wonder if very high capacitance with very big caps, and thus longer charge time, could lead to 'bass sounding slow or blurred' (despite that sims may show less ripple and lower noise)
Thanks for your posts.
I confess that without some good measurement equipments it is hard to do things and a lot of times we are shooting in the dark. For an example, I wish I had a better scope as the ones I have are a good analogue scope of 10MHz and a cheap digital of 100MHz.
I am not 100% sure if mains EMI filters remove the HF noise effectively, as they have fairly minimal amount of damping and may reduce noise at higher frequencies but create more noise lower down.
If there were generally no mains noise then why people reported lower bandwidth EI core transformers sound better than the higher bandwidth toroidals?
In terms of diode switching noise, I am pretty sure it exists and can be pretty bad. For an example, my toroidal power transformers have an additional 15V winding to be used to power up the softstart circuit. In my softstart circuit I used schottky diodes and capacitors across the diodes, which seem to be overkill. But one day I found that by using a separate small transformer for the softstart the sound was cleaner! Obviously, the diode switching pulses in the softstart, even with low current, are still capactively coupled via the power transformer passing through to the rail and to the output!
This time, I am planning to do it this way, in additon to using fast / soft recovery diodes and 22nF across each diode, 100nF + 1R to shunt the secondary to ground before the bridge (to further lower down the resonance frequency) and a snubber of around 18R (to be calculated and possibly measured) + 470nF to damp them all, hopefully. and after the bridge, power and ground plane with multiple low ESL caps, followed by LRCLRC.
Subjectively, the previous L=2uH in the CLRCLRC has been proven to be much better than CRCRC alone, and this time L=60uH is still much better than CRCRC alone. I am just not sure if this tiem the 60uH is better than the 2uH due to possible saturation of the ferrite cores.
With regards to high feedback PSRR of the amp, yes in lower frequencies it is fantastic. It is not good enough to be relied on once it gets to higher frequencies.
I confess that without some good measurement equipments it is hard to do things and a lot of times we are shooting in the dark. For an example, I wish I had a better scope as the ones I have are a good analogue scope of 10MHz and a cheap digital of 100MHz.
I am not 100% sure if mains EMI filters remove the HF noise effectively, as they have fairly minimal amount of damping and may reduce noise at higher frequencies but create more noise lower down.
If there were generally no mains noise then why people reported lower bandwidth EI core transformers sound better than the higher bandwidth toroidals?
In terms of diode switching noise, I am pretty sure it exists and can be pretty bad. For an example, my toroidal power transformers have an additional 15V winding to be used to power up the softstart circuit. In my softstart circuit I used schottky diodes and capacitors across the diodes, which seem to be overkill. But one day I found that by using a separate small transformer for the softstart the sound was cleaner! Obviously, the diode switching pulses in the softstart, even with low current, are still capactively coupled via the power transformer passing through to the rail and to the output!
This time, I am planning to do it this way, in additon to using fast / soft recovery diodes and 22nF across each diode, 100nF + 1R to shunt the secondary to ground before the bridge (to further lower down the resonance frequency) and a snubber of around 18R (to be calculated and possibly measured) + 470nF to damp them all, hopefully. and after the bridge, power and ground plane with multiple low ESL caps, followed by LRCLRC.
Subjectively, the previous L=2uH in the CLRCLRC has been proven to be much better than CRCRC alone, and this time L=60uH is still much better than CRCRC alone. I am just not sure if this tiem the 60uH is better than the 2uH due to possible saturation of the ferrite cores.
With regards to high feedback PSRR of the amp, yes in lower frequencies it is fantastic. It is not good enough to be relied on once it gets to higher frequencies.
Why are we talking about inductors in a thread about capacitor size?
By what mechanism can a filter create noise?HiFiNutNut said:
I'm not sure whether to call this 'belt and braces' or tilting at windmills or straining at gnats. Have you definitely established that for some reason either you are unusually sensitive to (alleged) HF noise, or that your mains supply is unusually noisy? All these extra components are as likely to create problems as solve them. Where is the careful analysis or measurement?
Tinitus, the opposite will happen, but only possibly be audible with a low feedback circuit. Many guitar amps use very small PS filter caps in order to get "sag" as an effect. Over a period of a second or so, there will be an audible increase in distortion as the B+ sags down a bit, if the amp is being pushed hard.
HiFinut, you are trying to be a hardcore perfectionist with meager test equipment and limited knowledge. Much of what's said in forums is wrong. Many audiophiles believe all kinds of things that aren't true. Look at other similar amps to determine a default cap size if the math is confusing. A typical stereo 100watt SS poweramp will have about 10,000uF PS filter caps. If the supply is bipolar, you need two of them. Bypass them with 0.1uF caps. Diode switching noise is real and is largely fixed by using 0.01uF caps across each diode (use fast switching diodes too). When you include a choke, it might be better, but also is now a 2nd order filter that could cause a resonance at some frequency unless you know what you are doing. Using a second RC stage of filtering may be a better way, if you think you need that. I've never needed an EMI filter on the AC. I put 3kV 0.01uF ceramic caps from each side of the AC line to ground right where it comes into the chassis, and call it good. Using a separate power supply for the softstart circuit will not affect the sound at all, unless you believe it will. Make sure you understand how to do star grounding, but tie the transformer ground to the large cap grounds together seperately, then run a wire from there to the star center, which will also be the one place the circuit grounds tie to the chassis. Also, tie the AC green wire ground to the chassis separately, right where it enters the chassis, not to the star center.
HiFinut, you are trying to be a hardcore perfectionist with meager test equipment and limited knowledge. Much of what's said in forums is wrong. Many audiophiles believe all kinds of things that aren't true. Look at other similar amps to determine a default cap size if the math is confusing. A typical stereo 100watt SS poweramp will have about 10,000uF PS filter caps. If the supply is bipolar, you need two of them. Bypass them with 0.1uF caps. Diode switching noise is real and is largely fixed by using 0.01uF caps across each diode (use fast switching diodes too). When you include a choke, it might be better, but also is now a 2nd order filter that could cause a resonance at some frequency unless you know what you are doing. Using a second RC stage of filtering may be a better way, if you think you need that. I've never needed an EMI filter on the AC. I put 3kV 0.01uF ceramic caps from each side of the AC line to ground right where it comes into the chassis, and call it good. Using a separate power supply for the softstart circuit will not affect the sound at all, unless you believe it will. Make sure you understand how to do star grounding, but tie the transformer ground to the large cap grounds together seperately, then run a wire from there to the star center, which will also be the one place the circuit grounds tie to the chassis. Also, tie the AC green wire ground to the chassis separately, right where it enters the chassis, not to the star center.
Last edited:
This thread has been really great with excellent discussions but also appeared to have reached the end when Cootee completed his Excel Spreadsheet.
While most points make sense, when it was close to the conclusion it almost implies that the more capacitance the better the PSU will be, which may well be true but I think it is a diminishing return.
My view is that once a certain amount of capacitance is reached, which can be accurately calculated via the formula given by Cootee, and which is not necessarily a lot, there is very little to be gained by adding more capacitance. At that point, adding more capacitance will have very minimal effect on amplifier power / rail sag / clipping, etc.
However, adding more capacitance beyond that point will serve the purpose of reducing rail ripples. So I guess that if the thread is to continue its discussions it would need to link the capacitor size with the PSRR of the amplifier, not just with amplifier power.
The PSU needs to perform the duty of providing power / current to the amp, and this has been discussed extensively. The PSU also needs to provide a different function, i.e. PSRR, and this has not been discussed extensively. Both has to do with the capacitor size.
Throwing in more capacitance will definitely improve PSRR, but this is not a very effective way to achieve that goal.
I brought in a few topics: (1) Rail fuse can have a detrimental effect on the PSU, reguadless of the capacitor size; (2) An additional heavy duty, passive filter for the front end of the amplifier (or a regulator) is a far more effective way to improve PSRR comparing to increasing the reservoir capacitor size, i.e. adequate C for the OPS stage only but heavy duty C for the IPS and VAS; and (3) Using some chokes in a CLRCLRC PSU instead of a C or CRC PSU will be much better in filtering high frequency noise coming from the mains and from diode switching.
I believe all of those have got to do with "Power Supply Reservoir Size".
Anyway, I feel that I am unwelcome here, so I have now decided to leave. Please accept my appology if I have been a pain.
Obviously, I have been trying to learn from you guys here and I confess that I have learnt a lot from your replies to my posts. Thank you for helping someone like me at the beginner level. I would like to especially thank abraxalito, twest820 and andrewT whole heartedly for helping me to understand many things I did not understand. You guys are great and simply amazing!
Regards,
Bill
Last edited:
Thanks for the questions. Mains filters use high Q capacitors / inductors with minimal damping which can react to the mains line as well as the transformer parasitics that may well create some resonance peaks below the filtered frequencies. The noise above MHz may well be filtered out but impedance peaks may be created in the hundred kHz region if any LCR tuned circuit is formed with the parasitics.
Most power line filters (and the power lines themselves) have these resonances. They are part and parcel of what you are doing. Effective filtering of power line noise is complex, especially since the line is an unpredictable impedance that is constantly changing.
Do not homebrew a powerline filter unless you really know what you are doing. Its not the steady state condition but the power line "event" that will burn your house down. There can be an enormous amount of energy at that connection and parts that were not designed for it can easily do really bad things. We surge test filters with a 6KV 3KA surge to confirm that nothing bad will happen. Most audiophile parts are charred ruins after that. I did destroy an array of 10 high power Caddock resistors with this test. An array of 5W carbon comp resistors survived the surge just fine. The caps designed for powerline applications ("X" caps etc.) are internally fused by their design.
Do not homebrew a powerline filter unless you really know what you are doing. Its not the steady state condition but the power line "event" that will burn your house down. There can be an enormous amount of energy at that connection and parts that were not designed for it can easily do really bad things. We surge test filters with a 6KV 3KA surge to confirm that nothing bad will happen. Most audiophile parts are charred ruins after that. I did destroy an array of 10 high power Caddock resistors with this test. An array of 5W carbon comp resistors survived the surge just fine. The caps designed for powerline applications ("X" caps etc.) are internally fused by their design.
This thread has been really great with excellent discussions but also appeared to have reached the end when Cootee completed his Excel Spreadsheet.
While most points make sense, when it was close to the conclusion it almost implies that the more capacitance the better the PSU will be, which may well be true but I think it is a diminishing return.
My view is that once a certain amount of capacitance is reached, which can be accurately calculated via the formula given by Cootee, and which is not necessarily a lot, there is very little to be gained by adding more capacitance. At that point, adding more capacitance will have very minimal effect on amplifier power / rail sag / clipping, etc.
However, adding more capacitance beyond that point will serve the purpose of reducing rail ripples. So I guess that if the thread is to continue its discussions it would need to link the capacitor size with the PSRR of the amplifier, not just with amplifier power.
The PSU needs to perform the duty of providing power / current to the amp, and this has been discussed extensively. The PSU also needs to provide a different function, i.e. PSRR, and this has not been discussed extensively. Both has to do with the capacitor size.
Throwing in more capacitance will definitely improve PSRR, but this is not a very effective way to achieve that goal.
I brought in a few topics: (1) Rail fuse can have a detrimental effect on the PSU, reguadless of the capacitor size; (2) An additional heavy duty, passive filter for the front end of the amplifier (or a regulator) is a far more effective way to improve PSRR comparing to increasing the reservoir capacitor size, i.e. adequate C for the OPS stage only but heavy duty C for the IPS and VAS; and (3) Using some chokes in a CLRCLRC PSU instead of a C or CRC PSU will be much better in filtering high frequency noise coming from the mains and from diode switching.
I believe all of those have got to do with "Power Supply Reservoir Size".
Anyway, I feel that I am unwelcome here, so I have now decided to leave. Please accept my appology if I have been a pain.
Obviously, I have been trying to learn from you guys here and I confess that I have learnt a lot from your replies to my posts. Thank you for helping someone like me at the beginner level. I would like to especially thank abraxalito, twest820 and andrewT whole heartedly for helping me to understand many things I did not understand. You guys are great and simply amazing!
Regards,
Bill
Bill,
As you noted, the thread had basically sputtered out. So you were and are certainly welcome to inject some new life into it. I have been following along, sporadically, and have learned a few things from the discussions you sparked. So, thanks for that.
I know what you meant about PSRR, but to be accurate, PSRR is really a property of the system that is being powered by the power supply, not a property of, or something that can be affected by, the power supply itself. So increasing the reservoir capacitance doesn't actually improve PSRR. It just lessens any ill effects due to the PSRR of the system being powered. Lower ripple from the power supply would mean that the system being powered could get away with having a worse PSRR, without decreasing performance in that respect.
You are certainly right about the "diminishing returns" of increasing reservoir capacitance. There is a definite trade-space to be navigated, there.
I probably meant to explore decoupling capacitance some more, in this thread, but got busy with other things. Decoupling capacitance is almost exactly like reservoir capacitance that is right at the point of load. It supplies the fast transient currents best, because there is minimal conductor inductance and resistance between it and the point of load, since it is extremely close to the point of load. At the same time, it prevents some ripple, by not forcing the transient currents to flow through the (presumed) physically-longer power and ground rail conductors' inductance (and resistance). Without decoupling caps, the changing current would induce a voltage across the conductors themselves, since for an inductance v = L di/dt (note that the voltage's amplitude depends only on the rate-of-change of the current). So even though the inductance of the power rails might be quite small, a fast-changing current could still induce a significant voltage across it. Those induced voltages could also affect other circuits being supplied by the same PSU, so the decoupling capacitors "decouple" their part of the circuit from the rest of the system, to keep those types of effects local, so they can't mess up other circuits. Hence the name, "decoupling capacitors". At the same time, because of the rail-conductors' inductances, without decoupling capacitors the current might not be available exactly when demanded by the load, which would result in some distortion in the reproduction of transients. Anyway, the needed speed and size characteristics of the transient currents enable calculating the required decoupling capacitances, and the maximum distance they can be, from the point of load.
I'm with you on rail fuses, too. Terry Given made a good point about those, earlier in this thread.
But to deal with diode switching transients, you really should use an actual snubber network, to minimize the effect at its source. A snubber is primarily a resistor, which is matched to the characteristic impedance created by an LC circuit, in order to optimally damp the LC's resonance, absorbing the energy that might have caused peaking or ringing. In many cases, a capacitor is added in series with the snubber resistor, so that only the unwanted frequencies can get to the resistor, which lowers the power dissipated in the resistor to a reasonable range, if necessary. But the capacitor is not necessary, for the snubber resistor to damp the resonance. I know that a lot of people use only a capacitor across each diode. But that is not a good approach. It simply shifts the resonance downward in frequency. Anyway, in a linear power supply's rectifier, the capacitance involved in the LC resonance is usually mainly the capacitance formed in each diode, just as it turns off. And the inductance is associated with the transformer, plus a little from the rest of the circuit. Rather than needing to know the values of that capacitance and inductance, here is a post with a practical method for choosing the optimal snubber resistance, and the value of the series capacitance, for a snubber for an existing circuit: http://www.diyaudio.com/forums/powe...lm-caps-electrolytic-caps-30.html#post2828689 . Exactly the same method can be used to find the optimal value of the termination resistance for a transmission line, such as a digital buss. In that case, often no series capacitor is needed.
Cheers,
Tom Gootee
Last edited:
Sound advice! I would not dare to build mains filter myself. I bought half a dozen of them, which came in cased.
It is interesting to read "I did destroy an array of 10 high power Caddock resistors with this test. An array of 5W carbon comp resistors survived the surge just fine."
Bill,
Just some more random responses and thoughts:
If you use planes with cap arrays, you would want those to be the LAST part of the chain, I would think. Otherwise, you are throwing away a nice source of very low impedance that could be presented directly to the point of load, which, with careful physical layout (in the connections from the planes to the points of load), could serve as both reservoir and decoupling caps, and be much better than any typical single-conductor power and ground rails.
About diode switching noise, see my previous post. But also, perhaps when you changed to the small transformer for the soft start, you changed the LC resonance, which changed the diode switching effects. But note that that's just a wild guess.
I am not an expert on using inductance or resistance in a linear PSU filter. But if you decide that you do need an inductance, and it's only in the low uH range, you should consider just winding your own air core coil. That would eliminate the hysteresis distortion from, and the possible saturation of, a core. There is a good air core coil calculator, at Pronine Electronics Design - Multilayer Air Core Inductor Calculator .
By the way, your oscilloscopes should be sufficient for most uses, in this realm. But for high frequencies, i.e. more than a MHz or two, and definitely over 10-20 MHz, you do have to start worrying about the length of your probe's ground connection, and exactly where it gets connected. Otherwise, you might tend to see a lot of stuff that really isn't there, or is different than it appears. In many cases, too, the impedance of the probe and its ground connection can affect the circuit's behavior, just by being connected. Good probes, like Tektronix, should have available accessories that enable using extremely short ground connections. For really sensitive circuits, you might need an FET probe, or an amplifier, ahead of the probe, that has extremely high input impedance, so it won't significantly affect the circuit's characteristics just by being connected to it.
Cheers,
Tom
Last edited:
The surge has a peak power of 18 MW which can do a lot of damage. This is the peak energy that will be found inside a building from a lightning strike. Outside the building its much greater but the wiring reduces the energy (as every outlet in the building arcs over). It required for UL approval to be able to withstand that surge.
Few audiophile power strips are built for this either. And without testing you don't really know what will happen.
One little known issue with commercial line filters is the shunt to ground caps. Thise will dump as much as 1 mA of line current into the ground. If you build them into a box of yours that box will potentially cause no end of mysterious hums in a system if the ground (housing) is tied to the chassis and the internal audio ground.. If you float the ground the chassis will be at 1/2 line voltage. The medical rated filters are much better and don't have the line to ground and neutral to ground caps. I would not use the filters and move the function upstream to a power conditioner if possible.
Few audiophile power strips are built for this either. And without testing you don't really know what will happen.
One little known issue with commercial line filters is the shunt to ground caps. Thise will dump as much as 1 mA of line current into the ground. If you build them into a box of yours that box will potentially cause no end of mysterious hums in a system if the ground (housing) is tied to the chassis and the internal audio ground.. If you float the ground the chassis will be at 1/2 line voltage. The medical rated filters are much better and don't have the line to ground and neutral to ground caps. I would not use the filters and move the function upstream to a power conditioner if possible.
Bill, You are not a pain, and sorry if I came off sounding like you were. I'm glad to help you with any question. Check out my website if you're interested. I've uploaded many webpages on various audio projects that I've done in the last decade. Some of them go into great detail on the engineering aspects, both electrical and on acoustics. http://www.spiritone.com/~rob_369 http://www.spiritone.com/~rob_369This thread has been really great with excellent discussions but also appeared to have reached the end when Cootee completed his Excel Spreadsheet.
Anyway, I feel that I am unwelcome here, so I have now decided to leave. Please accept my appology if I have been a pain.
Obviously, I have been trying to learn from you guys here and I confess that I have learnt a lot from your replies to my posts. Thank you for helping someone like me at the beginner level. I would like to especially thank abraxalito, twest820 and andrewT whole heartedly for helping me to understand many things I did not understand. You guys are great and simply amazing!
Regards,
Bill
Last edited:
Gootee,
Thanks for your posts.
(1) For decoupling, I think the whole model can be simplified to the impedance model at the output device pin. In LTSpice if we can do AC analysis at the point of the output device pin, it will reveal everything we need to know about decoupling (add in parasitics in the model). How much to decouple and how to decouple depends on the PSRR of the output stage as well. The solution is in the printed circuit board layout. I posted a picture of the OPS circuit board layout in relation to this a couple of pages back, and it is now developed into this: http://www.diyaudio.com/forums/solid-state/237130-grounding-minimizing-current-loops.html
(2) To deal with diode switching transients, I mentioned in my last page "in additon to using fast / soft recovery diodes and 22nF across each diode, 100nF + 1R to shunt the secondary to ground before the bridge (to further lower down the resonance frequency) and a snubber of around 18R (to be calculated and possibly measured) + 470nF to damp them all, hopefully."
(3) cap arrays - I have found that buying a large number of small caps is not more expensive than buying big caps for the same capacitance value, therefore I will almost go all way using small caps on power / ground planes.
(4) Small inductor - thanks for the link. It is the way to go. I have illustrated that for a very low cost low value inductor it beats the best caps by many folds above 100kHz as far as filtering is concerned. Imagine to achieve the result of the 2uH as in my previous simulation we would need perhaps over 100,000uF and more.
Regards,
Bill
Thanks for your posts.
(1) For decoupling, I think the whole model can be simplified to the impedance model at the output device pin. In LTSpice if we can do AC analysis at the point of the output device pin, it will reveal everything we need to know about decoupling (add in parasitics in the model). How much to decouple and how to decouple depends on the PSRR of the output stage as well. The solution is in the printed circuit board layout. I posted a picture of the OPS circuit board layout in relation to this a couple of pages back, and it is now developed into this: http://www.diyaudio.com/forums/solid-state/237130-grounding-minimizing-current-loops.html
(2) To deal with diode switching transients, I mentioned in my last page "in additon to using fast / soft recovery diodes and 22nF across each diode, 100nF + 1R to shunt the secondary to ground before the bridge (to further lower down the resonance frequency) and a snubber of around 18R (to be calculated and possibly measured) + 470nF to damp them all, hopefully."
(3) cap arrays - I have found that buying a large number of small caps is not more expensive than buying big caps for the same capacitance value, therefore I will almost go all way using small caps on power / ground planes.
(4) Small inductor - thanks for the link. It is the way to go. I have illustrated that for a very low cost low value inductor it beats the best caps by many folds above 100kHz as far as filtering is concerned. Imagine to achieve the result of the 2uH as in my previous simulation we would need perhaps over 100,000uF and more.
Regards,
Bill
Good thinking. So you mean Y caps can cause trouble! The mains filters I have got all Ycap values to be very small comparing to other mains filters I have seen. But I still think it is a better idea to put the filtering outside the amplifier. What power conditioner do you recommend? I guess you are not referring to those that reconstruct the sine waves or use isolation transformers?
No you did not sound like that.
I like your open baffle speakers. I am an open baffle speaker nut. But that is out of the topic.
Both X and Y caps are rated for specific functions on direct AC connections. Look them up to be sure. I was using X cap has a generic descriptor.
It all depends. The ones I have been involved with do not have the audiophile stamp of approval but they do really work. They would be the Monster AVS2000 and HTS7000 (which has 2 isolation transformers). The lower products in the line have less stuff in them. The active regenerators are the right solution for the wrong problem I think. There is lots of buzz about current limiting but if you actually measure the voltage drop on different products its not significant while playing music. Good filtering can be very disappointing if all of the aerie space is removed (hash riding along with the music). Be careful what you wish for.
For any supply there's an RC filter formed by the source impedance to the mains and the bypass cap. If this filter's time constant is longer than the mains frequency then the answer to your question is yes, AM modulation at the bass frequency can develop on the supply rails---it's especially obvious when the bass frequency is below the mains. For example, fuses which blow around 100mA can have around 10 ohms cold resistance. Put, say, a 10,000uF cap behind it and the fuse+cap RC filter corners around 1Hz. Most builds probably use fuses closer to 10 milliohms but with the trafo windings and such it wouldn't surprise me if some of the more extreme supply builds with >100,000uF have problems with this.
I get the feeling that some are forgetting that the speaker and amplifier get their current from the last stage of capacitance.
The charging circuit and extra filtering before that last stage of capacitance does not respond to speaker current demand. These respond to a drop in voltage of the last capacitance stage.
The more impedance that is inserted between the mains and the last capacitance stage the less the speaker will draw from the earlier supply capacitances.
Make sure the LAST capacitance can supply the speaker demand.
Only then can you start thinking about adding some extra filtering to attenuate the 50/60Hz and kHz and MHz and GHz ripples.
Yes, our increased use of wireless in the home gives us a lot more in the GHz region than we will ever get from distant broadcast transmitters. Ott shows us that a cellphone 1metre away from our gear, is much worse than a 50kW transmitter just 1mile away.
The charging circuit and extra filtering before that last stage of capacitance does not respond to speaker current demand. These respond to a drop in voltage of the last capacitance stage.
The more impedance that is inserted between the mains and the last capacitance stage the less the speaker will draw from the earlier supply capacitances.
Make sure the LAST capacitance can supply the speaker demand.
Only then can you start thinking about adding some extra filtering to attenuate the 50/60Hz and kHz and MHz and GHz ripples.
Yes, our increased use of wireless in the home gives us a lot more in the GHz region than we will ever get from distant broadcast transmitters. Ott shows us that a cellphone 1metre away from our gear, is much worse than a 50kW transmitter just 1mile away.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.