Headache

[Hunka Hunka Burning Love]

Season premiere ... long wait and still no noodity.    Even in "Marvel's Iron Fist" the two lead actors had a bit of an interesting go at it.    I think I might have saw part of a nipple, or maybe it was wishful thinking... 

Edited March 29, 2017 by Hunka Hunka Burning Love

[Hunka Hunka Burning Love]

Maybe this book might be helpful...

http://www.diystompboxes.com/pedals/schematics.html

[esaj]

Experimental motor-controller for the robot:

Or, just the board for it... I rushed on the design over the weekend and milled a couple of versions (rushed too, with slightly too high feed rates ). While the milling quality is OK (definitely not the best, but at least all the 0.5mm traces are intact and looking good), I played around using 60-degree bits, and didn't realize the ground plane got cut on a couple of places. Guess I'll need some ugly jumpers   There's some rough edges here and there, but I'd prefer not to sand this, as those thin traces can get easily removed in the process (been there, done that ), and jumping them afterwards might be problematic, as they're covered by SO16-chips.

Plus, I'm not 100% certain whether it actually works even if I make no mistakes 

Edited April 4, 2017 by esaj

[steve454]

4 hours ago, esaj said:

Experimental motor-controller for the robot:

Or, just the board for it... I rushed on the design over the weekend and milled a couple of versions (rushed too, with slightly too high feed rates ). While the milling quality is OK (definitely not the best, but at least all the 0.5mm traces are intact and looking good), I played around using 60-degree bits, and didn't realize the ground plane got cut on a couple of places. Guess I'll need some ugly jumpers   There's some rough edges here and there, but I'd prefer not to sand this, as those thin traces can get easily removed in the process (been there, done that ), and jumping them afterwards might be problematic, as they're covered by SO16-chips.

Plus, I'm not 100% certain whether it actually works even if I make no mistakes 

That is a piece of art.  I don't know what it does or what it is for but it is great.

[Hunka Hunka Burning Love]

Here @steve454 I'll put a face on the circuitry.  It helps to see what @esaj is up to.  

Secretly hoping he's ordered this on AliExpress for the robot shell actually...

Edited April 5, 2017 by Hunka Hunka Burning Love

[esaj]

At times I feel that I haven't learned much anything over the last about 1-1.5 years of fiddling around with this stuff... but a funny thing, last night while I was going through my workspace and cleaning up shelves, I started putting failed or otherwise obsolete circuits into a box. Just a quick picture of the of contents, missing a bunch of boards... from these, I've already stripped any easily reusable electronic and mechanical parts, like socketed ICs, some connectors, TO-220 heatsinks and such:

In total (all not shown in the picture) I have a good two-three dozens (at least 20-30, maybe more) boards that are basically trash now (either didn't work as planned from the start or were just some testing/otherwise became obsole), some of which represent tens of hours of work from simulations through breadboard testing (not always, I've sometimes been brave enough to just skip that, although it's not always a good thing ) to layout / board design to milling (not for the vero/strip-boards, of course ) to soldering and final testing. Not to mention all the stuff that never was actually built on boards, just tested on simulations and/or breadboarding. They, or at least some of them, could be seen as failures, but I (like to ) think they were just necessary steps in the way to learn this stuff, and the financial input on those isn't that much. In my personal opinion, the time invested has been worth it so far, although I'm still a very far cry from someone who really knows this stuff.

Of course on top of those, I've got a bunch of circuits and boards that work as meant and are currently waiting to be finished, mostly still waiting proper encasing (like the reflow-oven control-board & the SSR, I'm still slightly worried about the mains AC-part, although the SSR is self-encased and the rest is just mechanical wiring ), or already encased and in use (like the continuity meter, motor controller for the CNC, the ATX-power supply, the function generator, the battery powered PSU, calibration unit for the desk-multimeter etc...) I won't list them all here, those of you have had the interest and patience to read my ramblings probably are already familiar with them, or they weren't "important" enough to write about.

I had a talk with a friend tonight about the current guitar pedal circuitry, although, he didn't exactly know much about electronics, but tried the pedal built into a breadboard and really liked it (especially the fact that it has the "softness" control and doesn't go all the way into "full distortion" when not needed), and was really interested after I explained what I still want to improve... looks like I'll have to build at least two or three pedals (three if I want one for myself  ) , barring of course that I get around to finish the design   

Edited April 9, 2017 by esaj

[Rehab1]

9 hours ago, esaj said:

I've sometimes been brave enough to just skip that, although it's not always a good thing ) to layout / board design to milling (not for the vero/strip-boards, of course ) to soldering and final testing. Not to mention all the stuff that never was actually built on boards, just tested on simulations and/or breadboarding. They, or at least some of them, could be seen as failures, but I (like to ) think they were just necessary steps in the way to learn this stuff, and the financial input on those isn't that much. In my personal opinion, the time invested has been worth it so far, 

Why do you have to live so far away? 

[esaj]

1 hour ago, Rehab1 said:

Why do you have to live so far away? 

Why does it matter?

[esaj]

I managed to encase the prototype distortion board for testing purposes, currently it has four different distortions based on different diodes for clipping (LEDs, silicon diodes, 1P60 "supposedly germanium-diodes, but could be just Chinese Schottkys" and a schottky diodes),  plus a fifth different type of distortion, if you count the op-amp overdrive. It looks like I'm coming down with fever, so I'll keep this short.

I recorded 5 tracks with different distortions on the computer, although I have very little experience with audio editing (using Audacity), so I just ended up laying the tracks on top of each other and cutting from one track to the next (and on some places, laying two tracks on top of each other)... the song is not the kind that plays well with heavy distortion, yet I left that there too, and the timing isn't very exact all the time (should have used a metronome ), so sometimes it feels like the song "shifts" a bit jumping from one track to another... probably should also have learned to use crossfades between the tracks instead of abruptly cutting from one track to the next . Maybe next time, if such comes. The end result could be probably called "somewhat annoying" at least at times , but the idea was just to show a few different types of distortions the pedal can make (actually, I didn't play around much with the soft-setting or tones in this one, so at least the 1P60- and schottky-tracks sound somewhat similar, although if listening to the original tracks, they do have their differences). Anyway, here it is:

 

You'll probably recognize the song...I also did some parts of "Fear of the Dark" earlier, but ran into troubles both playing it (I can't keep up with all the parts anymore, although I've played it a kazillion times in the past ) as well as mixing it, so just threw it away in the end.

 

Edited April 16, 2017 by esaj

[freetheoranges]

SoundCloud?

[Rehab1]

On 4/9/2017 at 8:41 AM, esaj said:

Why does it matter?

Just would be nice to have hands on advice on the acm project!

[Rehab1]

2 hours ago, esaj said:

You'll probably recognize the song...I also did some parts of "Fear of the Dark" earlier, but ran into troubles both playing it (I can't keep up with all the parts anymore, although I've played it a kazillion times in the past ) as well as mixing it, so just threw it away in the end.

Stairway to Heaven! A classic! Your distortion tracks sound excellent!

[Hunka Hunka Burning Love]

 Heeey that's quite good and cool effects too!

On 2017-04-09 at 6:41 AM, esaj said:

Why does it matter?

We could all come over to visit, jam, and check out all the electronics doodads!  Afterwards we can go head out for some Lutefisk!

[esaj]

57 minutes ago, Rehab1 said:

Stairway to Heaven! A classic! Your distortion tracks sound excellent!

 

46 minutes ago, Hunka Hunka Burning Love said:

 Heeey that's quite good and cool effects too!

It still has lots of details to work out, but it's getting there, slowly...

 

46 minutes ago, Hunka Hunka Burning Love said:

We could all come over to visit, jam, and check out all the electronics doodads!

Not that much to see, unless you like visiting junk yards in general

46 minutes ago, Hunka Hunka Burning Love said:

 Afterwards we can go head out for some Lutefisk!

I had to look up what the hell is lutefisk    Never had any myself, if you do, tell me if it's any good, looks disgusting 

 

[Hunka Hunka Burning Love]

I love junkyards!!!   Okay @Rehab1, we gonna take a road trip to Finland to visit!  Scratching lutefisk off the bucket list!  Maybe it's more of a Norwegian thing. Eh @Vik's?

Once you get your KS16, we can all go out for a roll in search of the closest KFC for some fried chicken!  

Although maybe you're not missing anything there... 

http://www.dailymail.co.uk/news/article-3948516/KFC-customer-finds-MAGGOTS-wriggling-inside-chicken-drumstick.html

https://www.quora.com/Why-is-there-no-KFC-in-Italy-Sweden-Norway-Finland-Luxembourg-Switzerland-Maldives

Edited April 16, 2017 by Hunka Hunka Burning Love

[Rehab1]

42 minutes ago, Hunka Hunka Burning Love said:

  Okay @Rehab1, we gonna take a road trip to Finland to visit

Road trip to Finland.... No way! Gassing up the G650! What's your nearest airport, YVR?  I am assuming @esaj's nearest port is either HEL or EFHK.

As for KFC and Marvin's maggot filled drumstick, it sounds like a MacDonald's hot coffee spill litigation gimmick! Anyway I prefer the food at Chef and Sommelier.  

Edited April 16, 2017 by Rehab1

[esaj]

1 hour ago, Hunka Hunka Burning Love said:

I love junkyards!!!   Okay @Rehab1, we gonna take a road trip to Finland to visit!  Scratching lutefisk off the bucket list!  Maybe it's more of a Norwegian thing. Eh @Vik's?

Once you get your KS16, we can all go out for a roll in search of the closest KFC for some fried chicken!  

Although maybe you're not missing anything there... 

http://www.dailymail.co.uk/news/article-3948516/KFC-customer-finds-MAGGOTS-wriggling-inside-chicken-drumstick.html

https://www.quora.com/Why-is-there-no-KFC-in-Italy-Sweden-Norway-Finland-Luxembourg-Switzerland-Maldives

We might need a "bit"  bigger batteries considering that map to reach the nearest... Maybe we can haul a trailer with extra packs behind?   Or just take Rehab's private jet.

 

26 minutes ago, Rehab1 said:

Road trip to Finland.... No way! Gassing up the G650! What's your nearest airport, YVR?  I am assuming @esaj's nearest port is either HEL or EFHK.

JYV / EFJY, about 25km north

[Vik's]

10 hours ago, Hunka Hunka Burning Love said:

I love junkyards!!!   Okay @Rehab1, we gonna take a road trip to Finland to visit!  Scratching lutefisk off the bucket list!  Maybe it's more of a Norwegian thing. Eh @Vik's?

Totally norwegian and it is disgusting

[Vik's]

18 hours ago, esaj said:

JYV / EFJY, about 25km north

Just whats @Hunka Hunka Burning Love needs with a 9b1 battery! 

[esaj]

Experimental motor-controller for the robot:

Or, just the board for it... I rushed on the design over the weekend and milled a couple of versions (rushed too, with slightly too high feed rates ). While the milling quality is OK (definitely not the best, but at least all the 0.5mm traces are intact and looking good), I played around using 60-degree bits, and didn't realize the ground plane got cut on a couple of places. Guess I'll need some ugly jumpers   There's some rough edges here and there, but I'd prefer not to sand this, as those thin traces can get easily removed in the process (been there, done that ), and jumping them afterwards might be problematic, as they're covered by SO16-chips.

Plus, I'm not 100% certain whether it actually works even if I make no mistakes 

 

I finished assembling this new controller for the self-balancing robot this weekend, but all didn't go as planned:

In that picture, the other side is still missing the mosfets and the motor phase-wires, but everything else is there. Didn't post a schematic, but basically there's just a Cd4053 analog switch IC between the Arduino PWM-pins and other pins selecting the motor direction (actually which side of each half-bridge gets the PWM-signal), that's connected to a HIP4082 H-bridge mosfet gate-driver (two of each, because there are two separate H-bridges for the two motors). The gate driver is the thing that actually switches the mosfets on and off, and handles raising the voltage high enough on the high-side mosfets (although it needs an external diode and a capacitor on each high-side), so the component count is much less than on the other designs I had before. Rest of the components on the driver-side are just capacitors, resistors and diodes, limiting gate currents and protecting from high-voltage spikes and such. On the right side of the board, there's an Arduino Nano (on the other side of the boad), a 5V regulator and some bypass capacitors. The ugly jump-wires are there because I needed to connect the ground-plane around 

I wrote a small software for testing that the motor drivers work as intended, simply running the motors back and forth, and it worked fine in the test bench:

Not too much space for the heatsinks again, so I had to bend them a little (the heatsinks can't touch or the drains of the mosfets get shorted together ). They didn't heat up enough for me to feel any heat when touching them, and quick measurements showed that things should work as intended.

The breadboard with pots for tuning the PID-values and the battery (a 3S LiPo, mostly because the HIP4082 gate-drivers need 10V minimum) plus some wiring is still missing at this point. The wires going to the middle-"shelf" are for the MPU-6050 -gyro/accelerometer on a break-out board, that I've hotglued in place.

I got it running and was trying to get the PID tuned for it to stay balanced for maybe about an half an hour or slightly more, until suddenly... *SNAP* and a puff of smoke comes from under the board. The Arduino Nano still stayed powered (don't remember if the other motor still kept running also), so I was pretty sure it was the gate-driver IC that gave up.  The LiPo was behind a 5A fuse (it's the RC-kind with separate balancing connector, so the battery itself has no protections whatsoever ), so no need to panic.

After dismantling everything and removing the board, it was actually a CD4053 analog switch (I've used them to select the high/low -side mosfets in the H-bridge, plus they also saved my ass since I couldn't run the traces in single layer directly to the HIP4082 without using a ton of jump-wires) that had burned a small hole in the middle of the case. There was about 6 ohms of resistance between the power supply-pin and the ground of the chip, so it was pretty clear what had given up.

Not sure why the chip blew up though. It's only using the 5V power line (shared with the Arduino and the CD4053's, coming from an LM1117 linear regulator), has it's own 100n bypass and the input/output pins are only connected logic-level input/outputs (and the unused ones are grounded), so there isn't much current going there and the voltages shouldn't bounce around a lot (CD4000's should be good for up to around 20V). I suspect I may have damaged it during assembly (overheating during soldering or ESD), but can't be sure. Once I get around to replace it (not tonight anymore, need to head for bed ), I'll see if it does it again, if it does, there's something wrong with my circuitry in general... The other motor has the same circuitry (with slightly different layout though), but that still seems to be working.

I'm not even pissed, as this thing was practically designed off the top of my head without too much breadboard testing, pretty much just skimming through the datasheets of the chips, so the fact that it even worked in the first place was nice

 

 

 

Edited April 23, 2017 by esaj

[esaj]

The large capacitors for the spot-welder bank finally arrived... although about three months later than they were originally supposed to.

The Colt-lighter, AA- and AAA-cell are there just to give some scale, these things are larger than what you usually see, not to mention heavier. I ordered 12 pieces, although I used 10-cap bank (470,000uF) in the original simulations. Now, I just need to find the time & energy to finish the charge-circuitry and to build the actual bank...   Oh, and the piece of copper clad with burn marks on it, on the upper left corner of the picture is a piece I've used for testing on the smaller-scale prototype   Although much lower capacity/current/energy (at 38,200uF, it held less than 1/10th of charge of the "final" bank), it was already enough to burn the copper off from a small area at 12V. Made some welds too, but as I had no actual control circuitry for the discharge, it tended not to weld very well (either burning a hole/cutting the piece to weld or not welding at all)   The final thingamabob's supposed to be microcontroller-driven, so I can control the welding time precisely.

A picture of a (poor) weld on some earlier testing.

For carrying the high currents and keeping the circuitry resistance otherwise low, I have some 16mm² silicone-cable where connections need to be made (that's about 5AWG?). I also recently came upon some copper-shunt blocks or ground rails or whatever they're called used in household mains cabinets (basically solid brass or copper blocks with some screws and holes for attaching cabling). Those might come in handy. Originally I planned to solder the capacitors directly into a 105um copper-clad (3 times thicker copper-layer than usual boards), to make a thin but wide (100mm) conductor, but I might run into problems fitting all the capacitors around it (and the whole thing's supposed to go into a box, in case the caps should blow up ). If that won't work, I'll have to solder the legs directly into the thick cable. For the welding prongs, I planned to use old soldering iron tips, although pure copper would probably work better, if I can find something suitable.

 

The circuitry is otherwise designed (and partially even tested ), but I've still got some fine-tuning to do on the charge-control side. The 12V and 5V voltages come from a computer ATX-power supply, so I won't have to worry about that.

The actual gate-driver has more push-pull -drivers in parallel to really kick some current into the gates. This is the stage driving 5 x 75NF75's:

Nevermind what the Rise in the stats says, the trigger is positioned so that it measures only the ringing... Plus I don't think I had any actual load on the mosfets here. Anyway, from memory, the actual risetimes were something like 150ns or thereabouts. Soo.. hopefully the mosfets don't blow up when the pulse comes even with larger currents (the momentary power dissipation can hit several hundred watts per mosfet, the slower it rises, the worse the heating)   Probably need something like a 15V transil/zener there to prevent the gate voltages overshooting too.

Edited May 3, 2017 by esaj

[esaj]

Got around to finish the charging-side design and building the board. Or actually two boards.

This is the pretty much finalized design (D2 is actually SK510 5A Schottky on the real board, D8 is STTH4R02B "ultra-fast" recovery diode, C14 is left out from the board). The fuse resistance was measured in testing, when I was finding out why I was getting much less current (like about 2/3rds) from the actual board than on the simulation, which turned out to be that I had forgotten to take the fuse resistance into account 

The one on the earlier pictures was overly complicated when it came to the gate drive, so I simplified it down a bit (it could be made even simpler though, but I need that ENABLECHARGE -input, so the microcontroller has control over the bank charging). The basic idea is to measure the voltage drop over the R9 sense-resistor (a 100 milliohm 5W ceramic in the board, although measurement showed it's closer to 105 milliohms in reality), and compare the voltage drop (caused by current) to the constant voltage divider-value from R10 & R11. LM358 requires the inputs to be kept at least about 1.5V away from the VCC (12V), so I've used voltage dividers for that. I made an adjustable version at first, but decided to make the latter board with constant value.

The maximum current the source gives out is determined simply by Ohm's law (I = U/R). The U (voltage) is the voltage drop over the sense resistance and R is the sense resistance (fuse + sense resistor) itself.

Maximum output current = voltage drop over sense resistance / sense resistance. Knowing the wanted maximum output current (say, 4A), the voltage drop needed over the resistance of fuse and sense resistor (let's just round it to 170 milliohms here) then would be:

U = R*I  =>  U = 0.170ohm * 4A = 0.68V

With input voltage (Vcc) of 12V, the voltage against ground is 12V - 0.68V = 11.32V. If the op-amp would allow rail-to-rail inputs (ie. the inputs can go as high or as low as the Vcc and ground), it would be enough to just take the other input straight from between the sense resistor and the P-channel mosfet, and set up a suitable valued divider to feed 11.32V on the other op-amp input. But, as LM358 isn't a rail-to-rail input op-amp, the voltage needs to be divided down further. Basically what is wanted is

(Input voltage - sense resistance * maximum current) *  R14_R21 -divider ratio = input voltage * R10_R11 -divider ratio.

Sense resistance * maximum current is the value calculated above (0.68V). I haven't taken input offset voltages and such of the op-amp into account, as not a real precise output current value is needed. I solved this by picking (pretty much arbitrarily) the divider ratio for the R10_R11 -pair first and calculating the resulting voltage:

Voltage divider output = voltage divider input * lower resistor / (lower + upper resistor)

12V * 180k / (180k + 51k) = 12V * 0.799221 = 9.35065V

Far enough from the VCC for the op-amp input. Then, knowing that, the ratio for the R14/R21 -pair can be solved:

(Input voltage - sense resistance * maximum current) *  R14_R21 -divider ratio = input voltage * R10_R11 -divider ratio

<=> 

R14_R21 -divider ratio = (input voltage * R10_R11 -ratio) / (input voltage - sense resistance * maximum current)

Basically it becomes the voltage from the R10_R11-divider divided by voltage from behind the sense resistor. Using the actual values calculated above

R14_R21 = (9.35065V) / (11.32V)  =  0.826029

The input for the op-amp should still be far enough from the VCC (about 10.5V at highest) even when no current is flowing and the voltage at the sense is at input voltage, how much does it give? 12V * 0.826029 = 9.91..V. Far enough. So, the voltage divider ratio should be 0.826029. Arbitrarily picking the lower value of 200K, the upper resistor should be

ratio = lower resistor / (lower + upper resistor)

<=>

upper resistor = lower resistor / ratio - lower resistor

=>  200k /  0.826029 - 200k = 42.1222k

The closest I had was 43K, so I went with that... Well, in reality I tried numerous different lower-resistor values to find a pretty close match from standard E24-values, before ending up with that, but anyway... . Of course it could be done the other way around (pick a divider for the R14_R21 -pair and then calculate the divider for the R10_R11 -pair). Or just use graphs or spread sheets or whatever works for you   There are probably neater ways to calculate these, but this is just how I went along.

Since the value is slightly "off" (42.1222k vs. 43k), it will give a bit different current, about (12V-9.35V / (200k / 243k)) / 0.17ohm =  3.76A, and also real world effects of the circuit will come into play with the real thing. But for me, this was close enough.

The actual way it works with the op-amp in this kind of circuit is fairly simple. When the current rises high enough, the voltage drop over the fuse & R9 will cause the negative input of the op-amp to (try to) drop below the constant voltage from the R10 & R11 -divider (the threshold value being calculated above). The op-amp will react by changing it's output, which controls the voltage at the mosfet-gate, starting to limit the current by forcing the op-amp conductivity down (the op-amp resistance/conductivity is controlled by the gate voltage). Pretty basic feedback. For real world application, note the capacitor C16. It's there to stabilize the op-amp, otherwise it will begin oscillating around the set value as it "overshoots" the corrections in both ways (phase shifting/delay). 100n was chosen as value pretty much based on simulations, in the real board it produced a stable enough output:

The higher repeating spikes are likely switching noise from the atx-supply giving the "ENABLE CHARGE"-signal, didn't measure them further as they seemed to have little to no effect. The main thing is that the gate voltage doesn't oscillate a lot, which would cause the output current to oscillate around the set value and not really stay (more or less) constant.

 

Two versions of the board. The one on the left is the earlier with adjustable output current (the trimpot seen near the upper left corner), the right one is the final version.

I did the first version yesterday, but made a stupid blunder: the two inputs of the op-amps are reversed as Kicad's amplifier-symbol has them in reverse vs. LTSpice. My own fault, should have been more careful, I was just copying the circuit from LTSpice to Kicad schematic without thinking it much...   that lead to a nice few hours of debugging the "oddly behaving" circuit   To test it further, I had to build a sort of an adapter that connects the two of the inputs correctly.

The board on the right side was built today, there's no trimmer anymore to adjust the current. The actual board gives a little over 3.5A constant current, whereas in the simulation it's about 3.76A (pretty much spot-on with the calculations above). Close enough for usage, probably it's just the tolerances of the real components, values changing when things warm up, input voltage not being completely "stiff" (ie. it may drop with load) and such

During testing, I used a 0.3ohm 50W power resistor as load, which is a bit troublesome, as the actual current would drop as the capacitor bank gets charged, but with the resistor the current stays at the maximum value as long as the output is enabled. I had to run the tests in short intervals, running the circuit maybe 5-10 seconds at a time, because the mosfet gets really hot fast. Consider that at 3.5A, the 0.3ohm resistor is dropping about 3.675V, the sense resistor + fuse take about 2,0825V. Other parts in the circuit have some drops too, let's round them to neat 6V in total. This still leaves 6V to drop on the mosfet, and with a continuous current of about 3.5A, it comes to a total of 21W of power to dissipate. The datasheet states that the junction-to-sink thermal resistance for greased flat surface is typically around 0.5 degrees Celcius / W. The heat sinks are from Aliexpress, so I don't have any datasheet for them, but similar looking heatsinks typically have about 1 C/W -thermal resistance from the sink to ambient. With a total thermal resistance of 1.5 C/W, 21W power dissipation should raise the junction temperature to 31.5 degrees above ambient, which doesn't sound that bad, but with a (cheapo) infrared thermometer, I got values close to 100C at the heatsink after running the circuit for a while (10 seconds or so?), so probably it could actually get much higher, especially as the ambient temperature around the sink also starts to rise. In the real usage, the circuit is only run for a couple of seconds at a time when the bank charges (even less when the bank isn't completely discharged), so this shouldn't become an issue. It if does, I'll need to add some more cooling. Mosfet with a lower Rds(on) won't help, as the actual minimum resistance won't matter (the mosfet always has to drop enough voltage to keep the current on at the set point if it tries to raise higher).

 

The schematic from KiCad. If you compare the U1A to the schematic in LTSpice, you'll notice that the + and - inputs are in reverse  Also, LM358 is a dual op-amp, but I only needed one op-amp in this circuit (yeah, could have used a single opamp too, but I have a ton of those LM358's laying around), so the B-unit (U1B) is just terminated ( see for example http://www.electronicproducts.com/Analog_Mixed_Signal_ICs/Amplifiers/Properly_terminating_an_unused_op_amp.aspx ).

 

Layout from Kicad.

Copper-sides of the two boards in comparison, older one on the left, new one on the right. The first version didn't have any mounting holes, so I added those. I also removed some extra copper (wasn't really necessary, but just wanted to try ). Had a text there, but most of it was gone with the copper removal too . The layout of most of the surface mount -components had to be redone, because the op-amp inputs needed to be reversed, so I did it pretty much from scratch in that part. The never board is still missing the transient suppressor (or just zener), I don't think it'll be needed here. Another empty place near the top-right corner is for an optional capacitor to further filter out power supply noise from the voltage divider setting the maximum current, if needed.

The large D-PAK -component (that one was a bit hard to solder with "normal" cheapo soldering iron) is a 4A / 16ns ("ultra-fast", have to love the marketing bullshit in the datasheets, it's not just fast, it's ultra fast  ) recovery diode, used as a flyback/freewheeling diode in case there's a negative voltage spike hitting backwards from any stray inductances when the bank is charged or discharged. I still need to figure out what diodes and where I need to put on the discharge-circuit to prevent this. Normally, the small stray inductance won't matter much, but when you have a fast high current pulse, even small inductances can cause severe back-EMF, and I need to check how to protect the discharge mosfets. Also, while there are equations and calculators available to calculate parasitics for wires and such, I have no idea whether the 32 x 0.5mm² conductors inside the cables should be treated in such calculations... as a single 16mm² -conductor, although the surface is not exactly round, or what. Probably should just try different methods, and pick the "worst case" (highest inductance) and go with that. Also I have no idea on the inductances of the other parts of the circuit...

That should cover the bank charging-side. I still need to finish building the actual bank (I'm still wondering some details there on how I should attach the capacitors, reverse voltage protection, flyback, self-discharge...), as well as the actual final discharging circuit (bunch of mosfets with some heavy duty heat sink(s), protection components and a big push-pull -array to switch the gates), and the controller (basically an Arduino Nano + some extra components and software). I made some prototypes of those too earlier, waiting for the caps to arrive, but probably still want to think them further. The footswitch I already have, as well as a couple of extra ATX-supplies (I mean to power this thing from the 12V -line) and an old computer case.

 

 

 

 

Edited May 7, 2017 by esaj

[Rehab1]

2 hours ago, esaj said:

The large D-PAK -component (that one was a bit hard to solder with "normal" cheapo soldering iron) is a 4A / 16ns ("ultra-fast", have to love the marketing bullshit in the datasheets, it's not just fast, it's ultra fast  

Strong work!!

I am sure there is a disclaimer on the box that states: "This soldering iron is not approved for  esaj"

[esaj]

5 minutes ago, Rehab1 said:

Strong work!!

I am sure there is a disclaimer on the box that states: "This soldering iron is not approved for  esaj"

  I think it comes mainly down to lack of experience of soldering components like that, the DPAK -package has a flat metal back:

I don't know if the back-plate actually needs to be completely soldered on the board, it might be enough that it makes contact with the board (for conduction and heat dissipation), might be enough just to solder it from the side at the tip of the backplate that remains visible. I ended up spreading a thin layer of solder on the pad, then pressing down the package while heating the pad from the side for a while, and then removing the iron and keeping the component pressed until the solder cooled down and solidified again...

As for the iron, the reality is that I'm still working with the same cheap CXG 936d (about $20 ) I bought over a year ago:

For most work, it's just fine, although I definitely wouldn't call it "high quality", it's not exactly a good sign that AC mains voltage goes directly into the handle and when using high temperatures (400C and above), the handle starts to warm up. Not sure if the plastic parts might start to melt if held at high temperature for a long time, I've had it at around 310-350 for hours, at least that hasn't been too much, but I had another iron where the part near to the tip started to melt after a while when set at 450. Also I've run into many situations where a hot-air station would make things just sooo much easier, especially removing all kinds of SMDs and soldering small-pitched ICs... Probably should just give in and order a proper station 

 

[Hunka Hunka Burning Love]

I keep telling ya... cheapo $60 hot air rework station!    I use mine all the time,  and I haven't even had a use yet for the hot air side....