Induction brass annealer redux

[ottsm]

@BillK55 Good stuff. I very nearly bought a larger version of the little aquarium pump that's on the various BoM lists for this very reason. I note also that this is an area where the EZ Anneal did a better job: a larger reservoir for starters and, by the looks of things, a more powerful separate pump. (If you browse the Alphacool website you will see a few of their parts.)

In the end I just went with the flow ('scuse the pun) and stuck to the little guy. I don't have a lot of space in my enclosure. What I did do was minimise any narrow tubing. In post 1 or 2 of this thread Gina shows a transition method to go from 3/8" ID tubing to 1/4" ID. I routed everything with 3/8" ID tubing and merely slipped a clamp-width section of 1/4" ID plastic tubing over the end of the 1/4" copper that passes across the induction board. (No need for the separate copper transitions.) It looks like the EZ Anneal may have gone for thicker coil tubing as well. Also they appear to have a longer coil, i.e. more turns, which would provide greater induction (more turns = more henrys) and quicker annealing times, potentially meaning less heat transfer from the hot cases to the fluid in the surrounding coil.

@ottsm probably has the best data on the cooling system. He is actually measuring the flow rate: 5.5 gph according to his video. He is also measuring the temperature of the coolant (albeit with a thermistor strapped to the plastic tubing rather than something with better heat transfer). I'm not sure which pump he is using and of course his results will depend on how he has coupled it all together plus his fan setup.

At the end of the day, however, much of the heat in this system isn't really in the coil. Unlike a computer system where one is trying to suck heat from, say, the CPU, here the hot part is ejected from the system through the trap door. It releases most of its heat sitting in a tray somewhere. We have two elements which require some cooling: (1) the heat generated from the induction board capacitors and diodes and (2) heat transferred to the coil from the induction process. (1) is dealt with by fan circulation of air. Re (2) annoyingly the cheapo induction boards are very poorly laid out and pass copper tubing straight over the top of the hot capacitors**. If annealing times are relatively quick the hot case is ejected before being able to transfer a lot of heat over the air gap between it and the coil. So there's less reliance on the liquid cooling system. At the end of the day, if the coil isn't getting too hot you're golden.

I added Insultherm to my coils for two reasons: one, I could get a tighter coil with bare copper tubing (no enamel insulation) without fear of shorting a coil loop and, two, to reduce heat transfer from the heating case to the coil. I reckon it is worth the extra several bucks. (It also looks good.)

I still think a better place for optimisation is the induction board itself. It's not a complicated board and is about as basic as it gets for this application. I'm not surprised that the EZ Anneal folks seem to have made their own.

** I am going to wrap some Kapton tape over the copper tubing in this area.

The pump I am using is a little tricky to figure out but it appears to be the following;
http://www.lg-motor.com/chanpinzhanshi/414.shtml
I'm using the 12Vdc version with a max lift of 4 meters which is listed at 0 flow so this is dead head. The max flow possible at 0 lift is 3 L/min. My concern at the time was pump cavitation but it hasn't been an issue as of yet. I was using a different pump in the video and was getting above 5 GPH but this pump is getting about 4.5 GPH. I switched pumps because I cracked the housing of the first pump by accident.
I will say that the flow meter is operating at the low end of its range, I have not done a bucket check to see how accurate it is or if I even got the equation correct. The following code calculates every 10 seconds, the "Flow_Total" is incremented by an interrupt elsewhere in the logic. Note the text following a "//" is a comment not actual code. To get 1380 pulses per liter we take F=23*Q or 23*60=1380 pulses/Liter

// Calculate Flow Rate
if(Flow_Calculation.check() == 1){
Flow_Rate = (Flow_Total / 10.0) * 0.68913; //Flow meter; 1380 pulses/Liter or 5224 pulses/gal, gph = pps * 3600/5224 or .68913
Flow_Total = 0;
}

If the pump of GinaEric is listed at 1.9 meter and getting 2GPH then I'd say the equation is close with the pump above at 4 meters and 4.5 GPH. To get it to work it must be mounted with the label end up to get the maximum number of pulses.
https://www.amazon.com/dp/B01D44N41U/?tag=accuratescom-20
If you look at the specs the low end is listed at 0.3 L/min, I am operating way below that range. I probably wouldn't have even tried it given that information. At the time I got the flow meter I didn't really have any idea what the flow rate would be. My main goal was and is to interlock the system down if I sense the flow is low. If the flow drops off and because I am calculating the flow every 10 seconds it may actually get one or two pieces of brass through but eventually it will stop before anything gets too hot. I intentionally stopped flow once before and kept putting brass through the system, that coil never did flow as much after that. I don't know if it melted some of the solder or the water flashing off could have coated the coil, either way the damage was done.

 

[BillK55]

My main goal was and is to interlock the system down if I sense the flow is low. If the flow drops off and because I am calculating the flow every 10 seconds it may actually get one or two pieces of brass through but eventually it will stop before anything gets too hot.

When you're on the low end of what the manufacturer says is the rated meter range, its not going to be linear, so you won't get an accurate flow, but it should be repeatable. If you set your alarm at half of your pulses per minute at what you know is normal flow, it should give you a reliable alarm to safely shut you down if something goes wrong.
I think I ordered a flowmeter similar to yours https://www.ebay.com/itm/1pc-Water-...Flowmeter-Counter-0-3-6L-min-New/300952068345 for $2.66. For that price everyone that is doing a microprocessor based build should add one. I ordered a little cheapo panel DC volt meter that I'm going to somehow rig it up with. I'm not going to try and calculate a flow, just get a reading I can use as a reference as I'm testing out my unit. Maybe later I can work out how to tie it in as a safety interlock.

Using these little pumps its best to try and keep them mounted below the reservoir for best efficiency. Centrifugal pumps never lift as efficiently as they push. With as much backpressure as we're dealing with pumping through the coil, you're not likely to get cavitation as long as you use pump connection sized tubing.

 

[ottsm]

I looked back at my code, I actually look at the flow total every seconds for the interlock trip and if I'm less than 4 pulses I kick out, the actually GPH reading for the display is calculated every 10 seconds. So I'm tripping below 2.75 GPH within a second.
The radiator I'm using has a built in reservoir and the pump is mounted directly on bottom of the radiator.
https://www.swiftech.com/MCRx20-QP.aspx?variation=177
I just did a check to see how long it took to fill 1 CUP. I ended up getting air in the system so this effected things some. I was also drawing out of another cup but still going through my radiator and then emptying into a measuring cup. The suction side height was changing so the head and thus the flow rate varied some.
It took 134 seconds to fill one cup which is (1 cup/134 sec * 1Gal/16cups * 3600 sec/ 1 hr) which is 1.68 GPH, which the flow reading probably averaged 2.00 GPH.
This is opposite one would have thought, slippage past the paddle and friction of the paddle I would have thought made it read lower. When the system is purged of air and on the closed loop it is very repeatable on what I read. I would have to do a longer test with the air out to get good answer. As for now it's in the ball park.

 

[BillK55]

I looked back at my code, I actually look at the flow total every seconds for the interlock trip and if I'm less than 4 pulses I kick out
I just did a check is 1.68 GPH, which the flow reading probably averaged 2.00 GPH.

I'm surprised that your 4 pulses/sec is below their recommendation, but anyways it sounds like you've got it covered.
Gina actually measured 100ml in 1 minute, which converts to 1.58gph, so you're running very close to her flow.

 

[BillK55]

s

The radiator I'm using has a built in reservoir and the pump is mounted directly on bottom of the radiator.
https://www.swiftech.com/MCRx20-QP.aspx?variation=177

With your radiator's spare inlet/outlet ports, it'd be real easy to hook up that additional pump I was talking about to recirculate through the heat exchanger, boosting the radiator's heat transfer efficiency and also get more flow through the coil. Actually having the reservoir integral to the radiator gives you even more potential for improvement than with the original GinaEric pump. As long as there's not something about your radiator's mounting orientation that would make reversing the flow through it a problem, then consider the following (if orientation causes some issue, then my suggested flow path will need to change up a little): assuming your current flow path is radiator to pump suction to coil and back to radiator, then you'd connect the new pump's discharge up to the side of the radiator that your current pump's suction is attached. The new pump's suction would connect to the side of the radiator where the outlet of the coil attaches.

 

[SGK]

Bill

Here is what your eBay seller sent me overnight


“We have send email to our related department to confirm it, and we will give you reply within 48 hours, please don't worry. But if you still have not received our message in 48 hours, please contact us again to get the solution.
Any questions, please let us know.
Best Regards”

 

[dskogman]

I got my Power supply from Totiwo on eBay. It does have a fan and has handled a few blown inductors without a problem. They also sent it DHL so it took about a week to get here.

 

[dskogman]

The heat in the induction PCB board is the limiting factor with the current design. The water cooled coil really does not get hot at all.
With a case triggered instead of timer driven system the duty cycle can get very high not allowing enough time for the capacitors time to cool. It’s not a problem with a 600 watt PS and some delay time but with 850+ watt and continuos feed it is.
I’m running 284’s at 2.5 seconds with a 1k watt PS and 50 amp PCB and have burned it up once.
I’ve added temp sensors to the PCB and more fans and heat sinks and it helps but some water cooled heat sinks is really what it needs.

 

[BillK55]

But if you still have not received our message in 48 hours, please contact us again to get the solution.

Lol yes that was same auto-reply I got from them 10 days ago. I just replied to their auto-reply. Maybe they receive emails via the return boat to China.

 

[SGK]

The heat in the induction PCB board is the limiting factor with the current design. The water cooled coil really does not get hot at all.
With a case triggered instead of timer driven system the duty cycle can get very high not allowing enough time for the capacitors time to cool. It’s not a problem with a 600 watt PS and some delay time but with 850+ watt and continuos feed it is.
I’m running 284’s at 2.5 seconds with a 1k watt PS and 50 amp PCB and have burned it up once.
I’ve added temp sensors to the PCB and more fans and heat sinks and it helps but some water cooled heat sinks is really what it needs.

What he said. The heat generated by the induction coil is largely ejected with the case through the trap door.

 

[SGK]

So easy to miss the little bit circled in red. The Sestos timer manual implies that it is already held at 0V internally. (As if pins 1 and 9 are connected when powered by DC.)

Took me about 2 hours to figure out why I couldn't trip the timer when inserting a case. And I did it the hard way - I only looked at your schematic after I had figured it out. FFS.

 

[BillK55]

The heat in the induction PCB board is the limiting factor with the current design. …….
I’m running 284’s at 2.5 seconds with a 1k watt PS and 50 amp PCB and have burned it up once

Your reference to a 50A ZVS brings me back to the theory behind how we're wanting to control the annealing process. I had a question about your choice of 50A rated ZVS, but before I ask I wanted to make sure I really understood how we wanted to control. Yes I know the ultimate goal is we want to hit 750 at the shoulder, but I'm talking about what's important in how we control the annealing circuit.
With the basic 600W GinaEric the ZVS and coil, if the case is inserted too far, it can pull more amps than our PS can give if we run at full 48V, so we limit our power consumption to prevent overloading the PS by a combination of dialing down the voltage and adjusting the case insertion. We never come near the 20A or 1000W limit of the ZVS
Along comes the RSP-750-48 750. Its got slightly more power (125% of the 600W) and again doesn't have the ability to exceed the 20A or 1000W rating of the ZVS. Its big plus is it has current control ability up to its 15A max while automatically maximizing the volts. I understood from one of Gina's posts that this brings us close to eliminating having to play with case into coil insertion, other than with the extreme small or large cases you just do a "standard" case insertion, and then find the highest amp setting that gives you not too fast of an annealing time. Does it sound like I have it? Please point it out if there's something I missed.

 

[dskogman]

You are correct for a timer driven design. There is enough built in delay to keep the duty cycle down and the board time to cool. I’m not sure that if you use a case triggered mechanism and an auto feeder that that you could not overheat it over time. Someone else may have more info on that.

If you want to anneal faster you need to make your coil smaller and use a larger power supply and inductor board. You can easily go over 20 amps for a few seconds. That coupled with a case trigger and fast feeding can overheat the board in 20 cases or so without lots of extra cooling. I now have temperature sensor data displayed on the LCD and slow down or stop for a few minutes if it gets too hot. The PCB actually cools down pretty quickly when it’s not running.

 

[GrocMax]

The heat in the induction PCB board is the limiting factor with the current design. The water cooled coil really does not get hot at all.
With a case triggered instead of timer driven system the duty cycle can get very high not allowing enough time for the capacitors time to cool. It’s not a problem with a 600 watt PS and some delay time but with 850+ watt and continuos feed it is.
I’m running 284’s at 2.5 seconds with a 1k watt PS and 50 amp PCB and have burned it up once.
I’ve added temp sensors to the PCB and more fans and heat sinks and it helps but some water cooled heat sinks is really what it needs.

EDIT- Mine is the 'old' 1000w/20A board. I bought two thinking it was a high probability I'd blow one, but its been rock solid.

I've torture tested mine running at 15.7A for close to 15 minutes continuous, peristaltic high pressure capable coolant pump, and 4 heat sinks on the bottom of the PCB between the capacitor leads, and a temp sensor activated PWM speed controlled pretty big 60CFM+ fan on the induction board. Can do multiple hundreds of cases as fast as possible, no overheating issues of either induction board or coil.

I could not get the proper coolant flow out of 1/8" .032" wall 101 copper tube with the vane pumps, went to a positive displacement peristaltic unit that can build lotsa pressure if it needs to. Thinner wall (.014") coil tubing would definitely help cooling but I wanted to use pure 101 copper .032" wall was all I could find.

 

[SGK]

Along comes the RSP-750-48 750. Its got slightly more power (125% of the 600W).....its big plus is it has current control ability up to its 15A max while automatically maximizing the volts.

It can "constant current" up to 110% of rated current ability - if you have access to 5.5V as a control voltage. We tend to just have 5V available. So it doesn't sag current and doesn't exceed a programmable (dial-able) limit. In this application, if you switch the positive supply to the induction board you can only compensate for voltage drop over the wiring to the relay switch. And then only up to 0.5V drop which is neither here nor there in this application. I guess if you switch the low-side return from the load (to PSU GND) you can place your V-sense wiring at the load but again the compensation is only up to 0.5V. Just use short, thick wiring from the PSU to the induction board. (I haven't bothered to think about whether the ammeter shunt complicates things as I did not bother with voltage sensing since I don't think it adds anything to this application.)

I understood from one of Gina's posts that this brings us close to eliminating having to play with case into coil insertion, other than with the extreme small or large cases you just do a "standard" case insertion, and then find the highest amp setting that gives you not too fast of an annealing time. Does it sound like I have it? Please point it out if there's something I missed.

Not sure this is right. The inductive field is still strongest at the centre of the coil. So you still have to pick a point in the coil at which you align consistently your cases. Work out an annealing time based on that point and that case and apply it consistently.

The strength of an air core inductor is a function of the number of turns, the diameter of the coil and length of the coil

L= (d^2 * n^2)/(18d+40l)

where:

L is inductance in micro Henrys,
d is coil diameter in inches,
l is coil length in inches, and
n is number of turns.

Want more Henrys for a given copper tubing and number of turns? Wrap the copper tubing around a thicker pipe.

@dskogman even with a case triggered mechanism it is easy to add delay to slow things down to a minimum. Just add a greater or lesser pause after a case is detected (feeder stops immediately on case detection) before annealing begins. You don't want the pause at the other end with a hot case sitting in the coil for the reasons already discussed above - eject that heat through the trap door!


So here's a bummer... You go to great lengths to find nice quiet fans with plenty of CFM (my build uses two 88.6 CFM capable fans in an enclosure which is only 13 x 11 x 6") and control them via PWM so they only run as fast as necessary. Then you power up your PSUs and find the fans in the Mean Well RSP-750-48 make a friggin' racket... Grhh!

 

[BillK55]

I’m running 284’s at 2.5 seconds with a 1k watt PS and 50 amp PCB

Ok now that I've gotten sorta confirmation that I'm understanding the intended way to control the ZVS, I want to ask about your 50A model, what voltage do you run it at? Doesn't that model 1000W PS that you referenced top out at 40V? On one of the 50A ZVS boards I saw, the rated voltage was 40v, but they recommended not going over 36v. Its not always easy to tell watching people's builds on video to really know what voltage they're set at since the numbers are jumping, but one that I saw pretty clearly was the early GinaEric on page 1 of this thread. It was running at 43 volts. I think I recall another one running at 47 volts. Somewhere I read, and it may have been in this thread, or possibly somewhere else cause I've wandered all around in researching this, but there's a target frequency for the ZVS, and I thought somebody in this group actually measured what theirs was running at, and it was pretty close to the target. Also heating is more efficient for our brass cartridges if you're at that frequency. I'm not going to pretend to know how to calculate the frequency, but taking a wild guess, I bet volts is in there somewhere and shifting from 47 or 43 down to 36 seems like it would have a significant effect, but then you say you're running 284’s at 2.5 seconds, so something is working right!
The 50A ZVS board has the built-in fan and 8 vs 6 cap's that the 20A versions usually have, and the components are arranged a little differently. Looking at it up close can you tell if they've beefed anything else up to make it more reliable? Maybe wider or thicker copper in the circuit paths? I know on one of the failed boards I saw a picture of, it looked like the paths had burned and/or separated from the board. GrocMax's idea of attaching heatsinks and safety interlock temp sensors sounds like a good way to compensate, but I'm wondering what's the root cause. It doesn't look like there's that many components. Has anyone considered building a new board with design improvements to reduce heat buildup and then swapping the components over to it? Lol I suspect I'm wandering into that little bit of knowledge is dangerous (or makes me sound silly) zone.
I am planning on making by build case sensing based start, and an auto case feed, so yes I will be pushing the ZVS duty cycle a little, but I won't likely be running hundreds at a time to meet some deadline, however ya know how it is, I might like to occasionally show off. I just like to think all my options through and end up with a robust design. Oh yea, and cheap too

 

[BillK55]

I finally got a response back from Ebay seller ouyou2010 about their $62 48v 1000w PS. https://www.ebay.com/itm/142462540092 They say it HAS NO FAN. Now I almost don't want to believe them. if the pictures are right, the top and sides are closed. There are no pics of bottom and back for this model. The front appears to have an opening above the terminal strips that looks sufficient if there was corresponding opening in back for straight thru flow, but there is no picture of the back. WTF? Is there some application where people install banks of PS's like this and they're all ducted to a common ventilation supply?

 

[#40Fan]

https://www.aliexpress.com/item/100...69f4cea&transAbTest=ae803_1&priceBeautifyAB=0

Looks the same and has a fan on the back.

 

[#40Fan]

Here's another that the picture even matches.

https://www.aliexpress.com/item/Smu...69f4cea&transAbTest=ae803_1&priceBeautifyAB=0

Even states the same low working temp.

 

[SGK]

GrocMax's idea of attaching heatsinks and safety interlock temp sensors sounds like a good way to compensate, but I'm wondering what's the root cause. It doesn't look like there's that many components. Has anyone considered building a new board with design improvements to reduce heat buildup and then swapping the components over to it? Lol I suspect I'm wandering into that little bit of knowledge is dangerous (or makes me sound silly) zone.

The 'root cause' is a natural product of the circuit.

The only 'issue' with GrocMax's heatsinks is that they aren't where you really want them to be. Nonetheless they likely help a little at the margin. At $10, I decided to add them.

The circuit for these eBay models is very simple. 19 parts - less than my control board. The design is self-oscillating; there's no PLL control of the frequency of oscillation. It's a very basic circuit. Apparently they operate at about 100kHz which is a far higher frequency than commercial units. (I have not verified this.)

And, yes, the board could be laid out in a much better form. Give it a go. The EZ Anneal folks did. PCBs are cheap to produce. Pissing about developing a better version takes time and energy and had best be something you just get a kick out of doing, unless you are going to commercialise it.

"Marko" is this guy:

https://markobakula.wordpress.com/power-electronics/500w-royer-induction-heater/

He took Mazilli's flyback circuit and modified it to work as an induction heater. Note his comments about using the liquid cooling system to also cool the MOSFETs.