Any chance you might have one or both capacitors disconnected?· Here are just some "quick" scope images I am looking at...· ( Note look at channel #2 on all images, keep in mind my input voltage supply is set at 10V)
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Post Edited (Beau Schwabe (Parallax)) : 3/14/2007 5:54:39 AM GMT
The·5V regulator (2940) on both have a capacitor on the output.· The difference is only a 10mA current drain on the output of the 3.3V regulator (2937).
With a cap on the output of the 3.3V regulator the signal was steady and I could place a current level up to·333mA (10 Ohm resistor) and the output
remains steady.
·
·
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Post Edited (Beau Schwabe (Parallax)) : 3/14/2007 6:01:10 AM GMT
Peter, what you are saying makes sense when the load is being switched but for him to get a high average voltage would suggest that it is not even oscillating around the set point. The load on the regulator is low on power up and then increases fairly linearly if the resistance is small to the maximum, does not seem much of a problem for the regulator to me.
May I suggest two possible courses of action, send a part or part number for the relay to Beau for him to try out. Or post some decent pics of the set up.
I have two of those relays… The only combination that I can come up with is an improperly connected or damaged capacitor on the output of the 3.3V regulator.
With the cap removed from the output of the 3.3V regulator and a coil connected across GND and the 3.3V output, I see a 6.7V pk-pk signal on the scope when
looking at the 3.3V output. Note: two coils in parallel also show a 6.7V pk-pk signal
No, I just can't reproduce it. I measure a steady 3.337V without a load of any kind, and then 3.335V with both coils energized. Removing a cap on the 3.3V output causes the voltage to go down from
4.27V without a load to 1.24V with both coils energized.
TL,
A picture is worth a thousand words, can you post a snapshot of your board in action?
What are your test points that you are taking the measurement across?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Beau Schwabe
IC Layout Engineer
Parallax, Inc.
I may be a little thick, too. But, what I don't quite understand is this: "The circuit is pretty simple: 6 propeller pins are connected to one side of six different reed relay coils, and the other side was connected to ground. This resulted in >4 volts between Vss and Vdd when one or more propeller pins were driven high, or when one of the wires was disconnected from the propeller and connected directly to Vdd"
To me, that sounds like you were sinking the coil current (other end connected to Vdd) and that is not what I understand from the rest of the posts. Also, if this Vdd is the 5 V Vdd, then it seems natural that current flowing from +5V through coil can lift the Prop's +3.3 V when the pin is off or high.
So, my question is: Do you supply the coils from +5 V or even +9 V? Or do you actually connect to Vss (gnd) as some other posts indicate?
Added later:
OK, I think that I misread your posts. Anyhow - a thorough check of wiring wouldn't hurt. Do it with an ohm-meter (a low volt one) and see if there's a bad Vss connection or something.
Post Edited (Skogsgurra) : 3/14/2007 10:27:07 PM GMT
No grounding problems, Skogsgurra. That was the first thing we checked. I soldiered together an octopus and jumpered over the ground rail on the breadboard. Nothing.
No bad cap connections, Beau. I moved them out of the traffic jam at the top of the breadboard, and everything was still wonky. BTW, there is a 0.47 uF cap on the 9Vdc input as instructed on page 6 of the PEK setups docs at this link: www.parallax.com/dl/docs/prod/prop/PE-Lab-Setup-v1.0.pdf I left that off the schematic because I was tired.
Peter: The LED already does come on when the coil is energized by bring one of the propeller pins high.
Misc: The 5Vdc regulator outout is rock-solid. All the problems are happening on the 3.3 volt side.
Originator: I'm going to order some caps from digikey to implement the revised caps on the output, and also some extra voltage regulators in case one of the ones I have now is confused. And I'm also getting a pile of optoisolators and solid state relays to replace the coils. As an ME, of course, I prefer the reed relays because (1) they're mechanical, (2) I can think of them as valves, (3) they make a charming little "ping!" sound when they open or close, and (4) they're braindead simple to hook up. SSR's and optoisolators have to be treated more like "real electronics" and require a few* extra braincells. (*just two, but I don't have all that many to begin with)
Also, the schematic was laboriously created using Microsoft Word's graphics editor. Not recommended, but it's what I've got. That's why I didn't fix it when I discovered the 0.47 uF cap missing from the illustration. Used to have Visio Pro with all the electrical symbols, which I never needed until now, but it disappeared when IT reloaded my laptop hard drive.
I think we might have some SMD caps lying around here with about the right rating, but I'd need a microscope to find them. Just curious, but what are ratings of the caps in the propellor project parts kit? I couldn't find any parts list for it. ... OK, I found a meter with a cap reader, and it looks like 3 tantalum 100 uF, and a bunch of little blue ?0.5 uF? The meter didn't like the little ones. OK, back to digikey.
Found a 22uF electrolytic in the lab. Turns out we have a big Jameco display box up on the pegboard, but they're all electrolytic. Per the LM2940 spec I installed it across the 5Vdc output. Getting cramped in there.
Didn't help.
Originator: Do you suggest I should replace the 1000uF across the 3.3 Vdc output with a 10uF ?
I always thought that dusty old Jameco display was just full of resistors and diodes and lots and lots of empty boxes. But lo! and behold, there are two rows of electrolytics in lots of sizes.
I replaced the 1000uF with the 22 uF. Didn't work.
So what the hell, I tried putting them in parallel. Click! 3.3Vdc. Hooked up 5 relays just to be sure, and still got 3.3 Vdc. Any thoughts on why?
Thank-a-you! Now I have to go home and see if my 5 month old still recognizes me.
PLUS, the one 0.47uF cap across the 9Vdc input to the LM2940.
PLUS, 22uF across the 5Vdc output, which is also the input to the LM2937.
PLUS, 10uF across the 3.3Vdc output. Actually stuck into the breadboard near the reed relays because there ain't no more room near the voltage regulators.
OK! I removed the two 1000uF caps and everything is still stable. And I was able to run 5 relays without a hiccup.
You were right, originator! Thanks again!
Peter: The LED already does come on when the coil is energized by bring one of the propeller pins high.
-TL
Sorry, I misread your schematic, it was a little confusing as I saw a line run off down to the bottom and another off to the top and that sort of thing can trick the eye. Have you tried the free software "Kicad" for schematic and pcb design? It looks good.
The point of the leds was not simply for indication but for allowing the reed relays to connect up to +5V and still be driven from the Propeller I/O directly. This is because the leds are going to drop 1.6V minimum and so the I/O pin will work fine and the relays get their 3.3 volts that they need.
The trouble with electros (in general) is that they could have come from anywhere and be years and years old. Quality varies considerably and where one electro works another brand fails or fails later on down the track. They are never really much of a problem when they are used for filtering the input side but a great many pcb faults are actually due to a dried-out electros, I know from fixing them. Which is another reason that I choose tantalums.
Doesn't matter. I figured it out and it is a major DUH moment.
When I tested the 1000uF caps to make sure they were connected properly, I jumpered from their original locations to their new locations. And everything behaved the same.
So then I got the circuit stablized with the new caps installed wherever I could fit them. And when it worked without the 1000uF caps, I moved the new caps to the positions the old caps were in. And everything went wonky again.
Hmmmm....
I hunted around and found a broken jumper (shown at row 9 on the Parallax pix) between ground and the 1000uF cap on the LM2940 input.
OK, so why does the board pass continuity tests? The EEPROM should be ungrounded, but in fact, everything worked fine, and I was able to run all sorts of test programs without a hitch. I never had a single problem until I hooked up an inductive load.
Turns out the ground is being carried thru the EEPROM from the WP pin to the other side, but with a large resistance, resulting in my idiotic problems. I assume this resulted in an apparently very large ESR, destablizing the circuit. Maybe one of you sparky types can explain it in small words so I can understand.
And yes, the board works just fine with the original 1000uF caps now that I replaced the jumper.
*sigh* Instead of going home in victory tonight, I go home in shame.
TL You shouldn't think in such terms. Most everyone has a starting point, and most everyone has a similar story. Very often those extremely frustrating phases produce valuable learning that carries over to the next project. I am greatly saddened though that my parallel capacitor theory still needs a lot more work after all [noparse]:)[/noparse]
BTW that formula was just having fun! I have never heard of such a formula.
Post Edited (originator) : 3/15/2007 1:07:11 PM GMT
The problem here is that someone started thinking along lines of transients and oscillatory behaviour. And then got carried away to the point where a simple check with a DMM wasn't good enuff any more.
Happens all the time and - as someone said earlier - it is part of the learning curve.
Thanks for sharing your little adventure with us. We all learned from it.
Not just this thread but many similar ones alway's show though that there are many people out there
trying to help and get things sorted...
All the issues raised were very real cases of what could possibly happen.
To some degree you would expect somebody to have checked all the usual obvious things .. like loose
connections and other stuff that is hard to diagnose via a FORUM...
Yep your right .. you never stop learning and as long as you live you will ...
Keep them coming...
Great to see you got sorted - the great thing about a problem like that - it will never happen to you again - another debug possibility has now been added to your repertoire !!
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ 'Necessity is the mother of invention'
Comments
Any chance you might have one or both capacitors disconnected?· Here are just some "quick" scope images I am looking at...· ( Note look at channel #2 on all images, keep in mind my input voltage supply is set at 10V)
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Post Edited (Beau Schwabe (Parallax)) : 3/14/2007 5:54:39 AM GMT
Compare these two images...
http://forums.parallax.com/attachment.php?attachmentid=45888
http://forums.parallax.com/attachment.php?attachmentid=45890
The·5V regulator (2940) on both have a capacitor on the output.· The difference is only a 10mA current drain on the output of the 3.3V regulator (2937).
With a cap on the output of the 3.3V regulator the signal was steady and I could place a current level up to·333mA (10 Ohm resistor) and the output
remains steady.
·
·
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
Post Edited (Beau Schwabe (Parallax)) : 3/14/2007 6:01:10 AM GMT
May I suggest two possible courses of action, send a part or part number for the relay to Beau for him to try out. Or post some decent pics of the set up.
Graham
With the cap removed from the output of the 3.3V regulator and no coil connected I see a 1.2V pk-pk signal on the scope when looking at the 3.3V output
(as indicated here http://forums.parallax.com/attachment.php?attachmentid=45888)
With the cap removed from the output of the 3.3V regulator and a coil connected across GND and the 3.3V output, I see a 6.7V pk-pk signal on the scope when
looking at the 3.3V output. Note: two coils in parallel also show a 6.7V pk-pk signal
With a cap in place, both coils operate as expected and the output is steady.
(as indicated here http://forums.parallax.com/attachment.php?attachmentid=45889)
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
When you look at these scenarios with a voltmeter, do any of them resemble TL's measurements?
Andy
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Andy Lindsay
Education Department
Parallax, Inc.
No, I just can't reproduce it. I measure a steady 3.337V without a load of any kind, and then 3.335V with both coils energized. Removing a cap on the 3.3V output causes the voltage to go down from
4.27V without a load to 1.24V with both coils energized.
TL,
A picture is worth a thousand words, can you post a snapshot of your board in action?
What are your test points that you are taking the measurement across?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
To me, that sounds like you were sinking the coil current (other end connected to Vdd) and that is not what I understand from the rest of the posts. Also, if this Vdd is the 5 V Vdd, then it seems natural that current flowing from +5V through coil can lift the Prop's +3.3 V when the pin is off or high.
So, my question is: Do you supply the coils from +5 V or even +9 V? Or do you actually connect to Vss (gnd) as some other posts indicate?
Added later:
OK, I think that I misread your posts. Anyhow - a thorough check of wiring wouldn't hurt. Do it with an ohm-meter (a low volt one) and see if there's a bad Vss connection or something.
Post Edited (Skogsgurra) : 3/14/2007 10:27:07 PM GMT
Out of 3 boards I had, one would often oscillate on the 5V supply with NO stamp plugged in. When a stamp was plugged in, that one usually worked.
Per the LM2940 spec, a 22uF cap is required on the output to maintain stability. On the Super Carrier the 1 uF can be marginally stable.
Sorry I've lost track of the board that was oscillating on the 5V, and I had replaced C2 with a larger cap on all.
Letsee, going backwards...
No grounding problems, Skogsgurra. That was the first thing we checked. I soldiered together an octopus and jumpered over the ground rail on the breadboard. Nothing.
No bad cap connections, Beau. I moved them out of the traffic jam at the top of the breadboard, and everything was still wonky. BTW, there is a 0.47 uF cap on the 9Vdc input as instructed on page 6 of the PEK setups docs at this link: www.parallax.com/dl/docs/prod/prop/PE-Lab-Setup-v1.0.pdf I left that off the schematic because I was tired.
Peter: The LED already does come on when the coil is energized by bring one of the propeller pins high.
Misc: The 5Vdc regulator outout is rock-solid. All the problems are happening on the 3.3 volt side.
Originator: I'm going to order some caps from digikey to implement the revised caps on the output, and also some extra voltage regulators in case one of the ones I have now is confused. And I'm also getting a pile of optoisolators and solid state relays to replace the coils. As an ME, of course, I prefer the reed relays because (1) they're mechanical, (2) I can think of them as valves, (3) they make a charming little "ping!" sound when they open or close, and (4) they're braindead simple to hook up. SSR's and optoisolators have to be treated more like "real electronics" and require a few* extra braincells. (*just two, but I don't have all that many to begin with)
Also, the schematic was laboriously created using Microsoft Word's graphics editor. Not recommended, but it's what I've got. That's why I didn't fix it when I discovered the 0.47 uF cap missing from the illustration. Used to have Visio Pro with all the electrical symbols, which I never needed until now, but it disappeared when IT reloaded my laptop hard drive.
I think we might have some SMD caps lying around here with about the right rating, but I'd need a microscope to find them.
-TL
Didn't help.
Originator: Do you suggest I should replace the 1000uF across the 3.3 Vdc output with a 10uF ?
Post Edited (originator) : 3/15/2007 4:15:36 AM GMT
I replaced the 1000uF with the 22 uF. Didn't work.
So what the hell, I tried putting them in parallel. Click! 3.3Vdc. Hooked up 5 relays just to be sure, and still got 3.3 Vdc. Any thoughts on why?
Thank-a-you! Now I have to go home and see if my 5 month old still recognizes me.
-TL
Post Edited (originator) : 3/15/2007 4:19:34 AM GMT
Post Edited (TL) : 3/15/2007 4:24:07 AM GMT
The two original 1000uF caps.
PLUS, the one 0.47uF cap across the 9Vdc input to the LM2940.
PLUS, 22uF across the 5Vdc output, which is also the input to the LM2937.
PLUS, 10uF across the 3.3Vdc output. Actually stuck into the breadboard near the reed relays because there ain't no more room near the voltage regulators.
--- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- --- ---
OK! I removed the two 1000uF caps and everything is still stable. And I was able to run 5 relays without a hiccup.
You were right, originator! Thanks again!
-TL
Sorry, I misread your schematic, it was a little confusing as I saw a line run off down to the bottom and another off to the top and that sort of thing can trick the eye. Have you tried the free software "Kicad" for schematic and pcb design? It looks good.
The point of the leds was not simply for indication but for allowing the reed relays to connect up to +5V and still be driven from the Propeller I/O directly. This is because the leds are going to drop 1.6V minimum and so the I/O pin will work fine and the relays get their 3.3 volts that they need.
The trouble with electros (in general) is that they could have come from anywhere and be years and years old. Quality varies considerably and where one electro works another brand fails or fails later on down the track. They are never really much of a problem when they are used for filtering the input side but a great many pcb faults are actually due to a dried-out electros, I know from fixing them. Which is another reason that I choose tantalums.
*Peter*
When I tested the 1000uF caps to make sure they were connected properly, I jumpered from their original locations to their new locations. And everything behaved the same.
So then I got the circuit stablized with the new caps installed wherever I could fit them. And when it worked without the 1000uF caps, I moved the new caps to the positions the old caps were in. And everything went wonky again.
Hmmmm....
I hunted around and found a broken jumper (shown at row 9 on the Parallax pix) between ground and the 1000uF cap on the LM2940 input.
OK, so why does the board pass continuity tests? The EEPROM should be ungrounded, but in fact, everything worked fine, and I was able to run all sorts of test programs without a hitch. I never had a single problem until I hooked up an inductive load.
Turns out the ground is being carried thru the EEPROM from the WP pin to the other side, but with a large resistance, resulting in my idiotic problems. I assume this resulted in an apparently very large ESR, destablizing the circuit. Maybe one of you sparky types can explain it in small words so I can understand.
And yes, the board works just fine with the original 1000uF caps now that I replaced the jumper.
*sigh* Instead of going home in victory tonight, I go home in shame.
-TL
We are all glad that you were able to solve your problem.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe
IC Layout Engineer
Parallax, Inc.
BTW that formula was just having fun! I have never heard of such a formula.
Post Edited (originator) : 3/15/2007 1:07:11 PM GMT
The problem here is that someone started thinking along lines of transients and oscillatory behaviour. And then got carried away to the point where a simple check with a DMM wasn't good enuff any more.
Happens all the time and - as someone said earlier - it is part of the learning curve.
Thanks for sharing your little adventure with us. We all learned from it.
Not just this thread but many similar ones alway's show though that there are many people out there
trying to help and get things sorted...
All the issues raised were very real cases of what could possibly happen.
To some degree you would expect somebody to have checked all the usual obvious things .. like loose
connections and other stuff that is hard to diagnose via a FORUM...
Yep your right .. you never stop learning and as long as you live you will ...
Keep them coming...
Ronald
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Necessity is the mother of invention'
Even if we do have a "trapezoidal capacitive inverse ratio" to worry about [noparse];)[/noparse]
Graham