a higher supply voltage delivers larger current transients. The supplys Ri becomes less important and the energy stored in the supply caps grows with U square.
Regards
Try running in mud and see if you can still deliver transients.
Try running in mud and see if you can still deliver transients.
Maybe you should call Panasonic and let them know about this new "mud cap" of yours! Who knows you could be the next bill gates!
Here's my final power supply design for my amp.
Schematic:
Silk Screen:
Traces w/ Silk Screen:
The only thing missing is the reservoir board with the 8 x 10,000uF caps and the voltage switching circuit if i decide to use it
Schematic:
An externally hosted image should be here but it was not working when we last tested it.
Silk Screen:
An externally hosted image should be here but it was not working when we last tested it.
Traces w/ Silk Screen:
An externally hosted image should be here but it was not working when we last tested it.
The only thing missing is the reservoir board with the 8 x 10,000uF caps and the voltage switching circuit if i decide to use it
It’s not meant as a critic to you power supply, it’s your design and your amp, but maybe you should have a closer look at the bridge rectifier?
Hi Tony,
As you can see, this thread has not been visited recently. If you don't get a response, try raising the question on one of the active threads. Also, there is quite a bit of practical power supply information in my new book, Designing Audio Power Amplifiers. See CordellAudio.com - Home. There is also an active thread on the book.
Cheers,
Bob
You can include SLITS in the GND_PLANE, to
force currents to only flow where you want.
Making sketches of current flow is required before
you decide to add slits.
Standard copper foil (1 ounce/foot^2) is 1.4 mils thick,
and a square inch of coil fully contacted across the opposite 1" long
sides will have 1/2 milliOhm resistance.
Should that be carrying 5 amps, you get 2.5 milliVolts across "ground".
By the way, a 6" square also has 1/2 milliOhm resistance.
Get a pad of square-ruled paper (an old "quadrile pad"),
and enjoy the thinking process.
By the way, currents do not take the shortest path.
They choose to take all possible paths, proportional to conductance.
That's reality, unfortunately. That means currents spread out
all over the "GROUND PLANE" and will interact. Hence sketches of
current flows, and possibly some slits, are your friend.
force currents to only flow where you want.
Making sketches of current flow is required before
you decide to add slits.
Standard copper foil (1 ounce/foot^2) is 1.4 mils thick,
and a square inch of coil fully contacted across the opposite 1" long
sides will have 1/2 milliOhm resistance.
Should that be carrying 5 amps, you get 2.5 milliVolts across "ground".
By the way, a 6" square also has 1/2 milliOhm resistance.
Get a pad of square-ruled paper (an old "quadrile pad"),
and enjoy the thinking process.
By the way, currents do not take the shortest path.
They choose to take all possible paths, proportional to conductance.
That's reality, unfortunately. That means currents spread out
all over the "GROUND PLANE" and will interact. Hence sketches of
current flows, and possibly some slits, are your friend.
How to compute magnetic coupling from a straight wire into a rectangular loop.
Consider a wire 4" from a PCB, the PCB having a 4" by 4" GROUND path
around it (this could also just be a 4" * 4" loop involving ANY part of the
amplifier, including the input signal, or the differential_pair_and_feedback
or the rail-rail VAS. Could be signal, Vdd, or GND paths)
The magnetic coupling can be scarily big. Here goes.
Vcouple = MUo*MUr*area
------------------ * dI/dT
2*pi * distance
(I think this comes from merging BiotSavot + Amperes laws)
Lets make area 0.1meter^2 (that 4" thing), distance the same,
Have output be 20KHz at 10 amps peak. Then dI/dT = 10Amp * 2*pi*freq
or approx 1.28Million amps/second. MUo = 4*pi*10^-7 Henry/meter. MUr=1
Math
V = 4pi10^-7 * 0.1 * 0.1 * 1.28Million
-----------------------------------------
2*pi*0.1
V = 2 * 10^-7 * 0.1 * 1.28Million = 2*10^-8 * 1.28 * 10^+6
V = 2.56 ^ 10^-2
V = 25 milliVolts
25 milliVolts coupled from 4" away, into a 4" by 4" loop.
As I plan my own power-system for a no-thermal-distortion
amplifier, these numbers motivate me to really sketch out
all the wires, estimating the inductances, estimate the ringing
frequencies, and dampen ringing, and use twisted-pairs for
minimal area to minimize transmitter and receiver coupling
of trash.
Consider a wire 4" from a PCB, the PCB having a 4" by 4" GROUND path
around it (this could also just be a 4" * 4" loop involving ANY part of the
amplifier, including the input signal, or the differential_pair_and_feedback
or the rail-rail VAS. Could be signal, Vdd, or GND paths)
The magnetic coupling can be scarily big. Here goes.
Vcouple = MUo*MUr*area
------------------ * dI/dT
2*pi * distance
(I think this comes from merging BiotSavot + Amperes laws)
Lets make area 0.1meter^2 (that 4" thing), distance the same,
Have output be 20KHz at 10 amps peak. Then dI/dT = 10Amp * 2*pi*freq
or approx 1.28Million amps/second. MUo = 4*pi*10^-7 Henry/meter. MUr=1
Math
V = 4pi10^-7 * 0.1 * 0.1 * 1.28Million
-----------------------------------------
2*pi*0.1
V = 2 * 10^-7 * 0.1 * 1.28Million = 2*10^-8 * 1.28 * 10^+6
V = 2.56 ^ 10^-2
V = 25 milliVolts
25 milliVolts coupled from 4" away, into a 4" by 4" loop.
As I plan my own power-system for a no-thermal-distortion
amplifier, these numbers motivate me to really sketch out
all the wires, estimating the inductances, estimate the ringing
frequencies, and dampen ringing, and use twisted-pairs for
minimal area to minimize transmitter and receiver coupling
of trash.
BY the way, you'll be glad to know that
is a worse-case formula; many arrangements of wire+loop
will have considerably less coupling; should wire run exactly symmetrically
over the loop, e.g. exactly diagonally, or exactly from middle-of-east-side
to middle-of-west-side, there will be zero coupling.
That explains why "moving the wires" affects the hum or buzz or
sonic-quality.
is a worse-case formula; many arrangements of wire+loop
will have considerably less coupling; should wire run exactly symmetrically
over the loop, e.g. exactly diagonally, or exactly from middle-of-east-side
to middle-of-west-side, there will be zero coupling.
That explains why "moving the wires" affects the hum or buzz or
sonic-quality.
Cherry showed an example PCB layout that used the technique of equal numbers of loop and reversed loop to cancel the loop area and thus cancel the antenna effect. I think it was a Wireless World paper.
----Have any of you read the Blowtorch thread? SOLID BILLET ALUMINUM CHASSIS FOR ----$3,000!! Do any of these people realize that milling a part from a solid block has -------the structural value of soap carving?
Thick aluminum will shield even 60Hertz, with 9dB (a neper, or 1/2.718 ) of
attenuation for every SkinThickness (approx 1/2 inch). So will thick copper,
but who can afford it? An inch of aluminum provides 18dB, 1.5" 27dB.
The higher harmonics are even more reduced..
Those 1millisec rectifier surges, if viewed as 1/2 cycle of an interferrer,
would have 2millisec period, or 500 Hertz, providing nearly 3dB more
of skin depth attenuation. Per skin depth.
Having enjoyed this thread for last 3 days as I try to read it all,
I offer
----distance is our friend
---- local summation of high currents (close each high-current path
at a private part of GND system, as Bob/Andrew etal suggest)
----minimize area of loops, both Transmitter and Receiver loops
--- twisted-pairs
--- reduce edge-rates (use that 0.22 ohm and 100milliFarad, to keep
fast rectifier surges in the first bank of capacitor, away from low-level circuits)
---(my favorite) implement "local batteries" to ensure the fast-surges are always
provided LOCALLY, and thus we can sketch out the entire loop used by the
fast-surges and minimize that area; consider R*C*R*C*R*C filters, because of
ESR in the capacitors; in small-signal work I assume every cap has minimum
ESR of 1ohm, and use 100ohm Rs
Thick aluminum will shield even 60Hertz, with 9dB (a neper, or 1/2.718 ) of
attenuation for every SkinThickness (approx 1/2 inch). So will thick copper,
but who can afford it? An inch of aluminum provides 18dB, 1.5" 27dB.
The higher harmonics are even more reduced..
Those 1millisec rectifier surges, if viewed as 1/2 cycle of an interferrer,
would have 2millisec period, or 500 Hertz, providing nearly 3dB more
of skin depth attenuation. Per skin depth.
Having enjoyed this thread for last 3 days as I try to read it all,
I offer
----distance is our friend
---- local summation of high currents (close each high-current path
at a private part of GND system, as Bob/Andrew etal suggest)
----minimize area of loops, both Transmitter and Receiver loops
--- twisted-pairs
--- reduce edge-rates (use that 0.22 ohm and 100milliFarad, to keep
fast rectifier surges in the first bank of capacitor, away from low-level circuits)
---(my favorite) implement "local batteries" to ensure the fast-surges are always
provided LOCALLY, and thus we can sketch out the entire loop used by the
fast-surges and minimize that area; consider R*C*R*C*R*C filters, because of
ESR in the capacitors; in small-signal work I assume every cap has minimum
ESR of 1ohm, and use 100ohm Rs
Hi Andrew
I uploaded one of Dr Cherry's articles awhile ago.
Here's Part 2, which contains an NDFL power amp design, and discusses the PCB issue. The low-frequency group delay compensation scheme is interesting too.
Part 1, which discusses NDFL theory, was posted here.
Sorry about the file sizes, but at least I had the decency to post them as zip attachments so they won't all automatically open in your browser.
Perversely, when Jan converted the JPEGs to PDF to add to his website, the total size increased.
zips 1 & 4 show what I was referring to. But it's not the same paper. He explained the loop and inverted loop principle and also the group of 3 power rails running down the middle of the PCB.
I have built two pairs of power amps (a chipamp and a discrete amp) with the LF compensation shown in zip2.
I am not sure if I heard any or a big change/improvement.
I have not tested the LF responses to see if the effect is measurable.
I have built two pairs of power amps (a chipamp and a discrete amp) with the LF compensation shown in zip2.
I am not sure if I heard any or a big change/improvement.
I have not tested the LF responses to see if the effect is measurable.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.