I would worry about the Mhz region from the POV of shorting out any RF interference that happens to intrude. So, don't get fancy but work to avoid any resonance peaks, that's the key thing.
From the POV of getting the amp to work properly, and very specifically the FB side of things, a rule of thumb I would use is to make sure up to 200kHz is good, the 10th harmonic of 20kHz. The key thing is to make sure that where the PSRR of the amp starts to look sad, the higher frequencies, that the power supply itself is still exhibiting excellent, low, impedance.
Which is why I use regulation here to do the job ...
Yes, I'm not fussed about low frequency "wobbling" of the supply, it shouldn't be a problem, I just don't want it to sag beyond a certain point ...
One simple way of looking at it is considering the worst case, of a maximum amplitude 20kHz sine wave. If you know the PSRR for that frequency, and work out the supply ripple in your worst load for that signal, you then want to keep the ripple below the level where it will inject an error signal of say -100dB into the output. If 20kHz is OK then every other frequency should also be OK, unless your amp has bizarre characteristics ...
Frank
Last edited:
gootee liked the ref in post 1739 . The use of polyprop in small values to lower impedance of the power supply above 220k hz looks good. Grant it that will be adding an other pole and zero to the equation given the addition. If the multi resonance are spread out over the band a wider and net flatter curve can be produced. Turning the power supply in to a multi unit device much like a speaker crossover does with a multi-driver speaker. The curve from a 3577a would I believe show this . Sorry I do not have one at this time to show it off. 3 or 4 stage of progressively smaller caps. to flatten out the curve will work. To get really low a regulator the choose however when current is high it may not be that easy to do and have hold up long term .
Last edited:
If you use such it must be placed precisely at the point in the circuit where the low impedance at high frequencies is required. Say, across the legs of the power supply pins of the opamp -- otherwise, it's useless and may in fact create problems ...
Frank
As I understand it yes. The inductance of the board traces would set up a resonance circuit if the cap is to far away from the ic.
Long time since any chatter here ... 
This is a nice little, tight summary of what can be done about the inductance of power supply traces, here is about as good a place to link to it as any: Power Tip 56: Estimate PWB interconnect inductance.
Frank
This is a nice little, tight summary of what can be done about the inductance of power supply traces, here is about as good a place to link to it as any: Power Tip 56: Estimate PWB interconnect inductance.
Frank
Yes, the units for the left half of that table (and in the accompanying equation) should be mm, not cm, and the inductances/mm in the third column should be divided by 10. OR, the numbers in the first two columns could be divided by 10.
The inductances are lower than I expected. It's looking like decoupling capacitor distances from power pins would not be as much of a concern as I had thought, if a ground plane were used, especially with a board thickness of 0.06 inch (1.52 mm) or less.
Andrew,
If the rest of the world would just capitulate and go to Imperial units all would be right with the world again! Just kidding, it is funny that anyone still is having troubles with metric units, we have been using them long enough for that to not be the case even if we are still using Imperial units also. There isn't really any excuse anymore, I can move back and forth without a problem and we have been doing it for long enough not to make silly unit mistakes.
On another note, I would still love to request that all the information that Gootee developed here integrated into a simple computer program would go a long way to solving many problems of implementing the solution that he developed. A fill in the blanks program that would be as simple as using a Thiel Small program for figuring out an enclosure problem or a simple network solution. A few fields such as power output, impedance and input voltage and a result of transformer size and capacitor requirements and move on to the rest of the circuit. I guess I can only wish that someone who is a computer programmer steps forward and writes the routine to do this and uses all of Gootee's hard work.
Steven
If the rest of the world would just capitulate and go to Imperial units all would be right with the world again! Just kidding, it is funny that anyone still is having troubles with metric units, we have been using them long enough for that to not be the case even if we are still using Imperial units also. There isn't really any excuse anymore, I can move back and forth without a problem and we have been doing it for long enough not to make silly unit mistakes.
On another note, I would still love to request that all the information that Gootee developed here integrated into a simple computer program would go a long way to solving many problems of implementing the solution that he developed. A fill in the blanks program that would be as simple as using a Thiel Small program for figuring out an enclosure problem or a simple network solution. A few fields such as power output, impedance and input voltage and a result of transformer size and capacitor requirements and move on to the rest of the circuit. I guess I can only wish that someone who is a computer programmer steps forward and writes the routine to do this and uses all of Gootee's hard work.
Steven
Long time since any chatter here ...
This is a nice little, tight summary of what can be done about the inductance of power supply traces, here is about as good a place to link to it as any: Power Tip 56: Estimate PWB interconnect inductance.
Frank
Thanks for that link, Frank.
There is another version of the trace-inductance equation at:
Analog Devices : Analog Dialogue : PCB Layout
I also stumbled across this list of links to articles (mostly Howard Johnson stuff), which you might find interesting and useful:
http://www.sigcon.com/Pubs/pubsAlpha.htm
Last edited:
Actually, I am also a computer programmer, or was for at least fifteen years, mostly with the C Language (which was, literally, the language after the B Language). But there are a lot of variables and always multiple solutions. So making it simple-enough would probably be the biggest challenge.
P.S. Having had my electrical engineering education in the USA (Purdue U.) in the 1970s, I can say that we used metric units, almost exclusively.
Last edited:
gootee,
Thank you for the response. I think that you obviously have a very good grasp of the materials that you spent so much time developing and I appreciate that you are not only versed in electronics but also programming. I always left the program development up to my brother who is a Linux geek, I use to watch him write in basic, but I never took any of that stuff up. I bought him a book about Linux long ago and he never stopped once he went down that road. I would assume these days it is C++ that is being used, but I remember a girlfriend doing Cobal, Assembly and all these other now obscure languages back in the day.
I watched your development from the beginning to the end of the thread and just hoped you could make it user friendly for a guy like me who is not so well versed in all this though I am trying to learn. Ask me to develop a speaker and I am all over that, I get the mechanical side rather than the electrical side. I read as much and as fast as I can assimilate it, but it is a long trek from taking electronics in high school and the basics I have picked up along the way and where you and many others are. Self and Cordell and Colloms are all in front of me all the time, just not easy to keep up with the guru's here and learn all that is going on. Thanks again for all your hard work and time you put in here.
Steven
Thank you for the response. I think that you obviously have a very good grasp of the materials that you spent so much time developing and I appreciate that you are not only versed in electronics but also programming. I always left the program development up to my brother who is a Linux geek, I use to watch him write in basic, but I never took any of that stuff up. I bought him a book about Linux long ago and he never stopped once he went down that road. I would assume these days it is C++ that is being used, but I remember a girlfriend doing Cobal, Assembly and all these other now obscure languages back in the day.
I watched your development from the beginning to the end of the thread and just hoped you could make it user friendly for a guy like me who is not so well versed in all this though I am trying to learn. Ask me to develop a speaker and I am all over that, I get the mechanical side rather than the electrical side. I read as much and as fast as I can assimilate it, but it is a long trek from taking electronics in high school and the basics I have picked up along the way and where you and many others are. Self and Cordell and Colloms are all in front of me all the time, just not easy to keep up with the guru's here and learn all that is going on. Thanks again for all your hard work and time you put in here.
Steven
Steven,
Thanks for all of your kind words. I love Linux, too. I don't get a lot of time to program, these days. But I just might think about it the next time I design an audio power amp power supply. I think it would be good to try to somehow capture what has been learned, here, in a more practical and usable form.
Regards,
Tom
Thanks for all of your kind words. I love Linux, too. I don't get a lot of time to program, these days. But I just might think about it the next time I design an audio power amp power supply. I think it would be good to try to somehow capture what has been learned, here, in a more practical and usable form.
Regards,
Tom
Excellent lot of material, thanks for that, Tom ...
With regard to programming "The Knowledge", I still reckon the spreadsheet way is best, most people are comfortable with this mechanism, and it's always useful at every stage of its development
Frank
Hi to everyone for such a great thread, specially to Gootee for the extended analysis throughout the 170+ pages.
I'm actually going through page 25 and there are some things I just can't fully get, mostly based on the dispariry of opinions.
Sounds a bit like an overkill of the whole matter.
There is some weir concept -to me- of PSU "speed", and based on this speed is the whole sonic performance of the PA.
If I'm not wrong there is no such speed regarding electricity -of course I know there is a measurable speed value but I guess is irrelevant to consider it at audio level frequencies.
Straight to the point I can't get why a rail capacitor will delay its charge and not deliver it to the output devices, let say "on time". That the case, it should cause a serious rail sag at a rate of the audio frequency going through the amplifier. If you attach an oscilloscope to the rail and can't find serious at frequency rail sag then the previous concept of speed is BS.
We should agree the PSU capacitors will deliver their charge only based on their ESR and the impedance of the lines connecting them to the output devices. This 2 factors might hamper delivery but not "speed" which will be close to speed of light.
The other thing that left me thinking is the concept that too much PSU capacitance will have adverse sonic effect on the PA.
If I get the point correctly the whole capacitance calculation is made so that its value is correct for the amplifier running at full power (about 2mF/A). So the capacitance in excess theory states that if you increase total capacity way above this figure you will be loosing sonic quality for a too stiff rail value.
If this theory is correct, then even sticking to the optimum capacitance value (2mF/A) for full power, as you turn down the volume pot, the amplifier will start to loose audio quality as the capacitor bank in the PSU will be larger and larger for the given volume setting requirement. So even with the PSU capacitor bank being optimum for full power, it will make the amplifier have a low volume setting poor performance.
Something sounds weir about this. Please let me know where I'm wrong. I'm no EE.
Pablo.-
I'm actually going through page 25 and there are some things I just can't fully get, mostly based on the dispariry of opinions.
Sounds a bit like an overkill of the whole matter.
There is some weir concept -to me- of PSU "speed", and based on this speed is the whole sonic performance of the PA.
If I'm not wrong there is no such speed regarding electricity -of course I know there is a measurable speed value but I guess is irrelevant to consider it at audio level frequencies.
Straight to the point I can't get why a rail capacitor will delay its charge and not deliver it to the output devices, let say "on time". That the case, it should cause a serious rail sag at a rate of the audio frequency going through the amplifier. If you attach an oscilloscope to the rail and can't find serious at frequency rail sag then the previous concept of speed is BS.
We should agree the PSU capacitors will deliver their charge only based on their ESR and the impedance of the lines connecting them to the output devices. This 2 factors might hamper delivery but not "speed" which will be close to speed of light.
The other thing that left me thinking is the concept that too much PSU capacitance will have adverse sonic effect on the PA.
If I get the point correctly the whole capacitance calculation is made so that its value is correct for the amplifier running at full power (about 2mF/A). So the capacitance in excess theory states that if you increase total capacity way above this figure you will be loosing sonic quality for a too stiff rail value.
If this theory is correct, then even sticking to the optimum capacitance value (2mF/A) for full power, as you turn down the volume pot, the amplifier will start to loose audio quality as the capacitor bank in the PSU will be larger and larger for the given volume setting requirement. So even with the PSU capacitor bank being optimum for full power, it will make the amplifier have a low volume setting poor performance.
Something sounds weir about this. Please let me know where I'm wrong. I'm no EE.
Pablo.-
sex,
Thanks for your kind words.
Unfortunately, it's not quite as simple as 2 mF/A. Maybe for just the C-value calculations, you should skip ahead, to:
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-169.html#post3320547
but also read the link there, first.
Also, try the spreadsheet at post 1666.
Cheers,
Tom
Thanks for your kind words.
Unfortunately, it's not quite as simple as 2 mF/A. Maybe for just the C-value calculations, you should skip ahead, to:
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-169.html#post3320547
but also read the link there, first.
Also, try the spreadsheet at post 1666.
Cheers,
Tom
When people talk about PSU (or amplifier) 'speed' you are generally safe to ignore them. When people talk about capacitor series resistance, PCB trace inductance etc. you should listen to them, epecially when they use numbers too.sextaafondo said:
Too much capacitance wastes money and space, and may increase induction of charging pulses into nearby circuits or via poor grounding. There is no harm in having the rail voltage 'too stiff'; some people use stabilised supplies specifically to achieve this. However some slack may allow a conservatively rated amplifier to give more power on brief peaks.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.