1000W Discharger Project

[magudaman]

always overdesign That's my motto!

Here's my BAD (Big Ass Discharger)..... working on the gate driver and controll circuitry right now, but the mechanical side of it is done....

Designing for ~200A per module, and there's 6 of em..... still need to find some decent fans, these get hot! Each pair of modules protected with a contactor and fuse and each unit has it's own shunt. I know it's overkill, but we plan on testing some serious loads.

and BTW, thanks for the help John. We were considering load resistors too, but ended up going with driving them in their dissipative mode.

I am using the Sunon 120mm PSD1212PMBX fan. Actually I will be using 2 of the them. They are a 200CFM fan and are the most powerful 12v unit I have personally ever seen. I made a quick video here:



Keep in mind when I pull the camera away I believe that my phone had turned the microphone way down. They are amazing units!

Oh and they are also only 9.95! http://www.ebay.com/itm/Sunon-120mm-200CFM-12-Volt-DC-Fan-PSD1212PMBX-Powerful-/320752775803?pt=LH_DefaultDomain_0&hash=item4aae5af27b

[magudaman]

Ah I also got in my heat sink today too! I calculated this guy to over 1000 watts, so hopefully it holds true. It heavier than I though at around 10lbs. It also an interesting design. It is actually made up of pressed together segments that make up each fin. It also of course has holes in it already I will have to work around.



\
b


So I have been learning about op amps and think I am going to try to make one of Johns circuits from above. Once I test it out, then I plan on running 10 of them connected to my heat sink to get around 100w per unit. I am going to just electrically hot heat sink so I can avoid the isolation pads so I figure 100w is fairly reasonable.

It seems I might be better off getting away from the logic level fets now in that design. The op amp will boost the voltage required for a normal FET I believe.

[CamLight]

LOL...now THAT'S a heat sink!
I would guess it's probably good to a few hundred watts (with decent fans) and you've got some verrrry serious fans so 1000W might work! Do you have a thermal resistance number for the heat sink?

How were you going to mount the fans? For short discharges, you might not even need the fans since there's so much metal there that the sink won't heat up much for a while. But, after that you'll need both fans for sure.

For max heat efficiency, spread the FETs out equally and keep them away from the edges. Also keep them out of the middle third (lengthwise) of the heat sink as that will be the area receiving the least cooling The fans will be doing a lot less cooling there due to no air flow because of the hubs (assuming that the fans are mounted on top of the fins blowing down). I'd probably mount them in two rows of 5 FETs, each about 1/3 of the way from the long sides and spaced equally down the sink with the same space to the short edge of the sink as exists in between each FET. Your sink does have a great thick base though. That will help with heat spreading a lot so I'm probably overthinking the FET mounting pattern.

If you're powering the op-amps with +9V to +12V, then you can pretty well drive any standard MOSFET out there as you'll only need to drive them to their linear region (2V to 6V max) unless your specs for the setup require each FET to be turned on fully in order to pass enough current for the discharge. If you do need to turn them on fully, power the op-amps at +12V so you can get at least 10V at the output. It all depends on the op-amps.

Can't wait to see this thing working! And you'll get a free room heater too.

[magudaman]

Sadly no C/W, its surplus/used. When I ran the specs through here http://www.aavidthermalloy.com/thermal-tools/bonded-fin I believe I was above 1000w but this thing can always be dialed in afterward.

So I think I have a good selection of components and decided to go to the normal fet for the possibility of getting more resolution when driving from the PWM and the RC. I think I will feed the OP amps and Fets 9 or 10 volts from a Vreg since I'm not sure how much my 12v PSU I purchased will vary while under load. I'm not real assured of its quality either so the Vreg will be a level of protection.

Sticking with John's design I came up with these parts:

FET:
Alpha & Omega Semiconductor Inc
100V 150A AOT410L
.045 C/W
Decent DC performance
http://aosmd.com/res/data_sheets/AOT410L.pdf
$2.64 for 10

OP AMP:
Texas Instruments TLV2370-75
Rail to Rail input and output
$1.32 for 10


http://www.ti.com/lit/ds/symlink/tlv2370.pdf

Resistor:
TO-220 35w .05ohm
$2.64 for 10

These would be good to over 25 amps, so maybe I can make my device 250 amp now for single cell lithiums. I plan to dump their heat into the heat sink too for size and easy cooling.

I'm going to order 2 sets and try to put one or two together and do some testing before I buy a bunch again

[CamLight]

You're right about the possible power handling of that sink! It looks like you'll be able to hit around 1000W if you bring the FETs up to their absolute max operating temperature. Probably about 770W if you derate the max temp to 80% for reliability.

The AOT410L looks like a decent FET to use. The thermal resistance is about 0.85C/W (0.35 theta-jc + 0.50 theta-cs) and that's good, lower than the typical 1.0C/W or so for most TO-220 FETs.

I've lost track...are the TO-220 resistors going to be the current sense resistors or drain resistors to take some of the power dissipation away from the FETs? I'm only asking because if they're current sense resistors, you'll have to subtract the power they'll dissipate from what the MOSFETs could handle (if resistors and FETs share the same sink).

Looks good! Anxious to hear about how the testing goes.

[magudaman]

You're right about the possible power handling of that sink! It looks like you'll be able to hit around 1000W if you bring the FETs up to their absolute max operating temperature. Probably about 770W if you derate the max temp to 80% for reliability.

The AOT410L looks like a decent FET to use. The thermal resistance is about 0.85C/W (0.35 theta-jc + 0.50 theta-cs) and that's good, lower than the typical 1.0C/W or so for most TO-220 FETs.

I've lost track...are the TO-220 resistors going to be the current sense resistors or drain resistors to take some of the power dissipation away from the FETs? I'm only asking because if they're current sense resistors, you'll have to subtract the power they'll dissipate from what the MOSFETs could handle (if resistors and FETs share the same sink).

Looks good! Anxious to hear about how the testing goes.

Yah the sink is definitely an unknown. Once everything is working I be able to do more temp monitoring and tests then decided what the max wattage will be. It will be an easy change in my micro controller code to prevent me from going over the max wattage when setting the current. I also intend on having two temp probes that will shut down the discharge if the sink gets too hot.

Well originally the resistors were moving the heat away from the fets but now I going to use them as my current sense for your design. I realized putting them on the same heat sink put more watts into the sink but it all part of that shared 1000w number I am shooting for, 1000w total = FET watts + Resistor Watts + Wiring losses

I'm going to place that small order and hopefully next week I'll have some time to prototype a bit! I am also planing on making some boards on CNC machine since I'm going to have 10 individual little boards at some point. I'm no good at Eagle but I guess I gotta learn at sometime.

[magudaman]

Alright got in all my parts on Tuesday for prototyping but have been so busy I didn't get to play with them till tonight!

So I quickly wired up the schematic from John on a bread board minus the ref voltage and instead substituted a variable bench supply. I then used my broken CBA heat sink to hold my FET and 35w resistor. I fired up the circuit and with a bit of adjustments it seems to be working perfectly! I'll have to get the lab supply on the load side so I can confirm it really is holding a set current but it seems to be doing pretty good.

Its late and I'm quite tired but I had to post these up so here are some crappy photos:





So I did find something that may be a problem. I wanted to originally be able to get to around 90C to 100C on the heat sink before shutdown but I don't think these resistors I chose are going to be up to the task. Either I run a lower resistance or a lower amperage

http://riedon.com/media/pdf/PF2200.pdf

So it looks like at 100c these and all other through hole resistors are really only rated at 40%. So I think I may drop down the .02 Ohm for final design so more heat will be coming out of the FETS but still around 100W.

Anyways its still moving forward. I'll play some more tomorrow.

[CamLight]

So it looks like at 100c these and all other through hole resistors are really only rated at 40%. So I think I may drop down the .02 Ohm for final design so more heat will be coming out of the FETS but still around 100W.

Anyways its still moving forward. I'll play some more tomorrow.

Congratulations on that successful smoke test!

I may be misunderstanding your post but one big advantage to using thru-hole resistors, versus using TO-220 resistors on the same sink as the FETs, is that the ambient temperature for them is 25C or so. That's allow them to run at 100% power (which isn't recommended for any resistor, but you can do it). Only resistors being mounted to the same sink as the FET will have an elevated ambient temperature, which will severely limit their power handling.

[magudaman]

So it looks like at 100c these and all other through hole resistors are really only rated at 40%. So I think I may drop down the .02 Ohm for final design so more heat will be coming out of the FETS but still around 100W.

Anyways its still moving forward. I'll play some more tomorrow.

Congratulations on that successful smoke test!

I may be misunderstanding your post but one big advantage to using thru-hole resistors, versus using TO-220 resistors on the same sink as the FETs, is that the ambient temperature for them is 25C or so. That's allow them to run at 100% power (which isn't recommended for any resistor, but you can do it). Only resistors being mounted to the same sink as the FET will have an elevated ambient temperature, which will severely limit their power handling.

Well they do consider these TO-220 resistors also through hole as opposed to chassis mounted resistors. I smoked the resistors today while doing some more testing! Arg and say resistors because both are dead now. Since I found out that they were only really rated for 15w or less I put the two in parallel to give me around 15w per unit. I hooked up my icharger to provide me a load in wire cut mode and wanted to see how well the circuit would hold current up to about 40v. I should have used something like a current loop to double check the Icharger's current reading. It only varied from 2.3 amps to 1.9 amps at 0-10v then 1.9 amps all the way up to about 38v and then my resistors spit smoke all over me bleh!

So not really sure what happened there especially since I was running them at such low current at the time. So now I need to decided to I just go with the .02 35w TO-220 (which is rated for a maximum of 25a which could be a problem if I wanted to run 25a per device, maybe that is why these failed) or a .02 ohm chassis mounted 25w like this:



but I don't know its current limit and I would still be mounting it to my heat sink provided it fits and its more expensive than the TO-220,

or I just use a sense wire at .005 ohms and basically get away from trying to dissipate any heat in the resistor like this:



The reason I don't wanted to mount my resistors off the sink is size and more fans. I would like this device to be a small as possible and if I have use large chassis mounted stuff or big ceramics they take up a lot of room and then need additional cooling fans.

EDIT:
Talked to the guys that make that chassis mount one above, they basically told me that it is limited by the wattage rating. So any current should be find. So it looks like those may be my best choice if I still wanted to put some heat into those guys. They also seems to have 100% power rating to about 85c.

[magudaman]

Well I feel silly but I didn't check it and it turns out the FET died! That is why the resistors blew too! It failed in a short. So that is no promising either

[CamLight]

Well I feel silly but I didn't check it and it turns out the FET died! That is why the resistors blew too! It failed in a short. So that is no promising either

Ouch! Sorry you're having to deal with that. I had practically a jarful of dead FETs by the time I was done.

While you're replacing the FET, shorten all wires as much as you possible can and mount the FET gate resistors directly to the gate pin. That can help to reduce any oscillations that can fry the FET (even at very low power levels).

If you're still having problems after that (or to try everything you can now)...
- Have the op-amp's power/GND leads connect to the protoboard near the op-amp.
- Decouple the power/GND leads with a big cap (10uF min, 100uF is good), with short leads, before having short power/GND jumpers going to the op-amp.
- Add a decoupling cap (ceramic, 0.1uF or so) with short leads across the GND/VCC pins of the op-amp.
- Add 0.1uF and 1uF-10uF caps to the non-inverting pins of the op-amp (to GND near the op-amp's GND pin) to swallow any noise on those lines.
- Shorten the leads of all the resistors as much as possible and mount everything as close to the op-amp as possible. Really, really close. 1/8W resistors are great for this. The 1/4W ones can go vertical though.
- Power the circuit with a battery pack (short leads!) instead of a switching power supply and power the fan (if used) from the switcher. Join the switcher's GND lead to the pack's GND at the battery pack, not near the protoboard.
- Keep any fans, switcher power supplies, and other devices away from the protoboard and heat sink.

[magudaman]

Ouch! Sorry you're having to deal with that. I had practically a jarful of dead FETs by the time I was done.

While you're replacing the FET, shorten all wires as much as you possible can and mount the FET gate resistors directly to the gate pin. That can help to reduce any oscillations that can fry the FET (even at very low power levels).

If you're still having problems after that (or to try everything you can now)...
- Have the op-amp's power/GND leads connect to the protoboard near the op-amp.
- Decouple the power/GND leads with a big cap (10uF min, 100uF is good), with short leads, before having short power/GND jumpers going to the op-amp.
- Add a decoupling cap (ceramic, 0.1uF or so) with short leads across the GND/VCC pins of the op-amp.
- Add 0.1uF and 1uF-10uF caps to the non-inverting pins of the op-amp (to GND near the op-amp's GND pin) to swallow any noise on those lines.
- Shorten the leads of all the resistors as much as possible and mount everything as close to the op-amp as possible. Really, really close. 1/8W resistors are great for this. The 1/4W ones can go vertical though.
- Power the circuit with a battery pack (short leads!) instead of a switching power supply and power the fan (if used) from the switcher. Join the switcher's GND lead to the pack's GND at the battery pack, not near the protoboard.
- Keep any fans, switcher power supplies, and other devices away from the protoboard and heat sink.

Well I received my new set of parts earlier in the week and hooked up a new circuit this morning. I really like the new wire wound resistors and they are much smaller than I was expecting despite looking at the spec sheet. I shortened up my leads because I easily had a good 6-10 inches of wire on certain components in the breadboard design. I did some testing at 3.7V at 25 amps and let it run for a good while. Then 13V and 7 amps for a good while too. Then I jumped up to 52V and wanted to only do around 1.8 amps. As I ramped the current up to about 1 amp I actually heard my BMS make a sort of noise and noticed my circuit drop to zero amps. I then went to hook back up my 3.7v cell and then popped my 25 amp fuse. The FET died again. So I guess its time to put in a new fet and add some of your items above. If I don't get it working with my last two remaining fets and some of the changes above I may be making a new FET choice.

I am using a linear regulator to give me 10V now from a 12v battery. I had a 10uf on the input of that device and 1uf on the output.

Oh so one thing I noticed is I was hooking up the power to the op amp circuit then plugging in my battery which would create a bit of a spark. So I am assuming when no current is going the FET is basically turning on to almost full open. So it could be possible I was doing damage in that spark. Maybe its doing that because my adjuster line was not truly at zero volts or shorted to ground.

I was a bit confused as I am still learning the terminology, but you said add the capacitors to the non inverting lines of the op amp that go to ground. Isn't there only one non inverting pin 3 / input + ? So then leg of capacitor from op amp + and the other leg of cap to ground? If I'm not mistaken can't I put a cap on the inverted one too, since its just positive voltage from the shunt?

I looked at my fet spec sheet again and spec wise it still is a really great choice, but doesn't seem to be pulling the numbers at the higher voltages.


EDIT: It looks like I wouldn't want put the caps on the inverting pin / shunt sense, it would probably slow the op-amps response to changes.

[magudaman]

So the saga continues. I tried to eliminate a lot of variables today and get a working unit that can handle more than 40V. I hooked up the capacitors and busted out the scope to see exactly what was happening signal side and power side. What I found is the added capacitors didn't seem to change the signals what so ever. I tried even connecting them and disconnecting them while everything was live but would see now changes on the scope. But what I did see on the scope didn't seem right:

This is the signal going into the FET:




I thought that doesn't seem like a good thing so I connected to power side and once the current is ramped up I was seeing some pretty crazy peaks in voltage. At 25 amps I would see 3.7v peaking at almost 18V and the 13V batter peaking at 28V:



So it seems I was getting this oscillation that was causing some serious inductive voltage and that was why I was frying my FET over 40V since the peaks could easily break 100V.

So I figured why not try to filter the FET signal with a .1uf ceramic cap and see what happens.

Both sides then are a nice DC with no frequencies or oscillation. YAY!

Now time to ramp it up to 60V and see if she holds. I get everything wired up and start to ramp up the amperage. 0.2, 0.4, 0.6 POOF! The fuse spits smoke out at me. GRRRR no go. And of course I didn't have the scope setup to record what was happening on the signal or power side so I still have no idea what is happening in the upper voltage. Well I have one FET remaining but its not looking promising.

[magudaman]

Did a little reverse engineering on a CBA device and added a .1uf cap, 10uf cap, lastly a 10k resistor all from gate to ground. This time when I blew the thing up a I had the scope recording. Both sides were looking like nice smooth DC when for some reason the Op amp decided to jump up 3 volts on the gate. Not quite sure why. I got another thread running over on all about circuits trying to see if anyone else has some good ideas. I was going to fully reverse engineer the CBA dual op amp circuit and consider adopting their design. I am also going to go to a 200V fet. I found some that have basically better dc performance and can still pull 25 amps at 2 volts but with a beautiful .3c/w junction to case.

I also tested out a nichrom wire PWM device and it worked but I still need a way to make it so it not so nasty on my batteries. You can actually hear clicking from from the pack. So basically would require a giant inductor but then would also need fly back protection. That is a problem at single cell voltages.



Its kind of neat you can see it actually took 25ms for my 15 amp fuse to blow.

[Alan B]

A CBA schematic is available somewhere on Candlepower Forums. Probably a hand reverse engineered version. There is a lot of CBA experience over there. The hotwire guys run some pretty amazing power levels in their flashlights, and various lithium battery packs. Not quite ebike levels, but more than you'd probably expect. My hotwire regulator is good up to about 15 amps and 40 volts and they wanted higher voltage upgrades for it.

[magudaman]

A CBA schematic is available somewhere on Candlepower Forums. Probably a hand reverse engineered version. There is a lot of CBA experience over there. The hotwire guys run some pretty amazing power levels in their flashlights, and various lithium battery packs. Not quite ebike levels, but more than you'd probably expect. My hotwire regulator is good up to about 15 amps and 40 volts and they wanted higher voltage upgrades for it.

Thanks Alan I'll have to try to hunt that down.


So talking with the guys on all about circuits forum I got some answers to my problems. It turns out that me adding capacitors was bad. That was slowing down how quickly the op-amp could react to changes in current. This is bad when it starts to go into an over current situation and would blow it self up. So I took out the capacitors I had added. They had me change the resistor from output of the op-amp to the gate to a 300 ohm and the sense resistor to a 10K. The the most critical change was 100pf capacitor that runs between output and the current sense line. For the first time today I was able to get to 60 volts and 1.3 amps with out a bunch of smoke blowing in my face. I did add a bit of a variable to the mix though since I was all out of my original FET I had to use an IRL2910 (CBA Fet) and drop my op-amp voltage to 5V. Though as stable as it seemed I think it will be fine once I go back to 10V and get my 200v fets in. If all goes well I might even be able to do something like 150w per device with the new .3c/w junction to case number, if my big sink can handle it.

Now that this problem is sort of solved the next step is to try my hands at CNC'ing a PCB board. Then I have to make 10 of them. Totally stoked to keep this moving forward! It might even be really cool to end up with a totally modular setup where the user buys 150w blocks that tie into a central setup.

[Alan B]

Great progress!

What are you using for PC layout? I was looking at diptrace and they mentioned a way to convert a layout to a milling g-code. There's a free version of diptrace available. I haven't tried it yet but have heard some good things about it.

[magudaman]

I am using Eagle. It was fairly easy to learn and it looks like most board places like it too if I want to go that direction. Eagle has what is called PCB-Gcode add on that I still have to learn. It doesn't look very user friendly, it command line as far as I can tell. There is also a cool board place that apparently you can get boards for around $1.70 per square inch with no setup or anything: http://dorkbotpdx.org/wiki/pcb_order

[Alan B]

You might check batchpcb.com, they have good pricing and are Eagle friendly. They collect a bunch of boards onto a larger panel and send it out so it takes slightly longer but the quality and pricing is good. Looks like your suggestion has better pricing, don't know much about them though.

Eagle is common but a painful program to learn and use and expensive to do anything commercial with, so I keep looking for other choices.

[magudaman]

You might check batchpcb.com, they have good pricing and are Eagle friendly. They collect a bunch of boards onto a larger panel and send it out so it takes slightly longer but the quality and pricing is good. Looks like your suggestion has better pricing, don't know much about them though.

Eagle is common but a painful program to learn and use and expensive to do anything commercial with, so I keep looking for other choices.

It took me a little while to get it but once I built a circuit all the way through I feel I like I have it down fairly well. I feel like even though I just started with it I could build a simple circuit in about 5 minutes.

So I decided not to go the CNC route and instead tried for the first time making a board from scratch with the Toner Transfer method. Considering its my first board it turned out pretty good. In fact I made all the traces way too small for this method but they all ended up working. This is 2oz board. It was a struggle to solder up correctly since there is no silk screen and everything is extremely tight, I ended up with a bunch of bridges I had to find and clean up.






There was only one mistake that reared itself today and it turns out I had it wrong in my schematic. It was an easy fix on the prototyped board and my next batch it will already be fixed. I also got and attached the new 200v FET to my circuit and by golly it works beautifully!

Here is a little video of it at 100w at around 60V:



So now I should be able to make my 10 boards and order all of the components from digikey.

[bigmoose]

Don't know if you gents know about the ACDC_EL750B Load. Rated 750 watts 150 amps max, 55 volts max. It will do constant current, or constant resistance and is externally programmable. You can parallel them also. They are not in production anymore, but there are some floating around from time to time. I snagged 5 of them during the past year for a project that I finished up around Christmas. If a couple of these would help someone out, let me know.

[CamLight]

Eagle is common but a painful program to learn and use and expensive to do anything commercial with, so I keep looking for other choices.

I've used it for all my commercial stuff for years now (several dozen boards). While it was...umm...interesting, to learn IMO it's a great package for the price. I haven't found any other that gives you that much for the dollar without being just as tough to master.

But, many people do find it exasperating to use so I can understand that you're still looking.

[CamLight]

So now I should be able to make my 10 boards and order all of the components from digikey.

Great news, congrats!

[Alan B]

Comparing prices, diptrace is cheaper than eagle for similar capability, though comparisons are not exact.

Reports are that diptrace is easier to learn compared to eagle, and not a command line program with graphics glued on like eagle but an integrated graphical system. Diptrace supposedly has a better system for pad layout and multiple component placement with array feature.

One problem with eagle for a "for profit" but occasional user the pricing is pretty high. "Light" versions of both programs are under $100 but standard versions jump to $900 for eagle vs $350 for diptrace, and Pro versions soar to $1600 for eagle and $700 for diptrace.

I've installed diptrace to give it a shot. Free version is 300 pins, 2 signal layers, non-profit. Includes schematic, PCB layout, parts libraries and autorouter. G code converter to make pcb on CNC router is free.

http://www.cadsoftusa.com/shop/pricing/

http://www.diptrace.com/order.php?en=1


Nice progress on the discharger.

[CamLight]

Comparing prices, diptrace is cheaper than eagle for similar capability, though comparisons are not exact.

Looking forward to hearing what you think of DipTrace! I've heard some not-so-good things about DipTrace (in addition to Eagle) but it's such a personal choice for so much of it. Trying both programs, all of them actually, is really the best way to go IMHO.