|
|
Post by beatstick on Mar 24, 2021 5:16:58 GMT
While working on a custom module, I noticed that on my AE, GND on the bus cable has high enough impedance that it's causing GND to "sag" and ruining BUS CV.
Using a multimeter I measured ~0.3 ohms of resistance between the MASTER's GND female socket and GND on a bus cable. This creates a big issue. As an example scenario, imagine that the system is drawing 200mA, and BUS CV is 4V at MASTER. BUS GND will end up "sagging" by 40mV, causing BUS CV to appear as 3.96V at the modules. That's only 1.7mV away from being a completely different note!
There's only a margin of +-41.7mV around each note (1000mV / 12 notes per octave / 2).
This is at 200mA load. If the load increases the problem gets worse. (And yes, the example scenario above is exactly what I'm measuring in real life on my AE: 3.9642V on BUS CV, 4V at MASTER).
I could compensate for this using the CV trim on the MASTER module, or calibrating each module, but that poses two problems: 1) The CV female header on MASTER won't match BUS CV anywhere else. 2) It won't help if the load varies at all. Besides all that, GND sagging isn't exactly ideal in general.
Is this normal? Is there supposed to be that much impedance on BUS GND?
I'm going to keep poking around my AE to see exactly where that impedance is coming from, but would appreciate feedback. It probably isn't the bus cable itself, since that's only 0.2Ohms per meter. I'm suspecting there's something weird with that board that the bus cables attach to. |
|
|
|
Post by keurslagerkurt on Mar 24, 2021 7:42:33 GMT
Hmm, interesting, and pretty much why I haven't touched DIY'ing anything that needs precision or V/Oct: too little room for errors!  When I'm home I'll try measuring how much Ohm's I get. Although with such rather small resistances, my method of measuring (IE put a cable in the bus somewhere and touch it with my multimeter pin) might introduce a too high error with contact-resistances and such? |
|
|
|
Post by beatstick on Mar 24, 2021 21:40:29 GMT
Thank you!
Yeah, measuring resistances that small is challenging so I've switched to just measuring voltages. Here are some new measurements on my system:
Everything relative to MASTER's female socket labelled GND, MASTER in top row, 1st and 2nd horizontal units. EX330 multimeter in autoranging mode.
CTRL1 I/O shield: 0mV
BUS GND from top external connector: 18.8mV, 18.8mV BUS GND from where MASTER's bus connects to the splitter board: 18.8mV, xxx
BUS GND from 3rd horizontal unit on top row: 22.9mV, 23.9mV BUS GND from 16th horizontal unit on top row: 25.5mV, 25.5mV BUS GND from 7th horizontal unit on bottom row: 31.6mV, xxx BUS GND from bottom external connector: 24.5mV, 24.5mV
I tried to measure the GND connection at the MASTER's bus header, but it's quite trickery with the cable attached. I got ~1mV.
Because I only saw ~1mV at the MASTER's header, the impedance probably isn't in MASTER itself. But yeah, I can see impedance coming from the MASTER's bus cable, the other bus cables, and the splitter boards as well.
I guess those bus cables just aren't high enough gauge for this application? At 26 or 28 gauge they should only be 0.2Ohms/m, divided by two because there's two GND wires. *shrug*
|
|
|
|
Post by lukylutte on Mar 25, 2021 6:14:42 GMT
This could explain why I can us two expression pedal/CV generator at once on one system? Maybe not related at all, so sorry in advance  I modified 2 expression pedals to generate CV From this discussion.However even if I plug each one from an other module into two different one. Both pedals affect both modules. Example: Exp 1; +5v & grd from Master module - VCO pitch Cv Exp 2; +5v & grd from USB power module - LPF CV If unplug one the other works without problem but If if both are plug it's just madness where both affect both so become not useful... |
|
|
|
Post by keurslagerkurt on Mar 25, 2021 9:16:16 GMT
Thank you!
Yeah, measuring resistances that small is challenging so I've switched to just measuring voltages. Here are some new measurements on my system:
Everything relative to MASTER's female socket labelled GND, MASTER in top row, 1st and 2nd horizontal units. EX330 multimeter in autoranging mode.
CTRL1 I/O shield: 0mV
BUS GND from top external connector: 18.8mV, 18.8mV BUS GND from where MASTER's bus connects to the splitter board: 18.8mV, xxx
BUS GND from 3rd horizontal unit on top row: 22.9mV, 23.9mV BUS GND from 16th horizontal unit on top row: 25.5mV, 25.5mV BUS GND from 7th horizontal unit on bottom row: 31.6mV, xxx BUS GND from bottom external connector: 24.5mV, 24.5mV
I tried to measure the GND connection at the MASTER's bus header, but it's quite trickery with the cable attached. I got ~1mV.
Because I only saw ~1mV at the MASTER's header, the impedance probably isn't in MASTER itself. But yeah, I can see impedance coming from the MASTER's bus cable, the other bus cables, and the splitter boards as well.
I guess those bus cables just aren't high enough gauge for this application? At 26 or 28 gauge they should only be 0.2Ohms/m, divided by two because there's two GND wires. *shrug*
I will try measuring some voltages this weekend when I get my system out! (busy week rn) But I would think that when you measure another ground point with no module attached there that you should not measure any voltage? Like there should be no path at all between 5V and ground? So the voltages you measure have to come from the other modules somewhere right? I guess I'm just missing something important so these could be very stupid questions! |
|
|
|
Post by beatstick on Mar 25, 2021 21:19:30 GMT
No stress! I appreciate the help.
So, what I'm doing is measuring the voltage difference between the female header GND socket on the MASTER module (which I call MASTER GND) and GND at various points around the AE. For example, the voltage difference between BUS GND at an open module slot and MASTER GND.
This is a measure of how much GND is "sagging" throughout the AE. This sag is caused by impedance on GND, which can come from the BUS cables, the BUS board, and the MASTER module itself.
If I put a multimeter between MASTER GND and a BUS GND from some open module slot, then I'm taking a measurement on the circuit formed by all the little impedances on the path from that BUS GND pin back to the MASTER module. Here's a falstad circuit to illustrate:
www.falstad.com/circuit/circuitjs.html$ 1 0.000005 10.20027730826997 50 5 43 5e-11
v 96 288 96 112 0 0 40 5 0 0 0.5
g 96 288 96 304 0 0
r 96 288 192 288 0 0.69616
r 192 288 272 288 0 0.016
r 272 288 352 288 0 0.032
r 352 288 416 288 0 0.032
r 352 288 352 112 0 24.88
w 96 112 352 112 0
r 208 368 320 368 0 10000000
w 320 368 416 288 0
w 208 368 48 368 0
w 48 368 48 288 0
w 48 288 96 288 0
x 116 266 181 269 4 10 MASTER\sBUS
x 211 266 259 269 4 10 BUS\sboard
x 302 264 322 267 4 10 BUS
x 374 264 394 267 4 10 BUS
x 243 343 289 346 4 10 Multimeter
x 370 188 402 191 4 10 Module
x 114 190 155 193 4 10 MASTER
o 8 64 0 4098 0.625 0.00009765625 0 2 8 3 GND\sSag
38 2 0 0.008 3 MASTER\sBUS
38 6 0 1 200 Module\sLoadThe MASTER module is represented by a 5V source. The MASTER BUS cable, the BUS board, and pieces of the BUS cable, are approximated as resistors. An installed module is approximated as 25 Ohm resistor, causing ~200mA of load. Finally the multimeter, connected between MASTER GND and a BUS GND, is represented by a 10Meg resistor.
This is simplified, but illustrates where the GND sag is coming from. The installed module is causing ~200mA of current to flow through all these connections: MASTER BUS, BUS board, and BUS. You can inspect each of those resistors and see the voltages they drop. So it's easy to see that the multimeter is sensing the sum total of all those voltage drops. It's also useful to note that your intuition about the circuit was correct: the second BUS resistor which represents the open bus cable that we're measuring from, is not contributing to this voltage drop. The voltage drop is caused by the connected modules dumping their current through the GND path.
I added a slider to MASTER BUS so we can see that GND sag is proportional to the impedance of that (very important) path. I also added a slider to Module, so we can see how the GND sag is also proportional to load. The more power the modules are drawing, the worse the problem gets.
As for my AE, I tried replacing my master bus cable with jumper wires and saw an improvement of 10mV everywhere. So I suspect I might have bad IDC cables for some reason (my AE is secondhand). I've ordered some replacements so we'll see how much more improvement I can get by replacing all of them.
I'm still kinda worried that this is going to be a fundamental problem with the AE design, though. If I set the load to 500mA, the BUS board to 8m (roughly what it should be), the MASTER BUS to 16m (what we'd expect from an ideal 6" 28AWG IDC cable), and the BUS to 32m (12" of ideal IDC cable) we're still going to see 27mV of sag. That's not even considering a 3 or 4 row AE which would have much more impedance.
But I'm learning as I go here, so maybe I've got something wrong or missed something obvious.
|
|
|
|
Post by robertlanger on Mar 26, 2021 7:44:05 GMT
That's an interesting investigation; TBH I never made this measurements in this precision range. Don't forget, in the beginning, AE was meant as a modular noise machine, not even with 1V/Octave control; but grew in the direction of a serious modular then.
So this is super helpful to improve things for the future, I'm thinking especially about the upcoming precise VCO module.
Maybe the resistance of the connectors contributes to the overall resistance? Worth a check...
btw: The practical use of 3-4 row systems showed no noticeable impact :-)
For the simulation you chose 200mA for each module which is way more than the AE modules need; most modules are <10mA, some around 30mA
My 3 cents so far...
|
|
|
|
Post by robertlanger on Mar 26, 2021 7:50:59 GMT
Regarding oscillators and other precision CV related stuff: These should not be dependent on the supply voltage at all anyway; for this purpose, a local regulations/stabilization is highly recommended. I'm currently digging into the world of DC converters , especially for this reason ;-)
|
|
|
|
Post by beatstick on Mar 26, 2021 8:51:54 GMT
Oh man, thank you for popping in, robertlanger  I love my AE, and I hope my comments here didn't come off as criticisms! It's a ton of fun to play with. This is mostly just me learning, sharing, and trying to get my little custom module to work "perfectly". Such is the way of a digital engineer playing in an analog world. Of course; I was simplifying the simulation by using a single load resistor to approximate a rack of modules. Real life would be a bit more complex, with the load spread across the row's IDC. But the MASTER's IDC and the bus board will both see the full 200mA (or whatever the entire rack is adding up to), so those voltage drops are quite problematic.
Yeah, I'm not worried about the supply voltages; that's definitely each modules' responsibility. But GND sagging is a big issue for CV, since there's nothing an individual module can really do about it.
|
|
|
|
Post by keurslagerkurt on Mar 26, 2021 10:52:24 GMT
Very interesting discussion, thanks a bunch beatstick for the deep explanation and circuit example!! I understand what you mean now. |
|
|
|
Post by beatstick on Oct 28, 2022 23:20:31 GMT
Small update.
As an experiment, I designed a "Power Plane" replacement for the ribbon cable (see images below). It's a long, two layer PCB that can slot right in where the ribbon cable would be, with cutouts for the hex standoffs, and the usual 2x5 headers. By using a PCB I can dramatically reduce the resistance on the power lines, enough so that the GND sag shouldn't be a problem any more.
To verify I ran two tests. In the first test I used my bench to supply 5V on one side of the board, and put a 440mA load near the other end. Then I measured the difference between a GND at each end. On a ribbon cable the difference was 16.64mV. On my power plane it was only 4.19mV. See images below.
For the second test I used a MASTER module to supply power. Otherwise the setup was the similar. 24.7mV vs 9.77mV.
The test isn't perfect, but it was just set up quickly to double check my math.
The target is to keep GND sag well below 40mV, which this Power Plane almost certainly does for typical loads. That's the good news.
The bad news is that it isn't very usable. I had originally intended for it to be a drop in replacement for a ribbon cable. It does mechanically fit, all the modules connect to it fine, and in fact installing modules with it is a lot easier than the ribbon cable. You can just slot modules right in, no more fiddling with the springy ribbon cable. But there was one fatal flaw in my design ... removing modules. There's not really any points on a typical module where you can grab to "pull" them out of a socket. With ribbon cable, you can gently grab a knob or something to drag them out enough to then disconnect. But the Power Plane is, well, a solid plane ... So you have to unscrew all the modules, remove the whole set of modules at once, and then you have access to pull modules out of sockets.
Oops.
So, I don't know, it's useful for my needs where I'm working with higher power, custom modules, and trying to wrangle useful CV signals. You can use it without screwing the modules in. My friend does that, and he found it useful for a having a "prototype" rack where modules can be quickly swapped around.
I might ponder over a redesign to somehow make it better.
|
|
|
|
Post by dizzeesatchel on Oct 28, 2022 23:26:55 GMT
Sounds like it would be plenty sturdy enough that you could just use screws on the far left and right of the rack, then it's only four to remove? Lifting out the whole row doesn't seem too bad to be honest |
|
pol
Wiki Editors
Posts: 1,277
|
Post by pol on Oct 29, 2022 9:13:20 GMT
Small update.
As an experiment, I designed a "Power Plane" replacement for the ribbon cable (see images below). It's a long, two layer PCB that can slot right in where the ribbon cable would be, with cutouts for the hex standoffs, and the usual 2x5 headers. By using a PCB I can dramatically reduce the resistance on the power lines, enough so that the GND sag shouldn't be a problem any more.
To verify I ran two tests. In the first test I used my bench to supply 5V on one side of the board, and put a 440mA load near the other end. Then I measured the difference between a GND at each end. On a ribbon cable the difference was 16.64mV. On my power plane it was only 4.19mV. See images below.
For the second test I used a MASTER module to supply power. Otherwise the setup was the similar. 24.7mV vs 9.77mV.
The test isn't perfect, but it was just set up quickly to double check my math.
The target is to keep GND sag well below 40mV, which this Power Plane almost certainly does for typical loads. That's the good news.
The bad news is that it isn't very usable. I had originally intended for it to be a drop in replacement for a ribbon cable. It does mechanically fit, all the modules connect to it fine, and in fact installing modules with it is a lot easier than the ribbon cable. You can just slot modules right in, no more fiddling with the springy ribbon cable. But there was one fatal flaw in my design ... removing modules. There's not really any points on a typical module where you can grab to "pull" them out of a socket. With ribbon cable, you can gently grab a knob or something to drag them out enough to then disconnect. But the Power Plane is, well, a solid plane ... So you have to unscrew all the modules, remove the whole set of modules at once, and then you have access to pull modules out of sockets.
Oops.
So, I don't know, it's useful for my needs where I'm working with higher power, custom modules, and trying to wrangle useful CV signals. You can use it without screwing the modules in. My friend does that, and he found it useful for a having a "prototype" rack where modules can be quickly swapped around.
I might ponder over a redesign to somehow make it better.
I like the idea of this, 2 main appeals - add/removing modules easily (I don't screw mine in as I can't decide on the "perfect" layout) and putting the sockets on like you have where there isn't a module so the bus CVs/voltages are available (more)  . If you can't pull a module off in the case, would something mechanical like a piece of wood support the PCB enough to make this work? He asks hopefully.... |
|
|
|
Post by rockysmalls on Oct 29, 2022 12:21:34 GMT
beatstick i like the idea a lot... personally i'd be happy with a small one of these strips with some kind of simple power input that just holds modules temporarily together on my desktop. the plug in/out wouldn't be a problem "sans case"
|
|
I love my AE, and I hope my comments here didn't come off as criticisms! It's a ton of fun to play with. This is mostly just me learning, sharing, and trying to get my little custom module to work "perfectly". Such is the way of a digital engineer playing in an analog world. 
. If you can't pull a module off in the case, would something mechanical like a piece of wood support the PCB enough to make this work? He asks hopefully....