My first Evaporator Design :) 'The Daisy'

[Gray Mole]

Well since the unit that's using the evap is here already, not a lot of point in a new thread for it.

So a few pics of the case...





Not too bad. I don't make cases for a living, or even out of choice, but I've never liked to send a unit without a case. If it's local, not so bad, but my luck with shipping means a case is a really good idea.

So I've insulated already, it's just the wiring left to do and a few little bits of trimming to make the panels fit just right.

I did another set of load tests just to make sure there wasn't a slow leak that didn't show up right away, or a leak that could develop with high pressure or low temp. Seems fine.

So when the case is done, wiring done 'ready to go' I'll get a set of load test pics as well for the differing loads and where the evap stands.

Already liking it though. The thin based chambers combined with the mass of the plate areas around them really work well. The temp down and the temp up are pretty fast all around. I kind of like that.

On LN2 it's great to have a heat buffer, but in the refrigerated heads I like evaps that get to where they're going relative to load more quickly. Sure makes load testing a faster process as well.

Anyways, would take a video to show that part of it, and not really up for that just yet. Don't really know if that's something truly worth doing anymore.

Will update tomorrow anyways. Would be nice to have pretty load test pics on a pretty unit, instead of the fugly half done ones I keep getting the pics of.

Gray

[n00b 0f l337]

Very nicely done case! Any pictures of its framing?
I've been trying and working with just welding up 1/2" steel square tube, but it then always comes down to how to cut the panels and such. End up paying elsewhere to have them cut and that's sad and takes forever.

[Gray Mole]

Pics of the framing's just plain ugly.

All I did was chop and weld extensions into a mach chassis. I got stick before for the 'ghetto' looking work so I'll just post the nicer stuff on this one. It's the same, almost to the inch of what the quad gpu case is.

But in future (and since I have no mach chassis left to chop) I'll likely go all aluminum.

Probably mostly box section, 3/16 with rails on the base should be strong. I'll look at what's there and see where it goes.

I don't 'offer' cases much, but hate sending anything out without something.

I guess I should make up a template though, for the basics. Just wish there were still cheap cases out there. Dimas was great for that in the UK, but not sure of the shipping to here.

Using offcuts and scrounging the best of what I can for a decent price isn't bad for the end cost but it's a pain. Working out ok so far but it's case work, not really fridge work.

But the panels are easy to cut here thankfully. Put a finer ALU blade on the radial arm saw. Getting the cut is just too much here, and takes too much planning ahead. I can knock of a basic case that's not perfect but fine in not too much time I suppose.

If someone said 'Perfect' then I'll do what I can but it just costs more to do, and takes a lot of time to play with 'til it's perfect. At least, when you reweld a cruddy mach frame is it. Start with box, angle and thick plate and it's a lot easier to make the edges perfect but all that metal adds up.

Gray

Oh right, not a lot to show off I guess, but the basic wiring's done and it all works.



I'll wrap it all 'harness-like' and then put split conduit where it's needed. Plug at the back for standard PC cord, switch and display up front. At least the display seems close to calibrated this time. I'll have to double check but it's around where it should be. That's 1 out of 3 or 4 on this batch so far. I got 10 this time, was sick of running low on those, so we'll see if the first were just the anomoly.

Very happy that the mount and insulation hold up fine to longer running without load. No condensation, not even any cold spots to the touch anywhere. Larger insulation and stuff if it's cascade of course but for the regular SS even low load long term shouldn't be an issue.




Gray

[Hondacity]

sweet case, it should be named "razor blade"

[n00b 0f l337]

Youch indeed, didn't even think of that.
Buyer will be careful I'm sure, no need to worry.

Damn my lack of time. I've got a few things I want to do so bad it hurts. I have a good bender that just needs a spot and place.

Gray have you considered just doing two U pans? One for top and sides, one long one for bottom/front/back? Might give you that clean look you want without all that much extra effort

[Gray Mole]

I would have, but the mach chassis was there and I used the side plates for other stuff anyways. Should show you the air compressor mod I did.

So it was half to make a case, half to use it up since otherwise, just sits...

And yeah, a lot of sharp edges on it though I haven't done any final prep work before painting. Won't change a lot though. Rounding the panels can look good. It can also look awful. Square edges just look clean.

If it was staying 'natural' and just getting buffed and clearcoat then for sure, RaZoR would be a cool name for it. But I'm pretty sure he'd like black. didn't ask though. To me it looks like the industrial kitchen fridges and freezers. Or a medi fridge.


Gray

[n00b 0f l337]

It looks good is what it looks like.

You've re-invigorated me to give the miny case a try. Was ordering steel for new work tables before, but I guess now I'll have to add some 1/2" stock to the order.

[Gray Mole]

Evap basics...

Long it's been thought that an evap should be, and stay, very cold. That is should be slow to cool down with load. That it's better if it does.

I got to use a number of the @itor evaps when they were around and learned a lot about how an evap should best work from that.

1. Mass is good, but too much mass in the wrong place is bad for an evap. Any evap.

2. An evap that warms up quickly with load to it's highest point is the best kind.

3. The closer you can get your cpu and evap temps under load, the better the design.

1. Mass...

The worst place you can have it is the base. If your base is thick and heavy, the heat movement slows like crazy. If your base is too thin it'll warp when you make it or won't stay flat when you use it, so a 2mm base may not be such a great idea. I went with 3.5 and would go to 3mm, but would be nervous about thinner, only for that reason.

Ideally, you have a thin base, but some copper (preferably thin enough to also transfer heat quickly) attached to it. The design I came up with has that as much as possible for just a drill press. A drill mill could do better but it's also expensive to get done and take many passes if you want a deep channel and plate. For that, a stepper is easier but your plates are on the column. So all your heat transfer has to go through that column. Attach the plates to the base, you'll transfer that heat faster.

2. Speed of heatup...

If an evap get's hit with a high load and doesn't warm up, then it's not getting the heat. We're not using LN2 here, so temp control by mass isn't relevant. Enough mass and surface area to give heat transfer but not enough to slow heat movement is the goal. It's what Ssilencer/@itor was trying to show us ages ago.

If your evap stays cold while the cpu warms up then the base is too thick or the supporting mass (plates, column etc) is too thick/heavy.

The ideal is to have an evap that gets as much heat into it, and out to the refrigerant. When you remove the load, it gets down to it's ultimate cold temp as quickly as possible. When you apply a load it get's to it's warmest level possible.

ie. You have -50 unloaded. You apply 300w from a cpu, you get to -25c on the evap within 1-2 seconds. The cpu temp is as cold as it can be as soon as it can be. When you remove the load, the evap returns to that -50c within 3-5 seconds, or as quickly as it's mass allows.

3. Temp differential...

A colder evap under load isn't always good, and in fact can indicate your design is inefficient.

Say you have a 300w load from a cpu. The cpu is running at +15c with that load. The system is a -50c unloaded evap. The 300w load brings it to -25c. That gives you a differential of 40 degrees.

Now you have the same cpu. You have a new cooler. It runs at -45c unloaded. 300w brings the evap to -20c. The cpu runs at +10c. The new evap you have holds a 30 degree differential.

So your evap is 5c warmer under load, but it's able to transfer heat faster and keep the cpu colder. It's a warmer unit, but can keep the cpu colder because of the better evap.



------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

So that's my take on the whole thing. And I'm not the first, Ssilencer was really big on the @itor evaps, and for those exact reasons. I'm not going to try to convince anyone his were better or mine are better or whatever. Only that when you use and analyse your evap's ability to do the job, take all that into account.

A cooler that's made well, holds -40 under load, but can't get the cpu better than +5 or +10 under that load, means either the paste sucks, the mount sucks, or more often, the design of the evap...well kinda sucks.

Edit: Steppers are incredibly limited. The design dictates that all of the heat has to go through the central column before it can be transferred into the plates up top. So if the column is too thick or the plates are too thick, it's harder to get the heat out.

@itor got around that limitation by making custom cutters and having super thin plates and a very thin column which was about the best at moving heat through it.

The reason I went to a 'direct plate to base' setup is because it's cheaper to get that kind of result that way. The best stepper with standard cutters has 2mm plates with 4mm channels, since 4mm is the smallest standard cutter. @itor had 2mm or less channels and plates. His column was also very thin. Heat could move very well through it. The evap was really light, very low mass.

Just trying to pass on what he did to get where he wanted to be with it. The design here is how I got to where I have. Other methods can be just as good or better, when keeping the basics in mind.

If you have too much mass in the way of heat, you're often only really using the base and maybe one of the set of plates on the column for 90% of your heat transfer. So the rest is almost wasted or at least irrelevant copper mass. A design that uses as much, or all, of the mass in the evap is going to be the best you can get.

When you measure an evap near the top and see -30 under load, but measuring close to the bottom shows -10, it's a very clear indication that the upper mass of the evap is just wasted space. If you measure and find only a 5 or 10 degree difference, you're starting to get a much more efficient and effective design.

Temp differential within an evap can be a clear indicator, and often less prone to outside variables like paste or mount integrity.

Not attacking anyone's design, just saying that that's the keystone to the unit's performance and one place we can do better.


Gray

[sdumper]

I agree with you on mass..except that less is better to some extent.

[Gray Mole]

Not so much the 'less is better' thing. Just enough to do the job, but not so much that it slows down the flow of heat. Funny enough one of the primary weaknesses I see in the stepper design is the lack of any real mass on the outside of the base. It's all central. In a 990x cpu or anything where the cores are really spread out, it can show a weak result in the outer cores because there's no mass there to support them.

I see idea mass as being spread out to all areas possible within the evap to the limit of ideal mass and design factors.

A really heavy evap could be good ONLY IF if has the ability to transfer heat quickly.

The differential between cpu (or load test block) is the best clear indicator of mass usage and speed of heat movement. More load is always going to create a differential. Getting that as low as possible is the target for any evap.

A lot of evaps are 'top down' with the captube. That's a pretty good way to run it, for the sake of that little bit of pre-evaporation. The central chamber in the Daisy is for that, so it doesn't just spray in the liquid and have it take that small amount of time to actually start boiling.

But the top down evaps are the ones to best show weakness too. The effect of slow heat movement can be seen and measured most easily in them.

A probe at the top of the evap, on the return 1" back, and on the base itself, can show everything you need to know about how well the evap is doing. Then comparing the cpu to evap temp at high loads.

If the evap top is really cold, but the base is really warm, then it's not using much of the upper plates for anything at all.

If it's close to the same temp top and bottom, as well as the refrigerant pressure/temp matching as closely as you can get it, then your evap's heat transfer and efficiency is confirmed.

With too little mass, heat can bypass the evap. With too much, it can't get through it. So finding a mass that's matched to the max. load your system will deal with is the target and it's not 'easy' by any means.

I've gotten a bit close, but I'd like to get to the 30 degree differential that's my current target for around the 350w mark. Playing with mass, plate thickness, plate/evap height, etc. may get there, may not.

But as far as a good result for a really inexpensive cost, this is probably as good as I'm going to come up with.

I hope if I go to a larger width, same height, very slight mass increase but a 7 instead of 6 chamber design but the same drill press method, maybe I can get the differential lower.

But the outer thickness I think is a bit too much and that's a tough one to deal with short of going to a milling setup and that's the cost factor again. So I'll see. I have 7 more blocks ready to make up, and I may see if some really minor location changes will do anything.

Nice thing about the comparitive testing is that it's not the gas or compressor that really make a huge impact, it's simply the difference of the evap and cpu/load block to look at, so a small change in temps isn't going to skew the results. The difference will remain the same (or close) if the temps aren't too far from where this test took place.

[Gray Mole]

Quick add on to this...

Increasing the width of a stepper and locating the mass of the central column can really decrease stepper limitations with spread out/high load cores. Piotres has done with with his and seems to have some good results with it.

@itor's evaps were the best around, but I do think that with current high end cpu's that have a very spread out core mass, they would show similar limitations.

Moving designs forward to account for cpu design changes is an important part of evap design.

Case in point...

I patched up a shipping-damaged unit for someone that had a chilly1 head. The guy had a 980x. It consistently showed a 2-5c warmer temp on the outer cores at idle. That effect was dramatically increased with load.

On older cores where they were located centrally, to where the internal column of a stepper or spiral design's mass would be wouldn't show that weakness.

The 38mm/1.5" evap does cover those outer cores, but the outside edge of a stepper or spiral has little mass to assist heat transfer.

So locating some mass to the outer edge, or increasing the surface of the contact base of the evap, and moving the mass closer to that area, should equalise core temps.

In my setup that's easier with the smaller size, since the outer ring is a heat transfer plate. That's why I wanted a maze/mill design and can settle for a chambered drill press design. Because the plates are part of the base, and because the mass is relatively spread out across the entire area of the evap.

At this point, I'd like to make a unit that will cool a 6 core intel. So I can see if that issue is now countered with the same 1.5" evap. If not, then a wider base is a requirement for universal compatibility.

That, and I'd like to find out if the same issue with heat transfer is there. The unit I repaired was for 350w capacity, 400w was a push. Top down captube, it would cool the upper portion of the evap but the base was much warmer, and core temps were running very high. A 60 to 70 degree differential was there. Awful.

So either an evap limitation, a mount limitation, or a serious underestimation of the cpu wattage when overclocked.

I'm told people are experiencing similar issues with the very high load of 12 thread 6 core overclocked cpu's. That's one of the reasons I came up with this design, to see about getting past that.

The chilly1 head was top down, and this is 'base direct' captube feed. So the 'cold top/warm base' issue shouldn't be there. It'll hold load, or it won't. No getting tricked by partial load holding of a top down setup.

Gray

[Gray Mole]

As thoughts come to me, I'll keep adding to this...

In the olden days, we used tube-in-tube heat exchangers for cascades.

I tried them before taking a break from cascades, and the results were often something like this....

No load, HX output temp stage 1 -50c. Stage 2 condensing temp HX out -45.

200w, HX output temp stage 1 -35c. Stage 2 condensing temp HX out -20c.

So a 15c difference in refrigerant temp to stage 2 gas condensing temp.

Now, I use plate heat exchangers, as do most guys.

No load, HX output -50, stage 2 condensing temp out -48c.

350w, HX output -30, Stage 2 condensing temp out, -28.

So at max load, 150w higher than the old cascade, 8 degree colder actual condensing temp.

A plate exchanger is more efficient at transferring heat. The HX is warmer overall because it's actually getting more heat out of the thing it's cooling.

@itor evaps used thinner plates, thinner columns, low mass. They used the same principle as a plate heat exchanger to do a better job than other evaps at the time.

The thing we fight is that adding more plates doesn't help so much, since it's only the base that's connected to the cpu heatspreader.

So instead of adding more plates to take heat through a very bottlenecked column, I've gone to having plates connected to the base.

There's many ways to get there, but the most important thing to remember is that the target is a better cpu temp, NOT a better evap temp. Getting a colder evap is only useful if it can get the heat out of a cpu.

An evap is a heat exchanger. The only temp that matters in that instance is the thing that's being cooled. Otherwise it's an impressive number and not a lot else.


I think that's about all I can think of to add to this. It's a tough one to get into, since the idea that a warmer, more efficient evap can be better than a colder, less efficient one is something that's not really focussed on. 'Colder is better' is the mantra, just remember that's only true when you talk about the cpu (or anything you're trying to cool)

The whole idea is the best cpu temp you can manage with the best evap heat transfer you can come up with.

This design is my own, but you're welcome to use it to try to make one, whether it's the same or a bit different based on what you think might be better. It's a good starting point, according to my testing so far. Cheap as it get's, and simple to do. So there should be room for improvement.

Gray

[Gray Mole]

Well, a whole bunch of load testing. Not so much about the unit, more about the evap, but pics of shiney painted unit are never a bad thing

So I loaded it up to 200w and let it sit for a half hour or so. Make sure it's all hot and get the most accurate unit temps I can. The evap relationship doesn't change really from a hot system to one with a cool compressor though.

Worked up to 350w, then back down in 50w increments to see the differential at loads. Like I'd said, not really something that anyone can reproduce, nor is it an accurate reflection of cpu temps. Unless someone had the same block with similar heat cartridge with internal K sensor, it's unlikely to get the same thing.

But anyone wanting to can set up thier own method. A sensor drilled into a load block and testing of multiple evaps will still give a comparison. Once you test a variety you can see where one evap may be better than another, and look for reasons why. One evap may be better at lower load, one higher. Hard to judge accurately until you've tested a few on one load/temp test station.

I would guess, since the block I use isn't huge but it's far more massive than a cpu, and same with the cartridge, that the cpu temp would be closer on average in temp to the evap. This is simply my way of analysing evap efficiency and performance. The load on block can always test an evap temp, but it's more specific to test the evap to block temp to see how it's performance is, regardless of what unit it's on, or what temps it's holding.

That said, the pics...




No load, coldest it gets, just a pretty picture




Just after the testing, just pics of it done. After the load testing, takes a minute to get back to -50 as they do. -45c is it's immediate temp, get's there moments after the load's taken off. I'd call that it's natural unloaded temp. The -50c thing, and most units, the cold 'money shot' isn't really to give an idea of what it'll really do in use, but it's nice to look at



So first, unloaded



200w, -31c, 25 degree differential



Just a closer look, hard to get pics with the display so far from the meter



300w, -26c, 31 degree differential



350w (just over) -25c, 37 degree differential



250w, -27c, 31 degree differential. May have been an anomoly, didn't expect it the same as the 300w test. Could be right though. Didn't give it a huge time to settle on that one. Could be worse or better but only by a degree.



150w, -33, 19 degree differential



50w, -40c, 5c differential



And where it settled just barely afterwards. Within 10 seconds the temps balanced (block and evap) and I like that. Mass enough for good transfer and little enough to get the evap to be as cold as it can be at a given load fast.

As shown, the display's around .5 to 1 degree warm this time. no big deal but taking that into account, it's where my meter's reading were before in the shorter testing, or close.

So basically I have a 'first round' to try to beat if I can with the next one. I'll make minor changes as I go, but it's likely I'll stick to this basic and cheap design until I get another set of blocks that are larger.

When/if I move then I may set up more involved evaps, but honestly I'm liking the results so far

I'm reading a marginally higher low load result, but the evap is really small/short and I've read the side closer to the top on most to compare. Now that it's shorter and much lighter, I'm reading closer to the base temp than before.

So I do hope this one sees a cpu soon. I would like to see the real cpu temp result, since the comparisons are good but the cpu temp is the best way to represent if it's a step forward or not.

Funny how even a small 50w load has the impact it does. But I've seen a cpu that's a 4-6 core pull the cpu temp and the base evap temp, so this is getting me closer to a better idea of the effect of even minor load.

I have another week with it, as it turns out. I might (if I can get to it, other things to get done) try to locate a probe on the back of the evap, and the neck. I was going to leave that since I may have been sending it off tomorrow. But if it'll stay, could be worth getting that data for reference. Knowing the actual refrigerant return, evap back temp, and evap side closer to the base, could give a bit more info on what's going on and more to compare for later revisions.


Gray


Gray

[nigelke]

I actually like this one

My knowledge of an evap is not that big sow I don't really understand everything but it looks great

[piotres]

Nice idea Gray . And that's sth new and different which is most important .

But it need to be CNC made to look good .

[Gray Mole]

hehe

You're right Peter it's not pretty but it's also so cheap to make that it's hard to justify.

I guess when I factor in the time, it's not a lot different in cost to having it machined.

As I go, and set up specific saws, grinders and the like to make the cuts and parts more 'machined looking' it might be more impressive (or at least, less fugly) it'll help.

But I guess most of the uglier stuff's inside, and doesn't really affect how it works.

The worst part is cleaning up. When you have it made up at a machine shop it's really clean. When you grind away at things to make it fit, you get a lot of shavings and loose bits, that's the harder part. If even one bit gets into the return it can be trouble.

So I slowly chip everything out, make sure there's nothing to break loose.

But thank you it's not totally new, parts of the design have been in other things before me. Dimas' old evap had the plate/chamber setup same way with the captube going into a plate you had to wedge in and preferably braze.

I've seen drill press evaps with similar kinds of ideas.

I AM hoping soon to make another cascade and try it. Now I have an idea of the difference evap to heat cartridge on SS temps, I'm curious where it'll be with the cascade temps. I know it's always worse, but curious how much.

I think that it's the combination of the @itor and Unseen's evaps that made me go in this direction the most. @itor's being super light and everything thin. Unseen's maze evap that was short and the maze was the plate/column heat transfer.

I didn't really 'steal' much of any of it, just that it kind of made sense to do it that way. I suppose there's only so many ways to do it with this design.

It's not as ugly as most of the DIY evaps I've seen but now that I have the layout, and I know it doesn't suck I can start setting up what's needed to make it easier (and prettier) to cut it up.

[sdumper]

Regardless of the fugly look its great to see something new tried and its almost impossible to make it look professional without a mill to assist.

[Gray Mole]

Thanks

Wanted to say in case there was any doubt...

This, as well as the other stuff I keep posting about designs or tuning or whatever, it's all fine to use.

I mean, I like to think of it as 'my design' since it's cool to have your own design I guess, but I like doing the research, testing, etc.

So if anyone wanted to use it, or use the basics with their own take on it, that's fine. Can call it a 'freevap' if you want. I like to think of it as more of a stepping stone for now, til something better comes along.

Hope it's enough info on what i did to be able to replicate it if anyone wanted to.

Can't really say it's better than anyone else's evap. The numbers are pretty good though, and it's just really easy to make. So people are welcome to use it if they like


Gray

[teyber]

Nice job dude

When we going to whip up a full package

[Gray Mole]

I was afraid someone was going to ask that

I'm not really big on selling evaps. There's still people that do, and that's more their thing than mine.

If I had to I could start putting together some if someone really wanted one, and couldn't make it. Honestly I'd rather people 'stole' my design and made it than start offering evaps for sale.

It boils down to what you want from it. I like to make and test the coolers, and the parts that go into them. I can't keep them all and it's silly to rip apart a nice cooling when you're done, so I sell them.

Custom stuff I avoid when possible because of the pressure involved, and as always the chance that it's not exactly what they wanted or thought it would be. When you make it and sell it, it's what it is.

So to make and sell evaps isn't what I want out of it. More 'If I have to, I will' but I don't really do the phase change as a money thing. Making stuff specifically to sell isn't my bag. I have a job, I don't want another one

But I have a bunch of raw blocks and when I've cleared up the work I'm finishing off, I'll do a 'run' of them, and post some pics.

I've set up a couple more things to make them more refined looking, but this run won't involve having specific jigs for the crosscutting and stuff.

I just want to have evaps. Running low and running out just sucks, can't play around if I don't have them. Getting into the 'full package' pics and stuff, you end up with PM's about buying them.

If it got busy, I'd be stuck making evaps all the time, and not being able to focus on playing with the coolers.

But when I get the 'run' done I'll make up all the plates and mounts for all the blocks I have here. Should be pretty enough I guess.

I won't say 'NOT FOR SALE' since I think that sends the wrong message. I'm not proprietary, don't want to make people think I mean 'You can't have one unless you get a whole cooler' or anything, that just sounds like I'm trying to push coolers.

But I'd really rather not be an evap supplier if I can avoid it That's a bit more like work and less like fun.

If I do find a good machine shop or my friend actually comes through to make it all up for me for a decent price, that could change I guess since I wouldn't have to commit the time to it.

But that would mean the cost would go up a lot. Since my hobby time is sort of 'free' I don't have to account for it like you do at a job or business.

Still as long winded as ever. I could have just said 'soon' huh?

My postcount is only just over 2000 but if you split them into paragraphs, you'd probably end up with 100k

[mytekcontrols]

If I had to I could start putting together some if someone really wanted one, and couldn't make it. Honestly I'd rather people 'stole' my design and made it than start offering evaps for sale.

Well Gray I think I might just do that, or at least try something like it on my CryoBUG.

I like the design a lot. It's simple and it looks like it would have good heat exchange with the head (or should I say face). But first I need to stop tweaking with my design and get it nailed down. So at a later date I will gladly steal your design.

[n00b 0f l337]

Mytek, considering what you've given to the community so far, when that day comes tell me where to ship one to you.

[sdumper]

Nice job dude

When we going to whip up a full package

The bigger question is when are you going to get your business going again Teyber
You were making some kick butt products before you shipped off to school

[mytekcontrols]

Mytek, considering what you've given to the community so far, when that day comes tell me where to ship one to you.

Thanks Nol, but the version I have in mind would be quite a bit different in the end. Remember I'm not looking to cool CPUs. So I would incorporate the basic internal design, but have an entirely different mounting scheme, as well as vacuum insulated thermal isolation for mounting up with a KF-50 flange.

[n00b 0f l337]

We'll see what I can do about that
And hey Teyber, when you home from school?