Micro spot welder ?

[Javert]

I will need to spot weld 0.5mm Cu wire to 0.5mm brass sheet and 1mm brass wire. Approx. 200 pieces per day, and so I'm thinking about a suitable spot welder or its construction.
I'm interested in this as a reference piece, the price is insane.
Parameters
OUTPUT VOLTAGE RANGE 0.10-5.00 V (0.01 V Intervals)
OUTPUT CURRENT RANGE 10.0-250 A
OUTPUT POWER RANGE 1.0-650 W
WELD TIME (EACH PULSE) 1-500 ms (1 ms intervals)

After 30 seconds of thinking, I came up with this solution.
Please don't laugh, I present it as food for thought for someone who deals with this issue further.

I'll take a 2.7V 3000F capacitor any as this and this , 4pcs (may be 6 or 8) serial and paralle (5,4V battery) connected directly to the copper strip 20x3mm.
I will use a MOSFET as a switching element on the low side, too direct connect to copper strip, any as IXTN660N04T4 40V 660A 0.85mOhm
I charge the capacitors with a voltage of 100mV to 5.4V and then the MCU turns on the MOSFET for the set time and blows for 1 to x00ms.
Is this very stupid?

 

[KlausST]

Hi,

I miss the current control / limit in your description.

Without limiter you risk
* uneven welding results
* destruction of the electronics / electrodes / wiring / brass sheet / Cu wire

Using the MOSFET as linear limiter may cause a lot of dissipated power. Be sure to stay within SOA.

Klaus

 

[Javert]

Yes, these are things I only thought about in the second minute.
The first thing I thought about was measuring the current
- Measure the voltage drop across the transistor as the largest resistance in the chain. Rejected, Rdson varies with temperature if it is 1 at 25C it is 1.5 at 150C.
- Some kind of HAL sensor, probably slow, I don't know if it can measure 10kHz (10 samples at a ms pulse), or if it would be possible to measure directly from the copper strip
- OK classic current shunt. I think it is reasonable to consider a hundredfold amplification of the processed signal. That is, I can use a 500A current shunt 75mV, something like this.
Yes, this is measurement, not current limiting
In fact, I don't even know if the machines in question can limit the current, i.e. if it is possible to set, for example, a 30ms pulse of 100A.
I just skimmed through the manual, but it looks like only time and voltage can be adjusted. Current max 250A and power max 650W are given by the construction (internal resistance). I could be wrong.
Back to my point
IXTN660N04T4 have SOA DC 200A to voltage 3,75 and 100A to 5,4V, for 10ms DC 500A to voltage 3,5V and 200A for 5,4V.
I boldly believe that two transistors in parallel should cover most of the above even with uneven distribution between transistors.
I cannot judge whether the current from the capacitors could destroy two parallel transistors in the event of a short circuit because I cannot estimate the resistance at the terminals and connections.
anyway, I can't think of how to limit such large currents with sufficiently small resistance

 

[KlausST]

Hi,

Some kind of HAL sensor, probably slow, I don't know if it can measure 10kHz (10 samples at a ms pulse),

according Nyquist then analog signal (HAL, before ADC) should be limited to less than f_fample/2.

would be possible to measure directly from the copper strip

Copper has an awful temperature coefficient. So a lot of error according temperature.

There are current measurement devices. Look at https://www.lem.com/en/current-sensors. Just as an example, for sure there are other vendors.

I have some LEM sensors to sell (NOS).

****
What I don´t understand: when using the capacitors, they will drop voltage according drawn current. This is because:
* ESR (static)
* charge state of the capacitors (increasing drop)

How do you charge / discharge the capacitors?

Klaus

 

[danadakk]

Some designs to look at (not your power levels, but general ideas) :

24 DIY Spot Welder Projects: How To Make A Pro Spot Welder

Making a spot welder takes a lot of work, calculations, time, money, and energy. It's not everyone that tries to make a spot welder by themselves that succeeds,

www.littleloveliesbyallison.com

Obviously problem complex, classic control is PWM and frankly R and L
to meet current SOA limits. Adding to the issue is electrode degradation,
work piece interface/contact conditions.

w/o some form of ballasting one thinks of less reliable operation as well.
What is duty cycle of machine ? In any event I would consult with manu-
facturer tech guys, who many are welder/cutter application experienced.
And they have access to device characterization data, which data sheets
typically a subset of, all important in this application area.

The welding/cutting guys I worked with as FAE always did test beds and beat the
crap out of parts to find failure points.

At 500 mS you are pushing the SOA chart for even larger parts than
you are considering for design. Even C pulsed behavior at these current
levels has to be an issue to achieve consistent results.

The force and displacement capability of the Pico machine I do not see
as anything but straight forward embedded sense/control design. Of
course the algorithm interaction with current control would be inter-
esting. Do folks in this area also monitor T at weld area ?


Regards, Dana.

 

[BradtheRad]

Current control can be by inductive impedance rather than resistive. Your setup can yield a pulse coming from a capacitor, which declines in a few mSec. Create a ramping up effect with a coil (say) 100 uH, 10 gauge wire, in series.

Alter Henry value by adding or subtracting turns until you have the waveform that lasts the correct time, at the correct current level for your task.

 

[Javert]

I put my idea on paper.
Yes, I put the transistors and the shunt resistor on the High side. At 5V it doesn't matter, and somehow it seems more natural to have the valve at the top. The rest is actually just a shot of measuring current and voltage plus MOSFET drivers.
It will probably still need a small MOSFET with a resistor to slowly discharge the capacitors if I have a higher than required voltage.
In general for spot welding. The key role is played by the welded material, its surface, the surface of the electrode and the pressure force. The pressure force of these small machines is usually 30g to 4000g. All of this cannot be objectively measured, and so, in my opinion, the time, tension, and thus the energy in the pulse, and especially its length, are set experimentally.
According to the manual, it seems that long 500ms pulses are used for preheating, typically a 500ms pulse with 0.5V, then a 35ms pause, recharge to 2V and a 5ms pulse for welding.
In the evening I will try to write to the manufacturer and ask what the maximum pulse energy is.

 

[danadakk]

Why not place ballast in each leg of switch fets and double them as current
sense ? Just a thought. Not sure how effective 250 uOhm would be as ballast,
a sim is in order I would think.

The IA across the arc, gut tells me due to current levels and stray L to add
some input protection to it from excess V.....?


Regards, Dana.

 

[dick_freebird]

I have a fat sheet metal spot welder. Its control is "by touch" /
"by eye". But it's all just a very low voltage. very high current
transformer.

There exist thermocouple welders, which might be suited. But
these tend to be meant for higher temp, less thermally conductive
materials.

There are constant current and constant voltage welding
processes. I couldn't say a priori which would be better
suited to the materials systems.

As it's a hetero-joint, have you considered a brazing approach?
Wee drop of brazing paste, put a rack of 'em in the oven, walk
away? Fixturing of course is key. But given the volume said, this
could be more repeatable, process-optimizable and well cheaper
in the end.

 

[KlausST]

Hi,

I only had a quick view:

* It seems that NPort+ and NPort- are floating wrt GND. This won´t work with the amplifiers and gate drivers.
* those capacitors in series may need a voltage balancing circuit. In doubt look for application notes or ask the manufacturer.
* You put a voltage divider and a capacitor at the output of the voltage sensor amplifier. I`d rather put them at the input. But both has pros and cons.
It depends on wiring, noise sources, intention...

Klaus

 

[danadakk]

To Klaus's point about series connection super caps :

https://www.eaton.com/content/dam/eaton/products/electronic-components/resources/technical/eaton-supercapacitor-application-guidelines.pdf

https://www.analog.com/en/design-notes/voltage-balancing-techniques-for-series-super-capacitor-connection-for-max3888689.html

https://www.embedded.com/strategies-for-balancing-series-connected-supercapacitors/

To correct the floating problem a high valued R connection from OpAmp ground back to floating node
should suffice to create a path for the OpAmp bias circuits. Note you have to examine CM limitations
of the OpAmp but thats one way of handling the issue. And of course a low bias current OpAmp allows
for the large valued R.




Regards, Dana.

 

[oscillationeer]

The common solder-pistol is an AC transformer with very low impedance output that
has intrinsic current limiting (in the primary winding and core material). So, if you can
make and position spot-weld contacts for your small joints, a simple trigger pull will power it.

Solder gun

If you want something more elaborate, an SSR can replace the trigger pull, with any
timing you want or need,

Electronics to control high current low voltage is an unnecessary complication.

 

[Javert]

Thank you all for your comments, I will respond to them later.
Now allow me a little reflection, from which I realized what a monstrosity I proposed..
The battery of capacitors designed by me will have a capacity of 3000F at a voltage of 5.4V and a resistance of about 280uOhm if the resistance of the connections is neglected.
It can fit energy W=0,5*C*U^2 W=0,5*3000*5,4*5,4=43740 J or W second.
For comparison here are several capacitive spot welders, they have stored energy between 100 and 1000 J .
Let's assume I'll be charging it with a 300W power supply, that's 5.4V 55.5A.It will take 146 seconds to fully charge.
Powered by 5V (37500J) 125s
Powered by 4V (24000J) 80s
Powered by 3V (13500J) 45s
Powered by 2V (6000J) 20s
Powered by 1V (1500J) 5s
Powered by 0,5V (375J) 1,2s
Powered bt 0,1V (15J) 50ms
You don't even need a regulated power supply for this, you just need a 5.4V power supply and a stopwatch. For fine tuning, small controlled discharge load.
Current conclusion. The switching timing must be not every 1ms, but at least an order of magnitude more accurate

 

[KlausST]

Hi,

yes, seen from the capacitor it´s all correct.

Where I physically/mathematically see the problem is:
When there is a series impedance/resistance it all shifts to "unknown".

If the "load resistance" at welding point is below 280uOhm, you get fast discharge, high current, but only low voltage. So low Watts for a short time. Low Joules.
Highest Watts you get when the load resistance matches the sum of the other resistances. Still rather short discharge time. So high Watts, low Joules.
If you go higher with load resistance you get lower Watts but higher Joules across the load.

In the end the Watts need to be limited not to immediately vapurize the welding metal. (Possibly at the contact point between electrodes and metal)
The Joules need to be high enough to melt/weld the metal in short time.

I have no idea how professional spot welders work. So if I had to design a spot welder, I´d first read about existing spot welders, how they work, what problems they encountered, hwo they solved the problems..... Just to avoid to waste time in re-inventing the wheel. As said: It´s may way ... you are free to go your own way.


I´m rather sure with your method you find a setup that gives suitable results for your situation.
But I also expect that .. after a month of not using - the first welding will be much different than the 10th welding (in a row) just because of electrodes oxidation, temperatures, and so on.
In technical terminology we call this problem "process stability".

Klaus

 

[D.A.(Tony)Stewart]

1. First estimate energy [J] per weld and then the resistance before and after. It may need a fixture with contact pressure.
2. Then using MPT theorem match the impedance to maximize power transfer with sufficient energy for a deep diffusion weld.
3. Then choose the weld process and method.

Forget about transistor switching capacity of thousands of amps. Most welders take time and 150A to penetrate and not detonate the surface with an explosive spray of vaporized surface particles. This reminds of the dumb time I tried to remove a staple near the breaker panel with power on and melted the tip of a heavy duty screw driver which vaporized copper on my plastic lens glasses in a flash.

I recall a pal of mine had a hand press tool with 1000:1 AC transformer to heat bond sheet metal with low v high current from mains in a couple seconds. It’s all about matching the energy levels and time durations for heat transfer.

 

[dick_freebird]

1. First estimate energy [J] per weld and then the resistance before and after. It may need a fixture with contact pressure.
2. Then using MPT theorem match the impedance to maximize power transfer with sufficient energy for a deep diffusion weld.
3. Then choose the weld process and method.

Forget about transistor switching capacity of thousands of amps. Most welders take time and 150A to penetrate and not detonate the surface with an explosive spray of vaporized surface particles. This reminds of the dumb time I tried to remove a staple near the breaker panel with power on and melted the tip of a heavy duty screw driver which vaporized copper on my plastic lens glasses in a flash.

I recall a pal of mine had a hand press tool with 1000:1 AC transformer to heat bond sheet metal with low v high current from mains in a couple seconds. It’s all about matching the energy levels and time durations for heat transfer.

This sounds like my spot welder, Harbor Freight's finest. But way overkill for sub-mm wire. and wire-to-wire stainless 1/16 was iffy (skill level perhaps, but also easy to overpower the "workpieces".

Copper presents its own, somewhat opposite problems. SS is low thermal condivtivity, copper high. Welding copper takes a lot of local power density in the plate but the 0.5mm wire?
Start small and work up I guess.

Gut description of spot welding - the weld starts with moderate R from point contact / weak arc This is a good thing (until it isn't). Once the metal "yields" and has a point-melt, then you apply pressure and cut power. More power after melt is just going into a short and heating the surrouns. It seems to my hand and eye, a time-domain dance.

I think that a voltage sense at the tips looking for voltage to collapse, could be your signal to cut power (some dwell time, perhaps).

 

[D.A.(Tony)Stewart]

If you have a CC power supply , you can measure the mV drop for cold pressure using Kelvin Resistance 4 point method of not sharing by not sharing test points for injecting current with a DMM to measure voltage.

Then I^2*R*t=E energy process can be estimated using a constant power method for the pulse time. An exponential voltage may or may not match this constant power ideal.

Here is a step-down transformer method with what appears to be http://www.sce.carleton.ca/courses/sysc-3503/w17/notes/rec.pdf

 

[dick_freebird]

Having "stuck" a lot of MIG wire trying to weld sheet metal,
I think constant voltage and a time-gate might be a way to go.
In that process I'd see the wire stick if given too much feed speed
and then just sit there (turning red hot if long enough, otherwise
just stuck and grab the wire cutters). Basically acts like a spot weld
except for the extra resistance.

And maybe that could be another thing to try - like a stiff voltage
power supply and a cold high wattage incandescent bulb as both
limiter and indicator. Start hard, end soft, peak/hold.

I think it's time to try options empirically. There's been plenty
suggested, to investigate. I think theory (despite that there are
welding theorists) fails early when you get away from mainstream
processes, materials and fit-ups.

Still think that your time would be better spent making a broadside
brazing jig and let the oven do the work. Consider this especially if
N wires on one plate, in position, is the end goal and one or more
failures results in scrappage. In my limited experience, every joint is
a roll of the dice and I'm not talking "7". At "volume", consistency
is key and attaining it as a weldor is not an afternoon at the keyboard.

 

[KlausST]

Then I^2*R*t=E energy process can be estimated u

Good idea ...

then go one step ahead:

Calculate the energy by the integral of I * V over t.
Since I is known and constant you may omit it and just integrate V over t (as long as I is within regulation)

Then use a comparator and compare the integral with an adjustable value (representing the energy in Joule) as switch OFF condition.

Should be simple and should give more precise welding results.

Klaus

 

[D.A.(Tony)Stewart]

@KlausST Exactly what I meant. Once a slow weld energy is measured , the ideal discharge time constant may be enough with a capacitor time constant if 1 mohm <R< 1 ohm under pressure cold and C = 3000 F

The thick copper electrode might stick if too thin.

There must be sufficient avalanche energy, pressure and time to diffuse all materials evenly.