JP spot-welder, FET-switched, timed adj. pulse

[Arlo1]

Just get them from the same batch. You could test them, but there is more margin for error in that process than just using ones from the same batch.

I disagree. If you can set up a test system to make sure you measure them at the same voltage applied to the gate while they are the same temp then its quite accurate.

The way I do it is to turn them on about 1/2 way and measure the resistance then write the number on the fet and put them in groups with the closest numbers!

BTW I use a lab supply now...

[youtube]mYcjsMmaWzc[/youtube]

 

[riba2233]

Hello everyone,

I have bought a spot welder to Ribas, and everything worked well, 'till now, since the last week i'm having problems with the mosfets, i blow 2 of them 1 week ago and i replace them, now yesterday i just blow more 3, i don't understand why is that happening 'cause i'm doing everything normal (i'm welding 18650 samsung cells), does anyone know what's possible going on to blow the mosfets?


Thanks

People, please write what battery you are using for welding power, it makes tons of difference!! I've only seen mosfets blowing from too powerful batteries.

 

[Jorge Rocha]

People, please write what battery you are using for welding power, it makes tons of difference!! I've only seen mosfets blowing from too powerful batteries.

I'm using a 12V Pb battery, 90Ah 20H 720A.
But since i bought the machine i'm using this battery, the problem with the mosfets only started now...

Thanks for your answers guys

Hope you get better soon Ribas [], be strong.

 

[fechter]

It's possible that the FETs are somehow deteriorating with use and failing. If one goes or gets lazy, then the load is shifted to the remaining ones and they are overstressed. If one fails, I would suggest replacing the entire set rather than just the one that failed. Also make sure all the screws on the FET connections are tight. They could possible loosen over time.

Adding some resistance to the path would limit the current. Just use a longer wire somewhere.

 

[PPS]

Hi there everyone, I was watching the forums last year when the 12V spot welder was being sold.
It was temporarily unavailable at the time, and now with riba223 out of action, it looks as if it is once again.

Does anyone have a link for buying one of these types of welders, or info on building one myself - I'd prefer to buy one of riba223's units, or something similar, but is this going to be possible?

???

 

[pguk]

Today I installed a flyback diode on the welder. I used 2 IRFP4368 in parallel, which are TO-247 size 75V MOSFETs. I have 6 of auirf1324s-7p (24V rated surface mounted) as my switcher FETs, 75R gate resistors.


Here is the shot at Drains at turn off without the diode in circuit. We hit 27 volts for a period of 69.2us with ringing.



And with the diode in circuit. Time at 27 volts is reduced to 22us. Ringing is also attenuated quite a bit.

Thats a significant improvement - 69us down to 22us is less than a third. Suffice it to say I think I am sold on the flyback diode. I may invest in a TVS diode to see if I can eliminate that over voltage altogether.

Edit: same test setup with a 5KP12A TVS added:



A further improvement. Flyback voltage is now just a little over FET rated voltage (24V) for 7us. I think the TVS must be overloading because the spec says maximum clamping voltage should be 19.9V with a max pulse current of 256A. The breakdown voltage of this 12V device is between 13.3 and 14.7, I tested it at 14.2V. Which means anything more than a trickle charge will cause it to conduct.

 

[okashira]

Instead of FETs, use a nice Schottky diode, my avalanche is way shorter then that with flyback alone. With tvs diode , avalanche is basically none
Also make sure your flyback circuit is as short as possible.

 

[pguk]

I will try this, thanks okashira

 

[okashira]

Here is mine before and after.
went from 110us to 20us avalanche time. Mine is longer initially because i'm running long leads and 3,600 amps.

 

[pguk]

So I re read through this thread with my eyes opened to this diode solution, till I came to where you posted this at bottom of pg 15. Your photo of Test Setup for this scope shot looks to have a FET diode, and not your original Schottky diode? I have a TVS on the way.

Edit: No, it just looks like a FET usually looks.

 

[okashira]

So I re read through this thread with my eyes opened to this diode solution, till I came to where you posted this at bottom of pg 15. Your photo of Test Setup for this scope shot looks to have a FET diode, and not your original Schottky diode? I have a TVS on the way.

I have never used a FET diode.
The diode I am using now kinda looks like a FET I guess with its funky tabs.

 

[fechter]

I wonder if placing the welding cables close together would reduce that wicked flyback? If the welding cables are spread apart, they resemble a physically large, single turn inductor. If you lashed the cables together so they're touching most of the length, then when the magnetic field collapses, the EMF generated will be partially cancelled out by the field from the adjacent wire. More like a twisted pair or radio feedline.

Since the weld electrodes are right next to each other, no reason the cables couldn't be taped together most of the way.

 

[okashira]

I wonder if placing the welding cables close together would reduce that wicked flyback? If the welding cables are spread apart, they resemble a physically large, single turn inductor. If you lashed the cables together so they're touching most of the length, then when the magnetic field collapses, the EMF generated will be partially cancelled out by the field from the adjacent wire. More like a twisted pair or radio feedline.

Since the weld electrodes are right next to each other, no reason the cables couldn't be taped together most of the way.

I think that would help a little bit. But with flyback diode and TVS diode, the problem seems to be completely solved anyway.
Diode I am using now is like $3 and TVS is like $2.50.

 

[fechter]

I think that would help a little bit. But with flyback diode and TVS diode, the problem seems to be completely solved anyway.
Diode I am using now is like $3 and TVS is like $2.50.

I agree, but I'd be curious to see how it looks on your scope anyway. If just bundling the wires together solves the problem, that might be easier (though I would still use the diodes).

 

[flangefrog]

Have I got this right? As I understand it these are the important specs:

Schottky:
IF Forward Current: Higher is better
VF Forward Voltage: Lower is better
VRRM Repetitive Reverse Voltage: Above max battery voltage
IFSM Forward Surge Current: Above max current used

TVS:
VRM Working Voltage: Above max battery voltage
VCL Clamping Voltage: Below MOSFET Vds (Drain-Source Breakdown Voltage)
VBR Breakdown Voltage: Lower is better
IPP Peak Surge Current: Above max current used (but when used in combination with the schottky can probably be much lower)
PPP Peak Pulse Power: Higher is better
Polarity: Either but unidirectional is best

Good schottky diodes:
Maybe this one? Package probably not so great: http://nz.mouser.com/Search/ProductDetail.aspx?R=VS-MBR6045WTPBFvirtualkey61370000virtualkey844-MBR6045WTPBF
http://nz.mouser.com/Search/ProductDetail.aspx?R=VS-100BGQ015virtualkey61370000virtualkey844-100BGQ015
http://nz.mouser.com/Search/ProductDetail.aspx?R=VS-100BGQ030virtualkey61370000virtualkey844-100BGQ030
http://nz.mouser.com/Search/ProductDetail.aspx?R=VS-100BGQ045virtualkey61370000virtualkey844-100BGQ045

Okashira, which TVS diode are you using? This one seems suitable for 40V MOSFETS (IRFB7430PBF) with a 12V lead acid battery: http://nz.mouser.com/Search/ProductDetail.aspx?R=BZW50-15virtualkey51120000virtualkey511-BZW50-15
And this one for 24V MOSFETS also with a 12V battery, although it only has a 600A IPP: http://nz.mouser.com/ProductDetail/Vishay-Semiconductors/5KP14A-E3-54/?qs=sGAEpiMZZMuNo3spt1BaV%252bLr%252b0awzgKwzIIeWXFPTcY%3d

Edit: something I've learnt is that the TVS pulse current ratings as shown at electronics retailers (as well as working, clamping and breakdown voltage) are often incorrect. For example, the 5KP14A is rated at 216A, not 600A. Have a look at the datasheets for the proper values.

 

[flangefrog]

Now onto mosfets:

For a setup without schottky or TVS diodes the most important characteristic is the avalanche current (or avalanche energy?). Once you add those diodes it becomes a non-issue?
For a diode protected setup the mosfet selection isn't so critical but the RDS(on) and IDM are the most important - possibly also QG.
Is lower QG better for reduced switching losses? Or higher QG to reduce dV/dt?

Note that some characteristics are linked to the VDSS. If two mosfets have the same EAS but the first has a higher VDSS then the second probably has a higher IAR and will be better.

VDSS Drain-Source Breakdown Voltage: Above max battery voltage. Lower VDSS can mean better specs elsewhere.
RDS(on) Drain-Source Resistance: Lower is better to reduce heating
QG Gate Charge: Lower is better?
ID Continuous Drain Current: Higher is better
IDM Pulsed Drain Current: Multiplied by number of mosfets this must add up to more than the max current used.
EAS Single Pulse Avalanche Energy: Higher is better
IAR Avalanche Current: Higher is better
EAR Repetitive Avalanche Energy: Higher is better
RθJC Junction-to-Case Thermal Resistance: Lower is better so it can get rid of heat faster.
PD Maximum Power Dissipation: Higher is better

Some good mosfets:
http://nz.mouser.com/Search/ProductDetail.aspx?R=IRF1324PBFvirtualkey57370000virtualkey942-IRF1324PBF
http://nz.mouser.com/Search/ProductDetail.aspx?R=IRFB7430PBFvirtualkey57370000virtualkey942-IRFB7430PBF
http://nz.mouser.com/Search/ProductDetail.aspx?R=IRF1405PBFvirtualkey57370000virtualkey942-IRF1405PBF

 

[okashira]

I use the 5KP10A-E3/54 for TVS diode. You can choose 11v or 12v version for lead acid. Maybe double them up.
I have actually tried both of those mosfets you suggested. (other then the 1324)
They blow up pretty quick maybe last 2-3 welds.

the 1324 is the only fet that can last


This was without flyback and supression though. I am sure they would last longer with the diodes.

 

[flangefrog]

The highest breakdown voltage available with the 11V and 12V TVS diodes is 13.3V. Wouldn't this mean that when used with a 12V lead acid battery that is charged to 13.8V it would be passing current whenever the probes were touched together, even if the mosfets were off?

I wasn't exactly trying to recommend those mosfets, but I have read that they have been used successfully without protection diodes by others who are probably using lower current. My intention was to summarise all the information I've gleaned from the spot welder threads so I can get feedback on whether it's all correct.

Has anyone tried a metalized polypropylene cap installed where the schottky diode is? From what I've been reading about them they have a very low impedance and should help tame the inductive spikes. Some ebike controllers use them (or cheaper polyester ones) on the load side of the mosfets for the same reason.

 

[okashira]

The highest breakdown voltage available with the 11V and 12V TVS diodes is 13.3V. Wouldn't this mean that when used with a 12V lead acid battery that is charged to 13.8V it would be passing current whenever the probes were touched together, even if the mosfets were off?

I wasn't exactly trying to recommend those mosfets, but I have read that they have been used successfully without protection diodes by others who are probably using lower current. My intention was to summarise all the information I've gleaned from the spot welder threads so I can get feedback on whether it's all correct.

Has anyone tried a metalized polypropylene cap installed where the schottky diode is? From what I've been reading about them they have a very low inductance and should help tame the inductive spikes. Some ebike controllers use them (or cheaper polyester ones) on the load side of the mosfets for the same reason.

The datasheet should tell you typical reverse current vs voltage.
I use the 10v TVS diode okay and charge to 10.5V or so.

If you're not happy, use the 13v version. You may need another in parallel

 

[flangefrog]

The highest breakdown voltage available with the 11V and 12V TVS diodes is 13.3V. Wouldn't this mean that when used with a 12V lead acid battery that is charged to 13.8V it would be passing current whenever the probes were touched together, even if the mosfets were off?

I wasn't exactly trying to recommend those mosfets, but I have read that they have been used successfully without protection diodes by others who are probably using lower current. My intention was to summarise all the information I've gleaned from the spot welder threads so I can get feedback on whether it's all correct.

Has anyone tried a metalized polypropylene cap installed where the schottky diode is? From what I've been reading about them they have a very low inductance and should help tame the inductive spikes. Some ebike controllers use them (or cheaper polyester ones) on the load side of the mosfets for the same reason.

The datasheet should tell you typical reverse current vs voltage.
I use the 10v TVS diode okay and charge to 10.5V or so.

If you're not happy, use the 13v version. You may need another in parallel

Couldn't find that in the datasheet, but I understand now it's probably just a small current close to the breakdown voltage.

Avalanche current

In an effort to understand exactly what avalanche currents can be sustained without destroying the mosfets I've been looking at the Typical Avalanche Current vs. Pulsewidth figures in the mosfet datasheets. The mosfets I mentioned in my earlier post all have a 200A avalanche current rating for a single pulse of 100us when starting at 25C junction temperature and with an allowable rise of 150C. Change the pulse width to 110us which is what okashira was getting and it goes down to about 110A or 120A for the IRF1324PbF.

When starting at 150C and with an allowable rise of 25C it goes down to around 20A. The IRF1324PbF has figures for varying duty cycles and while it doesn't change much at 0.01 (I assume this is 1%?), at 0.10 (10%) it goes down to 12A.

Now assuming that the duty on these spot welders is lower than 1% (ratio of avalanche time to non-avalanching time) I think we can use the single pulse starting at 150C figures. This gives us about 120A total current allowed when using a 6 fet spot welder. With an 8 fet we get 160A.

With the schottky diode added the avalanche time went down to 20us. That gives us 480A for a 6 fet board and 640A for an 8 fet.

If the duty cycle could have been decreased to 8.5us with the IRF1324PbF then that would have allowed 1200A or 1600A for a 6 or 8 fet board respectively. Or getting it down to 12us with the other to mosfets would have allowed 840A with a 6 fet board or 1120A with an 8 fet.

Since the voltage will have to rise to above the mosfet VDSS to avalanche, I think the avalanche current will not be the same as the welding current. Instead it will be (welding current * (welding voltage / (mosfet VDSS * 1.3))). The 1.3 factor is from the datasheet and I think it's because the listed VDSS is minimum, not average. 31.2V is fairly consistent with the avalanche voltage okashira was getting. This means when welding with 2000A, a battery that measures 12V under load and 24V mosfets the avalanche current will be about 770A.

Have I been reading the figures and applying the values to our specific application correctly?

 

[Tesseract]

Greetings ES... Some of you might know me from DIYElectricCar; to the rest of you I apologize in advance.

Avalanche current

In an effort to understand exactly what avalanche currents can be sustained without destroying the mosfets I've been looking at the Typical Avalanche Current vs. Pulsewidth figures in the mosfet datasheets. The mosfets I mentioned in my earlier post all have a 200A avalanche current rating for a single pulse of 100us when starting at 25C junction temperature and with an allowable rise of 150C. Change the pulse width to 110us which is what okashira was getting and it goes down to about 110A or 120A for the IRF1324PbF.
...
Have I been reading the figures and applying the values to our specific application correctly?

Firstly, you don't want to avalanche FETs unless your goal is to create extremely fast transitions and/or lots of ringing - the appropriate solution is to use a freewheeling (or flyback) diode from the FET drain(s) to the positive rail (which needs to be decoupled with a decent amount of capacitance). Note that there must be as little inductance as possible in the loop between the input capacitance, the switch and the freewheeling diode to prevent overshoot/ringing during switch turn-off, and the freewheeling diode should be a Schottky type with a "single pulse" surge current rating higher than the expected peak current during welding, though the average current through it will be very low.

That said, if you already have one of these gadgets without a freewheeling diode then there is usually a spec in switching transistor datasheets for allowable avalanche energy (and if you are lucky it will state values for both single and multiple pulses with a duty cycle for the latter). And the amount of energy that needs to be handled during avalanche is described by the equation 0.5LI², where L is the total inductance of the welding cables (and of the positive battery cable(s), too, if there is no decoupling/input capacitance [hence why you need such]!) and I² is the peak welding current.

Inductance of a wire in free space is about 8-10nH/cm, so there won't be much here - probably about 1-2uH total - but the weld current is, of course, quite high and probably not knowable ahead of time as it depends on battery state of charge and all the various resistances in the circuit.

But let's say you have 1uH of stray inductance and 2000A of weld current; that gives you 2J (Joules) of energy stored in the stray inductance. If there isn't a path for this energy to freewheel it will create one, by causing an overvoltage breakdown of the FET (aka - an avalanche). Hence the FET(s) need to be capable of handling more than 2J of avalanche energy.

Attached below is a screenshot of a quick simulation of this circuit in LTSpice. The green trace shows the gate drive signal (1ms on, 1ms off, 8ms on, 1ms off); transition time is 10ns. The blue trace shows weld current - the rather lazy rise and fall time is the result of the L/R time constant formed by the total loop resistance and the cable inductance. Finally, the red trace shows the current through the freewheeling diode - and which more clearly illustrates why I only put one diode on the schematic...

The IRFH6200 FET is a reasonable choice, but I mainly selected it because it is one of LTSpice's default components, not because it is ideal for the job. If you look up its datasheet it says it can handle 780mJ of avalanche energy (ie - 0.78J) so 4 of them in parallel would be okay without the freewheeling diode as long as weld current is below ~2500A. Similarly, I didn't even bother with picking a specific freewheeling diode, mainly because the Spice primitive is good enough here.

Which reminds me - don't replace the freewheeling diode with a capacitor; that will just cause a lot of overshoot and ringing (in other words, instant FET destruction).

If the forum will let me I'll attach the LTSpice .asc file and those so inclined and capable can tinker with different things. I'd rather not have to give a lesson on using LTSpice, but I will point out that the first column of .param statements are what set the on time of each pulse as well as the delay time between pulses and the transition time from on to off or vice versa.

Okay, it wouldn't let me attach a .asc file so I will put it in my public dropbox folder for a while - ie, not indefinitely.

https://dl.dropboxusercontent.com/u/41803783/Double_Pulse_Spot_Welder_1.asc

edit: attempt to fix unattached attachment

 

[Tesseract]

Oh, yeah... Obviously I am interested in a cell tab welder (though mainly for welding copper wires to copper foils for switchmode transformers) and I figured someone on here had likely hacked something together better than that Chinese crap on ebay. I was also hoping that I could just buy one already made and then maybe tweak it as necessary for my specific application, but it seems that is not an option right now so I guess I'll have to make one myself.

And the actual reason for signing up here was to ask: What are you all using for welding tips and tip holders? :lol:

 

[amberwolf]

Okay, it wouldn't let me attach a .asc file so I will put it in my public dropbox folder for a while - ie, not indefinitely.

https://dl.dropboxusercontent.com/u/41803783/Double_Pulse_Spot_Welder_1.asc

I fixed it as an attachment to this post here:

by renaming it (filename).asc.pdf.
View attachment Double_Pulse_Spot_Welder_1.asc.pdf
just download it and rename it, removing the ".pdf" at the end.

BTW, the server had a hiccup, so your image doesn't show:
https://endless-sphere.com/forums/viewtopic.php?f=14&t=82335

 

[Arlo1]

Greetings ES... Some of you might know me from DIYElectricCar; to the rest of you I apologize in advance.

Avalanche current

In an effort to understand exactly what avalanche currents can be sustained without destroying the mosfets I've been looking at the Typical Avalanche Current vs. Pulsewidth figures in the mosfet datasheets. The mosfets I mentioned in my earlier post all have a 200A avalanche current rating for a single pulse of 100us when starting at 25C junction temperature and with an allowable rise of 150C. Change the pulse width to 110us which is what okashira was getting and it goes down to about 110A or 120A for the IRF1324PbF.
...
Have I been reading the figures and applying the values to our specific application correctly?

Firstly, you don't want to avalanche FETs unless your goal is to create extremely fast transitions and/or lots of ringing - the appropriate solution is to use a freewheeling (or flyback) diode from the FET drain(s) to the positive rail (which needs to be decoupled with a decent amount of capacitance). Note that there must be as little inductance as possible in the loop between the input capacitance, the switch and the freewheeling diode to prevent overshoot/ringing during switch turn-off, and the freewheeling diode should be a Schottky type with a "single pulse" surge current rating higher than the expected peak current during welding, though the average current through it will be very low.

That said, if you already have one of these gadgets without a freewheeling diode then there is usually a spec in switching transistor datasheets for allowable avalanche energy (and if you are lucky it will state values for both single and multiple pulses with a duty cycle for the latter). And the amount of energy that needs to be handled during avalanche is described by the equation 0.5LI², where L is the total inductance of the welding cables (and of the positive battery cable(s), too, if there is no decoupling/input capacitance [hence why you need such]!) and I² is the peak welding current.

Inductance of a wire in free space is about 8-10nH/cm, so there won't be much here - probably about 1-2uH total - but the weld current is, of course, quite high and probably not knowable ahead of time as it depends on battery state of charge and all the various resistances in the circuit.

But let's say you have 1uH of stray inductance and 2000A of weld current; that gives you 2J (Joules) of energy stored in the stray inductance. If there isn't a path for this energy to freewheel it will create one, by causing an overvoltage breakdown of the FET (aka - an avalanche). Hence the FET(s) need to be capable of handling more than 2J of avalanche energy.

Attached below is a screenshot of a quick simulation of this circuit in LTSpice. The green trace shows the gate drive signal (1ms on, 1ms off, 8ms on, 1ms off); transition time is 10ns. The blue trace shows weld current - the rather lazy rise and fall time is the result of the L/R time constant formed by the total loop resistance and the cable inductance. Finally, the red trace shows the current through the freewheeling diode - and which more clearly illustrates why I only put one diode on the schematic...

The IRFH6200 FET is a reasonable choice, but I mainly selected it because it is one of LTSpice's default components, not because it is ideal for the job. If you look up its datasheet it says it can handle 780mJ of avalanche energy (ie - 0.78J) so 4 of them in parallel would be okay without the freewheeling diode as long as weld current is below ~2500A. Similarly, I didn't even bother with picking a specific freewheeling diode, mainly because the Spice primitive is good enough here.

Which reminds me - don't replace the freewheeling diode with a capacitor; that will just cause a lot of overshoot and ringing (in other words, instant FET destruction).

If the forum will let me I'll attach the LTSpice .asc file and those so inclined and capable can tinker with different things. I'd rather not have to give a lesson on using LTSpice, but I will point out that the first column of .param statements are what set the on time of each pulse as well as the delay time between pulses and the transition time from on to off or vice versa.

Okay, it wouldn't let me attach a .asc file so I will put it in my public dropbox folder for a while - ie, not indefinitely.

https://dl.dropboxusercontent.com/u/41803783/Double_Pulse_Spot_Welder_1.asc

edit: attempt to fix unattached attachment

Not a bad first post.


I was going to try to sit down and put this all together.... But you beat me to it.
Yes you need something to absorb the freewheeling currents as close to the fets/diodes as possible. So you really want some caps and good ones right on the rails on the boards. Guys I don't know if any of you know but Tesseract comes from DIY electric car forums and he is a guru. I don't know if he can post publicly what he has done but he is instantly one of the top dogs on ES so listen to what he says!

Welcome to this little "electric bicycle" forum lmfao
-Arlin

 

[atomek1000]

Hey guys i've got a quick question. Did anybody had an issue with blown resistors going to mosfet gate leg? I have the 1-gen riba spotwelder with 6x AUIRF1324 and they seem to be ok but i heard a big puff and found out that 2 of this 820 resistors are blown up.

What are these resistors for? Should i just replace them and go on or do you recommend modyfing/checking something before welding again? I heard about diode mod but i have regular car battery about 700A and haven't had any problems for 1 year so i don't think this was issue of fly back current or was it?

Here are photos, i marked resistors with red circle:



Any help will be appreciated : )