Cheezeball DC Load: DL24P: Pump, or Dump ???

[Pukker]

Quite a buzz and a bang. Even the cat ran so hard in place that it didn't get moving right away Discharge was less than 90w and contact for cooling good.

Would you help a little bit of what parts I would order here to make it work at even 2/3 of the promised power?

(Attachment Link)

I think replacing the MosFet with an original one, ordering from an good supplier, and an 15V zener across gate and source is the best to start with.
I had blown my  MosFets (have PX100 and DL24) also, but after replacing no problems with reduced power. (80 Watts)
The complete DL24** designs are not very well designed for extreme powers and currents. Small screw connections, pcb traces and so on.
When talking about specifications the Chinese designers / sellers often see Double.

[Anomalia]

I think replacing the MosFet with an original one, ordering from an good supplier, and an 15V zener across gate and source is the best to start with.
I had blown my  MosFets (have PX100 and DL24) also, but after replacing no problems with reduced power. (80 Watts)
The complete DL24** designs are not very well designed for extreme powers and currents. Small screw connections, pcb traces and so on.
When talking about specifications the Chinese designers / sellers often see Double.

Ordered FDL100N50F and 15V zener, planning to increase cooling more.
Small fuse in load wires? Selected according to the type of load. Would that prevent damage?

[Pukker]

Ordered FDL100N50F and 15V zener, planning to increase cooling more.
Small fuse in load wires? Selected according to the type of load. Would that prevent damage?

Yes, a fuse is always safe. It wil maybe not always prevent you MOSfet,
but when your fet is going defect, that is mostly an shortage,
that can give much damage when you are testing high current capable batteries.
The zener is also OK, peaks (from the load itself) on the gate can break them.
FDL100N50F is power enough.
Hope you get him from stock, Farnell starts shipping in 2-2023

[SpottedDick]

Has anyone around got broken DL24s they don't want?

I've two here I blew the MCU on fucking around and ATorch are so fucking slow to respond. I can replace everything on this board locally except that fucking MCU.

Trying to see if they'll sell me a few preprogrammed since I doubt they'll share the binary.

[SpottedDick]

I have used the device (dl24p) at low power and voltage without any problems. I then thought about measuring the battery capacity of an electric bike.
The battery was 36v (+ 40v full)

Quite a buzz and a bang. Even the cat ran so hard in place that it didn't get moving right away Discharge was less than 90w and contact for cooling good.

Would you help a little bit of what parts I would order here to make it work at even 2/3 of the promised power? The voltage range I need is wide.
Probably a zener, what about a mosfet? Would it be FDH44N50. I read somewhere that the standard IRFP260N (certainly original) has also been used and the device has worked well since then.

(Attachment Link)

You're really pushing the design nearing 40V. I don't suppose you where logging the temperatures? If it was only doing 2A, I would suggest thermal issues. 80W is really the maximum of these with the default design.

[Anomalia]

You're really pushing the design nearing 40V. I don't suppose you where logging the temperatures? If it was only doing 2A, I would suggest thermal issues. 80W is really the maximum of these with the default design.

If I remember correctly ~ 85w / 2A. I first tried 1A and then raised 2A. After trying that nothing overheated, I left the room for a while and then started happening 
I didn’t log the temperatures, but I used a thermal camera, nothing looked too hot at the time.

I assume that after mods with better cooling you can discharge with a higher current at a lower voltage.

What is the safe voltage range, under 36v?

[Pukker]

What is the safe voltage range, under 36v?

I think (and used at higher voltage >60Volt) that the input voltage is not an great problem,
but the spikes on the gate (depending on the input voltage???) are more an problem and risk.
So a 15 volt Zener over gate-source is recommended.
Use the unit mostly to test batteries and powersupply with Testcontroller.
Normaly I don't go further than 80 Watts and I check the Temperature with Testcontroller,
wich I let shut it down when it goes too high.
For the price I like the DL24. When used with some attention it works fine.
You can do calibration, but even without doing that it was pretty accurate.
Maybe somebody need extreme currents and so on, but would and should you do that??
After replacing the Mosfet with an original one and adding an Zener, I like using the unit.
BTW. The earlier PX100, recent ones (version 2.4 and higher), with replaced MOSFet and Zener,
also working fine. But there are fakes on the market, which are dangerous and worthless.

[SpottedDick]

Am I right in saying the DLP24 seems to be using a solely software controlled gate which is reading back values from the power chip (marked ADC in a diagram above) and working off some fast PID like control?

This opposed to what we usually see on CC MOSFET designs which feed a voltage into an op-amp and use the shunt feedback to "balance" it?

[thm_w]

Am I right in saying the DLP24 seems to be using a solely software controlled gate which is reading back values from the power chip (marked ADC in a diagram above) and working off some fast PID like control?

This opposed to what we usually see on CC MOSFET designs which feed a voltage into an op-amp and use the shunt feedback to "balance" it?

Schematic was posted https://www.eevblog.com/forum/testgear/cheezeball-dc-load-dl24p-pump-or-dump/msg3516952/#msg3516952
Opamp is always used in the loop.
There may be some PID type control for constant resistance or power, as that requires constant measurement and adjustment.

[lordstein]

After 20 minutes at 20v and 4.4a I reached this temperature, it seems to be better designed this time.  it is the dl24mp-h version, the transistors seem to be used or recycled

[lordstein]

After change heatsink from amd pc cpu.after one hour

[stts]

I got one of these new Atorch units too and it didnt last very long. I tested my power supply with it. pulling an easy 1.9A at 52v. Putting a 100watt load on my power supply. The power supply didnt mind, but the Atorch load eventually quit. Temperature was about 45 degrees C and this Atorch had 4 of those Mosfets to share the load. But they eventually went to short. It was disappointing. It looked like Atorch had the problems figured out, with all these circuit changes. The picture shows it working at first. CC 1.9A was actually kept at 1.9A. But after awhile, a short crashed the voltage to very low and Current pegged at 2.05A. The heatsink also went cold. There was no fire or anything. The power supply was current limited so the shorted Atorch transitors were perfectly happy sinking the 2.05A that they were geting. If I was testing this with a battery, the 30 amp plus current surge would have sent a lightning bolt and Atorched my unit. I filed a complaint with the seller to see what they will do about this.    I looked inside the display unit and did anybody know there's a spot to solder a possible USB connector too? If the firmware is still loaded, maybe this unit can talk to a PC without a blue tooth. I have to wait till Aliexpress tells me what is to be done.

[thm_w]

Shame, I didn't go above 80W in my tests, but I'd assumed 100W might be ok.
If the 45C rating is from the display then its not very accurate, its just an NTC on the PCB as mentioned above. Real heatsink temperature is probably 65C, which still isn't that high.

Can you measure all of the large load FETs, are any shorted?
What are the part number of them?

[stts]

Well they changed heat sink and fan. But they still say its good for 150 watts. That heat sink looked alot smaller than the original one. But, with all the circuit changes, I figured that they must have made the drive section more efficient. The sink was hot to the touch, but its not like you get a blister touching it. So this is most definitely a disappointing outcome. I measured all 4 Mosfets from the back and its hard to say without actually taking them off. And I cant do that till I hear what Aliexpress will do about this. But being in parallel, and only getting 1.9 amps thru the failure, its possible one fet went down and is now sinking all the current. It looks like they each have their own gate drivers, so whatever went wrong, it could be limited to just one of the 4 circuits.

But one thing is for sure. The supply was only providing 1.9 amps. So a funny current spike is not the problem. And the supply is not capable of producing a voltage anywhere near 200volts. The Vds of these fets. So we are left with internal circuit reactive harmonics to produce voltage spikes, or cheap MOSFET fakes that cant go anywhere near the units rated current. The MOSFETs are IRFP260M. 4 of them. And something happens to blow them up. And its got nothing to do with current, since these are supposed to have no problems dealing with 1.9 amps. They don't look like fake FETs to me, so I am thinking about that high voltage gate noise that has been reported on earlier. Vgs of these has to be kept below 20v.  Spikes must still be happening. And its eventually causing the source and drain to short. The display resistance crashed from 25 ohms to like 0.8 ohms. Thats a pretty hard short.

The control section is a high speed digital circuit, so its noise activity could be whats bleeding into the gate control section. And the bluetooth being moved to the display had no effect in fixing this problem. Oh, and there is heat sink compound being used. I can see abit squeezed out from the edges.

[capt bullshot]

The MOSFETs are IRFP260M. 4 of them. And something happens to blow them up. And its got nothing to do with current, since these are supposed to have no problems dealing with 1.9 amps.

If one takes a look at the datasheet (there's the original IRF datasheet somewhere around, I've attached it for convenience), one can see this MOSFET isn't rated for linear operation at all. It's common but not widely known issue with certain generations of MOSFETs that they'll fail at low power dissipation (way below their max. ratings) when operated in linear mode. These MOSFETs are composed of a huge amount of "small" FETs on a die, and in some operating points they have a negative temperature coefficent. This leads to current hogging into a few FETs only, these get into a thermal runaway state and die. You'd have to decapsulate the chip and look for local overheat damage.
The SOA diagram gives some hints, for this particular rating the longest applicable pulse is 10ms. For a linear rated MOSFET one would find a line for DC here.
So it's pure luck if these MOSFETs are going to survive linear operation, some will, others won't. Some will stand higher power dissipation, others will die early.
The only way to ensure reliable operation is to use a proper linear rated MOSFET.

[Grandchuck]

This article has a good explanation:  https://eepower.com/technical-articles/understanding-linear-mosfets-and-their-applications/

[stts]

Hah, well that just figures. It didnt take me long at all to realize that there was a problem in Huston. So I cant help but wonder whats going on in these Chinese clean room factories. We sold off pretty much our entire electronics industry to the whole of the Asian world. And this sadly botched clever idea is what we are getting back. Well, it would seem that faulty part application is whats making this blow up. And even if only one of the 4 fets went bad, its a very bad idea to hook a lithium battery to this without better parts.

Thanks for this info. It was an eye opener. Indeed the "hint" burried in all that glowing description of these IR parts is easy as pie to overlook. I wouldn't even blame the chinese for missing it. But testing would have poped red flags all over the place. Aliexpress came back with a trick proposal. No refund and send them a video. They dont believe the pictures. And they didnt give me an address to send a video that will be many hundreds of megabytes. So the back and forth of rejections is in play. Ill make a video, but I reject that the refund be zero. And thats what the trick is really about. Get me to click accept while the refund is reduced to zero.

Aliexpress is full of tricks. Thats our punishment for giving away our tech industry.

[thm_w]

As I mentioned above, the FETs I received were 38N30's which have some specified level of DC SOA. You may have got unlucky with the IRFP260N's.
We don't know though if/why your FETs failed. Resistance of gate to source and drain to source will tell you if they are good, in most cases.

Use Windows video editor to reduce the length and size of the video to a minimum, or record at your cameras lowest resolution. As long as you can see the text on the LCD and whats going on, its good enough.

[stts]

Atorch finally responded and posted a mediafire link to post a video to. So I made a video at the lowest resolution that still looked clear. It was still 275mb, but the mediafire took it. Without stopping, I ran a kluge of ebike lights for a bit, then unplugged them and plugged in the Atorch. You can see the voltage on my supply short down to very low and the current to peg. Its current limited so my supply shouldn't blow up. And I grabbed the heatsink with my fingers and the fan wasn't even running. Got all that on one video, so it should be clear to Atorch that their unit is shorted out. Also videoed the transistor numbers on the back, so they can see what I have that shorted out. If they had better tranys for this unit, then its clear that I didn't get them.

So Atorch went and did a major redesign of their expandable battery load and its still got the same botched up problem. Makes you wonder what kind of engineers they got working for them. I know that when a Chinaman was in any of my classes, we all had to bust our butts studying because they ruined the grading curves. Aliexpress is a pain to get stuff from. One set of LED ebike lights was really Xmas tree bulbs hot glued in place when I opened the glued down back cover. The vendor sweared they were LEDs, like I was elstupido. Aliexpress refunded me. The freewheel sprocket I order with American threads was actually a freewheel with European threads. They are slightly bigger diameter, so I may just JBweld it on. It takes a month to get stuff from Aliexpress. And Aliexpress wont take my credit card any more. They didn't like the drivers licence, birth certificate, and social security card I sent them. Now they want me to post more credit cards. I think I already sent them too much of my personal info. They could make a clone army of me now and make me the arms dealer for the Russians. I think I will cancel my credit card and get new account numbers. Aliexpress is way more fishy than Amazon.

Here's a pic of Xmas tree light bulb Ebike light on the left. I lit the bulb up low so you can see the filament shape in the pic. Then LED light on the right. I had to send that to Aliexpress when the vendor swore that I got LEDs. There's plastic studs to screw down an LED PC board, but they hot glued those bulbs in instead. Chinese enginuety I guess.

[stts]

Atorch never did get back to me and Aliexpress jumped in and denied my claim of refund. They say there is no evidence that my unit is malfunctioned. They say if I can post evidence that it is malfunctioned, then they could reconsider their ruling. Aint that sweet. Can you guys tell from the pictures that this is shorted? And I sent them a video too. Well it looks like Ill have to make another video. This time of me hooking a battery to this Atorch. Then I can have a video of an explosion. Hard to say if that will change their mind...

[stts]

Been getting the run around on this Atorch. They said they needed a video as more evidence that the unit is defective. But their video upload button was defective. It acted like it uploaded, but my bandwidth meter said hardly anything was sent. And it wouldn't let me send it again. So I waited till the timer timed out and then they denied my complaint for lack of evidence. I filed an appeal and told them their complaint system is broken. Then posted a link where they could see my video. I didn't catch it on fire yet. I just fed it my 2 amp power supply and it arced and sparked rather well in the video. Just what a dead short would do. Even did it with the display unplugged. Shorts don't care if the computer is powered up. That got a reaction from Aliexpress. They offered me $11.50. Hah. Or a full refund if I sent it back to china. Bwahahahaha. Crazy huh. I told them I'm sending it back. I was fed up. I didn't yet know how I was going to send it back. This morning I got another email. They reconsidered again and are giving me a full refund since it will cost more to ship than the refund is worth. Geez. It took them long enough. What a hassle. Anyways, I had plenty of other things that I'm buried in, so it didn't really hold me up. I was going to use it to test ebike batteries but my battery order got cancelled. Seems they sold more than they had. So no battery and no tester. Back where I started. Ill wait till the bank says I got refunded before I figure out whats next.

[stts]

I got my Atorch running again. Only one of the 4 fets were shorted. The others were unleashed when the bad one was removed. And I got it sinking my power supply again but only at 80 watts. It blew up at 100 watts. I got a good look around at the unit. Its got separate op amp circuits running each of the 4 fets. So bigger fets could be used to make this a 2000 watt power sink. Like that IXTK90n25L2 fet that's good for 550watts in the SOA. But the current goes up huge too. So the entire current section would have to be disconnected and a high current equivalent would have to be built on a separate board. Complete with separate monster terminals. And monster heat sinks would have to be used also. I found a web page that calculated heat sink sizes for dumping watts. A 500 watt heat sink that is briskly fan cooled has to be 8x5x2 inches in size. And shaped properly so a point heat source will spread out to use the whole heat sink. Meaning a thick copper plate will have to be stuck on the bottom of a flat heat sink to spread the heat out. Maybe even with liquid metal. And then you need 3 more of these for 3 more fets to use all of the Atorches circuits. So its all quite possible on this model but the expense and extra work would be getting way up there. But it would be robust. Oh and a separate fan driver circuit would likely be needed to drive the 4 bigger fans that would cool these big heat sinks. Those circuits are the cheapest part of this expansion project. But most likely a necessity.

I used the heat sink calculator and figured the heat sink on my unit was only good for 100 watts. But there was only a dab of sink paste on the 4 fets. Only covering about half of the fet surface. So its pretty clear that the unit was not quite up to sinking even 100 watts. And that's why it failed after 2 days of use. So I got cool master paste on them now and the thermister right on the metal of the fet. And the temperature stable at 65 degrees. So the 3 remaining fets can still get this unit over 200 watts. With better heat sinks. I have several cpu fan sinks with copper inserts I never used. Ill drill a hole and mount the fets on those. Then make a separate board to mount those on. Since these fets are only good for 75 watts, these surplus sinks should be right in the sweet spot to get this usable. Then I run it for a week and see what happens. Its the most sensible plan with the price of things these days. There use to be a monster electronics surplus in town for 50 years. But the owner got old and closed it all down. Now we only got the Chinese. A huge hassle.

These are still not the proper fets for this kind of application. But 3 still work and I have what I need. So its worth the work for this situation. And being separately mounted, if they blow up with a battery, then they cant do much damage. Then Ill decide then what's next. I could still get better parts and only run them to the limits of what I have.

[yuhar]

What about the gate resistor, how much?
LM321 can drive 20-40mA maximum; this is not too much.

How much is the gate voltage? Gate treshold voltage of the FET!?  Has anyone checked the linear operation with an oscilloscope?

Large FETs have large input capacitance, need more charge to open...

Maybe this "slow", long open-close cycle (pwm signal) and the large induktance in the circuit generate high energy destroy the FETs?

Somebody thinking about snubber-capacitance? If we cannot make the switchig faster?   


Why is there a series diode in the circuit?
- Safety for reverse polarity connection?
- lots of loss on this smd diode, fuse maybe better? 
- remove the diode completely,  or using (anti)paralell to help the FET

[stts]

I dont see much problem with any of your concerns. This is not a switching circuit. Its a linear circuit controlled by a steady DC voltage to get a steady DC current. Both mosfets have similar capacitances for what we are doing. You turn on the transistor and it charges once. The smaller fet has 4nF the larger one has 23nF. The LM321 will have no problems charging them once. And in the linear region, the max voltage for max current of each is roughly 7 volts. We wont be commanding anything more. And why would you scope the linear operation? The data sheet clearly charts the linear SOA. So just tune your circuit to work within that envelope. If you plan to use these bigger fets to their maximum current capacity, you have no choice but to duplicate the high current section on a separate PC board with much bigger parts and wiring. You will need to choose a big big shunt and change Atorch parts to deal with the different shunt voltage drop. Just hunt down the resistor on the op amp that runs to the current shunt. I would replace it with a pot to make tuning adjustments. I don't know what the max wattage and current display will show, but you may have to tune your high output 2000watts to be commanded when the displays max is programmed. So if max is 1000watts on the display, then adjust the pot to command 2000watts out of the 4 Mosfets. Each one is good for over 500watts in the linear SOA.

Oh. And that shotky diode would likely need to be cranked up in amp capacity as well. As for fuses, I would fuse each of the Mosfets at nothing more than 90 amps fast blow fuse. If you have a mosfet short out with a battery being tested, the fuse will blow before the mosfet catches fire. And that just shows you that each fet wiring will have to be able to handle 90 amps. Clearly nothing on the Atorch board can handle that kind of current. So you see, it will be rather costly to be able to make this unit sink 2000watts. But I may be worth it to some people. Those Fets are about $35 each, the heat sink likely over $50, 4 fans add more, 100amp fuse blocks more, high current shotky more, big shunts more, big terminals or Jumper cables more. Hah. yup, jumper cables. And its not getting more portable either.

[yuhar]

- just ordered one DL24 board, I want to make some improvements.

I am a little bit confused after read this :
https://www.sevarg.net/2018/06/24/the-atorch-purple-fan-mosfet-destroyer/
but that is one other circuit...

- found answer "#65 mawyatt wrote" gate resistor is 1K...

- thinking to using IGBT has more SOA, and cheaper, than linear MOSFETs, antiparallel diode and fuse need.


To #11 ptluis: the ADC is RN8209C, - I think you already know.
Here is the english manual http://dev.ifra.jp/pdf/2018/0/1.pdf
the newer chineise: http://www.renergy-me.com/upfile/2019053116432630.pdf