High Voltage Forum

General electronics => Electronic Circuits => Topic started by: Anders Mikkelsen on May 02, 2021, 07:45:41 PM

Title: Testing and teardown of a commercial induction heater
Post by: Anders Mikkelsen on May 02, 2021, 07:45:41 PM
I recently found a CEIA Power Cube 45/900 for 200 dollars on eBay. This was just the power unit, but I already had a suitable workhead from a previous eBay purchase, and I thought it could be an interesting unit for a teardown. These heaters are interesting in that they are made to work at a pretty high frequency (750 - 1150 kHz) with low-inductance work-coils (in the 100 nH range). Their main application seems to be fine brazing and soldering of connectors, spectacle frames and other miniature parts.



The power unit is specified to deliver up to 45 kVAR and 3.6 kW running from single phase mains. The VARs are provided by parallel capacitor in the workhead, and the supply only provides the real power, up to 144 V and 25 A in this case.

The unit is pretty compact and dense, with a single pair of buttons to control the heating power and a two-digit seven segment display to show status codes and operating power. The front panel design is a bit unusual in that it contains the operating manual in English and French - convenient if not very aesthetically pleasing.

I cut the workhead cable at some point in the past to figure out how it was constructed. Four cord heavy litz wrapped in what felt like nomex paper, plus a shielded four-core signal cable, all within an outer shielding. So splicing it back together was not fun, but I did a half-hearted job using fibreglass sleeving, heatshrink and all too much kapton tape. Not very elegant, but good enough for testing.



The next issue to tackle was the work-coil, as the one that came with the head was an impractically long hairpin coil that was kinked and bent. I figured a two-turn 30 mm coil would resonate with the operating range of the unit, based on the measured 300 nF workhead capacitance. I wound the coil from 1/8 refrigeration tubing and silver-soldered it in place. Sweeping the head with an impedance analyzer showed parallel resonance at just over 800 kHz, a bit on the low side but within the range of the unit so I left it at that.



I wired up a control switch to the external control input and plugged it in. The display showed error A3 "Absence of water cooling". Turning up the water flow, the unit made a loud beep and the display showed "H1/99", ready to operate. Running the heater, it quickly became obvious that this thing is no joke. It reduces nuts and bolts to molten metal in seconds. Even solid M12 threaded stainless rod is easily molten in less than half a minute.


It looks like this one is worth keeping, so I bought a new mil-spec connector (MS3106A18-19P-RES) for the workhead cable and terminated it properly. The original cable was 3 meters long, so this leaves me with a complete workhead with a 1.5 meter cable, plus a spare 1.5 meter cable with a connector on it. The unit supports two heads so this is a good excuse for me to build a second workhead.



This concludes the testing, in the next episode we'll open it up and try to figure out how it works.
Title: Re: Testing and teardown of a commercial induction heater
Post by: davekni on May 03, 2021, 04:18:29 AM
Impressive what can be done with high frequency and the resulting ability to concentrate power into a small object!  Sure makes the little DIY induction heaters look wimpy.
Title: Re: Testing and teardown of a commercial induction heater
Post by: Mads Barnkob on May 03, 2021, 08:14:37 AM
Great job Anders. Getting such a powerful unit running again for such few money is a score!

You had me intrigued at "teardown", do you have pictures from underneath the lid? I would like to see the power electronics :)
Title: Re: Testing and teardown of a commercial induction heater
Post by: Anders Mikkelsen on May 03, 2021, 09:38:30 PM
Thanks! The combination of low tank impedance, high operating frequency and moderately high reactive power, gives it pretty decent power density.

I realized I forgot to add an image of the back of the unit, so I'll start with that to help give an idea how everything fits together.



From the left, we have standard mil-spec connectors for the two workheads, cooling water in, cooling loops for the workheads and cooling water out. A digital input is switched to ground to enable the heating for each head (only one can be used at a time), and a 0 - 10 V analog input can be used for power control. The unit also has an RS-232 port for monitoring and basic control.



Let's have a look inside. From the top, there's nothing really surprising. A two-stage EMI filter, with one stage enclosed in a magnetic shielding can, a potted toroidal transformer for control power, and the DC bus filtering caps on the main power PCB are also visible. The edge of the control board can be seen along the top of the image.



Looking from the side, the unit is built around a solid plate of aluminium on which everything is mounted. This plate also serves to transfer heat from the electronics to the water cooling block on which the transistors are mounted. Nice and clean design, but it does explain why this unit feels so heavy for its size.



The lower board contains the main power electronics section. There's a rectifier, some electrolytic DC bus caps a completely separate sub-board for the MOSFET full-bridge. The bridge has body diode isolation Schottky diodes and fast antiparallel-diodes. This makes sense, as it's easy to exceed the body diode recovery dI/dt rating on MOSFETs if the bridge current phase angle ever goes lagging, which it is bound to do under some operating conditions. Some modern devices are more robust on this front, but this is still a major concern in MOSFET-based resonant converters.

It's starting to look like this is an LCLR heater. Given the lack of a matching transformer in the workhead, it was either this or a current-fed topology. Measurements show that the DC bus capacitors are directly connected to the bridge, so LCLR it is. I made a quick sketch of the power circuitry, including the critical values.



Interestingly enough, it looks like they made their own DC blocking capacitor. It's a taped up stack of copper foils and mica, the yellow unit on the center-right of the bottom view. It's probably the biggest surface-mount capacitor I've seen, and it measures 50 nF.



The output transformer and matching inductor are both planar, made on what looks like a standard PCB process. The matching inductor is 3.2 uH and the transformer seems to be around 4:3, with a 90 uH primary magnetizing inductance. The matching inductor is placed before the transformer, so the output transformer has to carry the reactive power developed by the matching inductor. They are both bolted to the aluminium cooling plate using some fancy cast aluminium parts, pretty neat.



The final part of the power board is the output filtering and switching, hidden inside a shielding can. Some X- and Y-capacitors, a common mode choke made with a PCB used as a planar busbar, a relay for selecting which output to activate, and the output connectors.



Lastly, we have the front panel board, which is more-or-less what one would expect. An obscure but unremarkable 32-bit microcontroller, a CPLD, and a MOSFET full bridge to drive the gate drive transformers. The main microcontroller had the markings burned off by laser, but the part number was still readable. I'll honor their spirit by not mentioning the part number, but I don't think it would have made a difference even if it wasn't readable.

In the next post we'll fire it up and do some measurements.
Title: Re: Testing and teardown of a commercial induction heater
Post by: T3sl4co1l on May 05, 2021, 04:51:47 PM
Nice. I've been thinking about making a planar transformer for something like this.  Just that, without a product in mind or anything, it's very, very back burner... ::)

Tim
Title: Re: Testing and teardown of a commercial induction heater
Post by: Anders Mikkelsen on August 23, 2022, 12:47:05 AM
I forgot to follow up with measurements from the unit in operation.

 [ Invalid Attachment ]

Controls were pretty unremarkable. Frequency shift for power control as expected, waveforms were more or less textbook for the topology. The current waveform shows significant ringing with some levels of loading, but at that point I decided it was time to box up the unit again and put it to work doing what it was made for.

Title: Re: Testing and teardown of a commercial induction heater
Post by: oliv25 on February 23, 2023, 08:08:15 PM
EDIT: Translated by administrator.

Hello, my name is Olivier and I have to do some soldering tests on a power cube 64/900 but the problem is that it is faulty A6, it worked well a few years ago.
We took it out to do tests for a customer because the parts to be brazed deformed too much in our brazing oven.
When switching on, no display, a condo was out of order on the small plate on the front with the + and - buttons, it's a 220µf 16V, I replaced it, it lit up and communicates with the slave box .
I also had the A4 fault but it no longer appears for the moment.
I am new to electronics, a multimeter to take a few measurements, it may be difficult to find the problem.
However, has anyone ever encountered the A6 fault and how to solve it if there is a resolution?



Quote
Bonjour, je m'appel Olivier et je dois faire des essais de brasage sur un power cube 64/900 mais le problème est qu'il est en défaut A6, il a bien fonctionné il y a quelques années.
Nous l'avons ressortie pour faire des essais pour un client car les pièces à braser se déforment trop dans notre four de brasage.
A l'allumage pas d'affichage, un condo était HS sur la petite platine en façade avec les boutons + et -, c'est un 220µf 16V, je l'ai remplacé, il s'est allumé et communique avec le boitier slave.
J'ai eu aussi le défaut A4 mais il n'apparait plus pour l'instant.
Je suis novice en électronique, un multimètre pour faire quelques mesures, c'est sur cela risque d'être compliqué pour rechercher le problème.
Toute fois, est-ce que quelqu'un a déjà rencontré le défaut A6 et comment le résoudre si il y a résolution ?
Title: Re: Testing and teardown of a commercial induction heater
Post by: oliv25 on February 23, 2023, 08:11:31 PM
 [ Invalid Attachment ]
Title: Re: Testing and teardown of a commercial induction heater
Post by: DashApple on May 15, 2023, 08:17:44 PM
Thanks for the info here.

I had an attmept at making a custom head but it seems I ran into a failure . I built a head around a 200nF cap / Litz wire feed and such like the origional cable ; Head performace isnt great compared to the normal head.

My issue was everything was running fine, then the power was pulsing up and down ; crackle sound and a bang, upon looking a large resistor that is precharge from what I can tell had burnt out and after looking and tacking the bridge out ( Really nice its modular ) It seems one of the mosfets has failed and is split across its front. Other mosfets look intact but its hard to say what was taken with the faiure. I do hope it was the one mosfet.

They have the marking removed so I have no idea what mosfet to replace it with, any suggestions there ?

/>


 
Title: Re: Testing and teardown of a commercial induction heater
Post by: Anders Mikkelsen on May 19, 2023, 03:56:19 PM
Sounds like an opportunity to dive deep into a hardcore repair, if nothing else. This is where the fun starts.

Firstly, it's good to find out why it failed in the first place, so it doesn't happen again. My main suspicion is that the tank parameters are outside of what the unit expects. Since the capacitance is within the supported range of the unit, it could be that the inductance is too large for the given capacitor. Do you have a way to measure the resonant frequency of your tank circuit? It should ideally be well within the accepted 750 - 1150 kHz operating range of the unit. Also consider that loading the coil will modify the resonant frequency, generally upwards.

I had a look at the MOSFETs in my unit, but they also have the markings lasered off, so finding the correct part, or at least a suitable replacement, could be a bit tricky. The devices appear to be in a TO-264 package which rules out a lot of options, so that's a start. IXYS parts seem like a good candidate, as they are one of the few MOSFET manufacturers with a wide lineup in this package, part number prefix IXFK.

I'd bet more than one MOSFET is bad, at least the other one in the same bridge leg is likely toast. Also check the diodes if they are fine, the series blocking ones could have taken a beating but diodes tend to be pretty resilient so they could have survived. If the remaining two MOSFETs are fine, you can measure some parameters to help find a suitable replacement. A good start is to measure the Rdson (push an amp through the MOSFET when it's turned solidly on, and measure Vds). This along with the gate capacitance will allow you to get the sizing of the MOSFET right.

As far as performance goes, it looks like you have a lot of stray inductance between your capacitor and coil, maybe as much as in the work coil itself. This will lead to a lot of your tank VARs not being coupled to the workpiece.
Title: Re: Testing and teardown of a commercial induction heater
Post by: petespaco on May 20, 2023, 04:28:14 AM
Just a couple of comments from a guy who doesn't know much of anything about this particular induction heater:
1.   I think the resonant frequency generally DROPS as workcoil inductance increases.
2.  I think the resonant frequency generally Drops as a solid workpiece is inserted into the work coil, but it Increases if a hollow or tubular workpiece is inserted.

See this for a couple of examples (at about 9:26):
/>

3.  When attempting to select a Mosfet-  Somebody once told me to choose one with a Vds equal to at least pi times the DC supply voltage.  For example, my Chinese ZVS heaters run on 48 volts DC and they use the IRFP260N, which has a Vds Max of 200.

Don't take anything I say as "gospel" without verifying my comments---
Title: Re: Testing and teardown of a commercial induction heater
Post by: Anders Mikkelsen on May 20, 2023, 04:49:07 PM
Magnetic cores will increase the inductance at a coil at low frequencies, due to the high permeability of the material reducing the path reluctance. But skin effect at higher frequencies means that most of the workpiece doesn't see any magnetic field at all, so it doesn't contribute to increasing inductance. Since the eddy currents work against the external field, the effective permeability of conductive workpieces approaches zero, whereas without a workpiece the permeability is that of air, practically one. Above a few kilohertz depending on the size of the workpiece, any conductive material in the coil will reduce the inductance and therefore also increase operating frequency. The effect is more dramatic when the workpiece is a tube, but it still happens for solid workpieces.

A transistor peak voltage of pi times the supply voltage is specific to the inductance fed half bridge circuit with no conduction overlap, as used in the classic self-oscillating ZVS circuit. The tank circuit forces the voltage to be a sine wave, and since each transistor is on for half the time and the average voltage across the inductors has to be zero (ignoring ESR), then the waveform across the transistor is a half-sine with an average value of Vin. This waveform has a peak value of pi * Vin by definition.

The voltage rating here only needs to be higher than the DC bus voltage, since it's a voltage fed full bridge LCLR, supplied by rectified and smoothed 240 V. The DC bus voltage is typically around 340 V, and I bet the original transistors are 500 V rated as would be common in this operating voltage range.

On that note, I was playing a bit with the ZVS circuit to improve its performance, since the classic version has severe limitations on the work coil impedance, tank VARs and total power that it will process without expoloding. The circuit works well  for how simple it is, but overcoming its limitations takes a lot more effort. The critical changes I found was to use reverse blocking diodes in series with the MOSFETs along with a robust control circuit that allows phase-advanced switching and conduction overlap. Using more modern switching technology like SiC also helps a lot. I made a prototype but didn't have a chance to develop it further yet due to other projects taking priority. The circuit worked pretty well, but I didn't really push the limits beyond some 4 kW, running with 200 V input and just over 600 V on the MOSFET drains. I need to continue the work on this, because initial results were pretty promising. The only video I found of it is some shaky footage running at less than 2 kW.


 
Title: Re: Testing and teardown of a commercial induction heater
Post by: DashApple on August 24, 2023, 02:21:32 PM
Sounds like an opportunity to dive deep into a hardcore repair, if nothing else. This is where the fun starts.

Firstly, it's good to find out why it failed in the first place, so it doesn't happen again. My main suspicion is that the tank parameters are outside of what the unit expects. Since the capacitance is within the supported range of the unit, it could be that the inductance is too large for the given capacitor. Do you have a way to measure the resonant frequency of your tank circuit? It should ideally be well within the accepted 750 - 1150 kHz operating range of the unit. Also consider that loading the coil will modify the resonant frequency, generally upwards.

I had a look at the MOSFETs in my unit, but they also have the markings lasered off, so finding the correct part, or at least a suitable replacement, could be a bit tricky. The devices appear to be in a TO-264 package which rules out a lot of options, so that's a start. IXYS parts seem like a good candidate, as they are one of the few MOSFET manufacturers with a wide lineup in this package, part number prefix IXFK.

I'd bet more than one MOSFET is bad, at least the other one in the same bridge leg is likely toast. Also check the diodes if they are fine, the series blocking ones could have taken a beating but diodes tend to be pretty resilient so they could have survived. If the remaining two MOSFETs are fine, you can measure some parameters to help find a suitable replacement. A good start is to measure the Rdson (push an amp through the MOSFET when it's turned solidly on, and measure Vds). This along with the gate capacitance will allow you to get the sizing of the MOSFET right.

As far as performance goes, it looks like you have a lot of stray inductance between your capacitor and coil, maybe as much as in the work coil itself. This will lead to a lot of your tank VARs not being coupled to the workpiece.

Thankyou for the time with replying, sorry for such a delay in my response.

When this failure happened there was surging of the power been delivered to the work coil, this was apparent by the info displayed on the unit but before I could do anything the unit failed.

I do expect as you had said that the coil was outside the expected range / setup and this caused an error or the unit got mixed up during power delivery ; likely down to my poor tank setup.

I will at some point try and figure out the MOSFET needed and replace the full bridge ; Hopefully the gate driver side is all in working order ; I am still unsure why the pre charge power resistor failed the way it did , as if it had exploded as one end had become unattached to the PCB and the body is broken. This may be a bodge to replace as it sits behind the main filter capacitors.

Do you have any pointers on how to make this work head of this unit, I understand the need to keep stray inductance down as much as possible but I am unsure where to go with my current head. Power is feed by 2 pairs of litz wire and two twisted pairs for feedback. The leads to the work head are not twisted ; Would this make a difference ?

My other option and one I am leaning to is to eventually locate an original head and remake the bridge as then there should be no issues with the feed and tank head setup.

thanks again.
SimplePortal 2.3.6 © 2008-2014, SimplePortal