Modded Malectrics v4 spot welder - oscilloscope graphs interpretation request.

[pegaz666]

Hello all!
I’m not confident to draw conclusions myself hence this request for your help.

I have just built my malectrics v4 spot welder a bit modified as I’m using 2x FET boards in parallel instead of just one. So I have 12 FETs in total. On top, instead of using the original Infineon mosfets from the project I have used Fairchilds which are used in kWeld spot welder (I saw a higher peak power in their datasheet so I thought since kWeld is a reputable device it will make mine even “stronger” and better?).

To power the welder I use 3s CNHL Lipo 10.000mah 100C (200C burst). I may need to use something more powerful hence 2FET board as to do my 0.1mm pure copper + 0.15mm pure nickel sandwich it takes around 45ms of pulse time. I’d like to go lower with the time to reach ~20ms region.

Someone told me that one gate driver IC is already maxed out to fill 6 gates of original fets while I’m using here 12 fets so I should consider using a second MCP to drive the 2nd FET board independently so that each MCP drives one FET board. Of course I’d need to add extra capacitors for the 2nd MCP. That person wasn’t sure if the second MCP will be really needed for this application though so I was advised to do some oscilloscope measurements to justify that fact but after doing so I didn’t hear anything back from him regarding these graphs…

For your convenience I’m attaching datasheet for FETs used here and MCP gate driver:

kWeld FET:

https://www.mouser.pl/datasheet/2/308/1/FDB0105N407L_D-2312146.pdf

Gate driver:

https://www.mouser.pl/datasheet/2/268/mchp_s_a0001781682_1-2274787.pdf

Below are the results at same scope setting 1 FET board vs 2 FET boards:

25ns:



50ns:



500ns:



From the above i can see myself that for some reason Miller Pateau effect occurs just below 6V if using 2 FET board versus just below 8V when using a single FET board… No idea why such difference and which one is “better”?

The next observation is more obvious: for the gates to reach the full voltage it takes longer when using 2X FET board.
1FET= ~3500ns vs 2FET= ~5000-6000ns?
Again not sure if this is a “problem” in this application or not? I will be doing my welds 5 seconds apart or so.

Shall I be concerned about the above results using my welder with 2FET board and one MCP driver? Original creator said it’s okay for max 2 FET boards but the other person wasn’t so sure hence all the confusion… I just need to be sure that my welder is as reliable as possible.

Last question:
Did I do well by using the kWeld’s FETs instead of original from the project? (Here is a datasheet for the original Infineon mosfet for this project: https://www.mouser.pl/datasheet/2/196/Infineon_IRFS7430_7P_DataSheet_v01_01_EN-3363302.pdf )

kWeld FET:


Malectrics FET:

 

[pegaz666]

No one? ;(

 

[leffex]

I would agree with your assumptions and I don't know the difference why it takes longer. Charge time? However KWeld is really good device and I guess if you upgraded your malectrics then it is upgraded? If you were to use the other mosfets that malectrics recommended then the only difference with your device and a normal malectrics is the mosfets. I would check the mosfet fact sheets and se the resistance while in operation and put in the voltage you use in for comparison.

6v versus 8v. You want to use your spot welder with 20ms. So one is closer to that option than the other voltage?

 

[AllDIYtoyz]

I know this post is a little older, but what/where were taking the measurements from? Was the Probe connected to one of the mosfet gate legs? Or the MCP driver output?
I’m just curious as I would like to take some measurements of my setup..

THANKS

 

[Pogifej]

My Malectrics V4_mod version is finished at the board model level and I've completed some measurements and welding a 5S3P Makita power tool battery.

Power source: 6S1P NESSCAP 3000F 2.7V. Their series resistance is 1.4mΩ.
10A balancer with display, serious cooling DIY.
300A fuse DIY 0.22mΩ
6pcs FDB0105N407L-D I'll try to measure it. Expected value with two busbars is: 1.1mΩ.
2pcs copper busbar for FETs (Have to redesign)
2pcs electrode 7AWG 2x0.9mΩ (I plan to shorten it)

I purchased the empty pcbs and electrodes from the Malectrics webshop and assembled them myself (Half the price, much more work).
I also made the busbars for the supercapacitors and FETs myself.
I measured the output current with an AC clamp connected to an oscilloscope.
I haven't dared to cross the psychological limit of 2000A yet, but with 1800A, which I achieved with capacitors charged to 9V, I welded a 0.2mm NI plate with an active pulse of 15ms and a pre-pulse of 25%.
Each shot causes a discharge of about 50 mV from the capacitor bank.
Unfortunately, the finished battery will not be tested in the lawnmower until next spring.
Monitoring with a thermal camera after the completed battery, the electrodes warmed up to 50 degrees Celsius, and the 3V3 LDO on the Arduino board warmed up to 40 degrees Celsius, while the FETs and capacitors remained cold.
At 20ms serious inductive kickback is already visible, which the 3 TVS can still handle, but peace of mind is better, which is why I want to shorten the electrode wires.
As long as it's not boxed, I can still take other measurements if anyone is interested.
I used an online translator for the translation, sorry if it mistranslates from Hungarian.

 

[A-DamW]

I used an online translator for the translation, sorry if it mistranslates from Hungarian.

Your English translated from Hungarian is very clear and understandable.

I have a Malectrics V4 that I am still setting up.
I am going to power it with a 6S1P Maxwell 3000F ultracapacitor bank.

At 20ms serious inductive kickback is already visible, which the 3 TVS can still handle, but peace of mind is better, which is why I want to shorten the electrode wires.

Shortening is good.
Another thing you can do, is bundle/twist the wires from the power source close together.
And bundle/twist the electrode wires close together.
You probably know this, keeping the wires in close proximity to each other helps cancel out the opposing inductive magnetic fields.

 

[Pogifej]

Thank you for your react. I'll take some more pictures of my setup. My Fet busbar is directly connected to the busbar of the capacitor bank, so I think this part is fine. The connection points of my electrodes are far apart, so I can't twist them, but I can shorten them. I have a practical problem, after each shot my electrode sticks to the NI plate. Sometimes both of them.

 

[Pogifej]

Twisting the electrode wires was not a successful thing.
In the meantime, I gained new experience, I managed to burn out the bottom of a practice cell when the contact was not proper. This was the very first stress test for FETs, and fortunately they survived.
It is instructive to display the waveform of the output current (it's a shame that we don't see such a measurement from someone else), since the second half of the 25 ms pulse is no longer used, so it can be seen that 15 ms is sufficient.

 

[Pogifej]

I managed to measure the D-S_on voltage of the FETs (measured together with the busbars) with a through current of 1000A and an opening time of 5ms.
I don't like the result, which was 1.8V (since I'm using kWeld FETs, 1.1V would have been expected), so I have to rethink the material and size of the busbar.
The FET datasheet shamefully only gives the R_D-S_on value for 50A, so I'm just trying to follow the example of the design that achieves a better value.
This value limits the maximum current that can be extracted from the capacitor bank.
I was planning to try the stainless steel+copper sandwich technique, but first I have to fix this design flaw.