Some things I would do.
Make the track between the two star earths fatter.
Create a little distance between the tracks underneath R5 to reduce capacitance.
Move the track above U1 higher to also reduce capacitance from the track to R3.
Any tracks running in parallel next to each other are going to have a capacitance. On high frequeny boards this is actually used instead of capacitors sometimes.
Regards,
Mark
Make the track between the two star earths fatter.
Create a little distance between the tracks underneath R5 to reduce capacitance.
Move the track above U1 higher to also reduce capacitance from the track to R3.
Any tracks running in parallel next to each other are going to have a capacitance. On high frequeny boards this is actually used instead of capacitors sometimes.
Regards,
Mark
missing parts from PS circuit?
I notice that no one here seems to pay too much attention to the ground path on power supply diagrams, photos.
It has long been my habit to put a honking good sized torid coil in the return ground line from the load (amp, whatever) ...
Looking at: http://www.diyaudio.com/forums/attachment.php?s=&postid=946016&stamp=1150979128 ... a healthy coil might go right between the load and the "G" pad on the board.
The idea here being to reduce noise on the return ground path, closer to the load than the main caps. Why? Because this is where the electrons are. This is the usual source for ground loops and this is where RF noise can enter to make an otherwise cool amp run hotter.
No kidding.
I notice that no one here seems to pay too much attention to the ground path on power supply diagrams, photos.
It has long been my habit to put a honking good sized torid coil in the return ground line from the load (amp, whatever) ...
Looking at: http://www.diyaudio.com/forums/attachment.php?s=&postid=946016&stamp=1150979128 ... a healthy coil might go right between the load and the "G" pad on the board.
The idea here being to reduce noise on the return ground path, closer to the load than the main caps. Why? Because this is where the electrons are. This is the usual source for ground loops and this is where RF noise can enter to make an otherwise cool amp run hotter.
No kidding.
Hi,
now that your layout is close to what you can use for this project and the next few, can I suggest you squeeze the PCB to give a much more compact arrangement. At a guess you may be able to get all those components into about half the board area. Leave space at the edge (overhanging the edge) for IC sinks.
now that your layout is close to what you can use for this project and the next few, can I suggest you squeeze the PCB to give a much more compact arrangement. At a guess you may be able to get all those components into about half the board area. Leave space at the edge (overhanging the edge) for IC sinks.
Mark, I'm interested in your last suggestion about spacing the parallel traces. I guess it is unclear to me when you space traces more closely, and when you should avoid doing that. Maybe I'm thinking of ground loops and inductance vs. stray capacitance. The reason I ask is because if the design goal is to have more space between traces (or maybe avoid parallel traces altogether), it seems contrary to AndrewT's suggestion to make the board even smaller. Care to discuss? I'm away from my computer for a week or so, but I will update the PCB when I get back next week. Until then, I'd love to hear a continuing discussion about these issues. I'm learning a ton from you all.
AndrewT said:
Yes, this is true. However if you concider the pcb as a reactance plane then to avoid this, distance and also not having right angle traces (these behave as inductors at high frequency) etc are used to lessen the effects of this (or increase were specifically desired). So by having an overly small pcb the risk would be that all tracks become in close proxomity such that the entire pcb becomes a collection of capacitors and inductors. Also it not a bad idea to keep the diodes away from other parts of the circuit as these can be noisey. So I would not reduce the size because you can, only if you need to. This could be seen as extreme but there can be high currents in power supplies so I think these points are worth a thought.
Regards,
Mark (Zuus)
Bye, bye wall wart, hello solid linear supply
Your layout is quite good, well balanced for common mode rejection. Single sided boards = cost effective, reliable power supplies.
I might make two suggestions:
Increase the trace sizes (width) so that a lot more copper is left on the board. (This is more environmentally correct, more cost effective to produce, reduces resistance between components, reduces errors in stuffing and soldering and increases reliability.)
Consider adding an isolated double solder pad close to G (output ground) for the possible installation of a coil or other "hum bucking" component. (I always include a Ferite core coil of several turns in the return to ground path as it changes phasing of the power surges back to ground, improves "turn on" surge resistance and can contribute to even better regulation.
Good show :>)
Your layout is quite good, well balanced for common mode rejection. Single sided boards = cost effective, reliable power supplies.
I might make two suggestions:
Increase the trace sizes (width) so that a lot more copper is left on the board. (This is more environmentally correct, more cost effective to produce, reduces resistance between components, reduces errors in stuffing and soldering and increases reliability.)
Consider adding an isolated double solder pad close to G (output ground) for the possible installation of a coil or other "hum bucking" component. (I always include a Ferite core coil of several turns in the return to ground path as it changes phasing of the power surges back to ground, improves "turn on" surge resistance and can contribute to even better regulation.
Good show :>)
Hello linear supply / more on design
AndrewT: "Simply to point out that shorter traces with the same spacing will have less capacitance. ..."
Well AT is certainly right here when it comes to audio signal pathways to and from (I/O) of the signal processing components, etc. ... as stray capacitance and stray inductance all contribute to input and output impedence = of serious concern when DIY project building.
BUT ... when it comes to designing and building power supplies and line filters, stray or extra capacitance can be your best friend ... as extra capacitance on the power and ground lines means extra filtering = more total capacitance = increasing the "noise floor", etc.
DIY Lead Technician on this project, ezkadude: " ... As for the voltages, I want to end up with +/- 12 ... "
Everyone should note that EZKCDude's "isolated ground" design is easily adaptable to any split supply from an AC wall wart or poorly regulated DC wall wart. His layout and design could just as easily output filtered DC of a range of DC voltages from ~= +/- 5 to ~= +/- 24 VDC. The linear regulators, (+) LM317 and (-) LM337, are capable of regulating any input voltage of up to +/- 28 V (peak voltage if unregulated AC). ... any or all by mearly changing a couple of resistor values. The best efficiencies are for split supply output of +/- 9 through +/- 18 VDC ... See:
http://www.national.com/images/pf/LM317/00906301.jpg
http://www.national.com/ds/LM/LM117.pdf ... especially page 8, etc.
...
Notes in passing:
*Finding a decent AC wall wart as a transformer source should not be too difficult, but it may be ... For the +/- 5 VDC through +/- 18 VDC design, the wall wart can have just about any (peak) value of up to 50 or so Volts DC or ~= 50 Volts PEAK voltage AC. ( Example: http://www.halted.com/ccp11916-plug-in-adap-15vac-320ma-120vac-in-2-5mm-t35-15320-a000c-80558.htm ... 15 Volts AC is equal to 25.6 Volts peak alternating plus and minus (15 x 1.7) = 51 Volts peak to peak over all ... ~= $3 each ... which should work just fine.)
*Finding a 30 Volt DC wall wart is another matter entirely ... BUT since the EZKCDude's design is patterned after generally accepted "half wave" rectifier designs, the use of an AC wall wart should be acceptable. The above example has a power limit for cool operation of about 200 milliAmps ~= meaning an output limit for cool operation of about 100 milliAmps or 15 to 25 watts overall WITH some healthy heat sinks on the LM317 / LM337 "wings".
*Adding extra solder pads for "Snubbing" capacitors might be a nice touch.
*Adding Ferrite toridal coils to each output (+ & -) AND to return ground (on the left side of the board at V+, V-, G solder pads will "buck" any feedback "hum" from powered components .
*Twisting the output power leads, likewise, will increase "common mode rejection", always a good thing ... one full twist every two or three inches is good.
(UPS did deliver today and my orders for speakers just showed up ... so see ya all after the 4th.)
AndrewT: "Simply to point out that shorter traces with the same spacing will have less capacitance. ..."
Well AT is certainly right here when it comes to audio signal pathways to and from (I/O) of the signal processing components, etc. ... as stray capacitance and stray inductance all contribute to input and output impedence = of serious concern when DIY project building.
BUT ... when it comes to designing and building power supplies and line filters, stray or extra capacitance can be your best friend ... as extra capacitance on the power and ground lines means extra filtering = more total capacitance = increasing the "noise floor", etc.
DIY Lead Technician on this project, ezkadude: " ... As for the voltages, I want to end up with +/- 12 ... "
Everyone should note that EZKCDude's "isolated ground" design is easily adaptable to any split supply from an AC wall wart or poorly regulated DC wall wart. His layout and design could just as easily output filtered DC of a range of DC voltages from ~= +/- 5 to ~= +/- 24 VDC. The linear regulators, (+) LM317 and (-) LM337, are capable of regulating any input voltage of up to +/- 28 V (peak voltage if unregulated AC). ... any or all by mearly changing a couple of resistor values. The best efficiencies are for split supply output of +/- 9 through +/- 18 VDC ... See:
http://www.national.com/images/pf/LM317/00906301.jpg
http://www.national.com/ds/LM/LM117.pdf ... especially page 8, etc.
...
Notes in passing:
*Finding a decent AC wall wart as a transformer source should not be too difficult, but it may be ... For the +/- 5 VDC through +/- 18 VDC design, the wall wart can have just about any (peak) value of up to 50 or so Volts DC or ~= 50 Volts PEAK voltage AC. ( Example: http://www.halted.com/ccp11916-plug-in-adap-15vac-320ma-120vac-in-2-5mm-t35-15320-a000c-80558.htm ... 15 Volts AC is equal to 25.6 Volts peak alternating plus and minus (15 x 1.7) = 51 Volts peak to peak over all ... ~= $3 each ... which should work just fine.)
*Finding a 30 Volt DC wall wart is another matter entirely ... BUT since the EZKCDude's design is patterned after generally accepted "half wave" rectifier designs, the use of an AC wall wart should be acceptable. The above example has a power limit for cool operation of about 200 milliAmps ~= meaning an output limit for cool operation of about 100 milliAmps or 15 to 25 watts overall WITH some healthy heat sinks on the LM317 / LM337 "wings".
*Adding extra solder pads for "Snubbing" capacitors might be a nice touch.
*Adding Ferrite toridal coils to each output (+ & -) AND to return ground (on the left side of the board at V+, V-, G solder pads will "buck" any feedback "hum" from powered components .
*Twisting the output power leads, likewise, will increase "common mode rejection", always a good thing ... one full twist every two or three inches is good.
(UPS did deliver today and my orders for speakers just showed up ... so see ya all after the 4th.)
Hi FastEddy,
Capacitance = communication between two points, when this is by design then by applying your rules this will help matters. However one of the main functions of a power supply is to isolate the output from the input noise (i.e. the AC ripple and switching noise of the diodes (and RF)). Thus Capacitance (inductance) in the wrong place could go against this and dirty the ground/output.
I am not disagreeing, in fact I agree (totally actually), but there is a flip side to this statement.
Regards,
Mark
Capacitance = communication between two points, when this is by design then by applying your rules this will help matters. However one of the main functions of a power supply is to isolate the output from the input noise (i.e. the AC ripple and switching noise of the diodes (and RF)). Thus Capacitance (inductance) in the wrong place could go against this and dirty the ground/output.
I am not disagreeing, in fact I agree (totally actually), but there is a flip side to this statement.
Regards,
Mark
O.k. Fattened traces some more. Added a snubber (C9), but is this location o.k.? I was thinking about putting it in the "over" the first star ground in between the electrolytics. As for the ferrite core, do you mean ferrite bead, FastEddy? I have some of these:
http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail?Ref=79292&Row=561546&Site=US
Also, I'm not exactly sure where you mean to put them. After the second star ground, in series with G, for example?
http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail?Ref=79292&Row=561546&Site=US
Also, I'm not exactly sure where you mean to put them. After the second star ground, in series with G, for example?
Hi,
I think you have C9 in the wrong place.
The "snubber" should be in parallel to a diode (two diodes = two snubbers) and the traces should come straight from the diode leads.
The snubber consists of a series combination of resistor and capacitor. You could link out the resistor if you decide not to use it.
The dual holes (0.3 & 0.4inch?) for the smoothing caps need to be relocated slightly. Draw the body outline for each pair of holes and you'll find the cap body fouls the resistor and maybe C9.
I think you have C9 in the wrong place.
The "snubber" should be in parallel to a diode (two diodes = two snubbers) and the traces should come straight from the diode leads.
The snubber consists of a series combination of resistor and capacitor. You could link out the resistor if you decide not to use it.
The dual holes (0.3 & 0.4inch?) for the smoothing caps need to be relocated slightly. Draw the body outline for each pair of holes and you'll find the cap body fouls the resistor and maybe C9.
If you look at the schematic richie00boy posted on his website, the snubber goes all the way across the two diodes and electrolytic caps. That's what I did here. Does anyone else care to chime in their opinion on this?
Because this PSU is really intended to be mainly a "pre-regulator" for a DAC, I'm wondering if it is really necessary, anyway. I could put the "snubber" (whatever form it takes) on the DAC board, itself, right?
Because this PSU is really intended to be mainly a "pre-regulator" for a DAC, I'm wondering if it is really necessary, anyway. I could put the "snubber" (whatever form it takes) on the DAC board, itself, right?
Andrew, I added the snubbers you suggested, and moved C9. I also added additional silkscreen body outlines for larger caps. Thanks, for pointing that out. I think the board is now verging on too complex, but if these things help, the additional cost and effort of soldering is not such a big deal. I'm still wondering about the ferrites. Doesn't seem like there is much room left, so I may just leave those out.
Wall wart cleanup / stray capacitance
Zuus: " ... Capacitance = communication between two points, when this is by design then by applying your rules this will help matters. However one of the main functions of a power supply is to isolate the output from the input noise (i.e. the AC ripple and switching noise of the diodes (and RF)). Thus Capacitance ([and / or] inductance) in the wrong place could go against this and dirty the ground/output. ..."
Absolutely correct.
Of interest:
a) Several commercially available power supplies use double sided boards ... A few of these makers also leave as much of the copper clad on both sides as possible = extra wide traces, vast "ground planes", etc. These actually contribute to "common mode rejection" of unwanted noise, caused usually from external sources like RF, etc. There is a military spec regarding this in an attempt to reduce effectes of "EMP" and possible electronic interference from the "outside" in the battlefield. These slabs of copper ("planes" or "plates") are effectively parallel with a dialectric between (the non conductive board material) making the board itself a flat capacitor with reasonably good "common mode rejection". (Although EMF (EMP) radiation from certain incidents angles may actually negate this effect, random angles generally do not ... (Military secret = this makes the power supply a little more "stealthy" as well.)
b) Concerning trace widths, usually the fatter the better as in production, errors can creep in so big solder pads and fat traces are the norm in power supplies (but of course not in signal processing). In the case of double sided boards I have seen fat traces and plated through holes used to effectively "twist" conductor pairs over distances of just a few inches, further contributing to "common mode rejection" .... these fat, twisted traces act just like twisted wire bundles in cables ... and the fat traces seem able to accomplish this goal better than the slim ones.
c) Leaving as much copper on the boards as possible also contributes (modestly) to environmental concerns as less ferric-cloride is needed to remove the copper in the bath (a method still used in the Orient and in small batches here in the states) and/or when manufacturing by cutting copper from the boards (by laser or router, etc.), less scrap copper goes into the trash bin ... and reduces "speeds and feeds" in manufacturing time (time = money). This may seem like small details, but when making hundreds of boards it really adds up. .... and the design under discussion here has merit for extensive reproduction = possible market potential = a mass produced wall wart filter for general audio (and military and other) use == $$ = "Anything worth doing, is worth doing for money ..." - Alfred E. Newman (aka William Gates / Mad magazine)
Anyway, these kinds of things are learned by trail and error in the laboratory and shop and usually not discussed in classrooms ... so it is understandable that many here may not know ...
Zuus: " ... Capacitance = communication between two points, when this is by design then by applying your rules this will help matters. However one of the main functions of a power supply is to isolate the output from the input noise (i.e. the AC ripple and switching noise of the diodes (and RF)). Thus Capacitance ([and / or] inductance) in the wrong place could go against this and dirty the ground/output. ..."
Absolutely correct.
Of interest:
a) Several commercially available power supplies use double sided boards ... A few of these makers also leave as much of the copper clad on both sides as possible = extra wide traces, vast "ground planes", etc. These actually contribute to "common mode rejection" of unwanted noise, caused usually from external sources like RF, etc. There is a military spec regarding this in an attempt to reduce effectes of "EMP" and possible electronic interference from the "outside" in the battlefield. These slabs of copper ("planes" or "plates") are effectively parallel with a dialectric between (the non conductive board material) making the board itself a flat capacitor with reasonably good "common mode rejection". (Although EMF (EMP) radiation from certain incidents angles may actually negate this effect, random angles generally do not ... (Military secret = this makes the power supply a little more "stealthy" as well.)
b) Concerning trace widths, usually the fatter the better as in production, errors can creep in so big solder pads and fat traces are the norm in power supplies (but of course not in signal processing). In the case of double sided boards I have seen fat traces and plated through holes used to effectively "twist" conductor pairs over distances of just a few inches, further contributing to "common mode rejection" .... these fat, twisted traces act just like twisted wire bundles in cables ... and the fat traces seem able to accomplish this goal better than the slim ones.
c) Leaving as much copper on the boards as possible also contributes (modestly) to environmental concerns as less ferric-cloride is needed to remove the copper in the bath (a method still used in the Orient and in small batches here in the states) and/or when manufacturing by cutting copper from the boards (by laser or router, etc.), less scrap copper goes into the trash bin ... and reduces "speeds and feeds" in manufacturing time (time = money). This may seem like small details, but when making hundreds of boards it really adds up. .... and the design under discussion here has merit for extensive reproduction = possible market potential = a mass produced wall wart filter for general audio (and military and other) use == $$ = "Anything worth doing, is worth doing for money ..." - Alfred E. Newman (aka William Gates / Mad magazine)
Anyway, these kinds of things are learned by trail and error in the laboratory and shop and usually not discussed in classrooms ... so it is understandable that many here may not know ...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.