glad you have found a way to approach the sizing of your capacitors, tizman
Bill
I’m glad too! Thanks for the link to the Loesch article.
A question for some of you who are more technically minded . . .
I've seen more basic variations of the same type of calculator, posted as a high-pass calculator. I always assumed that the workings of a PS were somehow different, though, and that the same calculator wouldn't apply to a PS.
A high-pass involves a cap in series with the signal followed by a resistor to ground. For example when you have a coupling cap followed by a grid leak, or the output cap of a preamp followed by a resistor to ground.
But in the case of a typical PS, you have have a resistor (or choke) in series and a cap to ground.
Yet the calculator and the Loesch article seem to indicate that the result is the same even though the configuration of the Cs and Rs are different.
And, of course, the purposes are different. A coupling cap in series blocks DC and allows AC (the signal) to pass while in the PS a cap to ground reduces AC (PS ripple) and allows DC to pass.
This also raises the question of, what happens if the rolloff frequency of the C and R in the PS and the rolloff frequency of the C and R in the coupling circuit don't match?
If the coupling cap portion is configured to have a -3db point of 4hz, and the PS portion is configured to have a -3db point of 40hz, what's the result?
I've seen more basic variations of the same type of calculator, posted as a high-pass calculator. I always assumed that the workings of a PS were somehow different, though, and that the same calculator wouldn't apply to a PS.
A high-pass involves a cap in series with the signal followed by a resistor to ground. For example when you have a coupling cap followed by a grid leak, or the output cap of a preamp followed by a resistor to ground.
But in the case of a typical PS, you have have a resistor (or choke) in series and a cap to ground.
Yet the calculator and the Loesch article seem to indicate that the result is the same even though the configuration of the Cs and Rs are different.
And, of course, the purposes are different. A coupling cap in series blocks DC and allows AC (the signal) to pass while in the PS a cap to ground reduces AC (PS ripple) and allows DC to pass.
This also raises the question of, what happens if the rolloff frequency of the C and R in the PS and the rolloff frequency of the C and R in the coupling circuit don't match?
If the coupling cap portion is configured to have a -3db point of 4hz, and the PS portion is configured to have a -3db point of 40hz, what's the result?
According to the Loesch article, as the capacitance of a node increases, the DCR of the power supply feeding that node should decrease. This inverse relationship makes me think about the amplifiers in which my choices for a B+ node capacitance were arbitrary. Employing an extra large cap has repercussions to power supply DCR requirements that should be considered. I will practice more precision in my capacitor value choices moving forward.
They have completely different jobs to do. The coupling cap has to pass the full signal without alterations, the PSU has to pass clean DC POWER with low impedance.
The signal cap never should have the chance to charge and discharge significantly (no grid current allowed). The PSU caps can charge and discharge quite a lot and they do for sure when sustained musical passages happen and the amp is asked to deliver substantial power for some time
One key point that article is missing is that the performance of more complex amps than a Monkey is all about PSU timing! It is obviously more complex to achieve when one has to feed the amp with more than one supply. Addressing the PSU impedance and time constant of the plate supply only is not enough. ALL the time constants of the PSU need to be "right" and the starting key point is:
"The bias supply is as important as the anode supply".
This is often overlooked and I see lots of poor bias supplies around. This is not about spending significant more money.
When I read "Unfortunately, in my experience cathode follower do simply not sound "right" despite distortion (on the bench) is low and mostly coming from the power is tube", the conclusions are most of the time wrong. It's not the cathode follower the problem but the PSU as a whole which is not up to the task.
In my experience the original Audio Note amps designed by Kondo San just sound right to me because everything works as it should. Not saying they are perfect or the best or any other final judgement like that but surely they are well above average, to say the least.
So there are 2 things a power supply has to do:
1) When the amp is called to deliver substantial power for sustained time PSU voltages, including the bias (sic!), should change as less as possible. This is more obvious and, of course, is already described to some extent in that article.
2) The changes in PSU delivery need to happen at uniform rate. This is almost straightforward with a single supply, a lot less with multiple supplies typically found in more complex designs. In other words all voltages (plate, bias and screen if any) have to have the same constant, ideally (as much as possible in practice). Typical fixed bias supplies are often simplified to the bone and, where some care is applied, have relatively long time constant to minimize ripple. That's where the amp falls short of breath....and here is where regulated supplies can be very useful to get it right. Without regulated supplies sometimes it's not possible to get the PSU right for large sustained transients. So the circuit might look great, appealing or whatever but its PSU will never be up to the task and the amp will never give the best it can. But in most cases it is possible to address it.....
Last edited:
Power supply filtering works *exactly* the same as any other low pass filter. It presents a completely predictable source impedance to the amplifier that it's feeding, and a predictable amount of residual noise. Performance increases as both of these are decreased. There is no magic voodoo number that's correct. More is more. More than that is better. Beware of gurus.
All good fortune,
Chris
All good fortune,
Chris
Exactly.
That's why I wonder whether the calculated rolloff frequency should even be part of the discussion when it comes to determining cap value in a PS. Is that type of calculator even a legitimate tool for use in PS design? Its ability to calculate the recharge time of the cap seems useful, but I don't understand how the frequency part is applicable.
Again . . .
If the amp is built so that the rolloff frequencies conflict, what's the actual rolloff frequency? Is it 4hz or 40hz, or somewhere in between?
My technical understanding is somewhat lacking, but I think it's 4hz and that the PS doesn't actually act as a high pass filter since the configuration of the Cs and Rs are different.
A coupling cap / grid leak acts as a high pass filter. The calculator posted earlier gives the same results as a high pass even though a PS is configured differently. Here's a high pass calculator:
High Pass Filter Calculator
As Chris pointed out, the configuration of a PS acts as a low pass filter, not a high pass, as seen in this calculator:
Low Pass Filter Calculator
As you say, the main consideration is that the reserve energy stored in the cap is not depleted. And the lower the DCR, the less time it takes to recharge the caps. That's my understanding, anyway.
That's why I questioned the tiny 2.2uf cap used in the Decware, though my concerns are not informed by any formula or measurement. I've just never seen that low of a value used in any other amp.
Typically, I've noticed (mostly in vintage gear) caps in the 20uf to 30uf range being used to supply the input tubes. And I've never seen caps as large as he recommends used to supply the output tubes. Even in PP amps which (as I understand the theory) have more current demand fluctuation than SE.
Using a tiny 2.2uf cap may be adequate, I don't know. But it's safe to say that, given the same current demands, it would (using a gas tank analogy) get much closer to "running on empty" than a 15uf to 30uf cap. It may be more of a "belt and suspenders" thing but I'd much rather keep the tank over half full.
And, as others have pointed out, using the most demanding music (electronic dance music, 120bpm, with ultra-low frequency synth bass) is not a realistic parameter upon which to base your design decisions. Unless, of course, that's what you're using the amp for. Did his own designs use caps as large as he suggests should ideally be used?
In the first section he discusses various schematics and methods of driving a 300B but only the audio portions are presented. Do any of the these amps - some of them designed by legendary builders - use such large PS caps? I seriously doubt it. Did these designers even use a particular formula to determine PS cap size?
I also question whether the input tubes are subjected to the same demands as the output tubes. I don't think they are. If I'm correct, even if you accept the legitimacy of his formula for determining the cap value for the output tubes, this formula is pretty much useless when applied to the PS node that supplies the input tubes.
Would he suggest the same formula be used to design the PS in a stand-alone preamp? After all, the same frequencies pass through the preamp as do the 300B output tubes.
Just for grins I pulled up a schematic (published Aug '17) for what may be the last amp designed by the legendary Sakuma-san before he died. This was a monobloc.
It actually uses much more capacitance in the supply to the input / driver section than the supply to the output tube. The 5692 input and the KT-66 driver use a SS rectifier followed by 100uf and 33uf, then 33uf to the 5692 input and 33uf to KT-66 driver. The supply to the 845 output tube, which is tube rectified, uses a C-L-C-L-C with the caps rated at 8uf, 8uf and 10uf. Only 26uf total!
The last 300B design he published (Mar '17) takes a different approach. The same PS node supplies both the 6550 driver tube and the 300B output. The PS is C-L-C-L-C and the caps are 10uf, 33uf and 100uf.
It actually uses much more capacitance in the supply to the input / driver section than the supply to the output tube. The 5692 input and the KT-66 driver use a SS rectifier followed by 100uf and 33uf, then 33uf to the 5692 input and 33uf to KT-66 driver. The supply to the 845 output tube, which is tube rectified, uses a C-L-C-L-C with the caps rated at 8uf, 8uf and 10uf. Only 26uf total!
The last 300B design he published (Mar '17) takes a different approach. The same PS node supplies both the 6550 driver tube and the 300B output. The PS is C-L-C-L-C and the caps are 10uf, 33uf and 100uf.
Last edited:
I think the key to understanding his idea re. sizing the last capacitor for a tube is to realize that in a single-ended amp the power supply circuit loop is completed by the last capacitor in the power supply. The music signal goes through it.
The capacitor and the anode of the tube are in series, so yes, frequency roll-offs and time constants can be calculated. This is the AC impedance consideration he outlines.
Separately he describes the filtering and DC impedance issues in the design of the PS.
Bill
The capacitor and the anode of the tube are in series, so yes, frequency roll-offs and time constants can be calculated. This is the AC impedance consideration he outlines.
Separately he describes the filtering and DC impedance issues in the design of the PS.
Bill
To be fair, Decware used a 3.3 uF cap in that position in the SE84CS. Doing the math with the (usually) higher plate resistance driver tubes results in much lower capacitance requirements. It would appear from the Olsen ETF article linked by Bill Brown (if I am reading it correctly) that it should still matter what the capacitor value is for a driver tube for its power supply in a similar way that it matters for a power tube.
I asked for a guide/formula/procedure for sizing capacitor value to a tube’s B+ node. So far, all I have is the Loesch article that has anything providing a procedure or formula for working out this value. I’m totally open to other ways of working it out, but I haven’t seen any yet. What other people did in the past doesn’t necessarily transcribe to what is best in every situation. My understanding is that Sakuma-san had a holistic approach that resulted in a great sounding amp, but that did not lend itself to straight up circuit copying.
Last edited:
On second thought, is it 10% of the tube's plate resistance that should be used, or is it 10% of its' plate resistor value? For the 30 in my amp, if it is the plate resistor, 5.9 kOhm is used in the equation. This results in a 12.3 Hz cutoff, and a time constant of 13 MS. I'm not sure as it isn't covered in the article.
Also, what about the capacitors in the RCs before the node in question? If a node calls for 20 uF, can the capacitor in the node before be a 5 uF capacitor?
Also, what about the capacitors in the RCs before the node in question? If a node calls for 20 uF, can the capacitor in the node before be a 5 uF capacitor?
Last edited:
My thinking has been that earlier capacitances (as LC or RC sections) are where the AC is filtered out and "pure," noise free DC is generated. It is interesting to try different capacitor sizes in these positions with PSUD as you can see the DC on an initial large cap (sometimes goes quite high), resonances of RCs and LCs, and the effects on ripple.
The impedance of these sections (and of the total PS), then determine how quickly these capacitors and "charge up." With reasonably low impedance prior to the nodes capacitor, it can then be sized as Loesch does to ensure an adequate "DC reservoir" for that node. That can be looked at in PSUD with a current step- set the start time at a point where the PS has stabilized- tube rectifiers are fully "on" and all the capacitors are charged, then see how much the voltage sags at each node after it has been asked to provide more current.
So my thinking is filter, then maintain low impedance to quickly charge/replenish reservoir capacitors to provide DC to each stage.
Bill
The impedance of these sections (and of the total PS), then determine how quickly these capacitors and "charge up." With reasonably low impedance prior to the nodes capacitor, it can then be sized as Loesch does to ensure an adequate "DC reservoir" for that node. That can be looked at in PSUD with a current step- set the start time at a point where the PS has stabilized- tube rectifiers are fully "on" and all the capacitors are charged, then see how much the voltage sags at each node after it has been asked to provide more current.
So my thinking is filter, then maintain low impedance to quickly charge/replenish reservoir capacitors to provide DC to each stage.
Bill
Oh, and re. your last question. If Loesch is right, I think it is the plate impedance that is used in the calculations. He says in his example of the 300B that it is the 700 ohm plate resistance of the 300B rather than the reflected load it sees via the transformer. This of course will decrease the required capacitance of the reservoir capacitor of the driving stages with their higher (typically) Rp.
I think
I think
My doubts are about the calculator that tizman posted. He assumes it can be used, in conjunction with some of what Thorsten writes about, to optimize the value of the cap.
As I see it, the part of the calculator that shows the recharge time of the cap is interesting but inaccurate. And, I think the frequency aspect of the calculator is being totally misconstrued.
The text on the calculator page he linked states "Depending on the configuration you can use the RC filter to either filter out low or high frequencies. These are the high pass and low pass filters." But it doesn't show schematics that represent what a high pass and low pass filter looks like.
The calculators I posted earlier do.
High Pass Filter Calculator
Low Pass Filter Calculator
And if you compare the schematics you'll clearly see that the configuration used in a PS (the R in series and the C to ground) constitutes a low pass filter, not a high pass filter. While the configuration of a high pass filter is what you find with a coupling cap and the grid leak R that follows it.
Yet the discussion here has repeatedly claimed that the PS is a high pass filter which rolls off frequencies below a certain point. Take a look at the schematics. Which one looks like the PS?
Even the portion of the calculator that shows the recharge time is of limited use. Why limited?
Because when an amp is first powered up the caps charge fully. Then, after the music starts playing, the current demands reduce the energy in the cap and it is replenished by the voltage supplied to the cap.
But here's what the calculator says about the recharge time it shows: "Charging of the capacitor is an exponential process . . . The capacitor charge time is the time it takes for the capacitor to get charged up to around 63%. If you double the time, you get about 87%."
In other words, it's calculating the time it takes to charge a totally uncharged cap up to 63% and then to 87%. The amp, in contrast, starts with a fully charged cap which then loses some of its charge due to current demands and is subsequently replenished. These are two entirely different situations and, as a result, the calculated time is not accurate.
In the real world, it would be much more useful to know the time needed to replenish the charge of a cap that has only been partially depleted. Say you start with it at 87%, how much time does it take to recharge it from that point?
I'm not much of a math guy, but I assume someone here can extrapolate and tell us the % of charge after the calculated time doubles again and again. That might be more useful, especially if you can determine the % that the cap is discharged in real world use.
Thorsten's focus is on making sure the B+ cap that feeds the output tubes is large enough so that the energy it stores doesn't come anywhere close to being depleted, even when the most demanding music is being played. If music with continuously heavy bass is being played at 120 bpm, it's important to replenish the cap as quickly as possible.
As I see it, the goal is to strike a balance between having a capacitor with adequate storage and the ability to recharge it quickly enough. The lower the impedance of the supply is, the faster the cap can be replenished.
I, and other posters, have some question about the need to use such extremely demanding music to choose a cap size, unless that's the type of music the amp will be used for. But the goals of adequate energy storage and minimal recharge time are certainly valid no matter what music is being played.
But I would submit that the calculator tizman posted doesn't really help us to identify the optimal cap size.
I would point out two things about this article. First, it only addresses the B+ cap and the output tube section which it supplies, there is no mention of previous stages.
If previous stages played an important role, why doesn't he discuss their contributions and the parts needed to optimize their function?
In addition, he makes no mention of this type of calculator and he makes no claim that the parts he chooses have any effect on frequency response, either as a high pass or a low pass.
The schematic shows it does act as a low pass, but since it's not mentioned in the article, I assume that aspect is not an important consideration. Otherwise, one would think he would mention it.
When looking at the B+ supply and the output tube, Mr. Loesch looks at the cap value and the plate resistance of the tube. But if you look at the PS node that supplies the input tube, there is an additional element which is not present in the output section - the plate load resistor - which is in series with the plate resistance of the tube and is typically 3x to 5x the value of the plate resistance. Resistances in series are additive. Why would the tube's plate resistance be important but the much higher value of the load resistor be totally ignored?
I occurs to me that he also ignores the DCR of the OT primary but typically that's a much lower value than the plate load resistor used with an input tube.
There are a lot of extremely knowledgeable and even legendary designers out there and, from what I've seen, none of them go to extremes regarding PS cap sizes. I'm not aware that any of them use use the large cap values that Loesch suggests for the B+ cap or the tiny cap value used in the Decware.
Just to be clear, I'm not saying the cap size doesn't matter. I'm just saying I'd rather have a little more reserve in the tank than a 2.2uf or 3.3uf or whatever. It seems to me that using a somewhat larger value to supply the input tubes should have benefits and I don't see any potential downside.
My point in posting the comments about Sakuma-san's amps was that, since those two examples use very different approaches to the PS cap values, it would seem that he had no consistent approach or formula that he applied when choosing those cap values. On the other hand, other elements in his design philosophy can be seen to be consistently present in his various creations.
As I see it, the part of the calculator that shows the recharge time of the cap is interesting but inaccurate. And, I think the frequency aspect of the calculator is being totally misconstrued.
The text on the calculator page he linked states "Depending on the configuration you can use the RC filter to either filter out low or high frequencies. These are the high pass and low pass filters." But it doesn't show schematics that represent what a high pass and low pass filter looks like.
The calculators I posted earlier do.
High Pass Filter Calculator
Low Pass Filter Calculator
And if you compare the schematics you'll clearly see that the configuration used in a PS (the R in series and the C to ground) constitutes a low pass filter, not a high pass filter. While the configuration of a high pass filter is what you find with a coupling cap and the grid leak R that follows it.
Yet the discussion here has repeatedly claimed that the PS is a high pass filter which rolls off frequencies below a certain point. Take a look at the schematics. Which one looks like the PS?
Even the portion of the calculator that shows the recharge time is of limited use. Why limited?
Because when an amp is first powered up the caps charge fully. Then, after the music starts playing, the current demands reduce the energy in the cap and it is replenished by the voltage supplied to the cap.
But here's what the calculator says about the recharge time it shows: "Charging of the capacitor is an exponential process . . . The capacitor charge time is the time it takes for the capacitor to get charged up to around 63%. If you double the time, you get about 87%."
In other words, it's calculating the time it takes to charge a totally uncharged cap up to 63% and then to 87%. The amp, in contrast, starts with a fully charged cap which then loses some of its charge due to current demands and is subsequently replenished. These are two entirely different situations and, as a result, the calculated time is not accurate.
In the real world, it would be much more useful to know the time needed to replenish the charge of a cap that has only been partially depleted. Say you start with it at 87%, how much time does it take to recharge it from that point?
I'm not much of a math guy, but I assume someone here can extrapolate and tell us the % of charge after the calculated time doubles again and again. That might be more useful, especially if you can determine the % that the cap is discharged in real world use.
Thorsten's focus is on making sure the B+ cap that feeds the output tubes is large enough so that the energy it stores doesn't come anywhere close to being depleted, even when the most demanding music is being played. If music with continuously heavy bass is being played at 120 bpm, it's important to replenish the cap as quickly as possible.
As I see it, the goal is to strike a balance between having a capacitor with adequate storage and the ability to recharge it quickly enough. The lower the impedance of the supply is, the faster the cap can be replenished.
I, and other posters, have some question about the need to use such extremely demanding music to choose a cap size, unless that's the type of music the amp will be used for. But the goals of adequate energy storage and minimal recharge time are certainly valid no matter what music is being played.
But I would submit that the calculator tizman posted doesn't really help us to identify the optimal cap size.
I would point out two things about this article. First, it only addresses the B+ cap and the output tube section which it supplies, there is no mention of previous stages.
If previous stages played an important role, why doesn't he discuss their contributions and the parts needed to optimize their function?
In addition, he makes no mention of this type of calculator and he makes no claim that the parts he chooses have any effect on frequency response, either as a high pass or a low pass.
The schematic shows it does act as a low pass, but since it's not mentioned in the article, I assume that aspect is not an important consideration. Otherwise, one would think he would mention it.
When looking at the B+ supply and the output tube, Mr. Loesch looks at the cap value and the plate resistance of the tube. But if you look at the PS node that supplies the input tube, there is an additional element which is not present in the output section - the plate load resistor - which is in series with the plate resistance of the tube and is typically 3x to 5x the value of the plate resistance. Resistances in series are additive. Why would the tube's plate resistance be important but the much higher value of the load resistor be totally ignored?
I occurs to me that he also ignores the DCR of the OT primary but typically that's a much lower value than the plate load resistor used with an input tube.
There are a lot of extremely knowledgeable and even legendary designers out there and, from what I've seen, none of them go to extremes regarding PS cap sizes. I'm not aware that any of them use use the large cap values that Loesch suggests for the B+ cap or the tiny cap value used in the Decware.
Just to be clear, I'm not saying the cap size doesn't matter. I'm just saying I'd rather have a little more reserve in the tank than a 2.2uf or 3.3uf or whatever. It seems to me that using a somewhat larger value to supply the input tubes should have benefits and I don't see any potential downside.
My point in posting the comments about Sakuma-san's amps was that, since those two examples use very different approaches to the PS cap values, it would seem that he had no consistent approach or formula that he applied when choosing those cap values. On the other hand, other elements in his design philosophy can be seen to be consistently present in his various creations.
Last edited:
FlaCharloe: My posts with regard to the calculator assume that Loesch’s procedure is correct, but I don’t know that for sure as I, like yourself, am not a math guy. That said, no one who knows for sure has said otherwise thus far. The calculators posted and used in my examples work when applied to Loesch’s specific example of a 300B. Using the equations results in the same numbers that Loesch arrives at in the article.
Using the resistance after the capacitance in question would seem to imply that the tube’s plate resistance would be used in the equations in the case of the power tube. Following the same logic would imply that the plate resistor of the tube is the resistance used in the case of driver tubes.
Again, I don’t know for sure if I’m doing it correctly, or if the entire procedure is appropriate.
Using the resistance after the capacitance in question would seem to imply that the tube’s plate resistance would be used in the equations in the case of the power tube. Following the same logic would imply that the plate resistor of the tube is the resistance used in the case of driver tubes.
Again, I don’t know for sure if I’m doing it correctly, or if the entire procedure is appropriate.
"And if you compare the schematics you'll clearly see that the configuration used in a PS (the R in series and the C to ground) constitutes a low pass filter, not a high pass filter. While the configuration of a high pass filter is what you find with a coupling cap and the grid leak R that follows it.
Yet the discussion here has repeatedly claimed that the PS is a high pass filter which rolls off frequencies below a certain point. Take a look at the schematics. Which one looks like the PS?"
This is a point of misunderstanding. There is no doubt that the LCs and RCs in a PS are configured as high pass filters. That is their job. Ideally, we are trying to produce 0 HZ current (DC).
However, the final capacitor feeding a single-ended stage is in series (not parallel) with the stage, the music signal goes through it, this then a highpass filter. This was a hot topic here 10 or so years ago and lead to many different designs trying to get this cap out of the current loop (WE, differential, parafeed, etc.).
Check out the Lynn Olson ETF presentation I attached earlier. He does a pretty good job of explaining it.
Bill
Yet the discussion here has repeatedly claimed that the PS is a high pass filter which rolls off frequencies below a certain point. Take a look at the schematics. Which one looks like the PS?"
This is a point of misunderstanding. There is no doubt that the LCs and RCs in a PS are configured as high pass filters. That is their job. Ideally, we are trying to produce 0 HZ current (DC).
However, the final capacitor feeding a single-ended stage is in series (not parallel) with the stage, the music signal goes through it, this then a highpass filter. This was a hot topic here 10 or so years ago and lead to many different designs trying to get this cap out of the current loop (WE, differential, parafeed, etc.).
Check out the Lynn Olson ETF presentation I attached earlier. He does a pretty good job of explaining it.
Bill
I see where both Loesch and your calculator comes up with the same time that's needed to replenish the cap. But he says nothing at all about either high pass or low pass filters and their effects. Nothing. So I don't think the frequency aspect of the calculator applies.
The plate resistance of the driver tube may need to be used in the calculation but it seems to me that the value of the load resistor needs to be added to it. If the cap interacts with plate resistance of the tube, then it also must interact with the load resistor, since it's in series with the plate resistance and positioned between the plate and the cap.
I would guess that the demands placed on the input tube are vastly less than those placed on the output tube. But, again, it's not at all clear that Loesch's formula, which is based on B+ supply to output tubes, is even applicable to the input tube and its supply node.
I'm not sharp enough to dissect Loesch's article in detail. I just don't think that the particular calculator you posted tells you much of anything useful. If a standard electronic calculator could be used to explain his ideas more clearly, I would have thought that he would refer to it in his article.
Loesch says nothing about high pass or low pass rolloffs so that aspect of the calculator seems to be superfluous at best and misleading at worst. And, since some of your comments link the size of the input tube supply cap with a particular rolloff frequency, I would say that you have been misled.
Your comments assume the cap is acting as a high pass filter yet, as the schematics in the calculators I posted clearly show, a PS is actually configured as a low pass filter. So, my conclusion is that the frequency aspect of your chosen calculator is misleading and is confusing the issue.
And, as I pointed out, the part of the calculator that gives you the recharge time is also of limited use because it shows the time needed to go from 0 to 67%, while the time we're interested in is the time needed to go from (arbitrarily) 67% to 100%, or perhaps 87% to 100%. In reality, depending on the demands of the music, it's possible that the cap's charge may actually fluctuate between, say, 50% and 90% or something similar and it may not actually become fully recharged until the music stops.
Here's a calculator that also deals with cap charge time. The difference is that it tells you the voltage level that a cap is charged to after a given amount of time.
Capacitor Charge (Charging) Calculator
If you play with it you'll notice that the rate of charge is exponential, as the other calculator pointed out. Here's the results I got when I entered a supply voltage of 200v (which means the cap would be at 200v when fully charged), a resistor value of 250 ohms, and a cap value of 30uf. I then ran the calculator over a series of times, starting at 5mS and adding an additional 5mS with each step.
Total time --->Voltage
5 ---> 97.3
10 ---> 147.3
15 ---> 172.3
20 ---> 186.3
25 ---> 192.8
30 ---> 196.3
35 ---> 198.1
40 ---> 199.0
45 ---> 199.5
50 ---> 199.75
55 ---> 199.87
60 ---> 199.93
65 ---> 199.965
70 ---> 199.982
75 ---> 199.9909
80 ---> 199.9953
But, it seems to me, that in order to make it truly useful in determining an optimal cap size, you would need to be able to determine the voltage fluctuation that actually occurs when the amp is playing music. If you don't know how much charge you've lost, you can't really know how fast it can be replenished. That may be possible to measure with a scope or some other equipment (none of which I have) but I can tell you that I've put my multimeter on PS caps with music playing and it's not sensitive enough, or perhaps fast enough, to detect any fluctuation.
So the key elements that are important in determining if a particular cap size is effective are: how much is the fully charged cap being depleted when music is playing? . . . how much stored energy does it take to replenish it? . . . and is the impedance of the PS low enough to replenish the cap in a minimal amount of time?
Last edited:
I've read the Olson presentation. It has nothing at all to do with the question raised by tizman about the optimal size of the cap that supplies his input tubes.
Olson's point is that, in a SE amp, the cap, the tube, the OT, the cathode R and cathode bypass cap are all in the signal path and that they do, collectively, colorize the sound.
He goes on to discuss different circuit topologies and how the current loops in each differ. The advantage of the alternative topologies he presents are to remove the audio currents from the B+ cap, "leaving it to perform the much simpler task of filtering ripple from the PS".
But nowhere does he say that the PS creates a high pass filter or that the final cap in the PS should be different in an amp that's designed to reproduce the full range of audio vs one that designed to reproduce only high frequencies which, I believe, is one of tizmans avenues of inquiry.
Olson's presentation focusses on something entirely different than the Loesch article. He makes absolutely no mention of how different cap values influence the sound or the ability of the amps to reproduce challenging bass material.
Loesch's article is all about cap size and bass response, not about the characteristics of various topologies. And, since he's discussing the amp's ability to reproduce bass, if the PS created a high pass filter, then the presence of a high pass filter would seem to be a factor worthy of at least passing mention. But he also says nothing about a PS creating a high pass or low pass filter.
And, again, both the Loesch article and the Olsen presentation only talk about the output stage . . . and the focus of the recent discussion here - in this thread - is the size of the cap that supplies the input stage.
If you want to see what cap values Olson uses, you can click on the links that take you back to his home page and then look at the schematics of his amps.
The way he draws schematics is a bit confusing to me and there's also the fact that he couples everything with interstage transformers.
But the PS cap values he uses are almost all in the same range as what I see in other amps, anywhere from 40uf to 100uf. I say almost because he uses a c-L-C topology using a tiny first cap - typically .5uf to 1 uf. I've heard this referred to as a quasi-cap input supply.
But, again, the subject of discussion - in this thread - is the cap that supplies the input tube. If you look at the final caps that supply the input sections in Olson's amps, you'll see that they are quite conventional, often around 40uf. You'll also notice that his B+ supply caps are often around 100uf, so nowhere near what Thorsten recommends.
It would be interesting to get Olson's take on input tube cap size but it seems as though he hasn't been active in audio forums, or on the internet, in many years. The same is true for the other builders mentioned on his website.
Rather than debate theory and the question of whether the cap that supplies the input tube is part of a high pass filter, tizman can easily settle the issue.
I believe his amp is currently configured to reproduce the full frequency range and he's been listening to it with normal speakers, not using an active crossover and just tweeters. If that's still the case, he can just try a simple test.
Using an online signal generator, send the amp a relatively low frequency, say 50hz. This will just be a steady tone which will demonstrate only that the amp is capable of reproducing this frequency. It's not a dynamic, thumping bass drum since this is not a test of the cap's ability to provide enough current for such music. We just want to see if it can reproduce the basic tone. Presumably, it can.
Now remove the PS cap that that supplies the input tube and, leaving everything else as it was, replace it with a very low value film cap. Say, something around 0.1uf or .047uf. In a high pass filter, if you reduce the value of the cap, it creates a higher rolloff frequency. With a cap that small it should go much higher - perhaps as high as 800hz or 1k??
Now play the same 50hz tone, with the volume control set the same. If the cap is part of a high pass filter, the 50hz tone will either be attenuated drastically or, perhaps, even inaudible.
Tizman, please try this and post the results. Perhaps I'm wrong and it does form a high pass filter.
Last edited:
FlaCharlie: As I mentioned, I’m not sure if Loesch is correct, and therefore if the use of the formulas and calculators are correct. I know he’s far more experienced and familiar with the task at hand than I am, and that there must be a way to calculate the best capacitance value of a power supply node. If his method is wrong, and his calculators are wrong, what is the right way? I’m still waiting for someone to show me the right way to do it.
Based upon your arguments about the applicability of Loesch’s method to driver tubes, if you are correct, I would expect the capacitance required to be even smaller than previously calculated if you add the plate resistor value to the equation.
Based upon your arguments about the applicability of Loesch’s method to driver tubes, if you are correct, I would expect the capacitance required to be even smaller than previously calculated if you add the plate resistor value to the equation.