Headache

[Rehab1]

13 hours ago, esaj said:

The industrial look is nice, and clearly it's directed to people like construction workers etc, as it should withstand a 6-foot drop to concrete, and 2m/7 feet submersion for a while (5m/16 feet up to one hour when the small switches in front are turned to the "5m"-position, which seal speaker and microphone etc. holes) as well as the military specification demands for dust/shock/etc. -proofness, not that I have tested dropping or submersing it, nor do I plan to

That is one rugged phone. I could have used that durability on many occasions. Not sure how commercial protective phone covers would fair from a 6 foot drop onto concrete. 

 

13 hours ago, esaj said:

I was expecting to find at least a few dead panels, so no surprise here. I'd be a lot more worried if the heating wires inside the floor concrete slab would have been broken, as replacing those is hard (tear off all floor tiling, jackhammer the slab, remove old cables, install new ones...)

I vacillated back and forth about installing  a heated tile floor in my new bathroom  but decided against it just for that reason.. 

 

[Hunka Hunka Burning Love]

Didn't your mom tell you not to vacillate so much, or you'll go blind?  

Speaking of blind, I haven't seen @esaj in ages?  Where he be?  I'm regreting subliminally planting that idea to go to electronics school to further his passions.  We miss you man!  Come back to us!  First it's @Cloud vanishing into thin air like a cloud, and now @esaj has gone missing.  Or is he giving us the silent treatment?  There's gotta be some electronic gizmo or difficult to comprehend schematic you can share?

Edited May 9, 2018 by Hunka Hunka Burning Love

[Cloud]

19 hours ago, Hunka Hunka Burning Love said:

Didn't your mom tell you not to vacillate so much, or you'll go blind?  

Speaking of blind, I haven't seen @esaj in ages?  Where he be?  I'm regreting subliminally planting that idea to go to electronics school to further his passions.  We miss you man!  Come back to us!  First it's @Cloud vanishing into thin air like a cloud, and now @esaj has gone missing.  Or is he giving us the silent treatment?  There's gotta be some electronic gizmo or difficult to comprehend schematic you can share?

I think esaj is simply very busy. He posted in the moderator's section a couple months ago saying how busy he had been. So i think he is ok!

Edited May 10, 2018 by Cloud

[Rehab1]

On 5/9/2018 at 4:51 AM, Hunka Hunka Burning Love said:

Speaking of blind, I haven't seen @esaj in ages?  Where he be?

 

On 5/9/2018 at 11:51 PM, Cloud said:

I think esaj is simply very busy. He posted in the moderator's section a couple months ago saying how busy he had been. So i think he is ok!

I do miss @esaj‘s comments and interaction.

 

Edited May 13, 2018 by Rehab1

[esaj]

On 5/9/2018 at 11:51 AM, Hunka Hunka Burning Love said:

Didn't your mom tell you not to vacillate so much, or you'll go blind?  

Speaking of blind, I haven't seen @esaj in ages?  Where he be?  I'm regreting subliminally planting that idea to go to electronics school to further his passions.  We miss you man!  Come back to us!  First it's @Cloud vanishing into thin air like a cloud, and now @esaj has gone missing.  Or is he giving us the silent treatment?  There's gotta be some electronic gizmo or difficult to comprehend schematic you can share?

Nah, didn't go to electronics school or such, mostly just been busy & tired, and things got a bit hectic. I actually resigned from my job at one point, but then ended up taking a part-time contract.  Not much been going on in the electronics-side, the last thing I did was this in early Feb for a company year kick-off costume-party:

It's just a bunch of "chainable" RGB-LEDs (WS2818/"Neopixels") and a controller for them. I used some velcro strap to attach it around my arm, with the final version having the controller glued to the back of the board and power wires running under my shirt to a battery in my pocket... The board and the "faceplate" are milled, the black part in-between is 3d-printed.

Not much been going on since, someone did ask me for some help with building a speedometer for their wheel, but don't know if he ever finished it.

I did get a couple of 0-100V adjustable linear power supplies, although only one of them works:

The non-functional one does power up, the display works and allows adjusting the voltage etc, but when any current is being pulled from the outputs, the voltage drops near to zero. Could be that the current limitation is broken and activates regardless of where the actual limit is set to? It does help to have the identical working unit, so I can compare voltages and such between them.

The company who made these sometime in the 80's has been defunct since early 2000's, and there's no schematics available, so I have to do some reverse engineering to find out what's wrong with the other unit. There's a lot of wiring and 4 separate boards (+ large other single components) in these things, luckily they use just plain through hole (THT) -components and not really many integrated circuits, so poking around is easier than with highly integrated/SMD boads. Also the wiring is done very neatly and it seems most of the boards have contactors so it's possible to pull them out and put them back in. Still, it's not very fast to figure out things (and I haven't really even began yet) and I have to be careful because I'll have to power them up every now and then to measure voltages going up to 115V AC or more in there (around 230V AC at the input to the large toroidal transformer). Since I've so far been working around just with low voltage DC-stuff, it's a bit nerving  Also, the large capacitors (the big black blobs attached by white zip ties) hold the charge after power down and need to be discharged safely before poking my hands in there or measuring resistances or continuities.

Edited May 13, 2018 by esaj

[Rehab1]

52 minutes ago, esaj said:

Nah, didn't go to electronics school or such, mostly just been busy & tired, and things got a bit hectic. I actually resigned from my job at one point, but then ended up taking a part-time contract.  Not much been going on in the electronics-side, the last thing I did was this in early Feb for a company year kick-off costume-party:

Welcome back!! I thought we awoke a hibernating bear but you have been busy. Your custom LED logo is fantastic! You could easily sell them to member featuring a Gotway, Kingsong, Inmotion or Ninebot led logo.

 

52 minutes ago, esaj said:

Also, the large capacitors (the big black blobs attached by white zip ties) hold the charge after power down and need to be discharged safely before poking my hands in there or measuring resistances or continuities.

Toughen up Esaj! You know how much we enjoy burnt hands on this forum.  

 

Edited May 13, 2018 by Rehab1

[esaj]

That was a lot less dramatic than expected. Trying the faulty supply again, I could not get any voltage out from the outputs. I tried swapping the secondary board with connectors between the two units, but the working one was still working and the faulty one still didn't give out any voltage. Measuring around the outputs in both supplies, I found similar voltages with same settings on both units, all the way to the capacitor which (presumably at that point) was in parallel with the outputs. The most likely culprit was simply a bad contact at the output banana terminals.

Most of the time was spent figuring out how to dismantle the front panel and the switches and boards attached to it so that I could get enough space to get to the terminals and then prying the nuts holding the board in place as I didn't have a suitable socket for a wrench and there wasn't much space. 

The board finally off:

  

The banana terminals were simply connected by two screws on both sides of the board, pushing a lock washer against the traces. Over time, the contact has become worse, and when it originally dropped the voltage to near zero when I put any load on the outputs, the voltage was being dropped in the bad contacts. Apparently after that they didn't make any contact any more, as I didn't get readings from the output. I ended up simply soldering wires from the bananas and the middle connector (chassis ground) and put some foam plastic there to keep the board from making contact with anything. I then tested the supply with 100V idle, 50V / 0.1A load and 4V / 1A load (as I didn't have anything much more suitable in regards of power resistors or other load) and everything just worked. Also checked that the board or the parts near it don't get too hot to melt the foam plastic.

While nothing near the foam got hot, some of the components do run pretty hot: 

Power resistor and a couple of transistors in the secondary board:

Power resistors in the rectifying board:

There's also a large TO-3 -packaged power transistor that sits against the chassis, but it never got higher than 40 degrees. Didn't really run the tests for very long either, something like 15 minutes each.

I didn't take much pictures during the work (also I've noticed that the Cat S60's camera software, or maybe the camera itself, has horrible auto focus when trying to take close ups), but here's a couple that some people might find interesting:

Like I said earlier, the supplies don't contain any SMD (surface mount) -components, and also very little in terms of integrated components (not much need really) and a lot of things are made with discretes, except the board running the 7-segment panel in the front showing voltage or current seems to contain more logic chips. The round components in the middle and to right are actually OP07- op-amps, and the one on the left is MC1741-opamp in a (nowadays) uncommon TO-99 package:

I didn't try to reverse engineer the board really, since swapping them had no effect, so no need to look for a fault.

The more (or at least slightly) interesting bit is the backside of the board:

Back in the day, something like 70's and 80's I think, when PCB design and manufacturing was already more "industrial", small scale production runs and hobbyists still used boards that were drawn by (free) hand, sometimes directly to the film, then exposed the PCBs with UV and then etched and drilled them with their own hardware. It seems that these power supplies, while likely made in hundreds or thousands, were still too small scale (too costly) to make the boards in a PCB factory, or maybe they just had old school -designers who liked to design their boards by hand and the final boards are factory-made. Still looks a bit weird having accustomed to modern (usually) computer designed boards, and mind you that this is still an actual commercial product. You can find similar boards in lots of old devices dating to 70's and back, and some in newer products too. In DC/nothing really high frequency -stuff it doesn't really matter that much (to my knowledge) as long as safety distances for high voltage traces etc. are maintained.

Now that I have a working (up to) 100V DC power supply with current limits, I can test stuff that's supposed to run off of wheel batteries (16S or 20S) without needing to use actual batteries. The originally working unit goes back to a friend that had bonked them from somewhere a while back, and dropped them off saying I can keep the faulty one if I can fix it 

 

Edited May 13, 2018 by esaj

[Hunka Hunka Burning Love]

Looks like a board I made once in electronics class.  Except we used these little rolls of thin black tape and mini black O-ring stickers.    I do recall the dark room for exposing and etching the boards too.  Fun stuff.  We used to fry diodes, resistors, capacitors, you name it with our bench power supplies.  Some things would glow red while others exploded.  

I thought you were making autonomous robots in some sort of class there?

Edited May 13, 2018 by Hunka Hunka Burning Love

[esaj]

21 minutes ago, Hunka Hunka Burning Love said:

Looks like a board I made once in electronics class.  Except we used these little rolls of thin black tape and mini black O-ring stickers.    I do recall the dark room for exposing and etching the boards too.  Fun stuff.  We used to fry diodes, resistors, capacitors, you name it with our bench power suppies.  Somethings would glow red while others exploded.  

I thought you were making autonomous robots in some sort of class there?

The course was between September and December, I haven't really touched the robots since  I've actually had those power supplies laying around since mid-February, just today got around to finally start figuring out what's wrong. Between there and now I haven't really done much anything.

[Hunka Hunka Burning Love]

It's time to go riding with your fellow Finlandic riders!  There sure seems to be a larger number in your part of the woods.  Get any ride time in with the KS16s?  Any luck getting the gf riding?

House still needing work?  2D mill collecting dust?  I thought you would be into 3D printing by now.  I still haven't got enough room here to even think about getting a printer.  I've got too much junk sitting around.

[esaj]

Trying to keep this short:

On 5/13/2018 at 11:45 PM, Hunka Hunka Burning Love said:

It's time to go riding with your fellow Finlandic riders!  There sure seems to be a larger number in your part of the woods.  

Not in this town, but in general it seems more riders are finding their way at least to the forums.

Quote

Get any ride time in with the KS16s?  

Yes, not really long trips so far, but I commute on it daily now and sometimes take some longer laps.

Quote

Any luck getting the gf riding?

Nope, I'm considering selling the KS16B, that's why I rode it all the way to empty tonight (or until I couldn't stand the constant beeping that starts when the voltage drops to below 50V or so). Might have to leave it sitting turned on next evening, as it'll likely bounce back up a bit, then let it sit off for a while before charging & measuring the battery capacity.

 

Quote

House still needing work?

Always

Quote

 2D mill collecting dust?

For the past few months, yeah, I'm building up parts list for an adjustable load (a project that I've meant to built for at least 1.5 years or so, never had proper cooling components and such)

Quote

 I thought you would be into 3D printing by now.  I still haven't got enough room here to even think about getting a printer.  I've got too much junk sitting around.

The printer too has mostly sat unused lately, I've also spent most of my PLA filament (something I need to order too) and left with a couple of kilos of ABS, which has been pain to print with. Usually my printing needs go hand in hand with the electronics projects, so kind of follows with the pause there... It's handy otherwise, but making any larger parts with it is sloooow (like 4+ hours print jobs) and sometimes it may need a few tries before getting things correct, so in addition to space, it also takes a lot of time to "find" the good settings. It's fun to watch though, well, maybe not for hours straight  

 

 

 

Edited May 15, 2018 by esaj

[esaj]

After months of hiatus, time to get back on the saddle...

I ordered parts to build a DC electronic load (adjustable constant current sink), mostly just DAC's, 1.024V voltage references, bunch of large (TO-247) power mosfets, some 10mOhm and 50mOhm power resistors for measurement and a 1 meter (about 3+ feet) block of aluminum cooling profile that can be cut to size as needed.

What an electronic load is, is a device that can draw current from some other device under test ("DUT") in a controlled manner. It's used to stress-test devices in a controlled fashion for measuring and such, or could be used just for draining batteries or whatnot. If someone read the final report of the robots built for the embedded course, I mentioned not being able to do "good enough" load-tests with the switching buck converters due to issues with not having a proper adjustable load. The HP 6632A can be used as a load, but since the downprogrammer allows something like 250 or 300mA more current to flow than set, testing it with lower loads became a hassle trying to set the output voltage slightly below the device output, so that the power supply sinks instead of supplying current. On the higher end, it's limited to +20V voltage and 5A current. I've been meaning to build a load since something like 2016, but never got around to it before, mostly because I didn't have suitable heat sinks at hand.

What I got was a Fischer SK-610 http://www.fischerelektronik.de/web_fischer/en_GB/heatsinks/A04/Extruded heatsinks for PCB mounting/PR/SK610_/index.xhtml :

This final profile picked was a bit of a compromise, there were some that have lower thermal resistances, but in the end, not all Fischer-models were available from TME, some of them were much more expensive and this one has grooves and threads for M3 screws already in place, meaning I don't have to start drilling and threading holes separately. 2.5-5.5 C/W thermal resistance isn't that bad, and it could be made better by adding some fans to blow on the heatsinks.

 

The actual "load" -part of the system is somewhat simple, I haven't drawn up simulations or schematics for that yet, but here's the general idea of an adjustable constant current sink:

An N-channel mosfet is the device taking the brunt of the heat. Below the mosfet, there's a resistor R1 (1 ohm in this case), that's used to "sense" the current coming from the DUT (V2 PSU in this case). The gate of the mosfet is controlled by an op-amp, with one input coming from top of the measurement resistor, and the other coming from a separate voltage source setting the desired current. How it works is that the current passing through the R1 resistor will cause a voltage drop over the resistor, let's say that the current passing through it is 1 amp (1A), then the voltage drop over the R1 would be:

U = R*I   =>   U = 1ohm * 1A = 1V

The inverting ("minus") -input of the op-amp gets this same voltage, 1V. Op-amp changes its output voltage based on the voltage difference of the inputs. If the voltage from the V1 Control Voltage -source on the non-inverting ("plus") -input of the op-amp is higher than the voltage measured over the resistor, it starts to drop the output voltage, forcing the mosfet to start to restrict the current flow through it. If the control voltage is higher, the op-amp output voltage will raise, allowing more current to flow through the mosfet. This way, a feedback-loop formed from the resistor and the op-amp will control the mosfet gate so that it passes the amount of current required to cause the same voltage drop over the resistor as the control voltage. In this case, it's simply 1V = 1A.

The above picture lacks some details, like for example if made as simply as here, the mosfet control could start oscillating as it under- and overshoot the set target, it needs some more components between the op-amp and the measurement resistor and the mosfet, but the general idea is exactly the same.

I plan on using 1.024V references and a DAC (Digital-to-Analog -converter) before the op-amp to set the desired voltage programmatically from an MCU and a 50mOhm resistor, so the actual proportion will be 1V = 20A. The DAC's have a resolution of 12 bits, that is, the control voltage can be set in 4096 different steps from 0 to 1.024V,  so in steps of 0.00025V or 0.25mV. Each 0.25 millivolt step equals 5mA, so the current can be controlled from 0 to slightly over 20A (1.024V / 0.050 ohm = 20.48A)  in steps of 5mA. Using different resistor value, the proportion can be changed (higher values for smaller range, but more precision, smaller values for higher range but less precision).

In hindsight, I maybe should have ordered 100mOhm or 500mOhm resistors, as the heating power over the mosfet will get quite large with higher voltages, as explained below. Paralleling multiple modules, I could still get quite high maximum currents without sacrificing resolution.

 

When mosfets are used as switches, like in our wheels, they are just turned "on" (fully conducting) and "off" (not conducting), with as fast change between these states as possible. Generally as small as possible internal resistance is wanted, so that they don't overheat due to the power loss (heat) caused by the that internal resistance when current flows through. In this case however, we want to do the exact opposite: use the mosfet to cause a power loss, loading the DUT with a set current draw. This means that majority of the voltage from the DUT is dropped over the mosfet. As an example, 10A running through an internal resistance of 20 milliohms (btw, the brand-wheels usually use mosfets with internal resistances in single digits in milliohms) causes a power loss of:

R = 20mOhm 

I = 10A

P = U*I  (Power equals voltage times current)

U = R*I  (Voltage equals resistance times current)

We could first calculate the voltage drop and then the power, but as a short-hand, placing   U = R*I   to   P = U*I   yields   P = R*I²    =>

P = 0.02ohm * (10A)² = 2W

Not that much really, 2 watts. Of course this doesn't take into account the momentary higher power loss during turn-off and turn-on in the wheels, but that's irrelevant to the electronic load. 

When the voltage drop over the mosfet becomes bigger, the wattage shoots up fast. Let's say we have a 12V power supply under test, and we draw 10A from it, dissipating the power as heat in mosfet. Some part of that 12V will be dropped over the measurement resistor in the constant current sink, in case of a 50 milliohm measurement resistor, using the U = R*I -equation, 0.5V will be dropped over the resistor. That leaves 11.5V drop over the mosfet. How much power is being dissipated in the mosfet?

P = U*I   =>   P = 11.5V * 10A  = 115W

Ok, that's a different ballgame now. 

Originally, I was looking for 200V TO-247 -encased mosfets for this, but the model I had earlier picked was out of stock. Instead, I went with 100V / 57A (continuous) -rated IRFP3710PbF ( https://www.infineon.com/dgdl/irfp3710pbf.pdf?fileId=5546d462533600a401535628fd1a1ffa ). I was mostly looking at the casing, maximum current and voltage, and price when picking these. They've got an internal resistance of 25 milliohms, which is of no concern, since the op-amp shouldn't fully "open" the mosfet anyway, so the "resistance" of the mosfet is higher in use. The only case where it's fully conducting is if the DUT cannot produce enough current to cause any limiting to occur, in which case the internal resistance is already meaningless.

Now, the datasheet of that mosfet says that the maximum power dissipation is 200W. Like discussed earlier here in the forums (and maybe in this thread, I don't remember ), this is the theoretical maximum which the device could withstand, in practice, other considerations usually limit you to a much lower value.

The datasheet lists the thermal resistances as 0.75 C/W for Junction-to-Case (maximum), 0.5 C/W for Case-to-Sink (typical) and 62 C/W for Junction-to-Ambient (maximum). The operating temperature for the junction is -55 to +175 C. Since I'm using a heatsink, the relevant numbers are the Junction-to-Case and Case-to-Sink. Junction-to-Ambient -value is used for non-heatsinked device in free air. If not using a heatsink, and dissipating 115W on the device, the junction temperature would be

115W * 62 C/W = 7130C above ambient

That's one dead mosfet there, it would melt completely far before reaching that temperature. When using a heatsink, the total thermal resistance would be

Junction-to-Case + Case-to-Sink + (possible thermal paste resistance, but the datasheet in this case states that the Case-to-Sink is for "greased flat surface", ie. with thermal paste applied ) + Sink-to-Ambient

Using the values from the datasheet and the "worst case" 5.5C/W -thermal resistance from the heatsink (if using only a 4-5cm long piece of the block), that would be

0.75 C/W + 0.5 C/W + 5.5C/W  = 6.75 C/W

For 115W, that would mean   115W * 6.75 C/W = 776.25C above ambient

Still far too high, as the device will likely fail if the junction goes above 175C, and with a typical ambient room temperature around 25C, that means the upper limit is about 150C for the thermal resistances.

Using a longer heatsink block with 2.5C/W for the sink-to-ambient, we get

0.75 C/W + 0.5 C/W + 2.5C/W  = 3.75 C/W

How about now? 115W * 3.75 C/W = 431.25C

Still way too high. So, with this cooling, it's impossible to keep the device cool enough while dissipating that much power. In fact, it would require a thermal resistance of about 1.3 C/W to be able to dissipate 115W with 25C ambient temperature. Event then, it'd be running right at the limit, and if the ambient temperature around the heatsink went up (if it were enclosed for example, and the air surrounding the heatsink would linger around going up in temperature), it could still fail.

There are a couple of solutions, either limit the power dissipation to lower values or increase cooling, either by using better or larger heatsinks, or adding fans. I plan on using the cooling block I have now, and to use a somewhat conservative value of 6C/W for the thermal resistance. For ambient temperature, I use 40C (conservative value, likely it will be lower but having some safety buffer in case the ambient temperature starts to raise), leaving me with 175C - 40C = 135C buffer. With these values, the maximum wattage I could dissipate would be

135C / (6C/W) = 22.5W

So a far cry from the theoretical absolute maximum of 200W for the device (to my understanding, the maximum wattage in the datasheets is a theoretical value above which no amount of cooling will be enough to prevent the device from getting destroyed). But it's somewhat of a conservative estimate, as the actual thermal resistance is likely lower (using longer heatsink blocks per device) and I could still add fan or fans blowing on the heatsink.

What I've decided to do is to make the constant current sinks as separate modules, and placing multiple of them in parallel. The power dissipation per module will be monitored by the MCU using the voltage drop over the measurement resistor and the voltage from the drain of the mosfet. This way, the MCU can calculate the power from how much voltage is being dropped over the mosfet and how much current is passing through it, and limit accordingly to keep the wattage in check and prevent the devices from getting destroyed by overheating. Temperature sensors can be used to measure the ambient temperature and the heatsink(s). Each module will have its own fuse, so in case of a failure of a single module, it won't short circuit the entire DUT. The MCU will know that a fuse in device has blown simply by seeing that no current no longer passes through the measurement resistor (no voltage detected), or the modules themselves could have simple LEDs to show the status.

That's pretty much the basics of it, I do have some other functionality still in mind. Obvious stuff like display, inputs etc., but also a more special case of overvoltage protection, of which I've already made some design and simulations:

This looks a bit messy, and I haven't added any texts or such to help understand it. The basic idea is that as the mosfets are rated for 100V max, if I somehow managed to go beyond that, I want the input to shutdown so that the modules don't get destroyed. This is not a final design, for example the resistor values are something I just quickly picked and didn't give yet that much thought into.

The IRF7204 is P-channel mosfet (I plan to actually use IRF5210, but didn't have the simulation model) to which the DUT connects (V1 in this case, giving a sine-wave from 0V to 100V). P-channel mosfets work in the "opposite" direction of N-channels, meaning that the source and drain are reversed, and that the mosfet starts to conduct when the gate voltage is lower than the source voltage. R4 is here to simulate the actual constant current sink, so it doesn't actually limit the current to any set value in this simulation, but the idea was to see that the power gets cut when the voltage goes above a certain threshold.

The mosfet here is used as a switch, like in the wheels, basically fully conducting when the voltage isn't too high, and turning off if it goes too high. Starting from the top, R1 connects the gate of the mosfet to the input voltage of the DUT. If there were no other components around it, it would keep the gate at the same voltage as the source, meaning that the mosfet isn't conducting. When transistor Q1 is switched on (conducting), the gate is connected to -12V on the other end through the resistor R2 and transistor Q1. Alone, these act as a voltage divider, so the gate voltage would be about

V_gate = V_DUT * (R2 / (R1 + R2)),  where V_gate = voltage at the mosfet gate, V_DUT = voltage from the device under test and R1 & R2 are the resistor values

With values of R1 = 10k and R2 = 100k, this gives out a factor of 100k / 110k = 0.909090... or about 0.91

To get the P-channel mosfet fully conducting, the gate voltage would have to be around 8...10V lower than the voltage at the source. If the voltage difference is less, the mosfet is not either fully conducting (becoming a load device) or is turned off if the difference is small enough. Also, there's an upper limit to the difference, given as V_gs(max) in the datasheets. If the voltage difference between gate and source (V_gs) is larger than this, the device will get destroyed. Usually it's +-20V for most power mosfets. This is where the zener-diode D1 comes to play. A 12V zener is placed in parallel with R1 to limit the maximum voltage drop over R1 to 12V, meaning that the gate voltage can't go lower than 12V below source. If it tries to drop more, the zener will start to conduct more, and keep it there (of course if enough current would flow through the zener, it would overheat and die). 

But what if I want to test a device that has lower voltage, like 3.3V? How can I get the gate to 10V below source, if the source voltage is 3.3V? That's where the -12V comes to play. ATX-power supplies conveniently have a -12V line. Even when the DUT voltage is close to 0, there's enough potential difference there to fully open the mosfet.

The rest of the circuitry is the voltage monitoring and controlling the transistor Q1 to turn on and off the mosfet. R5 and R6 form a voltage divider to drop the (up to around 100V) input voltage to a suitable range. The resistor values were again picked quickly without that much thought. 220k on top and 6.8k below gives about 0.03 -division factor, so 100V becomes 3V at the divider. The capacitor C1 is just for filtering any noise and keeping the voltage steady. The voltage from the divider goes to the inverting input of comparator U2 (the symbol looks like a op-amp, but it's a comparator). A comparator works by turning its output on and off based on voltages at the inputs. In this case, this is an "open collector" -comparator, meaning that either the output is disconnected, or connected to the negative rail (-12V in this case). The other input of the comparator is connected to another voltage divider formed by R7 & R8 from the 5V line. This gives an input of 2.5V to the non-inverting input. C2 is again just for filtering (likely I'll need to check the filterings more closely later on, as the voltages from the ATX-PSUs have some noise in them).

The comparator output is pulled to 5V through R9, so when the output is off, the base of transistor Q1 is connected to +5V through R9 and R3, and starts to conduct, pulling the gate of the mosfet to a lower voltage (about 12V below the source), and allowing the mosfet to fully conduct. When the voltage from the divider R5 & R6 goes above the voltage of the divider R7 & R8 (occurs around 95V with the resistor values used in the simulation and dividers connected to ground), the comparator connects its output to the -12V line through an internal transistor. There's some current coming from the 5V line through R9 (which could be bigger, like 10k to make sure the voltage drops low enough), but the voltage at the output will be close to -12V. There's not enough voltage difference over the base-emitter -junction of the transistor Q1 for it to conduct, the transistor stops conducting, and the gate of the mosfet gets pulled to the source voltage over R1, stopping the current from flowing.

That's pretty much it. I still likely will fiddle around with the resistor values a bit (and tie the dividers to -12V rail also) and check their power dissipations as they may be required to drop close to 100V over themselves, while not trying restrict the current to too low (the capacitor leakages would then have some effect), also should the comparator fail for whatever reason, the downside is that this set up will then leave the mosfet open. An opposite function could be better (ie. if the comparator loses power for instance, the mosfet would close). Since I came up with this after already ordering the parts, I didn't include any P-channel power mosfets. The ones I do have with high enough voltage are TO-220 and have an internal resistance of 60mOhm, which means they will also heat up with higher currents. Luckily though, they also have their drain connected to the case backplate, so if placed directly before the N-channel ones, I can mount them to the heatsink without causing the heatsink to act as "bypass" for the P-channels (I won't use electrically insulating tapes or such there, so the entire heatsinks will sit at the same voltage as the drains, which could be closer to 100V). The power dissipation in these will be much lower than with the N-channels acting as loads, but the temperatures and power dissipation stil have to be checked to make sure.

Edited May 27, 2018 by esaj

[ir_fuel]

What's so funny about electrionics is that the basics are extremely simple. There are only a few formulae you use 99% of the time, and there are not that many "base" components. Yet once you start building stuff and the complexity rises the thing really explodes. And then there is theory vs practice. It can all be correct in theory, and once you build it it doesn't work because of some effects you didn't or couldn't take into account before. Really interesting, complicated and time consuming matters. I like it from a personal interest point of view (and I know a really tiny bit about it through reading/tinkering/Udemy course), but it takes A LOT of time to become good at this stuff. As a professional software engineer I am still a bit jealous about people that can actually build complicated hardware (then again, there are people that are jealous because I can solve some of my own problems easily by quickly writing a piece of software).

Need more time in a day  

[esaj]

3 hours ago, ir_fuel said:

What's so funny about electrionics is that the basics are extremely simple. There are only a few formulae you use 99% of the time, and there are not that many "base" components. Yet once you start building stuff and the complexity rises the thing really explodes. And then there is theory vs practice. It can all be correct in theory, and once you build it it doesn't work because of some effects you didn't or couldn't take into account before. Really interesting, complicated and time consuming matters. I like it from a personal interest point of view (and I know a really tiny bit about it through reading/tinkering/Udemy course), but it takes A LOT of time to become good at this stuff. As a professional software engineer I am still a bit jealous about people that can actually build complicated hardware (then again, there are people that are jealous because I can solve some of my own problems easily by quickly writing a piece of software).

Need more time in a day  

Yes, the basics are pretty easy to grasp, but things get more complicated fast with more complex circuit or when you need more accuracy/speed/whatever. Or RF (Radio Frequency) design, which to me is just "black magic"  The internet is full of tutorials, people's projects, papers and application notes, but if you like a "single source" for learning stuff, the best book I've come across is Art of Electronics, 3rd edition:  https://artofelectronics.net/

“Wow. Chapter 5 details every circuit artifact that I’ve encountered in the past 30 years in a thorough, pragmatic, and straightforward way. My only ‘twinge’ is that it discloses and explains (in glorious graphical detail and with real part numbers) many topics that I thought were my personal trade secrets. I love the plots. I know that it must take an enormous effort to collate all of the device characteristics. It’s worth the effort. The way the data is presented allows the reader to get terrific perspective on a lot of landscape in a single view. Nice work.” — John Willison, founder, Stanford Research Systems

Unlike your usual text book going over the basics, a few basic circuits and leaving lot of things unnoted, they go much deeper, explain things, compare performance with different components and on many occasions, use circuits from real world devices as examples. The entire thing's over 1200 pages long, and I can admit that I haven't read all of it, but as a text book on "general subjects", it's probably one of the best there is. If you get it, consider getting the e-book version, the actual physical book is very heavy and hard to read in bed  Then again, there are entire volumes written just about designing more specific circuits (like "Op-amps for everyone", that talks only about op-amps and their usage in circuits, for example).

Still, compared to software, electronics is much more time-consuming and relatively expensive, for anything more complex, in addition to components, you already need at least a multimeter and good power source for testing, possibly an oscilloscope. If you want to do something more "permanent" than just build circuits on a breadboard, add a soldering iron, and either perfboards/veroboards, etching tools or order boards from a fab house, or mill them at home with a CNC, like I do. A cheap basic multimeter + adjustable low voltage power source can be had for probably <100€, but if you want more quality and accuracy, the price starts to add up. My kit has grown over time, and currently consists of:

Soldering & board fabbing & building stuff:

• Multiple cheapo chinese soldering irons (most are 50W I think, price ranging somewhere between 10-20€ each)
• Aoyue Int968A+ 75W soldering / 550W hot air station (about 250€ with extras, like soldering/desoldering tweezers) 
• Soldering helping hands/PCB stands (a few tens of euros in total)
• Cheapo USB-microscope for soldering really small parts (about 10€)
• 400 celsius silicone soldering mat (maybe about 20€)
• Another soldering mat with ESD protections (maybe about 20€)
• Lots of tweezers (ranging from cheapo chinese 6-in-1 packs that cost <10€ to Swiss-made Ideal-Teks that cost 25€ each) and umm, "soldering manipulator tools", like these:
• 
• Home-made reflow oven made from a toaster oven, although I use it rarely (oven + part costs a few tens of euros)
• Desktop CNC with 240x180mm work area (around 250€ with shipping & import taxes + maybe 50€ total for some replacement motors and parts, drills and milling bits)
• 3D-printer, a cheap Prusa -copy I bought off from my cousin (150€ + few tens of euros for filaments)
• Lots of miscellanous tools, like pliers, screw drivers, wrenches, calipers, side-cutters, wire strippers... no idea on total costs
• Soldering pastes and wires, maybe around 100€ in total (250-500g spools of different sizes, multiple syringes of paste kept in the freezer)
• Lots of copper clad boards for milling PCBs from (<50€ for the total inventory, usually the boards cost about 1€ for 35µm / 1.6x100x160mm FR4 piece)
• IPA (isopropyl alcohol / isopropylene), bought in 1 liter cans (around 6€ per can from TME, around 8€ per 100ml if you buy it from pharmacy) for cleaning
• Probably all sorts of knickknacks and small tools I forgot 
• Secondary desktop computer just so that I have a display directly in front of me when soldering (so I can check the component placements in real-time from KiCAD and use the USB-microscope) and to control the CNC (basically "free", since I already had it before)

 

Testing:

• Multiple cheap & basic multimeters, typically around 1% accuracy (<100€ in total)
• Cheap LC -meter for measuring inductance/capacitance (about 15€)
• Home-made power supply with 3.3/5/12V lines and adjustable output, made from an old ATX-power source (part costs of maybe 20€, not including the ATX-PSU)
• DIY low frequency, single-channel oscilloscope kit (around 15€)
• HP6632A 100W (20V/5A max) adjustable linear power source (rack-mounted, industrial quality, bought used for 130€, list price is >1000€)
• HP34401A desktop multimeter, 6½ digit, 0.0035% DC, 0.06% AC basic accuracy (240€ bought used, list price is >1500€)
• Rigol DS1054Z 50/100MHz, 1Gsps 4-channel oscilloscope (about 400€, bought new)
• Bontronics 100W linear power supply (100V / 1A max), got for free, needed a small repair
• A few ATX power supplies (got for "free" from old computers / friends)
• Home-made 1Hz - 180MHz sine-wave frequency generator with amplitude/offset controls, based around AD9851-module (about 20€ in parts)
• Home-made frequency/pulse counter (that I'm still supposed to add to my workbench, part costs <10 euros)
• Home-made continuity meter built straight onto my workbench (really handy when soldering, it's "always on", has sharp tips made from worn 0.1mm milling bits, no need to fiddle with multimeters and has 90mV output, so it won't pass any P/N junctions or Schottky diodes and cause false alarms, part costs a couple of euros)
• Lots of breadboards and jumper wires, small wall warts, batteries, measuring probes and wires etc

On top of those, I've probably sunk a good ~2000€ on components, connectors, wires etc... The stuff pretty much fills my entire study, and just this week I went out and bought 60 meters of 2x4's and 2x2s to rebuild my entire table set with more sturdy and deeper desks for more table space. On top of money, I've spent a lot of time just reading about circuit design and such, and of course testing and building my own projects. No idea how much, likely way over 1000 hours, if not several thousands over the last about 2.5-3 years.

The way I've approached most of my projects, is to first figure out the basic schematic, then simulate it in LTSpice (free circuit simulator from Analog Devices) to figure out component values and changes needed, possibly breadboard it for measurements and testing, design the circuit layout in KiCAD, build the G-code for the CNC with FlatCAM (open source CAM-software specifically for PCBs) from the Kicad gerber/excellion-output files for milling and drilling and then mill & drill out the board on a tabletop CNC (controlled by bCNC). Clean the board, solder in the components & connectors, clean the residual flux and it's ready for testing. For less complex/not very high part count boards, I can start the design in the morning and have a soldered board in my hand by evening. For complex cases, and where my inventory of parts doesn't contain something I need, it can be weeks. Or I need to do multiple prototypes before I get a "good enough" device done. Just picking the "best" component somewhere can take an entire evening, comparing datasheets and prices across various options.

Edited June 30, 2018 by esaj

[ir_fuel]

Got plenty of tools here too (no CNC machine, have a few 3d printers), oscilloscope, power source, soldering station, a bunch of Arduino and Rasperry Pi, boxes full of all kinds components etc etc. 

Just lacking the time now. Especially since it is summer I've got other stuff to do when not programming (like riding an euc ? ) 

[esaj]

I've still been planning the load, and in the process, ordered more/better components three times in a row, despite not having yet built even a single prototype of the actual load circuits    So things have been progressing rather slowly there.

What lately has kept me bogged down (among other things, like usual, time and energy) is the lack of space. My study is messy, but even if it was tidy, I simply lack enough table space to more easily work with the large 4U rack case I have for the load, so I'd have to work on it hunkered down on the floor, and probably suffer from at least some back ache:

Actually, this is an old picture. There's more stuff with another shelve built on top of the left-side shelve/rack on top of it and a lot more boxes there.

The current tables are made from 15-18mm chipboard, and the long (160x60cm) table is already bent at the middle from the weight of rack with linear power and other stuff. I'm a bit worried it might snap with time  . The small desk for soldering station's 80x60cm, and the drawer between the tables is actually what's keeping them from collapsing. The corner table with the CNC can't be used for anything else, and the last table is just for computers, paper work and general junk. I still have the 3D-printer in the other room with all sorts of crap. I've got to sort this out.

So, I finally started to build the new desks I've been planning to do since at least last fall:

This is actually the second table, sized 160x80cm, 3cm thick chip board. Another table, 200x80cm and 2.7cm thick (melamine), is in pieces in the background against the wall. Third table top awaiting cutting (160x66cm, 2.5cm thick chip board, only meant to support a couple of computer displays and other light stuff) is behind all the other crap.

Originally I was just planning to screw a bunch of 2x4's together, slap the table top on the frame, then build separate shelves on top. A friend of mine, who works with wood more, told me that it would be a waste of time and materials, and wouldn't last. So, enter "mortise and tenon"-joints:

Lacking proper tools, skill and technique, it's been pretty slow to make the joints. The larger tables have 12 of these each, and the last one likely will have at least 12, if not 16:

I've added a "lock pin" of sorts that goes through the mortise & tenon. The actual table tops themselves sit on top of 20x40mm rectangular steel pipes:

I'm still wondering whether I should add a third pipe going on the middle, although the top seems very sturdy (easily handles my friends' weight, at 80kg or so).

Once I get this one finished, I still have to take down and move everything away, dismantle the old tables, bring the new one in in pieces, assemble it and then figure out all the wiring and lighting (the power feeds will come through separate automatic fuses, with additional safety isolation transformer for certain devices and such)... Then I'll have space in the garage to finish the other table, and only after I get that one in place, will I know the measurements for the last one   

 

 

Edited July 20, 2018 by esaj

[esaj]

Slow progress, but moving on... I've got the first table in, but it needs at least one more shelf. Still need to dismantle the CNC-table and all the component cabinets, the CNC itself and the stuff underneat to get space for the next table. 

I finished building the first shelf on the 200x80cm table today:

The basic structure's similar to the 160x80cm table, two steel pipes holding the table top, with an additional 2x2 running through the middle sideways.Had a bit of an "oopsie" at the backside pipe height, so I had to make the holes larger and prop up the pipe

 

The shelf's a bit makeshit, I originally had a bit different structure in mind and had already cut some of the pieces, so I reused some of them to make this. The shelf's high on this one (like it is in the smaller table too) as I want space under it, and in this case, the 3D-printers' going in the right corner, needing close to 70cm in height, so I lifted the shelf 75cm from the table top.

Should get more up to speed now, as my last day at work was today.

Edited July 27, 2018 by esaj

[Hunka Hunka Burning Love]

Don't you guys have like an Ikea nearby?    Should you be cutting stuff with powertools when your near vision is a little weakened by the recent laser light show?    BTW I think you may have measured the posts for the top shelf slightly unevenly.    Maybe use a measuring tape instead of eyeballing it?

Edited July 27, 2018 by Hunka Hunka Burning Love

[esaj]

10 minutes ago, Hunka Hunka Burning Love said:

Don't you guys have like an Ikea nearby?    Should you be cutting stuff with powertools when your near vision is a little weakened by the recent laser light show?    BTW I think you may have measured the posts for the top shelf slightly unevenly.    Maybe use a measuring tape instead of eyeballing it?

I'm running low on material, so I make do...  Actually, what I thought was to use wider um... "glued laminated timber board" or something along those lines for the top shelf, so it sits on top of of the shorter posts and is partially inlayed within the longer posts... if that makes any sense. Or then just put similar "front-to-back" supports as the current shelf and use the 30cm glued board.

[Esper]

Viewing your table @esaj it looks very well made. I didn't have time to view the entire thread going back to 2015, but from the last 4 or 5 posts it looks nice.
Me coming from carpentry background would only suggest one thing that would strengthen the table (but it wouldn't need it) is little angular cross braces. As everyone knows, the triangle is the strongest shape. You can do this with wood, metal pipes, a braided metal tension cable, etc. It would only be useful in very heavy side to side motions. Like if the table is fully loaded with a lot of weight and you wanted to move the table by yourself either pushing or pulling it along the floor.
You could also eliminate that by adding wheels.

Overall, well done!

Edit: One more thing. Cover that exposed metal end! That could hurt if you catch the edge.

Edited July 27, 2018 by Esper

[Hunka Hunka Burning Love]

I'm no carpenter so I can of course be more critical in these matters.  It basically looks like what happens when I don't bother to look at the Ikea instruction sheets and grab the wrong boxes on the way out.    And yeah add some triangles!  There's always got to be more triangles especially if you and GF get "jiggy wit it" on that new table to break it in.  You don't want to have an injury which needs detailed explaining at the ER.  

Recently I had to put together a Friheten, and let me tell you it is aptly named!  What the Friheten is this part and where does it go was what went through my mind repeatedly.  Darn Swedes and their diabolical heavy furniture!  

Spoiler

 

 

Edited July 27, 2018 by Hunka Hunka Burning Love

[Esper]

I didn't know you were asian @Hunka Hunka Burning Love. This is the first time I've seen your face. I don't know why you would hide it, you would make all the ladies chase you. 

[Hunka Hunka Burning Love]

  Just because someone posts up an Ikea how to video doesn't make them Asian just like how someone who posts up a shuffle dance video doesn't make them a hot leggy female.     Darn pesky Asians always get the hot females.  

 

Spoiler

 

 

Edited July 27, 2018 by Hunka Hunka Burning Love

[Esper]

Wow @Hunka Hunka Burning Love you are a pretty attractive young woman.

[Hunka Hunka Burning Love]

And for only $7.99 / minute I can be a pretty, young, Russian bride woman who can talk sexy to you in a hot sauna where I feel compelled to take my clothes off.