AMB Laboratories DIY Audio

[linux-works]

15ma is not right. maybe something is limiting you.

the last time I looked at current, I had a dmm in series with my lab psu feeding a 3 term reg (317 or 350 style). the dmm was on the 10A setting. I saw peaks of 200 or 250ma (roughly) during relay switches. and during bootup, it was 1.5A or even 2.5A. bootup can be harsh but every 3 term reg I've tried seems to tolerate it well and once the startup surge is over (2 secs, about) its down to 70ma or so, steady. the 250ma peak is the most I've seen and its usually far lower than this.

your numbers don't agree with what I'm seeing. I'm not sure why.

as for test points, I would solder short insulated wires (1" or so) onto the pin1 and pin2 5v wires. use those 2 wires as test leads for grabbers or aligators. and take ground locally, too, when you 'watch' things. the middle 2 pins are your best ground for taking readings.

the logic 5v should be steady and insulated from the relay stuff if you have your own private 'alt' power.

as for grounding, try to avoid any snaking or daisy chaining. think if there is a more direct way from some gnd connection to the real star ground. chaining is convenient but could also cause sharing issues. for now, for debugging, go 100% star grounding. the lcd1 gets a private black or green wire to its supply (s25 #1). the d1 gets its own private wire to ITS s25 (#2). the lcd1 gets a private red wire to s25-1(+). the d1 gets its own red wire to s25-2(+). the s25-1 and -2 are tied ground to ground via a ferrite. and try to keep red or black wires to 1foot or less, from lcd1 or d1 to its regulator.

see if that gets you some progress.

[Mercator]

The current readings I was taking inline between pin 2 on the D1 and the power supply. The start up reading of 352 mA may be lower because it isn't seeing the load on the LCduino just on the D1 and D2 since I've separated them. I don't know if my meter is quick enough to register the peaks on the relays if they are very short. I need to find the manual on the meter, it's a Protek 6500, and see if I can record the readings rather than just watch the meter as I turn the volume up and down.

I'll try to work out a more direct grounding scheme. This was intended to be a learning experience after all. I know in the stock version all the connections go to the ground plane in the A10 board. I'll see if I can emulate that. I put the ground loop breaker on the D1 board but I guess that doesn't help the D2 if I don't stack them. In the end I may stack the boards and use the extra room in the enclosure to move the s25 in closer to the boards. All of that will take a while. I'll let everyone know if it works.

[linux-works]

I have avoided the ground-breaker feature on the d1/d2 boards.

to me, it adds more complications than it could solve.

OT but slightly related: I first learned the value of 'structured wiring' when I had some car audio installation work done many years ago. I asked the installer what his trick was in eliminating alternator whine and noise and he said to run a *separate* wire from the battery to each item. even if one item is a few inches away from the other, do NOT be tempted! yes, you have a thick bundle of wires that *seems* overkill and redundant, but he was right and that was a good wiring design. local grounds (not long wires) but *separate* wires, non-shared. sharing can cause problems. sharing is evil

that lesson really sunk in with me. non-shared lines that run 'all the way' to the source. yeah, extra wire and extra clutter but its a good philosophy to follow. at least start there and if you can cut corners later and not have it get worse, that's fine; but I do like to start with 'structured wiring' and that means dedicated lines for power and gnd, for each thing.

and also remember: the power line up to the regulator is not nearly as important as the *short wire* that comes after the reg, to the load. think about your wall wiring: it snakes all around the house and its length does not matter (much); but once you go into the regulator, each inch can 'cost you' some current and regulation ability.

hope those tips can be helpful.

[Mercator]

I've spent the last few days making changes to the wiring on my project. I've moved the Sigma 25 into the same box as the LCDuino, Delta 1 and Delta2 and it's being feed by a 15VA transformer. All of the boards are grounded to a common point and there is a 5 volt wire for each board from the S25. Now when I plug in the power cord everything boots up fine. The inputs and outputs all switch correctly. The only problem that remains is the 32 to 31 DB volume jump.

Is there anything else besides insufficient power that could be causing the volume problem? Is there something that I could be looking for that could be causing an excessive drain on the power supply? Short of adding another sigma 25 is there anything that could boost power to the boards? For instance I have the two halves of the D1 board joined by a ribbon cable, could that be a bottleneck? I feel like I'm almost there but that I'm missing something simple.

[linux-works]

pics?

glad that moving the s25 closer helped. kind of figured it would.

as a final resort, a software timing change could be made. it could be that you would benefit from a longer relay pulse. we trimmed it as short as we could to cut down on the step to step delay, but if you needed more time to let the relays all sink in, you are welcome to play with the timing constants. get yourself a usb/ftdi cable, the source code and edit file 'volcontrol_defs.h' and change values for the value CLICK_DOWN_DELAY. try increasing it by 20% or even 50%. worst case, just to see if it helps at all, try doubling its value. rebuild the image (using the gui, hit the 'play' button) then download the code to the lcduino and see if that gets all steps to 'sink in' via the longer delay.

another thing you could do is boost the voltage to just the relays. use a variable supply (like 317) and tweak it to 5.2v or so. use that to power the 'alt' side of all the d1's. use regular 5v for the lcduino and let the lcduino send that 5v into the pin1 power, not the alt pin2 power.

[linux-works]

I wonder if this means anything: but I'm not aware of any a10 motherboard builders who have had problems. if you use the s25 on that board, using its thickness of trace and pcb path - things just plain work. the trafo feeding the a10 system isn't super critical, from the bug reports I've seen so far.

the a10 is meant for a single d1 board and a single d2 board. but its powered by 'just' a single s25 board.

beyond 2 delta style boards and I would say you should use a 2nd s25.

[Mercator]

I tried the "final resort" last night. I tried altering the CLICK_DOWN_DELAY and, if I remember the variable name, the CLICK_UP_DELAY. I tried 5,7, and 10 ms with no effect but it was fun trying.

[Mercator]

Woops,I hit the post button too soon.

Looking more carefully at the behavior of the volume it looks like there are two discontinuities as I raise the volume. The sound goes up until about 55 DB and then drops suddenly to very low volume. It then rises steadily again to 32 DB and then jumps up to a much higher volume. Watching the LEDs as the volume goes up usually one of two cells will light with each change but at the two discontinuities 4 cells will light up. I'm guessing this indicates it's trying to switch 4 relays at once. I can see why that might tax the power supply but I don't understand the jumps in volume. Is it just that not all of the relays are switching? If only one of the four is misfiring maybe I just have a bad relay. If both discontinuities involve the same relay that could be a clue. I'm hoping that if I understand the process better I might be able to track down the bug in the system.

[linux-works]

you understand it, I think; there are some transitions that require 1, 2, 3 or even 4 relays to change state. they are all being changed at the same time and it could be too much for the psu to handle. I was hoping that giving them more latch time would help but if its not helping, you really do need more power (less current drop; wiring still?)

at one time, I had code that would only toggle 1 or 2 relays at a time. it was more 'respectful' to the psu but it was audible when you changed vol at those points. the delay was too long since it would hit a few relays, reach its limit, then do some more - all during ONE vol control change event. it was easier on the psu but more 'jumpy' in sound so we stopped using that algorithm.

btw, when you changed values in the code, did anything change at all? are you sure you were able to edit the right values, compile and download the code? try giving that variable a much larger value just to be sure your code update procedure is working. try giving it a value 5x higher. you should really hear it being more clicky. if not, then something is not working in your edit/build/download procedure.

[Mercator]

Did anything change when I changed the code? That's a good question. I thought about that while I was changing things. I've debugged a fair amount of code in my time but never anything that involved a physical device. With software it was simple . You step through the code until it breaks and then fix it. I didn't detect any physical changes in the sound of the relays. They are always very clicky. I was listening for things to change more slowly but couldn't tell anything. This is what I'm editing. #define CLICK_DOWN_DELAY (10000). Is that the correct variable? If I increase it to 100,000 what should I expect to hear?