Analogue cell sealing machine

Wszechelektronik 11 Nov 2024 20:57 27 4806 Cool? (+16)

📢 Listen (AI):

TL;DR

A homemade analogue cell welding machine controller was built for welding cells into battery packs.
It uses a phase shifter and comparators to detect the mains peak, then a LM555 and optotriac fire the thyristors without a microprocessor.
The power supply uses a 10-20V 2W transformer, a single-pole rectifier, and a single-transistor stabiliser; the board was designed for easy milling.
The redesigned controller welded almost 300 good welds in a row.
The earlier version had timing instability and burned out one cell in 20-30, which triggered the redesign.

Generated by the language model.

.
I built a welding machine a few years ago, it was mainly to be used to weld cells into packages. I have a lot of cast-off components, so I'm trying to put at least some of them to some use. Unfortunately this is going with increasing resistance, but that's another topic. So the result is a rather unusual, processorless circuit, altogether simple and cheap, but not lacking in important features. Many welding machine controllers have a disadvantage, despite a stable pulse time, there is no control of the moment when the transformer is connected to the mains, and in the mains, as it is in the mains, the voltage changes all the time. As a result, the portion of energy supplied to the welding spot varies, which can result in burning out the cell or in too weak a weld, as well as overloading the network and tripping the protections. With this in mind, this design was developed:

.
Unfortunately, as it turned out after commissioning, there was a welding time instability, burning out one in 20-30 cells. A ghastly occurrence, fortunately nothing exploded, but you had to pry this burnt cell out of the pack, so overall a big bummer. I tried to find the cause of this instability, modified something there, but it didn't get much better until I finally put the circuit away for lack of time. I occasionally used this welder, with small packets the job usually worked.
Quite recently the need arose to weld a larger packet, I decided to redo the whole controller at the same time.
A schematic was created:

.
Maybe it will be useful for someone, also a simple chip without a processor, whether this is a disadvantage or an advantage I will leave to the readers to choose.
Untreated board, specially designed for my CNC woodworker. Simple copper layer cuts, significant milling inaccuracy allowed, easy post processing.

.
The new arrangement worked well, I welded the package, almost 300 good welds
in a row, except that... wanting to speed up the work, the driver was still in spider form, I made the board later. After fitting the new board, I did a few test welds, and it seems to work just as well, which is not at all obvious. Those who know know that spiders can work sensationally, and it is only the making of the plate that spoils the effect.

How it works.
The power circuit consists of a 10-20V 2W transformer, a single-pole rectifier, a single-transistor voltage limiter-stabiliser feeding the circuit. The AC voltage from the transformer via the divider goes to the phase shifter IC3A and further to the comparators that detect the zero crossing, Zero in this circuit is half the supply voltage. The phase shifter is there to be able to hit the peak of the sine wave, which is when there is the least current surge when switching on the high-power tray. Is this successful? Experiments show that this surge can indeed be reduced by adjusting the phase with potentiometer R25. A shifted zero is detected which indicates the peak of the sine wave. A pulse at the peak of the mains voltage appears at the output of the comparators. This signal via the S microstep triggers the LM555 timing circuit. Then it is typical, the optotriac switches on the thyristors, I wanted to use what I have, so here sit the Russian KY202h, but if I burn them all, you can of course insert other newer ones, the board is quite universal.
Power trafo standard from microwave, 4 coils I think 16mm, long time ago I wound it I don't remember exactly.

.
Spider testing platform.

About Author

Wszechelektronik wrote 100 posts with rating 157 , helped 1 times. Been with us since 2011 year.

Comments

Add a comment

Anonymous 12 Nov 2024 00:28

Isn't it simpler to time 1/4 of a period from going through "0" which should work out very close to max ? [Read more]

jarekgol 12 Nov 2024 01:35

Metering in such a system would be done on some RC anyway. As for the welders, the industrial ones support regulation of time (number of grid periods) and phase angle. Plus some ramping, repetition and... [Read more]

Anonymous 12 Nov 2024 08:38

According to the phase shift of current and voltage (the ULICU CIUL or LUIC principle), the voltage precedes the current in the coil by 90 degrees. Including a large transformer ideal at the peak of the... [Read more]

gulson 12 Nov 2024 09:51

Brave design, yes without a processor hmm. Out of curiosity why the choice without a processor, advantages/disadvantages? Do you perhaps know what caused the instability of the previous solution? Any... [Read more]

jarekgol 12 Nov 2024 10:10

@jewishmakaron does this apply to a transient such as switching on? After all, if you switch on at zero, no current will flow (and according to the principle it should automatically generate current). I... [Read more]

acctr 12 Nov 2024 11:30

. Does the phase shifter do something else? Wrong, the offset between current and voltage is not relevant here, but if anything it is the offset between voltage and magnetic flux. In the case... [Read more]

jarekgol 12 Nov 2024 12:34

@acctr so in theory it should matter at what point the trafo was switched off previously ? Was it in quotes "magnetised to N or S" ? [Read more]

acctr 12 Nov 2024 12:58

. Yes, it matters in practice too and that's where the impulse comes from. But switching on at the peak of the sine ensures that the core does not become saturated. Roughly speaking, it looks like this:... [Read more]

keseszel 12 Nov 2024 16:18

So a better solution would be to switch on the secondary rather than the primary.... , but currents..fact.... [Read more]

drunek 12 Nov 2024 17:01

Super project! I'm delighted that there are still some hobbyists who can create something practical from scratch, and without a microcontroller. I'll have a look at the schematic when I have a moment. It... [Read more]

metalMANiu 12 Nov 2024 17:55

. I appreciate with what ease you have managed to explain this :) . If the time during which the flux increases is so important, why not turn on the transformer e.g. 2 milliseconds before going through... [Read more]

acctr 12 Nov 2024 18:14

. Because this is closer to zero and the risk of core saturation increases. The peak of the sine, at least the 'real' one, is the same distance from the 'left' zero transition and the 'right' one, so... [Read more]

Anonymous 12 Nov 2024 20:42

A phase shifter as the name suggests shifts the phase - except that a timer may be more precise here at a lower cost. [Read more]

Wszechelektronik 12 Nov 2024 20:50

. If it's possible to achieve the intended functionality by combining a dozen or so simple components, then I guess it's worth taking advantage. Maybe it's out of habit, my adventure with programming... [Read more]

drunek 12 Nov 2024 21:01

Bottom line. What kind of thyristors should these be, i.e. for what anode current? I can see that a triac is unlikely to replace this. [Read more]

mfac 12 Nov 2024 21:31

This is about something else. I can't remember with details, but the issue breaks down into differential equations, the principle of continuity of current, the dphi/dt relationship etc. At start-up there... [Read more]

keseszel 12 Nov 2024 21:31

I was intrigued by the handle. A very ingenious solution, a total gimela ;-) . You have used copper wire, of some thickness.... I see that it's vaulted at one end, with the other ground down. How long... [Read more]

Wszechelektronik 13 Nov 2024 01:21

. The ones I've inserted have a max of 10A/400V, although in pulse they can probably handle a lot more, so far I haven't burned any. Maybe with a thicker winding there would be a problem. This arrangement... [Read more]

misiek84LG 15 Nov 2024 09:22

Super project, I will try to copy in the near future. How much 16mm cord so roughly needed to make 5 or 6 coils on a microwave core????? [Read more]

FAQ

TL;DR: For DIY battery-pack builders, this analogue MOT spot-welder controller fixed a failure rate of 1 in 20–30 cells and then delivered "almost 300 good welds" in a row. It uses zero-cross sensing, a phase shifter, an LM555, and two thyristors to fire near the mains peak, reducing transformer inrush and making weld energy more repeatable without a microcontroller. [#21299008]

Why it matters: Stable weld time alone does not guarantee stable weld energy when a transformer is switched at random points on the AC waveform.

Approach	Trigger method	Time control	Main thread takeaway
LM555 analogue controller	Peak-targeted via phase shifter	Continuous pulse timing	Worked reliably after redesign
RS flip-flop + CD4017 controller	Peak-synchronised start, then counts 50 Hz periods	20 ms steps	Good for preset times, less flexible
Zero-cross optotriac approach	Switches only at zero	Depends on external timing	Defeats peak-start advantage on transformers

Key insight: In a MOT-based welder, the turn-on point on the sine wave is a core reliability parameter, not a minor detail. Firing near the real voltage peak reduces inrush, lowers saturation risk, and makes welds more consistent.

Quick Facts

The control supply used a 10–20 V, 2 W transformer, a half-wave rectifier, and a single-transistor voltage limiter-stabiliser for the low-voltage electronics. [#21299008]
The author reported roughly 1 burned cell per 20–30 welds on the first controller, then almost 300 consecutive good welds after the redesign. [#21299008]
The installed thyristors were rated 10 A / 400 V max, with an added 350 V transil plus an RC snubber across the power stage. [#21300881]
For the MOT secondary, about 250 cm of 16 mm cable was said to be enough for 5 or 6 turns on a microwave transformer core. [#21304463]
The alternative counter-based controller offered 20 ms to 200 ms in 20 ms steps, then 200 ms to 2 s in 200 ms steps, using mains-period counting rather than an LM555 pulse. [#21308276]

How does an analogue, processorless cell spot welder controller work when it uses LM555 timing, zero-cross detection, and a phase shifter to fire thyristors?

It detects the mains phase, shifts that reference toward the sine-wave peak, and then starts an LM555 timing pulse that drives the power stage. The low-voltage section uses a 10–20 V, 2 W transformer, rectifier, comparators for zero-cross detection, and a phase shifter so the trigger lands near the mains peak. The LM555 output drives an optotriac, which then fires two thyristors feeding the MOT primary for the set weld duration. That gives simple analogue timing without firmware or a microcontroller. [#21299008]

Why does switching a MOT transformer at the sine-wave peak reduce inrush current and core saturation compared with switching at zero crossing?

Switching near the sine-wave peak reduces the extra magnetic flux that can push the MOT core into saturation. One explanation in the thread states that if you switch at zero, flux can rise for about 10 ms in the worst case, but at the peak it rises for about 5 ms, giving the key relation Φz = 2 Φm. That smaller flux excursion lowers the chance that residual core magnetism and new flux add up enough to create a large inrush pulse and blow fuses or stress thyristors. [#21299739]

Would timing exactly 1/4 of the mains period after zero crossing work as well as an analogue phase shifter for hitting the sine-wave peak in a spot welder?

Yes, in a clean 50 Hz system it should land very close to the peak, because one quarter-period is about 5 ms. Thread participants noted that metering would still rely on RC timing, so a timer can be simpler and cheaper than an analogue phase shifter. The practical reason to keep the phase shifter is trim: the author used potentiometer R25 to align with the real peak seen by the actual transformer and comparator chain, not just the mathematical quarter-period. [#21300568]

What likely caused the weld time instability in the author's first controller, where roughly 1 in 20 to 30 cells burned out?

The likely cause was interference from the power transformer coupling into the controller. The author said the first version showed weld-time instability severe enough to burn about 1 in 20–30 cells, and performance improved somewhat after placing a steel plate between the transformer and the board. He also noted that the original control concept was simulated in an old program that did not include electromagnetic interference, which explains why the bench result diverged from the schematic expectation. [#21300590]

What improvements were made between the first and second versions of the analogue cell welding controller?

The second version replaced the earlier control concept with a new schematic built around better mains-phase handling and then moved it from a "spider" prototype to a PCB. It kept the peak-synchronised firing idea, used a redesigned analogue controller, and then proved itself with almost 300 good welds in a row on a larger pack. The author also later added a 350 V transil alongside the RC snubber for extra surge protection on the thyristor stage. [#21300881]

Which thyristor ratings are suitable for a MOT-based battery tab welder, and why is a pair of thyristors preferred over a single triac here?

The author used thyristors rated at 10 A and 400 V, and reported no failures in his present build. He explicitly advised against replacing the back-to-back thyristor pair with a single triac, calling the two-thyristor arrangement the most fail-safe proven solution for this application. A thicker MOT secondary could raise pulse current enough to challenge the devices, so the 10 A / 400 V choice worked in his setup but still depends on transformer build and weld demand. [#21300881]

What is a transil, and what role does a 350 V transil play alongside an RC snubber in a thyristor-controlled welder?

"Transil" is a transient-voltage suppressor diode that clamps fast overvoltage spikes, sacrificing leakage margin for strong surge absorption. In this welder, the added 350 V transil sits beside the RC snubber to catch sharp switching surges that the snubber alone may not fully tame. That helps protect the thyristors from voltage spikes created by transformer switching and wiring inductance during short, high-energy weld pulses. [#21300881]

What is a random-trigger optotriac, and how is it different from a zero-cross optotriac in transformer-powered welders?

"Random-trigger optotriac" is an optically isolated triac driver that can switch at any phase angle, unlike a zero-cross type that waits for near-zero voltage. In a transformer welder, that difference is critical. The thread states the controller must use a random-trigger type, because a zero-cross optotriac would delay turn-on until zero crossing and cancel the whole benefit of peak-synchronised transformer energising. Zero-cross parts suit capacitive loads better, not MOT primaries. [#21308724]

How do you set and tune the phase shifter potentiometer R25 so the transformer turns on near the real mains voltage peak?

Set R25 by trimming for the lowest observed transformer inrush while keeping reliable weld start. The author said experiments showed current surge could be reduced by adjusting R25, so the practical method is: 1. monitor the peak-detection output or transformer behavior, 2. make short test welds while varying R25, 3. stop at the setting with the least harsh start and stable welds. This aligns the shifted zero reference with the real sine-wave peak seen by the actual circuit, not an idealised waveform. [#21299008]

What wire length is roughly needed to wind 5 or 6 turns of 16 mm cable on a microwave oven transformer core for a spot welder?

About 250 cm should be enough. That figure was given directly in reply to a question about winding 5 or 6 turns of 16 mm cable on a microwave oven transformer core. It is a practical estimate for planning the secondary, not a universal formula, so exact length still depends on the MOT window size, bend radius, and how tightly the cable can be packed through the core. [#21304463]

How should copper welding tips in a DIY cell welder handle be shaped, maintained, and equalized to prevent sticking and uneven welds?

Use copper tips with closely matched contact areas, then correct them often with light filing and fine abrasive paper. The author reported sanding once initially, then making small corrections every few dozen welds with a small file and frequent cleaning with P600 paper. He said sticking stopped after he equalised the tip surfaces. Both arms in his handle also spring independently, which helps keep pressure similar on both electrodes and improves weld symmetry. [#21300881]

Analogue controller vs microcontroller-based spot welder controller: which is better for cell welding reliability, simplicity, and ease of repair?

The thread favors analogue if you want simplicity, easy repair, and enough function without software. The author chose a processorless design because a dozen simple parts achieved the needed behavior, while another participant argued a microcontroller had no clear benefit here. The analogue design then proved itself with almost 300 good welds in a row. Microcontroller boards were mentioned mainly in the context of reported failures in some low-cost Chinese designs, not as a clear reliability upgrade. [#21300590]

What does the 'spider' prototype method mean in electronics, and why can a spider-built welder controller sometimes work better than the final PCB?

"Spider" prototype is free-form point-to-point wiring where components are connected directly in space, without a finished PCB, prioritising fast testing over mechanical neatness. The author first validated the redesigned controller in spider form and got strong results, then noted that such builds can work "sensationally" while the final board can spoil the effect. In high-current welders, PCB layout can change noise coupling, grounding, and electromagnetic pickup, so the cleaner-looking version is not always the better-performing one. [#21299008]

How does a 50 Hz counter-based controller using RS flip-flops and a CD4017 differ from an LM555 pulse-timed spot welder controller?

A 50 Hz counter-based controller measures weld time in mains periods, while an LM555 controller measures it as an analogue pulse width. The alternative design described in the thread uses two RS flip-flops, a phase-shifted comparator, and CD4017 counting so the weld starts near the sine peak and then runs for a preset number of full 20 ms periods. Its adjustable ranges were 20–200 ms in 20 ms steps, then 200 ms–2 s in 200 ms steps, which is simpler but less flexible than a continuously adjusted LM555 timer. [#21308276]

Why won't pure nickel strip weld properly in some MOT spot welders while nickel-plated strip works, and what transformer or secondary changes help?

In this build, the welder did not deliver enough output for pure nickel strip, but it did work on nickel-plated strip. The author said pure nickel welds did not succeed and suggested a thicker transformer wire as the likely improvement, meaning a lower-resistance, higher-current secondary. That matches the rest of the thread: MOT weld quality depends heavily on available pulse current, and marginal output shows up first on harder-to-weld materials before it appears on nickel-plated tabs. [#21300590]

Generated by the language model.