Extra volts worth it?

[Glyph]

Hey everyone,

I'm getting into TECs and i see plenty of them with Vmax of 15-24 volts.

A power supply that can do that is the ideal solution but I want to build a voltage booster circuit that can convert the 12volt ATX power line to the 12-24 range i need.

I'm well-aware of the issues with such a converter and that i'll need more current going in to make up for the increased voltage going out. And i'll need a strong ATX capable of providing such current.

I've estimated the circuit cost at ~$60. Cheaper than most dedicated variable power supplies. (ATX included)

My question is: is the extra cooling power from the extra volts worth the cost of the booster?

I know the answer lies in both science fact and personal preference so i'd like to know what you guys think.

[zipdogso]

My question is: is the extra cooling power from the extra volts worth the cost of the booster?

It depends....for starters if you want a TEC with long life you don't run it at Vmax.
Vmax is the epitome of INefficiency. The best operating point is around the 70-75% of Imax to correctly determine the voltage you really need the performance charts for the TEC in question. I have a reasonable chart for a 127 junction TEC and it is possible to extrapolate values on to it to give a rough indication for another 127 junction TEC. Just to get a value you could assume 70-75% of Vmax.
It is not considered worth running a TEC past this value so it could be argued that it may not be worth the cost of the booster to get the Vmax.

Why 70-75% ? Generally at higher percentages you are getting less cooling than the power of running the unit i.e. a COP (coefficent of performance.) of less than 1. The difference in heat pumped between 75% and 100% of Imax is quite small. It is an increase of approx 8% in heat pumped for an increase of 25% of input power !!

[Stigma]

zipdogso is right. Generally the less voltage you use the MORE efficient it is, in the sense that the ratio of watts consumed to watts pumped though is higher. Of course, to get a decent output to cool anything you can't exactly just run it at 5volts for a 24volt element. if I remember correctly it dosnt scale linearly either, so 12volts wouldn't give make the TEC draw half of what it would at 24v, but rather much less because the amperage it draws decreases with the votlage (check this info yourself though, since its ben a while since I actively used TECs).

In general, somewhere between 60-80% is the sweetspot, depending on how good your backbone cooling system is. if your using high-end watercooling with a large radiator your cooling system will be able to keep the TEC warm side cool and you can increase the voltage to get more cooling effect. If on the other hand your using low-end water or even just air, you might actually get decreased cooling effect as you go above say 60-70% since the warm side starts to get hot enough that a significant portion of the heat starts to seep back to the cold-side, thus offsetting the extra cooling you got from increasing the voltage in the first place. Having some way of regulating the voltage is thus a very useful thing to have when working with TECs since it will allow you to dial in to a sweetspot and optimize both your power-useage and cooling effect.

-Stigma

[Glyph]

I see, so regulating the power at the "sweet spot" for a particular setup is a better idea than running straight to Vmax.

I suppose the question now becomes: is regulation worth it for the current variety of TEC setups as opposed to running directly off ATX?

Let's assume we're pitting a straight "15v TEC with a 12v ATX" setup against the same setup with a booster/regulator/controller capable of tuning in a 5v - 15v range. And lets assume the cost of the regulator adds $60.

[Stigma]

Well first of all, you should never run at Vmax as zipdogso explains, because Vmax is the limit at where extra power actually becomes counter productive even if you had perfect heat dissapation (which is impossible to achieve), so the sweet spot will allways be a good bit under Vmax, largely depending on your cooling setup.

As for your second question, it is not possible to answer given the information we have. First of all, I don't know how you personally would put a price on a few degrees difference in temperatures. Secondly it it dependon on the heat output, which we don't even know from what you have told us so far. If you could get a regulator (that can handle that many amps) for such a low price however, then I would atleast consider it, because as I said having the ability to regulate the voltage will let you dial it in to what you need. Otherwise its a good chance that you will overshoot or undershoot by some significant amount. Having a way to regulate the voltage would make it very versatile (and also much more useful for any later projects where the specs of what your cooling may change).

-Stigma

[Glyph]

I think I see where you're going.

Vmax is a limit, not a target. The actual target is lower. (I just wrote down 5v-15v cause i got no idea what that target is, just that its was somewhere there.)

A 12v direct system is basically a coin flip between running higher or lower than the "sweet spot" of a particular setup. What is higher for one might be lower for another.

A regulator would be useful for reaching the target sweet spot if that spot happened to be far away from the original 12v point. Otherwise, not worth it. Problem is, knowing if you're there or not to begin with, need to get the regulator first to find out.

The additional value of the regulator is that any enthuisiast will be upgrading/changing their hardware eventually, probably a few times within the lifetime of the regulator. Having a regulator that can be tuned in for any future system will ensure all variations on the peltier setup will function optimally (atleast with respect to voltage sweet spots).

Am I right so far?
I just want to be aware the pros and cons before i take a go at it.


As for how i'm getting the regulator, i'm building it. (don't worry i know what i'm doing... i think ) It's cheaper than expected cause i'm cheating: I'm not building a full power supply that runs off main's voltage, it actually runs off the ATX-PSU, so i've essentially "transferred" the cost of the main's voltage components into the cost of the ATX (need a hefty ATX, but i got that). I just need to build a device that handles the lower voltage of the ATX output (12v) and the high amperage.

ah well, certainly a challenging project if i decide to go with it.

[Uncle Jimbo]

I see, so regulating the power at the "sweet spot" for a particular setup is a better idea than running straight to Vmax.

I suppose the question now becomes: is regulation worth it for the current variety of TEC setups as opposed to running directly off ATX?

Let's assume we're pitting a straight "15v TEC with a 12v ATX" setup against the same setup with a booster/regulator/controller capable of tuning in a 5v - 15v range. And lets assume the cost of the regulator adds $60.

Parts cost for the PWM controller I designed http://www.xtremesystems.org/forums/...6&postcount=57
is around $20 or so - it can control from about 3V to 24V depending on supply and can be controlled with a thermistor to maintain a specific temperature.

Another advantage is you go to Vmax (or whatever you set as the limit) only when required, most of the time you run at a much more efficient operating point.

[Glyph]

Interesting controller design!

The controller/regulator i want to build runs directly off the ATX but still provides the full range 5v-24v. I want the convenience of not having to run extra power cords into my computer. however to meet those requirements i need high frequency design, thus raising the costs considerably.


Has your controller been built and tested?

To me some of it doesn't make electrical sense, like the D1 diode seems to be in the wrong place, but if it works, it works. Is it supposed to be buck regulator topology?

[Serpentarius]

zipdogso is right. Generally the less voltage you use the MORE efficient it is, in the sense that the ratio of watts consumed to watts pumped though is higher. Of course, to get a decent output to cool anything you can't exactly just run it at 5volts for a 24volt element. if I remember correctly it dosnt scale linearly either, so 12volts wouldn't give make the TEC draw half of what it would at 24v, but rather much less because the amperage it draws decreases with the votlage (check this info yourself though, since its ben a while since I actively used TECs).

In general, somewhere between 60-80% is the sweetspot, depending on how good your backbone cooling system is. if your using high-end watercooling with a large radiator your cooling system will be able to keep the TEC warm side cool and you can increase the voltage to get more cooling effect. If on the other hand your using low-end water or even just air, you might actually get decreased cooling effect as you go above say 60-70% since the warm side starts to get hot enough that a significant portion of the heat starts to seep back to the cold-side, thus offsetting the extra cooling you got from increasing the voltage in the first place. Having some way of regulating the voltage is thus a very useful thing to have when working with TECs since it will allow you to dial in to a sweetspot and optimize both your power-useage and cooling effect.

-Stigma

i'm curious, would it best if purchase 320w TEC and run it at 70% ... bout 224w??
or maybe stacking some? though i dun understand how stacking works yet ...

[zipdogso]

i'm curious, would it best if purchase 320w TEC and run it at 70% ... bout 224w??

No...sorry Serpentarius you have got it a bit wrong.
When we talk about the 70% of Imax ideal operating point that does not equal 70% of the heat pumped.
Look at the picture in the thumbnails the heat pumped is the vertical axis and the Amps are along the bottom. Ignore the different lines for now. Pick a point about 3/4 of the way along the bottom axis (ie 75% Imax). Trace that point up to the top line (red.). Then where they intersect trace that point to the vertical axis. You see ? the point of 75% Imax only drops the maximum heat pumped by about 12 %. If you look at the 3 lines below it at the same point you will see they all drop by roughly the same amount, I reckon the 4 top lines drop between 8% and 12%.
So what are the lines ?
Well they are the difference between the hot and cold side (delta-t) the top one assumes 100% excellent cooling and the difference between the sides is 0C so as soon as heat is pumped it is removed from the hotside, in the real world it can be difficult to achieve. The lines below indicate a difference of 10C each so the blue is 10C, green 20c etc.

Don't read too accurately on this chart it is only for rough estimates but it is good enough for you to see where you are going wrong.

or maybe stacking some? though i dun understand how stacking works yet

Stacking is not a good idea at all....It is really only for when it is necessary to pull the temps down very low, but not much heat pump is required. when you get to lower temps (sub zero) the delta-max (max difference between sides.) drops and if your source of heat is not very hot it can be difficult to get the TEC to pump effectively so you increase the delta-max by stacking the peltiers. But....say you have a TEC with 100w heat pumped and 125w rejected and you put one on top of a another...you don't pump 200w you only pump the capability of the first TEC ie 100w but... the heat rejected is 250W !! You see where this going and why you can't do it to cool a CPU. This problem is highlighted if you study any TEC manufacturer that does multistage TEC's...they only produce multistage TEC's at very small sizes and very small watteges.

[zipdogso]

Interesting controller design!

Has your controller been built and tested?

Well I can't find the thread now but yes someone on this forum has a least one working version of this circuit.
If I can find time over the next 2 weeks, I should count for another one!

[Glyph]

i can see how it works. R3 probably slows the rise time enough of the mosfet that it can snub the inductive kickback from the toroid, not the most efficient way of doing things but its cheap.

As a suggestion to Uncle Jimbo: Diode D1 seems redundant, the internal body diode of the N-mosfet should handle any reverse current at these frequencies. So you can save 30 cents and chuck the diode.

[Uncle Jimbo]

i can see how it works. R3 probably slows the rise time enough of the mosfet that it can snub the inductive kickback from the toroid, not the most efficient way of doing things but its cheap.

As a suggestion to Uncle Jimbo: Diode D1 seems redundant, the internal body diode of the N-mosfet should handle any reverse current at these frequencies. So you can save 30 cents and chuck the diode.

Leuler has built a version - see http://www.xtremesystems.org/forums/...8&postcount=67

I have built 2 prototypes. The design uses PWM to control the voltage so it is pretty efficient. The first bridge (R4, Z2, T1, VR1, and R2) set the control parameters. The control line cannot go higher than .7 of the supply or the controller shuts down, so Z2 maintains the bridge supply voltage.

R1, Z1, and C2 maintain the chip supply voltage, which can't go over 12V. R3 is designed to reduce rise time. It could be left out but the reference design uses it so I did too.

As you say, if your MOSFET has an internal snubber, D1 is not needed, but since the design was set to use any old MOSFET, it seemed like a good precaution. The toroid provides a low pass filter.

Switching frequency is low - around 30Hz in this design. There is no need for high frequency with a TEC.

[Glyph]

From what i see of the circuit, the mosfet is going to be in its linear region for a small but significant portion of the switching cycle. Does it get hot during operation?

As for the diode, rest assured, EVERY N-channel mosfet with 3 pins has an internal diode by construction. the only ones without them are special 4 pin mosfets for signal applications.

As a suggestion for efficiency, it might be useful to connect the diode instead from the drain of the mosfet to the V+ line. With the anode to the drain and the cathode to the V+ line. I know this seems odd but what it actually does is return the extra energy stored in toroid back to the power supply. This doesn't change the operation of the circuit but reduces inductive kickback that might otherwise be lost into the mosfet. this is also protection against someone getting a REALLY BIG inductor that has enough kickback to fry the mosfet, instead that energy is safely returned to the capacitor and the power supply. You also squeeze a few more percent of efficieny without incuring any new costs. The mosfet diode will take care of the mosfet, and the "freewheeling" diode can now be used to help the efficiency.

As for my converter, since i want to also include voltage boost features, i have to put in the high frequency or my inductors and mosfets will fry from the current ripple.

If people would like, i'll post diagrams and pictures of my progress as i make it.

[Uncle Jimbo]

From what i see of the circuit, the mosfet is going to be in its linear region for a small but significant portion of the switching cycle. Does it get hot during operation?

As for the diode, rest assured, EVERY N-channel mosfet with 3 pins has an internal diode by construction. the only ones without them are special 4 pin mosfets for signal applications.

As a suggestion for efficiency, it might be useful to connect the diode instead from the drain of the mosfet to the V+ line. With the anode to the drain and the cathode to the V+ line. I know this seems odd but what it actually does is return the extra energy stored in toroid back to the power supply. This doesn't change the operation of the circuit but reduces inductive kickback that might otherwise be lost into the mosfet. this is also protection against someone getting a REALLY BIG inductor that has enough kickback to fry the mosfet, instead that energy is safely returned to the capacitor and the power supply. You also squeeze a few more percent of efficieny without incuring any new costs. The mosfet diode will take care of the mosfet, and the "freewheeling" diode can now be used to help the efficiency.

As for my converter, since i want to also include voltage boost features, i have to put in the high frequency or my inductors and mosfets will fry from the current ripple.

If people would like, i'll post diagrams and pictures of my progress as i make it.

You know your stuff - I'll try moving the diode and see what difference it makes.

The heat sink I used for the MOSFET was about like Leuler's - just something I salvaged from a blown up PSU, about 3 inches square. Running a 30A load, it never got hot or even particularly warm to the touch.

I would be very interested in seeing your voltage booster design - Getting a controllable high current voltage booser would be a handy thing.

[Glyph]

Here is the quick diagram of your modified circuit.

[repsol_23]

Leuler has built a version - see http://www.xtremesystems.org/forums/...8&postcount=67

I have built 2 prototypes. The design uses PWM to control the voltage so it is pretty efficient. The first bridge (R4, Z2, T1, VR1, and R2) set the control parameters. The control line cannot go higher than .7 of the supply or the controller shuts down, so Z2 maintains the bridge supply voltage.

R1, Z1, and C2 maintain the chip supply voltage, which can't go over 12V. R3 is designed to reduce rise time. It could be left out but the reference design uses it so I did too.

As you say, if your MOSFET has an internal snubber, D1 is not needed, but since the design was set to use any old MOSFET, it seemed like a good precaution. The toroid provides a low pass filter.

Switching frequency is low - around 30Hz in this design. There is no need for high frequency with a TEC.

I have a friend of mine willing to build this controller for me and he was wondering what the spec was for the toriod? Thanks in advance.

[zipdogso]

I have a friend of mine willing to build this controller for me and he was wondering what the spec was for the toriod? Thanks in advance.

Uncle jimbo told me it's value is not too important you can use one out of a power supply thats no good anymore, every power supply has at least one if is a double one i.e. it has 4 tail ends you can just bend the two redundant ones up out of the way.

Here in the UK I found a store called Maplins that sells what they call High current toroids rated at 6amps. When I mentioned this to Uncle Jimbo he said it wasa minimal value and they would probably take 10 amps. Well that's what I am using because I could not find a duff power supply, I bought 2 and as electronic components go they were not cheap, equivelent about $5 each that was 2 weeks ago - now I have got a duff power supply LOL !

I hope to build my circuit in the next week or so.

I assume you have noted the movement of the diode as shown in the circuit above !

[repsol_23]

Uncle jimbo told me it's value is not too important you can use one out of a power supply thats no good anymore, every power supply has at least one if is a double one i.e. it has 4 tail ends you can just bend the two redundant ones up out of the way.

Here in the UK I found a store called Maplins that sells what they call High current toroids rated at 6amps. When I mentioned this to Uncle Jimbo he said it wasa minimal value and they would probably take 10 amps. Well that's what I am using because I could not find a duff power supply, I bought 2 and as electronic components go they were not cheap, equivelent about $5 each that was 2 weeks ago - now I have got a duff power supply LOL !

I hope to build my circuit in the next week or so.

I assume you have noted the movement of the diode as shown in the circuit above !

I sure did, but unfortunately my skills are not up to par for taking on this kind of project. Fortunately a good friend of mine just got back from an internship yesterday and he is willing to see what parts he can come up with and put it together for me. Luckily he is an EE major here at school and has access to some of the things in the labs and what he doesn't I will need to order online I guess. He was just wondering about the specs for the toriod. Hopefully this can be put together in the next month or two (not wanting to rush him) so that I can just use one meanwell (500watt 24V) rather than the two I use now.

[leuler]

I've been revising my chiller somewhat and preparing to build a case for
it, so I haven't actually put the controller to use yet. Would it be worth
it if I reconfigured my controller? Or, given the fact that I am not skilled
at soldering, would it make more sense for me to leave it as it is?

[Uncle Jimbo]

I've been revising my chiller somewhat and preparing to build a case for
it, so I haven't actually put the controller to use yet. Would it be worth
it if I reconfigured my controller? Or, given the fact that I am not skilled
at soldering, would it make more sense for me to leave it as it is?

glyph knows his stuff - I agree with his change and recommend you move the diode.