I was wondering if it was resistance on the output board itself, but I must confess I have no idea how it is designed. You're probably at the point now where all you can do is start building something then ABX test it to see if it makes a difference, though what I'd probably do is incrementally add capacitors, you may find you don't need hundreds of uF.
How much capacitance does the amp have in it on power rails now? Please excuse if you mentioned it and I overlooked this detail. I'm wondering how you arrived at the number 200,000uF, is it what you decided to allocate price wise, space wise, estimation, or mathematical calculation?
How much capacitance does the amp have in it on power rails now? Please excuse if you mentioned it and I overlooked this detail. I'm wondering how you arrived at the number 200,000uF, is it what you decided to allocate price wise, space wise, estimation, or mathematical calculation?
That number was from an article by Pass; his suggestion for a power supply for a 1000 watt amp regardless of whether a conventional power suppply or switching mode power supply.
I can't rememeber what's in there already; frankly I'm very pleased the stock caps still work, considering I bought some of these amps in the 1970's.
I can't rememeber what's in there already; frankly I'm very pleased the stock caps still work, considering I bought some of these amps in the 1970's.
The two types of PSU have vastly different capacitance needs. You can hack out a bulk energy storage capacitor bank as we've already discussed, and it could be no more expense than the difference in price between the psu you have and an appropriately higher capacity PSU, but only because you already have that PSU... to start over from scratch, the ideal would be using a higher capacity PSU instead for the same total/retail cost.
I have to wonder why you need 1800W amp? Usually this is reserved for commercial use where the employer is footing the bill so the lower cost option is the one that takes the least time on the engineer's part. For home use, it will just make you deaf faster and break apart walls which isn't necessarily a good thing... but I digress, you can spend a lot of time tweaking capacitance for a SMPS, but in the end it is usually better to use the time finding a higher capacity PSU and selling the inadequate one. Same is true for linear psu, though there is a broader range of capacitance increase vs benefit before one reaches the point of diminishing returns.
Here's a crazy question: What if it is accepted that it will distort if cranked up all the way? My point is, even if you get rid of the current weakest link, then it will be something else instead, then if you get rid of that weak link, still the next weak link and so on. Tweaking tll perfection tends to be not a very effective use of time, in the end if one wants 1800W they start with an amp spec'd for more and underestimate, just as you have seen with yours where its performance already is.
Just a thought, pushing a design to it's limits tends to have negative consequences because they economize and build the whole thing within the design limit.
I have to wonder why you need 1800W amp? Usually this is reserved for commercial use where the employer is footing the bill so the lower cost option is the one that takes the least time on the engineer's part. For home use, it will just make you deaf faster and break apart walls which isn't necessarily a good thing... but I digress, you can spend a lot of time tweaking capacitance for a SMPS, but in the end it is usually better to use the time finding a higher capacity PSU and selling the inadequate one. Same is true for linear psu, though there is a broader range of capacitance increase vs benefit before one reaches the point of diminishing returns.
Here's a crazy question: What if it is accepted that it will distort if cranked up all the way? My point is, even if you get rid of the current weakest link, then it will be something else instead, then if you get rid of that weak link, still the next weak link and so on. Tweaking tll perfection tends to be not a very effective use of time, in the end if one wants 1800W they start with an amp spec'd for more and underestimate, just as you have seen with yours where its performance already is.
Just a thought, pushing a design to it's limits tends to have negative consequences because they economize and build the whole thing within the design limit.
Indeed, buying more of them is a more effective option. But I'm also a curious experimenter and like to DIY tinker, especailly when it makes an improvement. You should see my sport motorcycle...LOL. Hobbies.
I think I found some guys with the schematics. Sounds like they normally use 2 10,000 mfd caps. And they recommend replacement someday soon.
I think I found some guys with the schematics. Sounds like they normally use 2 10,000 mfd caps. And they recommend replacement someday soon.
Having designed a few SMPS in my time adding significant output capacitance is a very bad idea. You are liable to trip current limiting circuits as the supply will duty cycle limit after what it thinks the statup should take. In addition to this the controler for the supply is designed around a specific value of output capacitance and the supply could become unstable which will destroy anything connected to it. If the output of the supply drops under load your probobly saturating the output inductor and its time to get a supply with a better output rating.
It would got to 100% duty cycle...I don't really know whether it will continue balls-out or limit. I'm talking about really short bursts. What I'm more worried about is that I have no cnotrol over whether that energy is used to handle a longer burst or larger burst; if I added a lot more energy storage intended for handling a burst demand for a few milliseconds more that energy migh be misused on a larger burst of shorter duration which might be more damaging.
Yeah, you certainly could be right Kipman725. It's risky any maybe even stupid. It's not like I can do a complete re-design and subsequent reliability tests. Or I might be able to double or triple (or more) the capacitance without any problems beyond startup, which could be handled multiple ways. I'll have to start with a look at the schematic. If I do it, it might be important to be careful with the material I feed it; no static sparks to the input terminal, no touching the hot terminal to listen for 60 hz hum...
The issue I'm hopelessly curious about is the same as the one Pass mentioned. A conventional amp powered by a switching supply would benefit from a sizable bank of capacitors just as much as it would if it were powered by a conventional simple transformer supply. Electronics schools teach power supply design with sizing these caps appropriately to the switching frequency and noise. Then real designers find that they have to size them appropriate to the amp's audio signal which may be several more cycles, or at the lmiits the amp's bass response suffers. Then usually they add ten times as much because it sounds much better and doesn't cost too much more. Then sometimes they add 10000 times as much and make the final realizaton that there's a lot to be gained from designing to take advantage of the dynamic nature of music...that you can have a smaller supply still do a decent job of handling occasional short peak demands. My issue is to try to raise some awareness of the nature of the musical material, and make power supplies and amps with capabilities to suit the music instead of the switching signal. It's just so difficult to be thinking about a fully regulated switching supply and simultaneously consider its perfomance outside those limits. So people ignore the issue and limit its dynamic capabilities to within the fully-regulated realm. I like switching mode power supplies, and I also like class "D" amps. But their strengths in continuous ratings don't reflect their poorer performance on real material unless way over-sized. Than again when a swithcin supply or amp is big enough to operate with all peak burst demands completely within its continuoust rating we do like the fact its rails don't sag or compress the music. But a swithching amp followed by a conventional amp does not necessarily need to have this same limitation of no additional dynamic headroom. It seems to me we should be able to improve this short-time tone-burst spec without adding iron or weight by just adding comparativley light and cheap capacitance. This could be a special advantage of this combination of switching supply and conventional amp. It could be small and light and cheap as a switching-mode amp, and also have additional short-burts headroom.
Or it could blow the power supply and/or the amp. It's just something I've been thinking about...
Yeah, you certainly could be right Kipman725. It's risky any maybe even stupid. It's not like I can do a complete re-design and subsequent reliability tests. Or I might be able to double or triple (or more) the capacitance without any problems beyond startup, which could be handled multiple ways. I'll have to start with a look at the schematic. If I do it, it might be important to be careful with the material I feed it; no static sparks to the input terminal, no touching the hot terminal to listen for 60 hz hum...
The issue I'm hopelessly curious about is the same as the one Pass mentioned. A conventional amp powered by a switching supply would benefit from a sizable bank of capacitors just as much as it would if it were powered by a conventional simple transformer supply. Electronics schools teach power supply design with sizing these caps appropriately to the switching frequency and noise. Then real designers find that they have to size them appropriate to the amp's audio signal which may be several more cycles, or at the lmiits the amp's bass response suffers. Then usually they add ten times as much because it sounds much better and doesn't cost too much more. Then sometimes they add 10000 times as much and make the final realizaton that there's a lot to be gained from designing to take advantage of the dynamic nature of music...that you can have a smaller supply still do a decent job of handling occasional short peak demands. My issue is to try to raise some awareness of the nature of the musical material, and make power supplies and amps with capabilities to suit the music instead of the switching signal. It's just so difficult to be thinking about a fully regulated switching supply and simultaneously consider its perfomance outside those limits. So people ignore the issue and limit its dynamic capabilities to within the fully-regulated realm. I like switching mode power supplies, and I also like class "D" amps. But their strengths in continuous ratings don't reflect their poorer performance on real material unless way over-sized. Than again when a swithcin supply or amp is big enough to operate with all peak burst demands completely within its continuoust rating we do like the fact its rails don't sag or compress the music. But a swithching amp followed by a conventional amp does not necessarily need to have this same limitation of no additional dynamic headroom. It seems to me we should be able to improve this short-time tone-burst spec without adding iron or weight by just adding comparativley light and cheap capacitance. This could be a special advantage of this combination of switching supply and conventional amp. It could be small and light and cheap as a switching-mode amp, and also have additional short-burts headroom.
Or it could blow the power supply and/or the amp. It's just something I've been thinking about...
Last edited:
^ to some extent the larger burst is self-limiting by peak voltage, assuming your PSU voltage is within the design limits of the amp and they did design the amp to meet the rated specs. On the other hand, turn the volume knob all the way up on many amps and eventually they will fail... without having added capacitance.
I don't really agree that a conventional amp will benefit as much from the capacitor bank increase if comparing it to a non-regulated PSU like so many use for projects because of the relative simplicity and low part count, unless the switching PSU current capability is too low for the job.
It's sort of like trying to use a 1 gallon bucket to carry 1.5 gallons of water. You could put a hole in the bucket, connect a tube, strap a small extra bucket on the side of the main bucket, or just use the right sized bucket instead.
I don't really agree that a conventional amp will benefit as much from the capacitor bank increase if comparing it to a non-regulated PSU like so many use for projects because of the relative simplicity and low part count, unless the switching PSU current capability is too low for the job.
It's sort of like trying to use a 1 gallon bucket to carry 1.5 gallons of water. You could put a hole in the bucket, connect a tube, strap a small extra bucket on the side of the main bucket, or just use the right sized bucket instead.
Last edited:
On tube guitar amps, the input signal has huge dynamic range, and the amps may operate with the outputs way outside their linear range or with the power supply rail voltage sagging severely. I had a lot of fun playing with emotionally expressive sounds where playing louder would compress and distort the output. Like Roy Buchannan playing major against minor in question/answer phrases, while also playing clean against dirty. It gets emotive like a saxophone or a voice adding harmonics and complex distortion when played loud. Hi-fi reproduction has a fundamentally different objective. But now I know just enough to be dangerous.
It seems I've been thinking about this wrong after re-reading what people said here. The switching supply produces its full voltage output continuously and keeps up with current demand into reasonable impedance loads. Can't get beter than that. No amount of capacitor storage will raise that voltage for any additional voltage output headroom. DOH! It's only gong to make any difference at all if it's driving a low-impedance speaker load and runs out of current and the rails sag, or maybe for a really complex messy load impedance curve. Don't know what I was thinking. I've got to get used to thinking about these SMPS. And the loads I'm going to use these with should be OK. So like somebody said, I don't really have a problem. And I would never hear any "enhancement" and the change I propose would not improve the spec I was concerned about.
That all said, I sould re-cap someday and when I do I may use the best that fit in the space anyway.
Thanks for getting my head right.
That all said, I sould re-cap someday and when I do I may use the best that fit in the space anyway.
Thanks for getting my head right.
Last edited:
Hi All,
I totally agree witpman725.
I have already addressed this issue in the forum. in the case of smps, does not make sense at large output capacitors, apart from problems of instability or startup. you get a very minimal increase in performance. The reason is that all the smps, are in reality in the DC-DC converters. then the job of filtering is translated To the first cell filtration. For this reason, some very bad examples, we see around smps power outputs of 500-1000w, with huge capacity (1000-1500uF in high voltage cell). solves this certainly warrants burst and current output, but destroys the form factor (cosine of f). apparently against the law while selling smps. yes, pfc stage resolve it, But is real dirt achievement for high-end audiophile amplifiers.
Also, I still see that many people think the regulator in DC.
this is a big mistake, regulators must be seen in the AC way (if the psu is for audio). this requires a very fast response and does not play on the main output capacitors. This applies to any regulator, including eg. an LM337.
Regards
Roberto P.
I totally agree witpman725.
I have already addressed this issue in the forum. in the case of smps, does not make sense at large output capacitors, apart from problems of instability or startup. you get a very minimal increase in performance. The reason is that all the smps, are in reality in the DC-DC converters. then the job of filtering is translated To the first cell filtration. For this reason, some very bad examples, we see around smps power outputs of 500-1000w, with huge capacity (1000-1500uF in high voltage cell). solves this certainly warrants burst and current output, but destroys the form factor (cosine of f). apparently against the law while selling smps. yes, pfc stage resolve it, But is real dirt achievement for high-end audiophile amplifiers.
Also, I still see that many people think the regulator in DC.
this is a big mistake, regulators must be seen in the AC way (if the psu is for audio). this requires a very fast response and does not play on the main output capacitors. This applies to any regulator, including eg. an LM337.
Regards
Roberto P.
Last edited:
watch your specs
Just a word of caution. If you want to try high value smps caps, bear in mind that the usually used few K uF ones are rated 105 C. and can get pretty hot in use.It is likely that large values would draw more current (especially at start)
It struck me that having seen many smps's with cooked ones (especially in monitors where the resulting latch failure gives weird boot symptoms or failure) are a boon to manufacturers by providing built in obsolescence.I replace same value but higher voltage rating.
Just a word of caution. If you want to try high value smps caps, bear in mind that the usually used few K uF ones are rated 105 C. and can get pretty hot in use.It is likely that large values would draw more current (especially at start)
It struck me that having seen many smps's with cooked ones (especially in monitors where the resulting latch failure gives weird boot symptoms or failure) are a boon to manufacturers by providing built in obsolescence.I replace same value but higher voltage rating.
I have experience in that I repair very "high end" self-powered pro speakers.
Several of their psu designs use an off the shelf 320w 48v switcher followed by either two or four 22k uF 50v caps. We get about 7 years or so of daily use before the caps start bulging. As for the necessity of the caps, in their designs the speaker measures what we consider worse when one or more are removed. Certainly design dependent but there's my two pennys.
Several of their psu designs use an off the shelf 320w 48v switcher followed by either two or four 22k uF 50v caps. We get about 7 years or so of daily use before the caps start bulging. As for the necessity of the caps, in their designs the speaker measures what we consider worse when one or more are removed. Certainly design dependent but there's my two pennys.
They usually get hot in use because of ESR, equivalent series resistance. A large add-on bank of decent quality capacitors designed for low ESR as should be used in switching circuits, will have much lower ESR and thus will not only generate much less heat, but the vastly larger surface area will shed the heat all that much better.
It isn't the inital turn-on surge that contributes to any significance in the capacitor heating, BUT that surge current could fry something upstream of the capacitors so current limiting should be employed if it is done at all, which the OP seems now to realize isn't an ideal solution.
I should add a qualifier to the above, as with monitors passively cooled and enclosed in minimally sized metal shielding, the heat buildup from the other components can certainly contribute to capacitor demise, especially if they are mere milimeters away from switching transistors or coils 'sinking away heat to the PCB copper that they are soldered to, but generally in switching filter roles their death is due to either ESR being too high or an unstable electrolyte.
Last edited:
why do it
To go back to the original thread, we must consider the limiting circuit factor for high current hf transients. I havent done the math, but 1/2pi fC gives an approximate ESR very low at the switching frequency even for 20 uF.
These bursts would probably be well supplied by such a cap. The limit may lie in the supply transformer output or the output stage impedance
To go back to the original thread, we must consider the limiting circuit factor for high current hf transients. I havent done the math, but 1/2pi fC gives an approximate ESR very low at the switching frequency even for 20 uF.
These bursts would probably be well supplied by such a cap. The limit may lie in the supply transformer output or the output stage impedance
FWIW:
1. Linear Power supplies required large caps because of the very slow switching speed: 120 hz [rectified]. The low switching speed requires a large storage pool of energy to prevent excessive ripple.
2. Typical SMPS have switching speeds measured in Kilohertz, so a much smaller energy storage pool is required. Oversizing filter caps on a SMPS can increase regulation instability, causing ripple to increase. Typically a SMPS has the feedback tuned to the LC output filter to provide low ripple.
3. For low Ripple requirements, Caps with low ESR and low inductance is the key to reduce ripple. Resistance and inductance impledance will limit the energy transfer from the cap to the load. The lower the ESR and inductance that better the filtering response. Ceramic caps usually have the lowest ESR, but have very small capacitance values compared to electrolytic (wet) caps. Most wet caps have low ESR when they are brand new, but as time passes, the electrolyte evaporates increasing ESR. Bigger wet caps typically have a larger inductances which can act as a current choke limiting the amount of energy that it can release. So the become ineffective with SMPS switching at tens of Kilohertz.
Some designers have tried hybrid solutions using parallel sets of wet and ceramic caps to get the best of both. One of the active members here on this forum "EVA" had posted some results that contradicted this (if I recall correctly) . She posted this probably two or three years ago. I would run a search for output cap filters with her name "EVA" to get a good analysis.
4. If your looking to reduce ripple to nearly zero, take look at this white paper which discussed the use for coupled inductors to filter out ripple noise. A parallel winding on the SMPS output inductors that has a shunt to ground with a small low ESR ceramic cap will shunt the ripple providing a low noise output.
http://www.hamill.co.uk/pdfs/azrtatad.pdf
1. Linear Power supplies required large caps because of the very slow switching speed: 120 hz [rectified]. The low switching speed requires a large storage pool of energy to prevent excessive ripple.
2. Typical SMPS have switching speeds measured in Kilohertz, so a much smaller energy storage pool is required. Oversizing filter caps on a SMPS can increase regulation instability, causing ripple to increase. Typically a SMPS has the feedback tuned to the LC output filter to provide low ripple.
3. For low Ripple requirements, Caps with low ESR and low inductance is the key to reduce ripple. Resistance and inductance impledance will limit the energy transfer from the cap to the load. The lower the ESR and inductance that better the filtering response. Ceramic caps usually have the lowest ESR, but have very small capacitance values compared to electrolytic (wet) caps. Most wet caps have low ESR when they are brand new, but as time passes, the electrolyte evaporates increasing ESR. Bigger wet caps typically have a larger inductances which can act as a current choke limiting the amount of energy that it can release. So the become ineffective with SMPS switching at tens of Kilohertz.
Some designers have tried hybrid solutions using parallel sets of wet and ceramic caps to get the best of both. One of the active members here on this forum "EVA" had posted some results that contradicted this (if I recall correctly) . She posted this probably two or three years ago. I would run a search for output cap filters with her name "EVA" to get a good analysis.
4. If your looking to reduce ripple to nearly zero, take look at this white paper which discussed the use for coupled inductors to filter out ripple noise. A parallel winding on the SMPS output inductors that has a shunt to ground with a small low ESR ceramic cap will shunt the ripple providing a low noise output.
http://www.hamill.co.uk/pdfs/azrtatad.pdf
Tech guy:
1) No no no. In a "linear" power supply for a GOOD high-end audio amp the caps do NOT just filter the 60 HZ or 120 hz to remove ripple. Their purpose is mostly to compensate for the relatively poor voltage "regulation" performance of the unregulated transformer output. That's why they are sized hundreds of times bigger than necessary to reduce the ripple to insignifcant. The cap bank smooths and reduces very short-time rail voltage drops that would otherwise occur when musical peaks are reproduced with short high-current demands. For continuous tones, the caps make no long-term improvement, but for music they allow the amp to handle occasional huge current demands without making the supply rails drop much (which would compress the dynamic range at the least). That's why Pass recommends such huge values for those caps, much larger than necessary for the ripple. They're a LARGE power reserve. Think of them more as batteries than fiters, though of course they are both. If you think of them as filters, they are sized to the dynamic range and other characteristics of the peak and average levels in the material and to the power transformer's performance as its load changes. That's what I'm used to, a big bank of extra energy reserve. That's what got me thinking wrong in the first place. On a linear power supply, the caps do keep the rail voltages up for musical peaks. On a switching supply, the rail volages are always regulated to the same value, so the caps don't do anything unless the current demand is so huge the SMPS is outside its intended operating range, then the voltage drops quickly or the power supply protection kicks in.
2) Yes. But that wasn't the question. But I will keep it in mind.
3) Low ripple is not what I was asking about, but worth noting.
4) No, I wasn't asking about the ripple. My interest was in storing extra energy to handle short high-current demands. But I'm just going to re-wire my load for different impednaces and purchase more complete amplifiers. But I love a good white paper!
Is a huge sag in the rail voltage "ripple"? I wouldn't pick that word.
Again, in the past I could significanly improve the sound of most amps just by adding more filter caps. But a lot of that sonic improvement was not the additional short-term current capacity, it was the fact it met that demand without "significant" voltage drop in the supply rails. It's cheaper and lighter than a bigger transformer, and there's something to be said for optimizing your amp for the common characteristics of music.
Anyway, thanks all for the responses.
1) No no no. In a "linear" power supply for a GOOD high-end audio amp the caps do NOT just filter the 60 HZ or 120 hz to remove ripple. Their purpose is mostly to compensate for the relatively poor voltage "regulation" performance of the unregulated transformer output. That's why they are sized hundreds of times bigger than necessary to reduce the ripple to insignifcant. The cap bank smooths and reduces very short-time rail voltage drops that would otherwise occur when musical peaks are reproduced with short high-current demands. For continuous tones, the caps make no long-term improvement, but for music they allow the amp to handle occasional huge current demands without making the supply rails drop much (which would compress the dynamic range at the least). That's why Pass recommends such huge values for those caps, much larger than necessary for the ripple. They're a LARGE power reserve. Think of them more as batteries than fiters, though of course they are both. If you think of them as filters, they are sized to the dynamic range and other characteristics of the peak and average levels in the material and to the power transformer's performance as its load changes. That's what I'm used to, a big bank of extra energy reserve. That's what got me thinking wrong in the first place. On a linear power supply, the caps do keep the rail voltages up for musical peaks. On a switching supply, the rail volages are always regulated to the same value, so the caps don't do anything unless the current demand is so huge the SMPS is outside its intended operating range, then the voltage drops quickly or the power supply protection kicks in.
2) Yes. But that wasn't the question. But I will keep it in mind.
3) Low ripple is not what I was asking about, but worth noting.
4) No, I wasn't asking about the ripple. My interest was in storing extra energy to handle short high-current demands. But I'm just going to re-wire my load for different impednaces and purchase more complete amplifiers. But I love a good white paper!
Is a huge sag in the rail voltage "ripple"? I wouldn't pick that word.
Again, in the past I could significanly improve the sound of most amps just by adding more filter caps. But a lot of that sonic improvement was not the additional short-term current capacity, it was the fact it met that demand without "significant" voltage drop in the supply rails. It's cheaper and lighter than a bigger transformer, and there's something to be said for optimizing your amp for the common characteristics of music.
Anyway, thanks all for the responses.
"The cap bank smooths and reduces very short-time rail voltage drops that would otherwise occur when musical peaks are reproduced with short high-current demands."
An "unfiltered" Linear power supply (AC rectified from a transformer) will have a ripple that is SQRT(2) [~1.414] multiple of the transformer output.
Near the AC transformer zero-crossing, all of the current must be drawn from the filter caps, since the transform isn't suppling any noticable impact. Load current will continue to be supplied by the output filter caps until the transformer voltage is near or exceeds the cap voltage. As the Linear transformer output voltage climes, it provides current to the load and current to recharge the filter caps.
"No, I wasn't asking about the ripple. My interest was in storing extra energy to handle short high-current demands. But I'm just going to re-wire my load for different impednaces and purchase more complete amplifiers. But I love a good white paper!"
Thats part of what causes voltage ripple on the power supply output. When a load current demand suddenly increases. The filter caps make up for the loss when the transformer can't supply enough (or the transformer output voltage is below the cap voltage). The caps will supply current causing the cap voltage to drop. Fast load transients cause increased output ripple. The amount of ripple is dependant:
1. Switching speed. When ever the PS isn't providing sufficient voltage or current, it will be a source of voltage drop at the output. Filter caps (and the use of output inductors) will this source of Ripple noise.
2. Dynamic load current, where the PS can't supply sufficient current to prevent a voltage drop. Even monster size filter caps can't prevent this since the caps will have ESR and ESL that slows or limits the amount of current required by the load.
Typically in a SMPS switching at tens of kilohertz the delay in power supply voltage drop is much smaller. In a full bridge SMPS, the delay would likely be measured in nanoseconds (switching dead time + diode recovery time) . in a Linear PS, the delay is measured in milliseconds. Since the delay is much much shorter the filter caps need a much smaller capacity. Even an unrelegulated SMPS would need much smaller caps than a linear PS because of the faster cycling.
For loads with high dynamic current demands. it best to place decoupling caps very near to the load. Consider that PCB traces and hookup wire has resistance and inductance that will limit the power supply and the output filter caps to deliver power. Typically in a digital system that operates in multiple Mhz, decoupling caps are places next to the IC's power supply pins to ensure that there is sufficient power at the point of load. You can have the perfect ideal output cap bank, but it will not address noise if the traces are at a long distance from the caps. Traces will also leak noise into adjacent traces. decoupling caps at the point of load also attenuate some of this noise.
An "unfiltered" Linear power supply (AC rectified from a transformer) will have a ripple that is SQRT(2) [~1.414] multiple of the transformer output.
Near the AC transformer zero-crossing, all of the current must be drawn from the filter caps, since the transform isn't suppling any noticable impact. Load current will continue to be supplied by the output filter caps until the transformer voltage is near or exceeds the cap voltage. As the Linear transformer output voltage climes, it provides current to the load and current to recharge the filter caps.
"No, I wasn't asking about the ripple. My interest was in storing extra energy to handle short high-current demands. But I'm just going to re-wire my load for different impednaces and purchase more complete amplifiers. But I love a good white paper!"
Thats part of what causes voltage ripple on the power supply output. When a load current demand suddenly increases. The filter caps make up for the loss when the transformer can't supply enough (or the transformer output voltage is below the cap voltage). The caps will supply current causing the cap voltage to drop. Fast load transients cause increased output ripple. The amount of ripple is dependant:
1. Switching speed. When ever the PS isn't providing sufficient voltage or current, it will be a source of voltage drop at the output. Filter caps (and the use of output inductors) will this source of Ripple noise.
2. Dynamic load current, where the PS can't supply sufficient current to prevent a voltage drop. Even monster size filter caps can't prevent this since the caps will have ESR and ESL that slows or limits the amount of current required by the load.
Typically in a SMPS switching at tens of kilohertz the delay in power supply voltage drop is much smaller. In a full bridge SMPS, the delay would likely be measured in nanoseconds (switching dead time + diode recovery time) . in a Linear PS, the delay is measured in milliseconds. Since the delay is much much shorter the filter caps need a much smaller capacity. Even an unrelegulated SMPS would need much smaller caps than a linear PS because of the faster cycling.
For loads with high dynamic current demands. it best to place decoupling caps very near to the load. Consider that PCB traces and hookup wire has resistance and inductance that will limit the power supply and the output filter caps to deliver power. Typically in a digital system that operates in multiple Mhz, decoupling caps are places next to the IC's power supply pins to ensure that there is sufficient power at the point of load. You can have the perfect ideal output cap bank, but it will not address noise if the traces are at a long distance from the caps. Traces will also leak noise into adjacent traces. decoupling caps at the point of load also attenuate some of this noise.
outside the box
OK why not instead have a rechargeable battery in parallel with the output caps to make up shortfall current from the transformer? Should ease the sag problem since the "smaller" output caps will recharge quicker.
Would need battery internal resistance of the same order as cap ESR's.
OK lets get right outside the box. You say you want reserve transient hf current capacity. The added high value caps will supply this you say like a battery.
OK why not instead have a rechargeable battery in parallel with the output caps to make up shortfall current from the transformer? Should ease the sag problem since the "smaller" output caps will recharge quicker.
Would need battery internal resistance of the same order as cap ESR's.
...didn't say transient HF; transient yes (maybe 20 milliseconds), anywhere within audio-band.
Yes; but the ideal batteries you would need for a few milliseconds would be...caps. Mos batteries dump and charge even slower.
The problem with TechGuy's response is that's what they teach you in school for making a power supply with "great specs" that guarantees good performance right up to some limit. Here the objective may be a bit different; to improve the performance specifically for the pulse demands of music at relatively low cost. He's addressing a non-musical requirement. His #2 is ignoring the fact that the less-than-ideal solution of oversize caps successfully provides a useful function for non-switching audio supplies. The very fact they're coming into use means the rails are already sagging, and indeed they neither charge nor dump instantaneously. But they might sound better when pushing the supply to (past?) its limits. And there's the real issue...what you consider "to" its limits versus considering that same use "past" its limits and irrelevant misuse. I love that my SMPS followed by a conventional FET amp has a fully regulated supply and doesn't compress the dynamic range. On the other hand, the same FET amp with a more conventional supply and oversize bank of caps acceptably handles much larger short musical peaks.
Yes; but the ideal batteries you would need for a few milliseconds would be...caps. Mos batteries dump and charge even slower.
The problem with TechGuy's response is that's what they teach you in school for making a power supply with "great specs" that guarantees good performance right up to some limit. Here the objective may be a bit different; to improve the performance specifically for the pulse demands of music at relatively low cost. He's addressing a non-musical requirement. His #2 is ignoring the fact that the less-than-ideal solution of oversize caps successfully provides a useful function for non-switching audio supplies. The very fact they're coming into use means the rails are already sagging, and indeed they neither charge nor dump instantaneously. But they might sound better when pushing the supply to (past?) its limits. And there's the real issue...what you consider "to" its limits versus considering that same use "past" its limits and irrelevant misuse. I love that my SMPS followed by a conventional FET amp has a fully regulated supply and doesn't compress the dynamic range. On the other hand, the same FET amp with a more conventional supply and oversize bank of caps acceptably handles much larger short musical peaks.
- Status
- Not open for further replies.