Pi in a calculation.
The stage being class A, doesn't both tubes see the whole OPT primary impedance at all times at AC? My OPT is 5K P-P, should be easy to drive...
Just read about those, interesting... Is it marked "Made in Occupied Japan" ?
I don't recall ever having held a slide rule in my hand, let alone use one... However I still use my high-school scientific calculator in the workshop/lab... Had to change the batteries a few years ago after 35 years in there. They truly don't make things like they used to...
There are a couple of ways to analyze a Class A push pull output stage.
You probably know this, but let's entertain everyone, especially the Newbies.
To make calculations simple, connect the 6L6 tubes in Triode Wired Mode.
The 6L6 "triode-wired" plate impedance, rp, is about 1250 Ohms (actually closer to 1700 Ohms, but to simplify the discussion, 1250 Ohms).
Your 5k p-p primary is also 1250 Ohms from plate to center tap, and the other plate to the center tap is also 1250 Ohms.
Both tubes are always on in Class A.
We can easily find a Triode Wired Pentode or Triode Wired Beam Power tube at certain voltages and currents that has an rp of 1250 Ohms.
This thread uses 6L6, so we tweaked on the element spacings, and got 1250 Ohms rp.
1. "Parallel" operation analysis:
Class A operation only
Each "triode" drives 1250 Ohms, but one triode pushes up 1250 Ohms, while the pull triode pulls down 1250 Ohms.
The two tubes are Aiding Each Other.
So, consider two parallel 1250 Ohm triodes driving a single 1250 Ohm primary winding.
With 1250 Ohms rp in parallel with another 1250 Ohm rp, we effectively have 1250/2 = 625 Ohms driving 1250 Ohms primary winding.
That gives a damping factor of 2.0 at the primary, if we could have a lossless transformer, that would give a damping factor at the seondary of 2.0.
DF of 2.0, before global negative feedback is applied (or we could choose to Not use global negative feedback).
2. "Series" operation analysis.
Class A operation only (again)
Consider one 1250 Ohm plate driving one end of the 5k p-p winding, and the other 1250 Ohm plate driving the other end of the 5k p-p winding.
We have a series circuit, 1250 Ohms rp, 5k p-p Ohms primary and another 1250 Ohms rp.
That is 1250 + 1250 Ohms driving 5k Ohms total primary load.
5k primary / 2500 Ohms total plate impedances = 2.0
"Magically" we again have a damping factor of 2.0 at the primary.
(I never thought of this 'series mode Class A push pull' on my own, someone else pointed it out to me, I think it was Lynn Olson, thanks!).
One of the things I really love, is when you can use two completely different calculations, and get the same answer!
That is always Either both right, Or both wrong.
In my long carriers that were either "using measurements", or that "was measurements", I often tried to find 2 independent measurement methods that got the same answer.
Again, both right, or both wrong.
If we used the push pull 6L6s in Ultra Linear Mode with 40% UL taps, the rp might be about 4k Ohms for each 6L6.
Clearly, we would either need to use a 10k p-p or 12k p-p primary, or we would need to use either push pull Schade, Modified push pull Schade to driver cathodes, or Global Negative Feedback.
If we used the push pull 6L6s in Beam Power mode, the rp of each tube is about 22k to 33k.
Clearly, we would need to use either push pull Schade, Modified push pull Schade to driver cathodes, or Global Negative Feedback.
If you can stand it to read through all of the above, I say Thank You!
You probably know this, but let's entertain everyone, especially the Newbies.
To make calculations simple, connect the 6L6 tubes in Triode Wired Mode.
The 6L6 "triode-wired" plate impedance, rp, is about 1250 Ohms (actually closer to 1700 Ohms, but to simplify the discussion, 1250 Ohms).
Your 5k p-p primary is also 1250 Ohms from plate to center tap, and the other plate to the center tap is also 1250 Ohms.
Both tubes are always on in Class A.
We can easily find a Triode Wired Pentode or Triode Wired Beam Power tube at certain voltages and currents that has an rp of 1250 Ohms.
This thread uses 6L6, so we tweaked on the element spacings, and got 1250 Ohms rp.
1. "Parallel" operation analysis:
Class A operation only
Each "triode" drives 1250 Ohms, but one triode pushes up 1250 Ohms, while the pull triode pulls down 1250 Ohms.
The two tubes are Aiding Each Other.
So, consider two parallel 1250 Ohm triodes driving a single 1250 Ohm primary winding.
With 1250 Ohms rp in parallel with another 1250 Ohm rp, we effectively have 1250/2 = 625 Ohms driving 1250 Ohms primary winding.
That gives a damping factor of 2.0 at the primary, if we could have a lossless transformer, that would give a damping factor at the seondary of 2.0.
DF of 2.0, before global negative feedback is applied (or we could choose to Not use global negative feedback).
2. "Series" operation analysis.
Class A operation only (again)
Consider one 1250 Ohm plate driving one end of the 5k p-p winding, and the other 1250 Ohm plate driving the other end of the 5k p-p winding.
We have a series circuit, 1250 Ohms rp, 5k p-p Ohms primary and another 1250 Ohms rp.
That is 1250 + 1250 Ohms driving 5k Ohms total primary load.
5k primary / 2500 Ohms total plate impedances = 2.0
"Magically" we again have a damping factor of 2.0 at the primary.
(I never thought of this 'series mode Class A push pull' on my own, someone else pointed it out to me, I think it was Lynn Olson, thanks!).
One of the things I really love, is when you can use two completely different calculations, and get the same answer!
That is always Either both right, Or both wrong.
In my long carriers that were either "using measurements", or that "was measurements", I often tried to find 2 independent measurement methods that got the same answer.
Again, both right, or both wrong.
If we used the push pull 6L6s in Ultra Linear Mode with 40% UL taps, the rp might be about 4k Ohms for each 6L6.
Clearly, we would either need to use a 10k p-p or 12k p-p primary, or we would need to use either push pull Schade, Modified push pull Schade to driver cathodes, or Global Negative Feedback.
If we used the push pull 6L6s in Beam Power mode, the rp of each tube is about 22k to 33k.
Clearly, we would need to use either push pull Schade, Modified push pull Schade to driver cathodes, or Global Negative Feedback.
If you can stand it to read through all of the above, I say Thank You!
Last edited:
Well I did even before finishing my first morning coffee, a feat for me 😎 My understanding of the class A PP stage was the "series" version above, which seems more intuitive to me.
The amplifier topology is two-stage fully differential, implemented as basically two long tailed-pairs sitting on sand CCSes, IT-coupled. It is a redo of an amp I built a while ago, but that amp's input stage was choke-loaded and cap coupled. It sounded quite good but its regulator being salvaged from a previous 2A3 SET it burned down from over-heating... I was young and stupid at the time. Now things are different, I'm not young anymore lol
The output stage is indeed wired 40% ultra-linear, and with both stages CCS-driven the only way to implement GNFB is the "Schade" topology, only at the time I didn't know it had a fancy name, just calling it "plate-to-plate" feedback...
At the time I only put a single resistor from the each input tube plate to the corresponding output tube plate, whereas I see that true Schade feedback also puts a resistor from input plate to ground... Would that make my implementation a "modified push pull Schade" ?
I guess I can put in that second resistor now but I don't see the point as the input tube itself is a "varying" resistance, but perhaps its purpose is to provide a more constant feedback ratio over that varying output resistance. I have to run errands now but I'll dig further into that topology for sure.
Last edited:
The extra resistor to ground sounds like it does 2 things:
It Ratios the Schade negative feedback, as you have noted
It Changes the current paths from the output tube plate, to both the input tube plate and plate load resistor, and the extra resistor to ground.
It affects the DC level of the input tube plate.
It is also possible to use individual self bias at the input tubes cathodes (with the cathodes no longer connected to each other).
The feedback resistor has to drive a capacitor to the junction of a series pair of cathode resistors, the top resistor is bypassed with a capacitor, just like most self bias circuits.
The series RC feeds the junction of those two series cathode resistors.
It can be done without the series capacitor in the feedback path, but then the output stage plate voltage, the input stage self bias DC voltage is affected by those 3 resistors, harder to calculate the resistors to get both the self bias voltage correct, and at the same time get the negative feedback ratio you want.
With the series RC, the negative feedback drops to zero, at some low frequency.
This works for both single ended, and for push pull.
I built both Schade; and modified Schade series RC to the input stage cathode.
Unfortunately, for push pull which is driven by a single ended signal, the input stages have to have the cathodes connected to each other, and to the CCS.
That is necessary to make the input stage into a phase splitter.
For a two stage amplifier, using Both Schade negative feedback, and Ultra Linear, the gain is reduced quite a bit.
The four factor Tradeoffs, of Pentode/Beam Power mode; Ultra Linear mode; and Triode Wired Pentode/Beam Power mode are . . .
Output Power
Distortion
Damping Factor
Gain
It Ratios the Schade negative feedback, as you have noted
It Changes the current paths from the output tube plate, to both the input tube plate and plate load resistor, and the extra resistor to ground.
It affects the DC level of the input tube plate.
It is also possible to use individual self bias at the input tubes cathodes (with the cathodes no longer connected to each other).
The feedback resistor has to drive a capacitor to the junction of a series pair of cathode resistors, the top resistor is bypassed with a capacitor, just like most self bias circuits.
The series RC feeds the junction of those two series cathode resistors.
It can be done without the series capacitor in the feedback path, but then the output stage plate voltage, the input stage self bias DC voltage is affected by those 3 resistors, harder to calculate the resistors to get both the self bias voltage correct, and at the same time get the negative feedback ratio you want.
With the series RC, the negative feedback drops to zero, at some low frequency.
This works for both single ended, and for push pull.
I built both Schade; and modified Schade series RC to the input stage cathode.
Unfortunately, for push pull which is driven by a single ended signal, the input stages have to have the cathodes connected to each other, and to the CCS.
That is necessary to make the input stage into a phase splitter.
For a two stage amplifier, using Both Schade negative feedback, and Ultra Linear, the gain is reduced quite a bit.
The four factor Tradeoffs, of Pentode/Beam Power mode; Ultra Linear mode; and Triode Wired Pentode/Beam Power mode are . . .
Output Power
Distortion
Damping Factor
Gain
Last edited:
Good question, to which nobody here has an answer. Someone really should measure it some time.
You experts are all looking in the wrong place.big capacitors,usually electrolytic slow things down soundwise.answere use smaller value poly caps/clarity t2 awsome and use chokes with medium henries and low dcr. Cut core chokes sound best,think hasimoto or monolith magnetics. If you want level shifting powerfull bass use a choke input power supply.once you have heard the sound transformation you will never go bak to cap input filter.try and hear it for yourself.
Here's the passive part of the PSU so far. The input capacitor is there to adjust the output voltage for two possible scenarios, either with output stage B+ regulation or not. The 3H is 26 ohms DCR and the little 320mH is 10 ohms DCR. C1-C3 are paper in oil motor run caps.
Merlinb,
Your Post # 15 . . .
Unbalanced output tubes.
No, at least not to start.
Tubes as they age, can not avoid that, but you can minimize the problem with balance over time:
Everybody . . . Please spend the money for very well matched output tubes; and of the same manufacturer; and the same vintage of that manufacturer.
Most of you probably spent at least as much $$$ per output transformer as the $$$ you spent per output tube.
Use a good and reliable bias circuit. I prefer individual self bias with individual bypass caps; or a single resistor 'common' self bias (no bypass capacitor).
Individual Adjustable Fixed bias is good too.
Measure the amplifier with the new very well matched tubes.
Pay attention to 2nd and 3rd Harmonic Distortion; and 2nd and 3rd Intermodulation distortion; at moderate power levels.
Then check saturation at low frequencies, power output before clipping, and damping factor.
As the tubes age, if there is a major balance problem, or an equal but major tube aging emission loss, it will show up in saturation at low frequencies,
or power output before clipping, respectively.
Using un-balanced push pull output tubes, is just like racing an old Datsun 240Z that has unequal balance of the two carburetors.
When the tubes have a problem, replace them (or live with poor amplifier performance & stop complaining).
Negative feedback can only fix moderate amplifier problems.
It often is like a band-aid . . . when some medication, or surgery is needed.
Just my $0.03 (accounting for inflation) It was nice when it was $0.02.
Your Post # 15 . . .
Unbalanced output tubes.
No, at least not to start.
Tubes as they age, can not avoid that, but you can minimize the problem with balance over time:
Everybody . . . Please spend the money for very well matched output tubes; and of the same manufacturer; and the same vintage of that manufacturer.
Most of you probably spent at least as much $$$ per output transformer as the $$$ you spent per output tube.
Use a good and reliable bias circuit. I prefer individual self bias with individual bypass caps; or a single resistor 'common' self bias (no bypass capacitor).
Individual Adjustable Fixed bias is good too.
Measure the amplifier with the new very well matched tubes.
Pay attention to 2nd and 3rd Harmonic Distortion; and 2nd and 3rd Intermodulation distortion; at moderate power levels.
Then check saturation at low frequencies, power output before clipping, and damping factor.
As the tubes age, if there is a major balance problem, or an equal but major tube aging emission loss, it will show up in saturation at low frequencies,
or power output before clipping, respectively.
Using un-balanced push pull output tubes, is just like racing an old Datsun 240Z that has unequal balance of the two carburetors.
When the tubes have a problem, replace them (or live with poor amplifier performance & stop complaining).
Negative feedback can only fix moderate amplifier problems.
It often is like a band-aid . . . when some medication, or surgery is needed.
Just my $0.03 (accounting for inflation) It was nice when it was $0.02.
Michael,
Agreed. Use a choke input power supply.
That requires careful calculation of the secondary voltage, when it is under load of the chokes averaging effect (as opposed to cap input filters with terrible peak to average current, fast rise, harmonic producing, noisy ground loop, etc.).
When I design the B+ supply, I am very particular about the tradeoffs of B+ ripple, B+ ground loops, complexity, and reliability.
Solid state rectifiers, quality choke, cap, low Ohms series resistor, high UF cap (for the output stage), medium Ohms series resistor, and cap for the phase splitter/driver. The high uF cap for the output stage takes care of at least 90% of both the music transients, and high average music power.
As to 'speed' of the power supply, there is how long it takes a capacitor to recover from a series of brief, spaced, transient currents;
Versus from a 20 second long sustained heavy current to produce a full power 32.7 Hz organ note.
A Class AB push pull amplifier has different requirements for the brief transient, versus the low frequency 40 measures of the C' pedal tone.
Give me a 500uF cap, the brief transient will Not Budge the B+ voltage. There is nothing that needs to be recovered, so speed is not the issue.
If you are an Organ fan, that changes the B+ system design requirements completely. A whole new circuit, or it will not meet the requirements.
Performance criteria:
Pick a design that maximizes transient response.
Pick a design that maximizes long term voice coil heaters (and cone suspension warmers).
Pick a design that is somewhere in between those requirements.
System design changes depending on what is paramount . . . including cost, weight, parts count, size, and complexity/simplicity, etc.
Agreed. Use a choke input power supply.
That requires careful calculation of the secondary voltage, when it is under load of the chokes averaging effect (as opposed to cap input filters with terrible peak to average current, fast rise, harmonic producing, noisy ground loop, etc.).
When I design the B+ supply, I am very particular about the tradeoffs of B+ ripple, B+ ground loops, complexity, and reliability.
Solid state rectifiers, quality choke, cap, low Ohms series resistor, high UF cap (for the output stage), medium Ohms series resistor, and cap for the phase splitter/driver. The high uF cap for the output stage takes care of at least 90% of both the music transients, and high average music power.
As to 'speed' of the power supply, there is how long it takes a capacitor to recover from a series of brief, spaced, transient currents;
Versus from a 20 second long sustained heavy current to produce a full power 32.7 Hz organ note.
A Class AB push pull amplifier has different requirements for the brief transient, versus the low frequency 40 measures of the C' pedal tone.
Give me a 500uF cap, the brief transient will Not Budge the B+ voltage. There is nothing that needs to be recovered, so speed is not the issue.
If you are an Organ fan, that changes the B+ system design requirements completely. A whole new circuit, or it will not meet the requirements.
Performance criteria:
Pick a design that maximizes transient response.
Pick a design that maximizes long term voice coil heaters (and cone suspension warmers).
Pick a design that is somewhere in between those requirements.
System design changes depending on what is paramount . . . including cost, weight, parts count, size, and complexity/simplicity, etc.
Last edited:
Agreed re tube matching, all excellent points.
I like to use “Blumein Garters” (with upper resistors bypassed in the conventional manner) to help with bias current equalization as the tubes/valves age. Self correcting, reliable, does not require adjustment are the attributes I like (being inherently lazy) about this arrangement.
Downside is one loses/wastes an additional piece of the B+ voltage. Which needs to be accounted for in the PS design of course.
As with most things, no free lunch.
Another option would be CCS’s in the output tubes cathodes ala Baby Huey, but then one is in class A the whole time. Which is not a bad thing in itself. 😉
Re matched tubes, I like +/- a mA or so, as a good place to start in any case.
My 2 cents, ymmv etc etc