Induction heater project

[eskoilola]

Last time I visited home back in finland I did some brazing. Brazing is actually quite easy but to achieve the needed temperature with a butane torch took quite some time. During this time the otherwise nice component got an layer of oxidation on top of it. Not nice. Further more that is an open fire and it can be somewhat hazardous in my small shop. Doing this outside is a no-go when the weather is rainy or there is a lot of wind. So I decided to start a project that I am capable of completing.




So here is the first part of this project. It is a chinese 1.8kW induction warmer device. Funnily enough - these guys have strictly prohibited .... whatever that is - it does not say. Quite many hobbyists have fried these devices as they do not know that the power should NOT be connected gradually. There should actually be a PHYSICAL switch or really fast switching FET array to cut on/off the power to this device. It should be directly in the DC line. This is because the construction is such that it needs a "kick" to start oscillating. A mechanical switch with the contacts bouncing and creating a lot of ripple before they settle is actually a perfect solution to this problem. It can also be a relay. A fet array might work but it generates only one pulse which might not be enough to get this going. If the power is applied gradually the output FETs of this device will blow skyhigh. I have suitable transformer (2 x 24V 2000W) and a pair of suitable filtering capacitors (100.000uF/63V) along with the needed rectifier components. Should not be a problem.




The overall construction is OK-ish but some details are just too flimsy for that kind of power. This needs new connectors for the output coil. Connecting kiloamperes using long brass nuts is nuts. At those current levels those warm up and dissipate the power in a place where it should not be dissipated. I will probably use 1 mm thick copper "foil" on top of the PCB traces where needed and on the output use some tube adaptors which have their counterparts silver soldered on the output coil. This allows for easy coolant flow and way better connection for the output resonance circuit.




Other than that the construction seems to be quite nice. The blowers are needed to cool down the output FETs and the output tank circuit capacitors. It will be seen whether I need to change those to some real quality stuff. This kind of a circuit generates currents in excess of kiloamperes and that will fry any not-good-quality capacitor. Luckily I have some suitable capacitors back home. If I would need to buy those it would cost more than this device - including the postage.




The supplied output coil is flimsy. Pushing 1.8 kilowatts through that one will create losses like anything else. I will visit the local (local in finland) hardware store and get copper tube that is more to the job. This tube is actually a joke with 5 millimeter diameter and 0.7millimeter wall thickness.




Another item I purchased for this project is a water pump. This one is up to the task. It is a really good quality and can make the cooland really go around. 11 liters per minute with 0.5atm pressure. Quite impressive for such a small device. It states that the pump must not be run dry and that the polarity of the wires should net be changed. It is a brushless motor the pump propeller being the motor rotor itself. Water is used as a lubricant and coolant for the motor. Therefore no dry run. I probably need to direct part of the yield directly back to the radiator to allow more flow for the pump. Let's see how that goes.




So today I ordered the radiator for this project. It is a oil cooler for racing applications. Very tough device. The one I ordered has a cooling area 75x115 millimeters. the actual area is a lot bigger as the cooler is quite thick and there are a lot of cooling fins between the coolant tubes. This was quite affordable as it has strange 3/8 BSP threads. 55 degrees of thread angle ... LOL. It will be fun tho fabricate fittings for this one. Hopefully the threads are not left-handed .... The actual page in eBay is HERE




Last but not least I paid a visit to the saturdaily flea market. Today there was this Noga deburrer which had a very affordable price of 50 eurocents. The blade is long gone. Ordered a set of new blades from USA (the chinese are cheaper but they take forever to arrive). This tool is probably of the very first versions ever made by Noga. It has an aluminium shaft and it is still as good as new - if You forget the blade.

[awemawson]

I'll be watching this thread with interest. 

I decommissioned my 100KW induction furnace when I moved here 11 years ago, and it is still in the fireproof space I built specially for it gathering dust 

It's a big rack on castors about 1 meter square and 1.8 tall driving a crucible 'furnace body' that looks like a cement mixer. I have two - one that tilts and pours the metal, the other inverts and the mould gets clamped to the top so the melt is never exposed.

In addition it has a 15 kw cooler to chill the water that is pumped though the coils

It was because of this device I had a 160 amp per phase 415 volt feed brought in ! (previously I used a diesel generator)

[eskoilola]

Yes. I also somewhat believe that the turnout of this one might be interesting for those that want to do some hot stuff but do not have the correct (fireproof) setup for that. With induction heater the energy goes (is supposed to go) exactly where needed without setting the cardboard boxes around on fire.

One thing I need to rhink over is the electrical isolation of the coil. I have seen some pictures of coils having a fiberglass sleeve but I do not feel comfortable with that solution. I would rather mold the coil in some fireproof cement, alumina maybe,  in a way that the the coil is free to move outside that ceramic. I have some coil formers that might be suitable. Those are made of some really tough ceramics ... but I have the faintest idea whether those will survive the temperatures. One option is to get a readily formed alumina tube - another option is to make it myself - I have to renovate our garden grill anyway so why not make it of some impressive material. The quantities one must buy the castable alumina are not small.....

[awemawson]

When I rebuilt mine I had to have crucibles made to order from one of the (then) newly escaped Russian satellite countries - wasn't too pricey - I had a dozen made and ten are still on the shelf. They are totally embedded and not removable without chipping them out. Oddly I had some spam from the firm today which I deleted so I can't look back and see who they are !

[PK]

So I decided to start a project that I am capable of completing.

That's classic me...

[eskoilola]

Just reading the "Data Sheet" of the coil driver. Definitely a google translation. Here is an extract from the document. This part is warning about the slow rising voltage. Somehow I fin this hilarious:

You must pay attention to when using switch power supply, because the high-power switching power supply usually have soft start function, namely the output voltage is slowly rising, and if the voltage rise to 10V before connected induction heating circuit will be caused by lacking voltage lead to circuit can't afford to vibrate, so that the two MOS tube conduction and burned components at the same time, so be sure to switch the power supply voltage stability and then access the induction heating circuit.

Here is another:
The minimum input voltage of the module is 12V, it doesn't mean all the 12V power supply can be used, some novice buyers may use a smaller power supply. This module has specific voltage and current requirements, the voltage 12V current must be more than 10A power supply can be used, if the power is too low will lead to the power output is greatly reduced, when the voltage is lower than 12V, it will lead to two MOS at the same time conduction, will burn MOS tube;

After reading this it seems like the circuit does not need a kick to start - it just does not oscillate if the supply voltage is below 12 volts. So switching with a FET array should be safe. This is actually a BIG assumption but I am used on exploding FETs - they are actually quite fun.

I am thinking on how to control the output power of this device. Basically there are two options:
1. Control the on-time. Sort of slow-motion PWM.
2. Control the input voltage. Maybe 18V to 48V. I could make a BUCK regulator for that.

From those the first one is more appetizing as it does not alter the operation parameters of the circuit. It is also easier to do. Could do the switching from the low-side with a FET array thus avoiding usage of high-side FET drivers.

And here is another funny extract from the documentation:
When heating the metal objects do not to heat too large metal, otherwise it may be overloaded with power, please add a 35A fuse or ammeter and air switch will be better. (package does not inclues)

So it seems like a current protection should be applied. Apparently sticking in a copper bar of 50mm in diameter will be a no-no-no.
After laughing my ass off while reading the documentation I have decided to do this in a certain way ... will add a block diagram later today.

In case You want to read the complete documentation it is readable in THIS eBay location.

[eskoilola]

Here is my interpretation of an induction heater.




This device needs two separate power supplies. First, the high power supply delivering juice to the heater coil through the driver. The other power supply is for all other functionality like control logic, water pump and fans. The high power supply can be switched on/off by the control logic whereas the auxiliary power supply has a real user operated switch.

The high power supply needs a soft start circuit which consists of two relays and a resistor. This is because the transformer I have is a toroid transformer and that devil will blow the fuse every other time if cold connected to mains. The high power supply will be switched on only when needed.

The rectifier and smoothing is just a couple of full bridge rectifiers connected in parallel using resistors (0.1 ohm). I have a box full of 10 amp rectifiers and if I connect 10 of these in parallel that should be more than enough for this application. The smoothing consistsa of 2 capacitors size of a Chevrolet piston each worth of 100.000uF.

Voltage sensor in just a resistor divider. The current sensor I probably make out of a hall sensor glued on top of a wire. If the sensitivity is not enough just cut a gap in a ferrite toroid and stuck the hall sensor in there. The wire then goes through the toroid hole. These sensors are used to cut off the driver in case the voltage is too low or the current is too high.

The FET array switch is a bunch of FETs connected in parallel. I have another box full of 30A/150V fets. Having 10 of these in paralle should do it. With FETs there must be quite a lot of margin as the FET resistance grows when it heats up which again raises the temperature ....

The fans and the pump will be ran only when needed. The temperature sensor is used to switch the heater off if the temperature raises too high. The flow sensor is used to see that the coolant is really moving. The level sensor is used to check that there is enough coolant in the system. These should cover all probable scenarios of overheat, coolant blockage and other situations when cooling is not functioning properly.

The control logic will most likely be a MCU board having an Atmel AtMega64 on it. The user interface is power switch, current (average) and temperature meters, power adjustment, ontime adjustment, heat on/off buttons and indicator lights. Should do it. Will show the desighn of that one later.

[awemawson]

The coil needs to be closely associated with it's resonating capacitor bank - the circulating currents are enormous. Mine adjusts resonance frequency dependant on the charge in the crucible, but is nominally 3kHz

The tank capacitors need very careful specifying if they are to survive

Rather than hijack this interesting thread I have started another with some pictures of my 100kW furnace here :

https://madmodder.net/index.php/topic,12596.0.html

[eskoilola]

Rather than fiddling with hall sensors and trying to split ferrite toroid it might be better to measure the current before the beefy rectifier with a current transformer. The result might be even more accurate and the experiences I have from hall sensors are not that encouraging.

I made a hall sensor thing for one of my friends to measure the windmill output current and that sensor was far from accurate. The result depended on the orientation with earth magnetic field - this had to be compensated and in addition those ferrites have some sortr of a hysteresis so that after a decent current spike there was remaining magnetism which affected the result as well.

I have also been reading a lot of articles on induction brazing, induction annealing and induction hardening. Those articles are encouraging and it seems like it would be a good idea to equip this with some sort of programmable process control. I do not mean numeric entry but rather a set of knobs to set heat-, sustain and decay times and the sustain power.

I would be really interested  on Your comments on the user interface - with a bad one it is easy to convert a perfect tool into crap.

[awemawson]

If you look at one of the photos of mine you will see that power is monitored by current transformer on two phases (no need to measure the third) of the buss bars, and the voltage is monitored there as well so the microprocessor can do a simple calculation and display power rather than current.

[eskoilola]

If you look at one of the photos of mine you will see that power is monitored by ....

It was Your pictures that made me rethink the current measurement. They are actually quite helpful ... cannot reverse engineer the schematics of that controller though 



Here is a mockup of the user interface and some of my thoughts of the functionality.

Far left are the mains switch and the mains indicator light.

The three round knobs on the bottom are for adjusting the heat up, sustain and cool down times. The time range for each might be one hour.

The yellow buttons on top of the three  knobs are buttons/lights. When a particular phase is running then a corresponding button has it's light turned on. It would be possible to enter any phase just by pushing corresponding button.
The big knob is the power. Range is from zero to 100%

On top there are three status lights/buttons.

Temperature/cooling status. When running OK the light stays on and when in error the light blinks.

Next status light/button is current/voltage measurement. The light is on during normal operation and will blink when in fault.

The last status light/button is a generic fault. When operating normally the light is tuerned on and in error this blinks

When the heating process is running pushing any other than the phase buttons will abort the process

When the heating process is not running then pushing the coolant button will start the coolant system and sets the light blinking. Pushing again will stop the coolant system.

[awemawson]

It was Your pictures that made me rethink the current measurement. They are actually quite helpful ... cannot reverse engineer the schematics of that controller though 

Actually back in 2005/6 I did go a long way to working out that controller - I got it down to 'block' level and component level for some of it. I then managed to chat up a girl at CFEI who copied a manual for me - still got it somewhere, I'll dig it out and see how detailed it is - it was a long time ago and much water has flowed under the bridge since then 

[PekkaNF]

OT....maybe....

.... and the experiences I have from hall sensors are not that encouraging.

I made a hall sensor thing for one of my friends to measure the windmill output current and that sensor was far from accurate. The result depended on the orientation with earth magnetic field - this had to be compensated and in addition those ferrites have some sortr of a hysteresis so that after a decent current spike there was remaining magnetism which affected the result as well.
...

Do you mind opening up a new (short) thread for it? I'm interested how did you do it and how it went wrong. To my understanding shunt resistor/kelvin connection is most accurate (and problematic too), then current transfomer is prefered on mains frequency and HAL is used when those don't fit.

I have semidecent, pretty cheap Hal probe for oscillsocope and I suspect it uses the universal priciple of using HAL.to detect "zero" flux, ie. there is a compensation winding to generate opposite flux to flux generated by measured current:



https://meettechniek.info/instruments/scope-probes.html
https://www.digikey.com/en/articles/techzone/2017/aug/understanding-selecting-effectively-using-current-probes

This is something I haven't eperimented first hand..might be all BS.

Pekka

[eskoilola]

Do you mind opening up a new (short) thread for it? ...

The thread

[Pete W.]

OT....maybe....

.... and the experiences I have from hall sensors are not that encouraging.

I made a hall sensor thing for one of my friends to measure the windmill output current and that sensor was far from accurate. The result depended on the orientation with earth magnetic field - this had to be compensated and in addition those ferrites have some sortr of a hysteresis so that after a decent current spike there was remaining magnetism which affected the result as well.
...

Do you mind opening up a new (short) thread for it? I'm interested how did you do it and how it went wrong. To my understanding shunt resistor/kelvin connection is most accurate (and problematic too), then current transfomer is prefered on mains frequency and HAL is used when those don't fit.

I have semidecent, pretty cheap Hal probe for oscillsocope and I suspect it uses the universal priciple of using HAL.to detect "zero" flux, ie. there is a compensation winding to generate opposite flux to flux generated by measured current:



https://meettechniek.info/instruments/scope-probes.html
https://www.digikey.com/en/articles/techzone/2017/aug/understanding-selecting-effectively-using-current-probes

This is something I haven't eperimented first hand..might be all BS.

Pekka

There is another way that doesn't need the magnetic field sensor.  You detect the second harmonic of the input signal generated by the magnetic non-linearity of the core material.  Then you use that signal to servo the compensation current to drive the second harmonic to zero amplitude.  Very similar to how a flux-gate magnetometer works. 

[eskoilola]

The radiator (oil cooler) arrived today. It is exactly the size and shape I expected it to be. Very nice unit.




The threads might be BSP but at least they are not tapered. That would have made fabricating the adaptor a manly experience. I can see that the adaptor actually consists of two parts. The nut and the hollow piece that makes a seal with the taper. Must be made out of aluminium (which I hate) to avoid electrolytic corrosion.


Two 9 centimeter blowers side by side pushing air through that one should create cooling enough for the heater. I have also planned to create a water cooled heatsink for the rectifier as the water is already available. The rectifier produces almost 100 Watts of heat when operating at full power. Arranging a separate air cooled heatsink for those would consume a lot of space and would require yet another fan.


The fets have their individual heatsinks at the moment. I assume that those could also be water cooled. If the capacitors can survive with replacement air cooling then this would cancel the two fans on top of the driver.


Basically the water could be tap water as this is a low voltage driver but I will use deionized water in order not to collect calcium and iron (our water has much of that) and other salts that can be found in our tap water. Rain water might also do it. Lots of that during a typical finnish "summer".

[eskoilola]

This is my insight of the inputs and outputs of the controller.

The inputs are the usual stuff the coil presence detection being a little bit challenging. Other that that nothing new here.

On the output side the FAN might be controlled by medium frequency PWM which in turn would be controlled by the coolant temperature vs. ambient temperature. It is idiotic to blow air through the radiator if that air does not cool it down.

The Driver power will bw controlled by low frequency PWM - maybe 20Hz. Might be higher if the driver can cope with that one.

[eskoilola]

Some progress on this one.
Have now designed the MCU and PSU portion of the controller unit. The used MCU will be Atmel Atmega32L which has just enough support for this design.
Next part of this design will be the FET array driver, relay drivers (for the Mains) and other hardware interfaces.

In case someone is interested, the first page (the only page at the moment) is available as a PDF file here.

[eskoilola]

Some more progress on this one.
Have now designed the FET array driver, relay drivers, coolant pump interface and the FAN interface.

I have tried out several ways to drive the gates of a FET array. In this case the array will consist of 10 high current fets connected in parallel. The difficulty of driving these is the quite high capacitance at the gates which are all connected in parallel. In addition to this there will be the miller capacitance (feedback from the connected voltage) which does nothing to help the fets switch rapidly.
Most of the commercially available FET drivers do not come even close to the currents needed here. The gate current will peak at about 10 amperes. Fortunately this peak does not last very long after which the current is neglible. This behaviour of fets is one reason why IGBT components are so common in high current switching applications as they do not have such high capacitances.

The next portion of the design will be the analog interfaces.


In case someone is interested, the schematics are available at:
PDF Page 1

PDF Page 2

[eskoilola]

The final page of the control schematic for the induction heater is ready.

The last page contains the interface to AD converter including current, voltage and temperature sensing. In addition to these there are the 4 potentiometers to control attack, sustain and decay times as well as sustain level. Those are terms from some musical synthesizer - I just could not figure out better words for these. Hopefully the device does not make loud sounds - I like when a powerful device is doing it's stuff without any additional complaints.

Talking about machine noise. This one should become really silent. I do not know how much noise the coolant pump will produce but at least the cooler fan is the quietest I could find with enough airflow to keep things cool. The two blowers installed by the wise chinese on the driver will have to go. Those are of inferior quality and will make a very loud noise for sure.

If the capacitors on the driver PCB get hot I just simply replace them with better quality capacitors that have smaller losses and do not get hot. Unfortunately I could not find any references and/or datasheets for those capacitors so I really do not know what they are made of so I have to figure out that empirically.

The next step for this controller is to design the PCB. This PCB is quite easy as this does not contain any RF stuff and generally all electricity on this board will stay on the conductors wich would not be the case with high frequency, voltage or current. The high currents are elsewhere in this apparatus.

There is, however, one thing that I must be aware of when designing this device. There will almost certainly be a lot of RF radiation around. For this reason a double sided PCB is a must and the other plane should be kept as much as possible as a ground plane. There is already some isolation for the ADC ground which must be kept separate from the general ground. Further more, all sensitive cables must be shielded and grounded on the PCB. The high power supply will have it's ground connected to the PCB close from the FET array and only there. The wiring of this heater will be an interesting task as well.

For those who are interested about the schematics, those are available as 3 PDF files via the links below.

PDF Page 1
PDF Page 2
PDF Page 3

[eskoilola]

Now, the ratsnest has always an intimidating appearance. But I am not afraid - I have done hundreds of these.

[eskoilola]

So what's inside that flow meter.



Nothing much really. An impeller with a magnet. One hole to let the fluid in and another to let it out. O-ring to keep the fluid from getting into places where it is not supposed to get into. This is absolutely not a pressure vessel so it must be placed between the pump and the coolant reservoir. Being plastic I very much doubt that only water can be used as coolant - which is actually OK. The orientation of this device is the name-plate down. If oriented in any other way the sensitivity will suffer as a result of poor bearing alignment. Orienting name plate up might work but not as well as name plate down. I need to create fasteners for the sensor and also fittings for coolant in/out.



The inlet hole is rather small being under 3mm. Will open this up to be the same diameter as the output hole which is almost 6mm. This will of course make the sensor less sensitive but it is not my intention to actually measure anything. The intention is to see whether the coolant is circulating or not.

So far this sensor seems allright to me. It could use a better temperature tolerance but I assume it will not get destroyed with coolant temperatures below 100 degrees celcius.

[eskoilola]

Almost there !



This is the top side of the PCB. What is missing are the traces and planes on the bottom side. At this point the worst is over. I could distribute the components more evenly on the PCB but I am not sure whether this is actually a good idea. The bottom layer is void on components and at the moment it is unused so it should not be too painful to create the planes and supply rails there.

I changed the three separate air cooled heatsinks to an aluminium block which will then be water cooled.

I have a few questions for You guys and ladies:

- assuming that there are water holes inside the (10 mm thick) aluminium block, how would You create the holes?

- does anyone have experience on having the PCB done ... probably in china? This one should use through plated holes and those are beyond my capabilities.

[vtsteam]

- assuming that there are water holes inside the (10 mm thick) aluminium block, how would You create the holes?

Not sure how you want channels arranged, or how many but you can drill and plug as a single block, or make the block 3 piece -- a body and end caps. The body has the lengthwise drilled channels and the caps have the drilled manifold holes that tie them together. And an inlet and an outlet.

Or if you just plan on a single internal chamber, make it two piece, split lengthwise, and mill out the cavity.

[eskoilola]

The PCB and the electronic design is now as ready as it becomes doing it virtually. I will order this PCB from some company that is specialized in small prototype series. They are usually not very expensive and the quality is far better than I could ever reach at home.

The schematics, Eagle project, PCB design, and all files needed to produce the PCB are at Your disposal in case someone is interested and if this device materializes itself as expected. It is my hope that these are useful for someone - if not used directly but serving as an example of what or what not to do depending on how this turns out.

The files are here