Tír Na HiFi

rickmcinnis wrote:Randy,

Of course, the ground line is important but one must make a choice. The layout kind of determined itself. Easier to have the vias anchor the wires to the shift registers and that results in the wire exiting the back of the board, hence placing the batteries under the board which results in minimal length of the "hot" leads.

The wire to the positive end of the batteries are not quite 2.5 inches - grounds to the switch will average about seven inches, if that much. I figure the best ground point is the original input ground and that determines the length possible. It is far way from the batteries but there is not a suitable ground point anywhere else. I want something of substance to tie into and not a SMD pad.

If that "original input ground" is a signal ground then this wouldn't be the optimal point for your PS ground. The main principle in good grounding is to keep the PS ground return currents from polluting the much lower power signal return currents. So the optimum is to tie all signal grounds together in a single point - a star ground for signal, if you will & to tie all PS ground points together in a PS star ground. These two star grounds are then tied together.

Impossible to run twisted pairs. I am only attaching to the"non-ground" side of the of the shift register. Twisting is more of a factor for AC signals, not DC, anyway. I do not see any twisted pairs in your photo so I am baffled by this comment.

Again, we need to go beyond the definitions - the power going to shift registers or indeed any high speed digital ICs is drawn in fast current pulses. This can radiate RF noise & twisted the + & - wiring is not a bad practise. As to the actual need for this in practise - it depends on so many factors. I actually thought you were running both positive & ground wires to each shift register & that's why I was suggesting it might be a worthwhile configuration. The treatment of ground returns are the more important part of the PS setup

I cannot picture where you are attaching the copper tape. Is this coming from the input ground? I do not see any advantage to using foil/tape as opposed to stranded wire. The foil would have to be pretty wide to be the equivalent of a 16 gauge stranded wire. I am using some old WE wire that sounds really good as a speaker cable and I have some left over. Foil, alone, does not make a ground plane

I think it is important to have all power grounds meet in one spot and not be scattered about the board. Another one of those phobias.

Yes, agreed

jkeny wrote:

rickmcinnis wrote:Randy,

Of course, the ground line is important but one must make a choice. The layout kind of determined itself. Easier to have the vias anchor the wires to the shift registers and that results in the wire exiting the back of the board, hence placing the batteries under the board which results in minimal length of the "hot" leads.

The wire to the positive end of the batteries are not quite 2.5 inches - grounds to the switch will average about seven inches, if that much. I figure the best ground point is the original input ground and that determines the length possible. It is far way from the batteries but there is not a suitable ground point anywhere else. I want something of substance to tie into and not a SMD pad.

If that "original input ground" is a signal ground then this wouldn't be the optimal point for your PS ground. The main principle in good grounding is to keep the PS ground return currents from polluting the much lower power signal return currents. So the optimum is to tie all signal grounds together in a single point - a star ground for signal, if you will & to tie all PS ground points together in a PS star ground. These two star grounds are then tied together.

Sounds like a good idea star gnd for digital star gnd for analog and join the two at the battery gnds

Impossible to run twisted pairs. I am only attaching to the"non-ground" side of the of the shift register. Twisting is more of a factor for AC signals, not DC, anyway. I do not see any twisted pairs in your photo so I am baffled by this comment.

Again, we need to go beyond the definitions - the power going to shift registers or indeed any high speed digital ICs is drawn in fast current pulses. This can radiate RF noise & twisted the + & - wiring is not a bad practise. As to the actual need for this in practise - it depends on so many factors. I actually thought you were running both positive & ground wires to each shift register & that's why I was suggesting it might be a worthwhile configuration. The treatment of ground returns are the more important part of the PS setup

id be tempted to try both +/- and gnd from shift registers as it would be easy to disconnect gnd for the a/b comparisons

I cannot picture where you are attaching the copper tape. Is this coming from the input ground? I do not see any advantage to using foil/tape as opposed to stranded wire. The foil would have to be pretty wide to be the equivalent of a 16 gauge stranded wire. I am using some old WE wire that sounds really good as a speaker cable and I have some left over. Foil, alone, does not make a ground plane

I think it is important to have all power grounds meet in one spot and not be scattered about the board. Another one of those phobias.

Yes, agreed

not sure this is clear cut
with the audio pc i put in several and many separate ps and gnds on the same motherboard with nothing but improvement it was like the noise was travelling in the direction of the current rather than travelling the long route past/though other components

i find gnd to be a bit of a grey area, not as magic sink hole and not to be ignored

nige2000 wrote:

jkeny wrote:

rickmcinnis wrote: I think it is important to have all power grounds meet in one spot and not be scattered about the board. Another one of those phobias.

Yes, agreed

not sure this is clear cut
with the audio pc i put in several and many separate ps and gnds on the same motherboard with nothing but improvement it was like the noise was travelling in the direction of the current rather than travelling the long route past/though other components

i find gnd to be a bit of a grey area, not as magic sink hole and not to be ignored

Yes ground is complicated, indeed. Even with a solid, unbroken ground plane the return currents do not necessarily behave as expected - with lower frequency currents (typical audio frequencies), the return path follows the path of least resistance. As long as this ground plane has no breaks in it this will be fine but if there are tracks on the ground plane (used to solve the routing of some signals on the upper plane) then the currents will have to go around these & can cause unexpected results.

High frequency ground return currents follow a path of least INDUCTANCE which is very different to least resistance. In other words if a copper trace on the top plane bring current to a power pin the return current on the bottom solid ground plane will follow the path of the power copper trace on the top plane. The return current path on the ground plane will NOT trace a pathway that is confined to the width of the PS trace on top layer but will spread out from that line decreasing as it gets further away from that line. This again can have unexpected consequences.

In some ways 3D wiring can be better than pcbs in this regard - very defined return current paths!

rickmcinnis wrote:Randy,
Impossible to run twisted pairs. I am only attaching to the"non-ground" side of the of the shift register. Twisting is more of a factor for AC signals, not DC, anyway. I do not see any twisted pairs in your photo so I am baffled by this comment.

LMAO. So there is the stuff I know I should do, and then there's what I actually do. If I went for perfection, I'd never accomplish anything, so I've learned its better to throw things together, and just do the best I can. Otherwise I have a bunch of 1/2 finished projects laying around.

On this project, instead of twisted pairs of power and ground, I tried to tie the batteries to ground kind of close to the devices they would power. And I did it with a low impedance path. I'll admit I could probably do better if I tried a little harder.

My low impedance path is a piece of copper tape. I insulated the tape with kapton tape, and then folded it, so it doesn't look like a wide piece of tape, but it really is. For DC returns, cross section matters, or large gauge wire. For high frequency AC, you need surface area, so a wide piece of copper tape works better. I think my tape is 2 inches wide. Some of this is aimed at stuff in the Mhz and higher, but I apply to to digital audio stuff anyway, figuring it can't hurt.

And I'll admit I don't understand grounding as much as I should, I just do the best I can, with the effort I'm willing to put into this "hobby".

Randy

jkeny wrote: In some ways 3D wiring can be better than pcbs in this regard - very defined return current paths!

now ive a comeback
when someone says my build is messy
i can say its not messy its an extremely sophisticated 3D return current system :)

nige2000 wrote:

jkeny wrote: In some ways 3D wiring can be better than pcbs in this regard - very defined return current paths!

now ive a comeback
when someone says my build is messy
i can say its not messy its an extremely sophisticated 3D return current system :)

There ya go !!

jkeny,

The ground point I am using is where one would attach the two power transformers for a box stock SOEKRIS.

I have noticed others using the signal ground at the output and though this obviously will work it does not seem ideal to me.

AS far as the twisted pairs - if I really thought it would make a difference I would do it BUT when you look at the size of the batteries versus the width of the solder pads and the fact that I need to combine four power inputs for each bank I figured twisted pairs would require a WHOLE lot more wire and would make the installation very messy and complicated. Most of the time simplicity is better than the textbook's dictum in my experience.

I exaggerated the length of my ground wire yesterday. More like five inches including the switch. I do know there is no way to make any of the wires shorter than they are.

I love foil for wire as much as anyone but, again, there are constraints with dealing with someone else's circuit board. I could have used silver foil for the power from the batteries to the shift register but worry that they could be pulled off the pad and the pad with them. One could lay a large piece of copper sheet below the board and attach grounds to this But the sheet would be too far away from the board to act as a ground plane and again would just add complication without assurance of any benefit.

As with all things I say I hope to be proved wrong, would not bother me in the slightest!

Nigel,

Your 3D layouts are great for experimentation. I know you like to work fast while the ideas are still fresh in your head. On the other hand I deliberate for hours/days and by the time I put something in place you have come up with a better approach. You have ready to go modules available so you can put things together quickly.

I have to make the pieces to follow your lead so while I am doing it I am given the opportunity to try to put them together as close together as I can with the benefit of knowing what you have found works so the time spent in minimizing the install is justified. If I was to strike out on my own my work would look just the same as yours. I figure if you find something that works the least I can do is to attempt to work out some of the packaging.

ran silver wire from both +/- 3.3v at switch to shift registers, i had intended 16 runs to gnd too but i got lazy so maybe thats for another day
definitely another improvement, transients are faster, more separation and clarity

looks a bit chaotic

wonder if theres a tidier way to do it

rickmcinnis wrote:jkeny,

The ground point I am using is where one would attach the two power transformers for a box stock SOEKRIS.

That sounds fine as long as the board's grounds are done correctly which I have no reason to believe they are not

I have noticed others using the signal ground at the output and though this obviously will work it does not seem ideal to me.

Sure

AS far as the twisted pairs - if I really thought it would make a difference I would do it BUT when you look at the size of the batteries versus the width of the solder pads and the fact that I need to combine four power inputs for each bank I figured twisted pairs would require a WHOLE lot more wire and would make the installation very messy and complicated. Most of the time simplicity is better than the textbook's dictum in my experience.

Yea, I agree - I was just pointing out that the high speed current pulses being delivered to digital ICs can benefit from twisted wiring - remember the return currents follow the path of least inductance & twisting the wires together achieves this lower inductance

nige2000 wrote:ran silver wire from both +/- 3.3v at switch to shift registers, i had intended 16 runs to gnd too but i got lazy so maybe thats for another day
definitely another improvement, transients are faster, more separation and clarity

looks a bit chaotic

wonder if theres a tidier way to do it

Very good - it could be a result of using silver wiring or a larger gauge (lower impedance) for power carrying wiring?

I've no theoretical reason why but I've heard the difference silver wire can bring.

I think the quality of this tweaked DAC is now revealing these small improvements which otherwise remain hidden by noise & other distortions in "normal" DACs

It might be interesting to hear if 16 separate ground wires to the shift regs makes a difference & then if twisted wiring is any different