The actual needs from a power supply are not easily put into a single catagory or two.
And it is usually 'relative' rather than absolute, so you can not just compute the ideal solution. Usually, more is better, unless you go to extreme excess.
Less will work, but not as well. It should be obvious, but many people think that they can get an equation for optimizing an analog power supply, but there isn't any.
And it is usually 'relative' rather than absolute, so you can not just compute the ideal solution. Usually, more is better, unless you go to extreme excess.
Less will work, but not as well. It should be obvious, but many people think that they can get an equation for optimizing an analog power supply, but there isn't any.
Technical specifications and Graphs are a good point to start. but manufacturers like to hide data especially the unfavorable one. most of time they use tricky language such as "subject to manufacturing tolerances" or show graphs taken with custom equipment rather than using standard euipment such as Audio Precision for instance. Not to mention that they never tell the exact way they positioned microphone, room size etc. The article states that tones are predictable and music is very complex. but He forgot to mention that tones are by far the easiest way to create a repeatable test to use on any kind of amp/speaker for comparison purposes.
The educated ear can spot subtle audio artifacts, colorations and so on. but most people don't have the training or musical education. that's why they like to rely on specs and graphs.
Listening to music is a subjetive experience and "subjetiveness" is just imposible to measure.
As I say before. Technical specifications and Graphs are a good point to start, but it is your taste, ear training and curiosity. what will ultimately rule in the long run.
Kind Regards
Alvaro
The educated ear can spot subtle audio artifacts, colorations and so on. but most people don't have the training or musical education. that's why they like to rely on specs and graphs.
Listening to music is a subjetive experience and "subjetiveness" is just imposible to measure.
As I say before. Technical specifications and Graphs are a good point to start, but it is your taste, ear training and curiosity. what will ultimately rule in the long run.
Kind Regards
Alvaro
Search back about a month or so right here, and you'll find a message in which I outlined the relationship of Joules to Watts, as described by Motorola in one of their app notes.
It's no hard science, just a relationship they established as a general rule of thumb. But, upon working it out, you'll find it does coincide quite a bit with what is being said here.
It never left me hanging out to dry.
Aw, heck, ...
Briefly, the say we will need 1-2 Joules of energy per every 10 Watts of dissipated power, depending on how easy or hard a load is.
The basic equation is:
Joules = V sq. * Farads,
Where "V squared" is the say plus power line of a symmetrical supply and assumes the minus side is equal to it, and V is the voltage of the line.
So, let's say your supply voltages are +/- 50V, and assuming you use say 22,000 uF capacitors, per line, this would be:
50 squared * 0.022 = 55 Joules.
Assuming worst case conditions, i.e. 2 Joules/10W, this would be good for some (55:2) 275 Watts of peak power.
This obviously means that before getting down to it, you need to define two things:
1. How optimistic (1 Joule/10W) or pessimistic (2 Joules/10W) you are, and
2. Exactly how much power do you want to have into what kind of load.
You know your power supply line voltages, so that's that.
So, if you want a say nominally 100/200/400 WPC amp, obviously, you look at the 400W number. In my case, since my power supply lines will be at +/-56V, 32,000 uF (22,000 // 10,000 uF) should give me around 100 Joules of energy storage, enough to cover even the 400W/2 Ohms in worst case scenario. That's per line, per channel, or a grand total of (32,000*4) 128,000 uF for the whole stereo amp.
ALL of the above calculations are related to one channel only. If you are using a single point supply for both channels, you'd do well to double the figures and multiply them by 1.2 at least, to take into account interactions between channels.
Briefly, the say we will need 1-2 Joules of energy per every 10 Watts of dissipated power, depending on how easy or hard a load is.
The basic equation is:
Joules = V sq. * Farads,
Where "V squared" is the say plus power line of a symmetrical supply and assumes the minus side is equal to it, and V is the voltage of the line.
So, let's say your supply voltages are +/- 50V, and assuming you use say 22,000 uF capacitors, per line, this would be:
50 squared * 0.022 = 55 Joules.
Assuming worst case conditions, i.e. 2 Joules/10W, this would be good for some (55:2) 275 Watts of peak power.
This obviously means that before getting down to it, you need to define two things:
1. How optimistic (1 Joule/10W) or pessimistic (2 Joules/10W) you are, and
2. Exactly how much power do you want to have into what kind of load.
You know your power supply line voltages, so that's that.
So, if you want a say nominally 100/200/400 WPC amp, obviously, you look at the 400W number. In my case, since my power supply lines will be at +/-56V, 32,000 uF (22,000 // 10,000 uF) should give me around 100 Joules of energy storage, enough to cover even the 400W/2 Ohms in worst case scenario. That's per line, per channel, or a grand total of (32,000*4) 128,000 uF for the whole stereo amp.
ALL of the above calculations are related to one channel only. If you are using a single point supply for both channels, you'd do well to double the figures and multiply them by 1.2 at least, to take into account interactions between channels.
I might be missing the point here . If a class B( AB ) amp we can tolerate 5 V of ripple at full power especailly if we under-drive the amp by lets say 8 V ( 55 V raw DC 47 V to driver stage ) . Then as long as the ripple rating of the caps is not exceeded ( factor of 1.5 in that seems wise ) we don't need all these Joules .
My instinct is to agree with Douglas Self and say if bass sounds more powerful with more caps than the ripple ratings demand then something else is wrong . Is it not " showing the engine off " at hi fi shows that dictates this ? Personally I think this is all wrong . If you fit higher voltage caps you might do yourself a favour . Lets say use 100 V at 63V . Nearly always they have better performance ( tan theta ) , if not their durability would be compromised . It is a bi-product . They are big and chunky when showing off the engine . I did this by chance with a Naim Nap 250 . We could get some very high grade 6800 uF 100 V BBH at a special price ( £18 ) . It was a stop gap repair . They have been retained as they sound much better . I am told the caps even reform and give higher values , no idea if it is true ?
If class A I suspect the same rules can be used . That is when life gets interesting . Suddenly and amp which seemed fine in class B has ripple hum . If so it always had hum , it was masked by the music . Hum modulated music is not to be recommended .
Most amps will survive a few seconds of class A , look for the ripple hum . Often it will caused be overly complex earthing ( too many caps ) . I never get this to my satisfaction .
2 x 10 000 uF domestic grade , 2 x 22 000 uF high grade , 2 x 47 000 Mundorf I think would be my recommendation . All have good HF performance and allow easier earthing .
Leave enough space in the box as one day it might be the amplifiers birthday and buy it some Mundorf . Although I do it often I do not feel comfortable about multiple caps , it's messy . An over-sized power transistor decoupler as I said before seems an opportunity . Why not 4700 uF if high grade ? It is often 220 uF . Then it is both decoupler and reservoir . If we do use multiple caps they should be clustered around the output transistors in good symmetry . This is how very cheap PA amps are made for cheapness . They are dreadful . However the PSU might not be . That temps me to fix one I blew up ( usual op amp and MOS FET's design ) , give it a new power amp section . Maybe put all the caps on two sections of a 4 layer PCB . No Star earth , just two unbroken layers .
I have been entered for a Spud amp competition in Germany . I have declined because it is stupid to design amps that simple . The input transformer would cost a fortune even if Sowter agreed to make it . If 2 x KT 88 ( GU 50 , EL34 ) and 1 x ECC85 I would enter ( that's just an extra ECC85 , I have a box full ) . However the best amps using as many 2N3055 as you like and only 2 x 1000 uF appeals to me . I would go 1000 uF 100 V and run the 2N3055 in a bridged bridge . As I said before a special prize for using LM741 ( replaced by a better op amp at the close of the competition ) . 2N4401/3 BC550 /560 BC327/337 BD139/140 MJE 340/350 MPSA42/92 2N5401/5551 would be the transistors allowed . I would place a 20 Kg limit and no switch mode PSU's . Fans allowed as long as they do not exceed a certain noise level . The amp must be able to run at 10 watts without fans . All amps must be able to run at better than 50% efficiency in one mode of operation . Switching to class A allowed . All amps stereo and shared PSU .
My instinct is to agree with Douglas Self and say if bass sounds more powerful with more caps than the ripple ratings demand then something else is wrong . Is it not " showing the engine off " at hi fi shows that dictates this ? Personally I think this is all wrong . If you fit higher voltage caps you might do yourself a favour . Lets say use 100 V at 63V . Nearly always they have better performance ( tan theta ) , if not their durability would be compromised . It is a bi-product . They are big and chunky when showing off the engine . I did this by chance with a Naim Nap 250 . We could get some very high grade 6800 uF 100 V BBH at a special price ( £18 ) . It was a stop gap repair . They have been retained as they sound much better . I am told the caps even reform and give higher values , no idea if it is true ?
If class A I suspect the same rules can be used . That is when life gets interesting . Suddenly and amp which seemed fine in class B has ripple hum . If so it always had hum , it was masked by the music . Hum modulated music is not to be recommended .
Most amps will survive a few seconds of class A , look for the ripple hum . Often it will caused be overly complex earthing ( too many caps ) . I never get this to my satisfaction .
2 x 10 000 uF domestic grade , 2 x 22 000 uF high grade , 2 x 47 000 Mundorf I think would be my recommendation . All have good HF performance and allow easier earthing .
Leave enough space in the box as one day it might be the amplifiers birthday and buy it some Mundorf . Although I do it often I do not feel comfortable about multiple caps , it's messy . An over-sized power transistor decoupler as I said before seems an opportunity . Why not 4700 uF if high grade ? It is often 220 uF . Then it is both decoupler and reservoir . If we do use multiple caps they should be clustered around the output transistors in good symmetry . This is how very cheap PA amps are made for cheapness . They are dreadful . However the PSU might not be . That temps me to fix one I blew up ( usual op amp and MOS FET's design ) , give it a new power amp section . Maybe put all the caps on two sections of a 4 layer PCB . No Star earth , just two unbroken layers .
I have been entered for a Spud amp competition in Germany . I have declined because it is stupid to design amps that simple . The input transformer would cost a fortune even if Sowter agreed to make it . If 2 x KT 88 ( GU 50 , EL34 ) and 1 x ECC85 I would enter ( that's just an extra ECC85 , I have a box full ) . However the best amps using as many 2N3055 as you like and only 2 x 1000 uF appeals to me . I would go 1000 uF 100 V and run the 2N3055 in a bridged bridge . As I said before a special prize for using LM741 ( replaced by a better op amp at the close of the competition ) . 2N4401/3 BC550 /560 BC327/337 BD139/140 MJE 340/350 MPSA42/92 2N5401/5551 would be the transistors allowed . I would place a 20 Kg limit and no switch mode PSU's . Fans allowed as long as they do not exceed a certain noise level . The amp must be able to run at 10 watts without fans . All amps must be able to run at better than 50% efficiency in one mode of operation . Switching to class A allowed . All amps stereo and shared PSU .
Last edited:
Almost. The correct equation is Joules = V sq. * Farads /2dvv said:
Cap size is set by things like required ripple voltage, transformer winding resistance, peak current draw, PSRR etc. The biggest demand on the caps' stored energy would come from a full level 50/60Hz square wave, as if it is aligned to the supply phase then all the energy has to come from the cap. For all other situations things are easier, as either the cap gets recharged partway through (lower frequencies, or not phase aligned) or the current is only drawn for part of the recharge cycle (higher frequencies).
Peak power out = V x I. Suppose we want V to droop by no more than 5%. Then 0.05V = I x 0.01 / C (assuming 50Hz mains and full-wave rectification). Rearrange this to get CV = I/5 . Now substitute I=P/V and get CV = P/5V. Finally 0.5CV^2 = P/10. So for 5% max droop on a phase-aligned 50Hz square wave you need 1 joule of stored energy for every 10 watts of peak output power (or 5 watts of mean sine wave power). I guess a calculation like this is where Motorola got their rule of thumb from.
Having done the calculation, I have no idea how relevant it is to real amplifiers handling real music.
My guess is that the Motorola writeup Mr DVV referred to simplified it for both rails : V sqr * Farads /2 times 2
(jules ?)
Factors of 2 might appear and disappear for various ways of looking at single and double supply rail systems. My calculation included the /2, and was intended to determine the storage in each cap for one rail of a double rail supply.
By the way, before anybody misunderstands, the '1J for 5W' estimate applies to 5% droop under worst case conditions with a 50Hz full-wave rectified supply. You need slightly less for 60Hz, and none at all for pure DC. 5% might or might not be a reasonable amount of droop. The 20% ratio is not a magic number!
By the way, before anybody misunderstands, the '1J for 5W' estimate applies to 5% droop under worst case conditions with a 50Hz full-wave rectified supply. You need slightly less for 60Hz, and none at all for pure DC. 5% might or might not be a reasonable amount of droop. The 20% ratio is not a magic number!
I'm very pleased you pointed to the correct energy calculation.
http://www.diyaudio.com/forums/lounge/200865-sound-quality-vs-measurements-123.html#post3057970
Especially at a public forum, formula should be precise (and correct units
).
http://www.diyaudio.com/forums/lounge/200865-sound-quality-vs-measurements-123.html#post3057970
Especially at a public forum, formula should be precise (and correct units
).I think even PCL 86 isn't allowed ! That would make it worth the effort .
Shame as my hotel room was paid for me .
Allow 3 bottles in stereo I would say . 2 x PCL 86 and 1 x ECC81 . 2 x Gu50 1 x ECC81 ( best I feel ) .
I remember at college the derived units were always single and without capitol letters , e.g. volt never volts . Joules was allowed as it was a fundamental unit . It was 40 year ago and I probably remember incorrectly ?
Last edited:
Report:
Well, we do experiments because we don't know the outcome.
Hypothesis: The break up region of the tweeters is causing my wife's sensitivity. (masked slightly or not by the differences in amplifiers)
Test, Built a decent speaker with two tweeters, one with pronounced breakup, one without. These are the Seas 27TBFC and 27TDFC tweeters.
Nearfield and farfield measurements confirmed virtually identical integration of the tweeters using the same crossover. Only the tweeter changed.
Based on the the Zaph SR-71 with R5 changed to 1 Ohm
Still using the Rotel amp.
Verified the Paradigm 20's were still "bad"
Verified my little side trip to some Fostex full range were really bad. Treble breakup was so bad, she could not judge the problem I was testing for.
Track one of The King James Version.
Metal dome. PASSED
Cloth dome: FAILED Not as bad as the Paradigms, but had the edge that causes her pain.
This is exactly the reverse of what I expected.
What I hear, is the metal is slightly more extended above 18K and there are some details in recordings I thought I knew well I had never heard. Measurements confirm this. Zaph's measurements give the metal an advantage in distortion, but not by much.
Conclusions:
It is not the breakup mode. (surprise)
Better drivers make better speakers. (no surprise)
At the listening levels we use, the very low crossover (1700 Hz) is not a problem with these drivers. (surprise)
Spin-off conclusion: My office Kef Q1's are a lot worse that I thought. They measure well, but in contrast, are lifeless. I knew they were not quite up to my Dayton/Vifa's.
Nice cabinets though. I might just recycle them.
I can think of a lot more tests. Rotel vs. Parasound vs. modified Hafler amps is of course the original question. I need to build the good cabinets to do that. My MDF awaits.
Well, we do experiments because we don't know the outcome.
Hypothesis: The break up region of the tweeters is causing my wife's sensitivity. (masked slightly or not by the differences in amplifiers)
Test, Built a decent speaker with two tweeters, one with pronounced breakup, one without. These are the Seas 27TBFC and 27TDFC tweeters.
Nearfield and farfield measurements confirmed virtually identical integration of the tweeters using the same crossover. Only the tweeter changed.
Based on the the Zaph SR-71 with R5 changed to 1 Ohm
Still using the Rotel amp.
Verified the Paradigm 20's were still "bad"
Verified my little side trip to some Fostex full range were really bad. Treble breakup was so bad, she could not judge the problem I was testing for.
Track one of The King James Version.
Metal dome. PASSED
Cloth dome: FAILED Not as bad as the Paradigms, but had the edge that causes her pain.
This is exactly the reverse of what I expected.
What I hear, is the metal is slightly more extended above 18K and there are some details in recordings I thought I knew well I had never heard. Measurements confirm this. Zaph's measurements give the metal an advantage in distortion, but not by much.
Conclusions:
It is not the breakup mode. (surprise)
Better drivers make better speakers. (no surprise)
At the listening levels we use, the very low crossover (1700 Hz) is not a problem with these drivers. (surprise)
Spin-off conclusion: My office Kef Q1's are a lot worse that I thought. They measure well, but in contrast, are lifeless. I knew they were not quite up to my Dayton/Vifa's.
Nice cabinets though. I might just recycle them.
I can think of a lot more tests. Rotel vs. Parasound vs. modified Hafler amps is of course the original question. I need to build the good cabinets to do that. My MDF awaits.
I have just completed a very simple tube amp . It has the option for parallel tubes ( ECC81 and EL 34 ) . It measures about the same with all ( 5 to 8 watts ) . I will push for more power later .
1 x KT88
2 x EL34
2 x EL37
1 x EL34
However the 2 x EL34 is far better . The tweeter sibilance is virtually non existent with 2 x EL 34 . The amp has a Sowter SA07 transformer which gives good square-waves at 100 Hz and 2.5 kHz . For a valve amp that's good .
Speakers are just cheap units with the Audax cheap tweeter . The sacrificial ones when testing . I have to say with such an amp they sing .
The amp is - 3dB at 50 kHz so not too filtered .
1 x KT88
2 x EL34
2 x EL37
1 x EL34
However the 2 x EL34 is far better . The tweeter sibilance is virtually non existent with 2 x EL 34 . The amp has a Sowter SA07 transformer which gives good square-waves at 100 Hz and 2.5 kHz . For a valve amp that's good .
Speakers are just cheap units with the Audax cheap tweeter . The sacrificial ones when testing . I have to say with such an amp they sing .
The amp is - 3dB at 50 kHz so not too filtered .
- Status
- Not open for further replies.