I didn't go into any more detail than Melcor did. And there is a reason they did it. As they make the things, I assume the intent is to reduce trial and error, and that their methods are good. If my post seems to have too much to say, tell me if you think that's still true after reading the design guide.
I think the main thing to remember is that it is not 1/4 of Imax, but 0.4 of Imax that is the optimum current for efficiency. Also, if you can get TEC's cheaply enough, aim for big Qmax in a small size, and power low. But not too low. 1/4 probably is too low, it starts getting inefficient again if you go that low.
I don't know much about TECs, how low will they operate? For instance, if I had an endless supply of dry ice, could I use that to cool the hot side and acheive even colder temps on the cold side?
If you had a load of dry ice the Carnot efficiency will be WAY better if you just cool the hot stuff directly with the dry ice. (Carnot efficiency being an ideal-by-laws-of-physics limitation. Compressor based fridges are way better than TEC's for that, and some have used CO2. Still do I think, in stuff like refrigerated trucks.)
You might be able to get colder temperatures that way, but CO2 will go to about -40 I think, and a TEC won't work so well that cold, so it will work for a sensor and excellent insulation, not so well for a hot-running set of diodes. For bigger loads, better if using CO2 to do it directly.
You could do that and the cold side temperature can be lowered even further than dry ice temperatures, but as the temperature drops the efficiency will drop and so you can not do this forever. Yet, with extreme TEC stacks with 6 stages , delta T's of over 135 C is possible (for very low heat loads).
Doctor,
The reason I recommend 1/4 Q max is because the COP (coefficient of performance) is a broad curve and if you undersize the device you will run out of capacity because the efficiency drops ever more quickly as you move above optimal. I am being conservative. I don't disagree with Melcor's guidelines. I also am not arguing with you. I think JR will enjoy experimenting and there are going to be many variables (like heat sink flatness and clamping pressure) that may turn out to be more significant than our discussion of the fine points of module sizing.
Ok. Hatchet buried. Wasn't waving it about exactly, I was just concerned that if people were encouraged to ignore my posts it might result in stuff getting asked when the answer was already there. Like that bit about sealed enclosure and cold trap. The question about silicon gel might never have got asked if my post had been read first...
Anyway, sure, I experiment too, but only if I have the stuff to try. Even now TEC's are expensive.. Cheapest way to brute-cool is to use a choice from the standard range of 127-couple TEC's all over eBay, for which the voltage relation is constant. Other reason for constant voltage is that 5VDC at high current is also cheap and easy these days. So Pmax becomes directly proportional to Imax in this context. All the rest relates to that. Figuring it this way makes it a lot easier.
Full control of set temperatures changes everything, but in this case I doubt that's what's wanted.
I did try to find a graph of V and I for a TEC but Googling that is surprisingly tough! They all seem to graph one or other with DeltaT, or Q, or COP, anything but the two together. I was interested in this because dropping one drops the other, BUT it's a passing interest because the cheap brute way to go is 5V at high enough current, and a 127 couple TEC chosen for Qmax 2.5 times whatever the amount to be pumped will be.
If calculated like that the leeway is fairly wide, it's a much less critical choice to make, so easy to satisfy with cheap stuff, and the efficiency will still be excellent, as Peltiers go. I'm not sure that picking a smaller high density one is a problem because it's running way below Pmax. It's actually harder to get adequately flat surfaces for bigger TEC's too, so a big one might not always help with effective thermal coupling. Big stuff tends to run into significant differences in thermal expansion too.
EDIT:
A bit of Googling tonight suggested that Laird Technologies now own what was Melcor. Their PDF guide contains soem text and diagrams from the original Melcor guide. The graph of Q to P is wider, easier to gauge relations from at lower power. The selection graphs are missing though! Instead there is a picture of a bit of software... Planters, do you have it? If I can evade their long form filling exercise to get a copy, I will.
Planters, I think we both got something wrong. I found a graph of V with I in a datasheet for a small Marlow TEC.. It shows a Vmax of 4 with Imax of 2, so Pmax of 8W. While not linear, the curves are close enough to it. Going with Melcor's 0.4 of Pmax, that's 3.2W, so looking for a convenient pair of values to get it finds 2.5V and 1.3A (3.25W). Now 1.3/2 is WAY over a quarter, it's way above 0.4 too! It's 0.65 of Imax (and roughly of Vmax too) when it's 0.4 of Pmax.
My error is in talking about 5V for the standard 127 couple TEC's to optimise efficiency. They have Vmax of 15.4V, so 5/15.4=0.32, way too low. I was doing ok earlier when I said about 8V, and even that's a tiny bit low. The idea of 5V operation for cheap easy power supply might still be great for TEC's that use Vmax=8.6V, all having 71 couples, but I doubt those are as cheap and widely available so it defeats the point. What MIGHT work is the standard 127 couple types running on 10V, IF it is easy to run two 5V supplies in series, but I bet the cheap ones are all ground referenced, not isolated outputs.
There may be an ideal graph of COP to find the optimal current but as COP of the system depends on all the other thermal coupling, I doubt it's worth trying that hard. The simple graph in all the guides I have seen might be enough. It suggests we can go lower than 0.4 of Pmax, but not a lot because after about 0.3something the Q drops away so hard it vanishes pretty much before I (as in, not me) does. But I want to vanish now too.
No, I think I'll stay with my recommendation. Take a look at some data from Tectech. I do not routinely source my coolers from them, but I have used their modules and they have worked as specified.
Ok, I have the data sheet, but here's my interpretation of it: (Going with the graphs for hot side temperature of 30°C.)
They have a V with COP graph, bottom right, and we need about 31°C difference with hot side about 30°C, pumping at least 36W (cold side will fall where it will, but will be about zero in this case if losses are low). The closest curve they have for that is the light brown one, DT=29.5°C, with V at a wide peak between 6 to 8V, (so I was close enough, I put it at 7.6 in my first reckoning), and 7/Vmax (15.7 for this TEC) is 0.45. As the relation of V and I is close to linear under load, squaring 0.45 is good enough, so 0.2 of Pmax.
I can see your point about running at a quarter of Qmax from the top left graph where the blue 6V line at 15W or so hits COPmax, about a quarter of Qmax (weirdly vindicating my 5V idea), but look back at the COP graph I started with, bottom right. At 6V or less the COP curve drops off quickly. By increasing to 7 or 8V you gain several times in stability what you lose in COP. And you won't lose much COP (efficiency) either. I think this is why Melcor are recommending 0.4 of Pmax, which is around 0.6recurring, 2/3, of Qmax!
Now I know it's a matter of choice, but I think that uprating as far as you do is risky, and expensive. I wouldn't do it. I'd uprate to no more than 2.5 to 3 times Qmax in a chase for better performance because any further the voltage falls too much for the required pumping and COP drops sharply.
The place for experiment is where you put it in post 17: measuring the temperature of the diode mount, and the base plate. Test the thermal coupling empirically, before selecting a TEC. Once you have that as good as possible, TEC selection can be designed, rather than buying more types than needed just to do trial and error. I know that tolerances are low, and empirical tests count, but TEC's are still expensive enough for most of us to have a need to understand the design and selection, just to cut those costs, and the experimentation time. Also, if empirical tests show an error, the calculations are our best shot at knowing the magnitude of the error, or even just whether there is one or not.
This interests me enough for more detail, and another post...
Planters, any chance your quarter-Qmax is a memory slip from quarter-Pmax? If so, it could work. (I also found a possible validation of your quarter-Qmax too, see last paragraph...) I'm assuming your observations are good, and my methods are good, so there should be a way to get convergence. I've found two.
First is that the optimal voltage for a 127 couple TEC is about 7V, backed up by data from your datasheet, my Googled datasheet, and Melcor's and Laird Tech's guides. If I put each data set through my methods I get that same 7V each time. 7/Vmax of 15.4 is a ratio of 0.45. Square that to get the ratio of Pmax, and it's 0.2, close to your quarter. So if you replace your Qmax with Pmax for this ratio, it works.
The second convergence error I found is based on the data sheet's rating of Qmax itself. That sheet you linked to, it states 76W, but that's only true for hot side of 70°C, Qmax falls to about 64W for hot side 30°. It's about 15% error.
The TEC choice in that datasheet you linked to is a good choice for cheapness, it's what most people would do as it meets the example's need, gets to about 0°C while pumping 36W or so with hot side 30°C on little more than 12V. Now, if we aim for efficiency using 7V we see that the COP is best at 0.25 of Qmax, but ONLY at DT around 30 to 40 °C, and this TEC only pumps about 16W (because Qmax for this 30°C hotside temp is about 64W). We need to uprate this by about 2.5 because then we can pump 40W and reach 0°C with same 7V for best COP, same DT, same hotside temp, enough for the 36W with a few to spare in case losses are more than hoped for. If you add the 15% error from the way Qmax is headline-specified in the datasheet, you get 2.875, closer to the 3-fold uprating limit I mentioned before, which would apply when looking at basic specs on web pages, etc, assuming they were based on datasheet headline specs at all.
Note one particular point... Once you take that ability to pump 40W, having uprated the 64W Qmax TEC 2.5 times to 160W, there actually IS a quarter Qmax that matches what you said, which makes what you said correct, in the context of Q/Qmax ratios in the NEW TEC. There is also an uprating from the OLD TEC of 2.5, not 4, from the smallest spec TEC capable of doing the job at the nominal 12V a 15.4Vmax TEC is intended for (the one most manufacturers would pick), to the one that can do it most efficiently with the same number of couples. So I am also right because the cheap ranges of TEC's all over the web range in jumps of Qmax and all have 127 couples to run on nominal 12V. There is also the roughly quarter-Pmax to watch out for, the one resulting from Vopt/Vmax for any TEC. I think we just have to be careful how we look at the same situation.
Reply With Quote
Originally Posted by steve-o
(Carnot efficiency being an ideal-by-laws-of-physics limitation. Compressor based fridges are way better than TEC's for that, and some have used CO2. Still do I think, in stuff like refrigerated trucks.)
