CCS ground points
One basic question about PCB layout of any audio amp.
I have, in principle, two ground points, going independently to an off-board star ground. One is the signal ground, and the other is the power ground. The signal ground is the ground reference for the NFB loop and the input signal. The power ground is used for the "lower" legs of decoupling caps. (My speaker's ground return does not come back to the amp PCB... it goes straight to the star ground.)
Which ground should the CCS ground points connect to? Isn't it a better idea to connect them to the signal ground instead of the power ground, considering that they're supposed to be carrying a steady DC?
One basic question about PCB layout of any audio amp.
I have, in principle, two ground points, going independently to an off-board star ground. One is the signal ground, and the other is the power ground. The signal ground is the ground reference for the NFB loop and the input signal. The power ground is used for the "lower" legs of decoupling caps. (My speaker's ground return does not come back to the amp PCB... it goes straight to the star ground.)
Which ground should the CCS ground points connect to? Isn't it a better idea to connect them to the signal ground instead of the power ground, considering that they're supposed to be carrying a steady DC?
Hi,
I think that a CCS that has no ripple from the Vrail can go to signal ground.
If there is significant Vrail ripple that the CCS is attenuating then probably safer to take it to power ground or to the other Vrail.
If a CCS is fed from a noisy Vrail then, adding an RC attenuator before the CCS may make it quiet enough to connect to signal ground.
A major modification to AmpGuru's GB150 is adding Vrail ripple attenuation to the LTP CCSs. Most builders report at least some sound quality improvement.
I have difficulty with this as well. I have asked in the past but got no sense or agreement.
I think that a CCS that has no ripple from the Vrail can go to signal ground.
If there is significant Vrail ripple that the CCS is attenuating then probably safer to take it to power ground or to the other Vrail.
If a CCS is fed from a noisy Vrail then, adding an RC attenuator before the CCS may make it quiet enough to connect to signal ground.
A major modification to AmpGuru's GB150 is adding Vrail ripple attenuation to the LTP CCSs. Most builders report at least some sound quality improvement.
Hi tcpip
You have to ask yourself what feeds signals to the CCS circuits, or decoupling capacitors, and what grounds are available.
For example, in input stage CCS circuits, the signal levels are in the region of 10 mA or so. This may be fed from a noisy power rail, but as the current is small the ground references can and should go to the quiet (input) ground. If you connected the CCS references to the noisy (power) ground, although the reference and source signals might both see the same noise, there will always be some modulation of the collector voltage and this could cause signal feed through into the CCS.
On the other hand the CCS to the VAS collector side has to reference to the noisy rail. But in this case the signal levels are higher and the relative impact of common mode signals should be reduced. But if you want to be safe, as is used by many designers, the VAS and its CCS (or mirror stage) should run from a higher supply than the output stage. Then it is possible to provide a quiet signal for this.
On the other hand power capacitors which are used to decouple the rails may simply transmit ripple voltages onto the earth line. In this case the correct ground is the noisy side (power) stage.
IN general the rule of thumb would be - noisy or large currents or voltages, noisy line ground, small signals - quiet ground.
cheers
John
You have to ask yourself what feeds signals to the CCS circuits, or decoupling capacitors, and what grounds are available.
For example, in input stage CCS circuits, the signal levels are in the region of 10 mA or so. This may be fed from a noisy power rail, but as the current is small the ground references can and should go to the quiet (input) ground. If you connected the CCS references to the noisy (power) ground, although the reference and source signals might both see the same noise, there will always be some modulation of the collector voltage and this could cause signal feed through into the CCS.
On the other hand the CCS to the VAS collector side has to reference to the noisy rail. But in this case the signal levels are higher and the relative impact of common mode signals should be reduced. But if you want to be safe, as is used by many designers, the VAS and its CCS (or mirror stage) should run from a higher supply than the output stage. Then it is possible to provide a quiet signal for this.
On the other hand power capacitors which are used to decouple the rails may simply transmit ripple voltages onto the earth line. In this case the correct ground is the noisy side (power) stage.
IN general the rule of thumb would be - noisy or large currents or voltages, noisy line ground, small signals - quiet ground.
cheers
John
Hi tcpip,
I'm sorry to have ranted about an old board layout version.
It's just that I believed to see some things in your way. Let me step through your reply and see what concerns may be left.
I know what you mean, but the DRC does only see clearance/distance issues it is aware of. As an example, the board can use different clearance settings for different pin/pad/hole types and different distance settings for different net classes such as small signal nets, supply nets, ground, etc.
Look at R28 and C12 (just as an example), they might be sufficiently spaced for the DRC with the standard settings, but are the copper traces really sufficiently spaced for a potential difference of tens of volts, as it's supply and ground? It will work this way, but does it withstand the environmental influences of the next decades?
Not much to worry in this regard with version 9, though. You improved it a lot.
Same thing here, I was merely commenting the distance between copper of "opposite" potential. The copper pour of V+ used to come alarmingly close to PGND over a lot of copper surface/shape length. You ruled that out now.
Other than that, there's nothing bad about copper pours in general, I just wouldn't use them for a power supply conductor close to small signal circuitry such as a base/source resistor.
I just happen to know the eagle symbols you employ - and how it comes out on a board made with them.
I understand the problems of space constraints. I always try to absolutely minimize the board size required for a given set of components. And that gave me the experience of boards that worked - but were just unsafe to mount and be left alone.
To me it is always of importance to stress safety in such projects. It's the power supply lines of a power amplifier, after all. If your fuse holders sit at the board edge and something touches the board edge, it will be connected to the power supply directly. One could now argue that there are fuse covers, but such a cover will protrude even wider over the board edge and will thus be more prone to physical stress, making the whole thing even more unsafe to try. Things like shorted (circumvented) fuses, unfused current drain, dangerous voltages at the speaker output, and such come to mind.
I'm not saying you absolutely can't do it that way - despite the fact that any safety agency would be happy to hold a mains or signal wire to the board edge and wait (!) - but you have a lot of board space to give away even in version 9. I can see a couple of ways to reduce upper layer occupation, to shorten signal traces and to compact the board size, without safety issues.
Some specs I find locally:
- Wima MKS 10u/63V, 8.5x18.5x26.5mm, lead pitch 22.5mm;
- Epcos MKT 10u/63V, 10.5x18.5x26.5mm, lead pitch 22.5mm;
- Futaba MPC71 0.1 to 1Ohm at 5W, induction-free, 14x5mm, lead pitch 9mm;
- Philips/Phoenix AC05 5W, wirewound, 17x7.5mm;
Unfortunately, I can't give Digikey (or similar) part numers, as I'm not familiar with their catalogues. I mostly order locally from small businesses or online distributors.
Futaba makes the MPC resistors (Eagle's 'RMPC71-5' in 'resistor-power'), but they are actually available from manufacturers from different asian countries.
Philips (and nowadays Phoenix) makes the AC0x resistors (package size 0817 in 'rcl'), they seem to be available worldwide.
But I was also after another aspect of capacitor size: the capacitance value. It does not neccessarily have to be 10uF, that's just what Slone uses for his input impedance and corner frequency.
For example, MKS type caps with 3.3uF are available from 7.5x7.5mm, lead pitch 5mm. But I can of course understand if you don't want to limit your possibilities to try different types and sizes.
Absolutely! No worries, that's what this hobby is about.
I usually gain the most knowledge about my layouts by just printing them in actual size, cutting the print to shape and trying the actual components on.
Of course - why not?
Eagle is also available in a commercial version without the autorouter module. But the free version contains all modules and "just" has the size limitation of single-sheet schematic and 100x80mm board size.
I do, too.
Sebastian.
tcpip said:
I'm sorry to have ranted about an old board layout version.
It's just that I believed to see some things in your way. Let me step through your reply and see what concerns may be left.
I know what you mean, but the DRC does only see clearance/distance issues it is aware of. As an example, the board can use different clearance settings for different pin/pad/hole types and different distance settings for different net classes such as small signal nets, supply nets, ground, etc.
Look at R28 and C12 (just as an example), they might be sufficiently spaced for the DRC with the standard settings, but are the copper traces really sufficiently spaced for a potential difference of tens of volts, as it's supply and ground? It will work this way, but does it withstand the environmental influences of the next decades?
Not much to worry in this regard with version 9, though. You improved it a lot.
Same thing here, I was merely commenting the distance between copper of "opposite" potential. The copper pour of V+ used to come alarmingly close to PGND over a lot of copper surface/shape length. You ruled that out now.
Other than that, there's nothing bad about copper pours in general, I just wouldn't use them for a power supply conductor close to small signal circuitry such as a base/source resistor.
I just happen to know the eagle symbols you employ - and how it comes out on a board made with them.
I understand the problems of space constraints. I always try to absolutely minimize the board size required for a given set of components. And that gave me the experience of boards that worked - but were just unsafe to mount and be left alone.
To me it is always of importance to stress safety in such projects. It's the power supply lines of a power amplifier, after all. If your fuse holders sit at the board edge and something touches the board edge, it will be connected to the power supply directly. One could now argue that there are fuse covers, but such a cover will protrude even wider over the board edge and will thus be more prone to physical stress, making the whole thing even more unsafe to try. Things like shorted (circumvented) fuses, unfused current drain, dangerous voltages at the speaker output, and such come to mind.
I'm not saying you absolutely can't do it that way - despite the fact that any safety agency would be happy to hold a mains or signal wire to the board edge and wait (!) - but you have a lot of board space to give away even in version 9. I can see a couple of ways to reduce upper layer occupation, to shorten signal traces and to compact the board size, without safety issues.
Some specs I find locally:
- Wima MKS 10u/63V, 8.5x18.5x26.5mm, lead pitch 22.5mm;
- Epcos MKT 10u/63V, 10.5x18.5x26.5mm, lead pitch 22.5mm;
- Futaba MPC71 0.1 to 1Ohm at 5W, induction-free, 14x5mm, lead pitch 9mm;
- Philips/Phoenix AC05 5W, wirewound, 17x7.5mm;
Unfortunately, I can't give Digikey (or similar) part numers, as I'm not familiar with their catalogues. I mostly order locally from small businesses or online distributors.
Futaba makes the MPC resistors (Eagle's 'RMPC71-5' in 'resistor-power'), but they are actually available from manufacturers from different asian countries.
Philips (and nowadays Phoenix) makes the AC0x resistors (package size 0817 in 'rcl'), they seem to be available worldwide.
But I was also after another aspect of capacitor size: the capacitance value. It does not neccessarily have to be 10uF, that's just what Slone uses for his input impedance and corner frequency.
For example, MKS type caps with 3.3uF are available from 7.5x7.5mm, lead pitch 5mm. But I can of course understand if you don't want to limit your possibilities to try different types and sizes.
Absolutely! No worries, that's what this hobby is about.
I usually gain the most knowledge about my layouts by just printing them in actual size, cutting the print to shape and trying the actual components on.
Of course - why not?
Eagle is also available in a commercial version without the autorouter module. But the free version contains all modules and "just" has the size limitation of single-sheet schematic and 100x80mm board size.
I do, too.
Sebastian.
semicon datasheet sources
IC-ON-LINE --- Datasheet Search And Download Site
Datasheet catalog for integrated circuits, diodes, triacs, and other semiconductors, view
FYI please
IC-ON-LINE --- Datasheet Search And Download Site
Datasheet catalog for integrated circuits, diodes, triacs, and other semiconductors, view
FYI please
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.