ATX to TEC Power Converter and regulator build

[Glyph]

The above circuit actually draws its power from the ATX-TEC boost converter. (SV1)

The fan header is isolated by optocoupler and only provides the voltage signal needed for the controller to adjust the converter.

That being said... Above circuit is untested , I personally i have no use for a fan header controlled unit so i probably never will test it.

While i am very confident it will work, a seperate problem is that it has an inherent flaw: if the fan header fails, or gets disconnected, the unit interprets it as "zero signal" and consequently believes it should apply no power to the TEC.... it doesn't take a scientist to figure out what happens next.

[memolipd]

The above circuit actually draws its power from the ATX-TEC boost converter. (SV1)

The fan header is isolated by optocoupler and only provides the voltage signal needed for the controller to adjust the converter.

That being said... Above circuit is untested , I personally i have no use for a fan header controlled unit so i probably never will test it.

While i am very confident it will work, a seperate problem is that it has an inherent flaw: if the fan header fails, or gets disconnected, the unit interprets it as "zero signal" and consequently believes it should apply no power to the TEC.... it doesn't take a scientist to figure out what happens next.

ahh true but this is where the genius of my idea to mount the TEC cold side to cool the water, completely seperate to any hardware. Therefore if the TEC packs in then your left with a standard watercooler.

the down side to this approch is that you cool the water (which is pretty difficult) instead of the processor directly, as a result you'll never get anywhere near sub-zero cooling on ur CPU but u will keep the water temp down when its needed.

just to make it even more pointless, I found the water temperature only increases by a single degree C when the system is under load and therefore the ambient temperature has a MUCH greater effect on the system temp than the TEC on/off.

having the TEC turn on and off as i do at the moment is probably the best solution for me due to it not affecting the water temps as much as i'd hope, as it currently only switches on when the water temp is wayyy too high. I have yet to reach this temperature -LOL therefore i have never had a need to use it and so im saving a bunch of cash.

A bigger TEC mght be the solution to the problem, I will investigate/upgrade my TEC when i next drain my loop. I then may have a good need for this circuit. In the mean time, GL

[Glyph]

Hi everyone, me again.


After tinkering around with the prototype (of the circuit defined in #43) under various load conditions i found that it isn't quite stable under low load conditions. For those of you that want to try undervolting your peltiers down to lower levels (like say 50% or 30% of vmax) the converter might not regulate properly. I find that it goes to an unstable "hiccup" cycle.

I found that by chucking out the synchronous mosfets and increasing the size of the schottky stability can be achieved even for very low load levels.

The snag with this approach:

Converter efficiency suffers. The schottky must be heatsinked. So much for my requirement that this was fanless...

now i think that with better (I.E. more expensive) circuitry that it might be possible to build a unit that works at both low power levels and still be fanless at high power levels.

What design criteria would you guys want in a converter? Fanless? or cheap?

Personally, i'm thinking that since ALL TEC setups must have great cooling they must also have great fans, I could design the converter layout so that its easily mountable behind one of the fans already used for TEC cooling. Since the converter can operate even at 70 degrees it can be mounted on the "exhuast" side of the fans and therefore incur no penalty to system cooling. Snag is that some people might think it looks UGLY.

=====
On a different note: I've been taking pictures of my progress, should i post those here? or in a new thread so i don't clutter the crap outta this one?

[repsol_23]

I think that this thread would be as good a place as any to place your pics. I for one would be really interested in seeing pics of your progress! Thanks.

Hi everyone, me again.


After tinkering around with the prototype (of the circuit defined in #43) under various load conditions i found that it isn't quite stable under low load conditions. For those of you that want to try undervolting your peltiers down to lower levels (like say 50% or 30% of vmax) the converter might not regulate properly. I find that it goes to an unstable "hiccup" cycle.

I found that by chucking out the synchronous mosfets and increasing the size of the schottky stability can be achieved even for very low load levels.

The snag with this approach:

Converter efficiency suffers. The schottky must be heatsinked. So much for my requirement that this was fanless...

now i think that with better (I.E. more expensive) circuitry that it might be possible to build a unit that works at both low power levels and still be fanless at high power levels.

What design criteria would you guys want in a converter? Fanless? or cheap?

Personally, i'm thinking that since ALL TEC setups must have great cooling they must also have great fans, I could design the converter layout so that its easily mountable behind one of the fans already used for TEC cooling. Since the converter can operate even at 70 degrees it can be mounted on the "exhuast" side of the fans and therefore incur no penalty to system cooling. Snag is that some people might think it looks UGLY.

=====
On a different note: I've been taking pictures of my progress, should i post those here? or in a new thread so i don't clutter the crap outta this one?

[zipdogso]

After tinkering around with the prototype (of the circuit defined in #43) under various load conditions i found that it isn't quite stable under low load conditions. For those of you that want to try undervolting your peltiers down to lower levels (like say 50% or 30% of vmax) the converter might not regulate properly. I find that it goes to an unstable "hiccup" cycle.

Would this really be a problem ?
The only way to use undervolted pelts is to multiple them so if they are in series you are back up to the sort of voltages your initially thinking of. You only undervolt to get the efficiency then multiply the pelts to get back up to the heat pump. A single undervolted TEC is near useless.

[Vinas]

Just put a fan on it. (=

[Glyph]

Ok here we go. the step by step worklog of my progress. Since some people find the PCB making process fascinating i've included that too.

First picture, Design of the board. I use EAGLE from cadsoft for this.

Second picture. Print out the board on transparency sheets. This allows me to use the UV etch option. The Toner transfer method of making PCBs cannot resolve the very fine traces i need for the surface mount components.

Third picture is placement of the sheet onto the board. The board is presensitized so UV exposure will induce chemical reactions that allow later steps to strip off the unwanted copper. I placed the sheet under a piece of glass to keep it even.

Fourth Picture is the exposure setup. I got a nice powerful UV lamp. Exposure itself only takes 2 minutes.

Fifth picture is the developer step. The board after exposure is immediately dunked into a mixture of sodium hydroxide and sodium carbonate in water. This removes the protective coating in the areas that were exposed to UV light.

[Glyph]

First picture is some fully developed boards.

Second picture is cutting the boards with a pair of scissors.

third is the etch step. The unwanted copper is etched off in a mixture of hydrogen peroxide and sulfuric acid. The mixture must be warmed to 60 degrees celsius for this to work.

fourth is a close up.

the fifth picture are some fully etched boards. I selected the board in the right since it was the best one.

[Glyph]

First picture is drilling. I used special carbide drill bits to drill the very small holes needed for some of the components.

Second picture is the pre-soldering step. Alot the components i'm using are surface mount and it its notoriously difficult to solder surface mount parts by hand. So i'm doing it by a reflow method. First, I need to apply solder to all the points where there will be surface mount components.

Before anyone starts screaming NOOB... i'm well aware the better method is solder paste. But solder paste has a limited shelf life and requires refridgeration. Not very appealing to DIYers that aren't doing surface mount on a regular basis. The Pre-solder point method is lower quality but can be done by anybody that already has a soldering iron and regular wire solder.

Third picture is all the solder points done. The amount of solder you need to put on each spot varies with the component. experience helps.

Fourth picture is pre-heating of the hot plate. The plate must be above the melting of solder to melt all the solder points and solder the components to the board. I prefer a temperature of 235 degrees celsius. I've almost reached it in this picture.

Fifth is reflow soldering of the board. All the surface mount components are placed on the board and the board is placed onto the hot plate. The solder points melt underneath the surface mount components and solder them to the board. Whole process is pretty quick and easy to do.

[Glyph]

First picture is a close up of the board with the surface mount components attached to the board. Pretty sweet.

Second picture is placement of the through-hole components on the other side of the board.

third picture is hooking up the board to the testing rig.

I know the board is missing two transistors and has some resistors added. The pictures so far where of the first prototype that didn't work. The sixth prototype that did work is shown in this picture. The process of making all the prototypes were the same with just some component variations.

Fourth picture is the waveform of the gate-driver. A nice square wave, which is what i'm after.

Fifth picture shows the circuit in action with a dummy load of some LEDs. The batteries are providing 11.1 volts going in and the converter is boosting the voltage to 24 volts indicated by the voltmeters.

IT WERKS!!!!!

....although not as perfectly as i'd like..... ah well, more prototypes to come.

[Glyph]

So there ya go guys.

It IS possible to build a boost converter for running TECs of an ATX... Sure i've toasted more money building these prototypes then i would have saved, but at least this is loads of fun


On a different note. Keeping the unit fanless is becoming exorbitant in complexity and cost. My next prototypes will probably require fans. I figure if anybody is HELLBENT on a silent converter they can always get a mosfet waterblock and cool the schottky that way.


New schematics coming soon.

[littleowl]

looking good Glyph.
so what voltages are you about to run at so far?

[Glyph]

all the way up to 41 volts.

[littleowl]

Nice!! what wattage range are we looking at then? How many Amps at what voltage?

[Vinas]

all the way up to 41 volts.

Great job, Glyph!!! Got a little excited with the solder I see! The end result is impressive, still! Looking forward to see how you'll enclose this. Assuming you were going to add a fan and enclosure.

[Glyph]

Nice!! what wattage range are we looking at then? How many Amps at what voltage?

the components i selected should in theory go up to 600 watts with an output limit of 30 amps. so at 40 volts this should be 15 amps. at 24 volts this is 25 amps. but at 15 volts it hits the output limit at 30 amps.

But there is no actual limiter circuit in the device. so if you drew 40 amps from it, it won't stop you. But the parts will overheat

I haven't been able to test the current capability fully since i don't have an ammeter that can go that high. i'm looking into purchasing one in the near future.


adding more capability is just a matter of adding more components and thicker copper traces.... but i don't think anyone has more than 600 watts of spare capacity in their ATX. Anyone needing more than 600 watts for TECs should get a seperate supply.

[Glyph]

Great job, Glyph!!! Got a little excited with the solder I see! The end result is impressive, still! Looking forward to see how you'll enclose this. Assuming you were going to add a fan and enclosure.

thanks alot,

The generous solder is because this is high-amperage, gotta use lots of solder to let those amps flow.

as for enclosure, I'll probably leave it open frame. it needs good airflow to keep things cool. also, letsay the computer case already has a fan the user can mount the device just behind one of them and not add any more fans to the system then they already have.

[Glyph]

First picture is the device after i put on the 12v connectors.

Second picture is the device connected up to the computer. The TEC wires would go into the green terminal block.

As you can see, to handle the high currents you need LOTS of connections

Do you guys think there is enough demand for such a device that i could sell it? I might consider it if it was viable.

[Vinas]

If you priced it right I think so. Those of us with extra ATX PSU's or mega supplies could find value here. So many people post about getting 24v out of an ATX supply that it's getting very old. Just put up a paypal or ebay site.

[Glyph]

hmmmmm.... other power supplies....

So far i was only thinking of converting ATX supplies to the right voltage but if people have other supplies of other voltages maybe they would want those supplies converted as well.

looks like an entirely new design is in the works. A universal power converter for TECs.

[memolipd]

hmmmmm.... other power supplies....

So far i was only thinking of converting ATX supplies to the right voltage but if people have other supplies of other voltages maybe they would want those supplies converted as well.

looks like an entirely new design is in the works. A universal power converter for TECs.

hell, who said anything about TECs why not go all out and call it: "the universal power converter" u can plug anything in and get anything back.

looking lush Glyph (even if it is more effort, money, time, skill and larger risk than buying a seperate PSU - it still looks fun tho!)

[Glyph]

hell, who said anything about TECs why not go all out and call it: "the universal power converter" u can plug anything in and get anything back.

looking lush Glyph (even if it is more effort, money, time, skill and larger risk than buying a seperate PSU - it still looks fun tho!)

heheh

other than TECS, i dont see much use for a converter in a computer. I've been told those 24v pumps dont actually run better at voltages above 12v.

Led lighting maybe? but they're all set to 12v anyway.

as for other non-PC markets, a meanwell is a better option for 'em

[zipdogso]

Ahhh...I see now, how you do surface mount ...interesting.
On account of the required hardware (although it would not be beyond the realms of reality to find some suitable heating device.) I will stick to "thru the hole" don't really build enough circuits to bother.

[Glyph]

Since im radically changing the design. My first thoughts are using the dsPIC30F2020 which is available in through-hole. all the components can be through-hole if you so desire.

Since i'm moving toward a heatsinked design rather than fanless, its a very highly likelihood the next prototype will be through-hole.

-----------
quick question to all: Should i keep posting updates here (like a blog) or only when i got working prototypes?

[THE JEW (RaVeN)]

Impressive work! It's not too often I get to see people fabbing their own boards so this was of particular interest to me.