DIY dehumidifier/air conditioner/heat pump water heater
 

I noted that even during the peak of Texas summer, the cold water is still too cold to shower with, especially in the morning. So why not use a heat pump to dehumidify the house (and provide a little cooling) in order to reduce the load on the main A/C units as well as provide "free" hot water during the summer?

My plan is to start out with a 1/2HP (or so) dehumidifier or window A/C, preferably one that uses R22 since I have quite a bit of ES22A. I'm thinking of replumbing the existing condenser as an additional evaporator in order to increase efficiency. But rather than just connect it in series or parallel with the evaporator, why not use an ejector like the A/C system in the Prius uses? ( 2010 Prius first with ejector-cycle air conditioning ) I'm not exactly sure how to design the ejector, can someone point me to some resources? (I'm planning on using a MIG welding tip for the nozzle unless it's going to be the completely wrong size.)

I'll have to add in a heat exchanger and a pump. I found the following two items:
Grundfos UP15-18B5 Open System Circulator Pump, 1/25 HP, Bronze, 115V, 1/2" Sweat Mount - eComfort.com
B3-12A 40 Plate Heat Exchanger 3/4" Male NPT 7.5"x2.9"
Is that pump going to be powerful enough for this application? It's expensive but so far, it's the cheapest good quality unit I have found. The heat exchanger is on the big side but that should just result in better efficiency. I'm going to build my own variable frequency drive for the pump. (The compressor, however, isn't going to be variable frequency since that doesn't add much benefit for my use.)

I'll use variable speed computer fans for the evaporator. (I could possibly use some fancy Deltas with Cindy Wu sensorless FOC, but that would be overkill so I'll most likely just use some Panasonics I have lying around.) I might also use a pair of radiators or heater cores and a fountain pump for precool/reheat in order to enhance dehumidification, but I would first get it working and add those in only if it doesn't work well enough without them.

And instead of using a TXV, I'm thinking of using a solenoid valve. But rather than the conventional solenoid valves where the pressure difference tries to close the valve (which would result in very unstable flow modulation), the pressure difference tries to open the valve like how some transmission valves work. For example, if the worst case pressure difference is going to be 200PSI, the spring would be sized such that it would begin to open at that pressure. By varying the voltage applied to the coil, the pressure difference that opens the valve (and thus the flow rate) can be varied. (I'm very good at electronics and microcontrollers, so don't worry about the control side of things.) The nozzle in the ejector provides some resistance in order to stabilize things.

Full speed ahead on your project! Sounds great.

My only caution would be that brazed plate HXs are really efficient, but given that you want to do an open loop project, a brazed plate HX will be prone to fouling, over time. This is due to the fact that in closed loop situations, there is a limited amount of 'foulants' (did I just coin a new word?) but in an open source the supply of foulants is unlimited.

Tube-in-tube (or similar) would be the correct choice here. Acquario and also Randen have successfully built their own tube-in-tube HXs... check them out.

-AC

I know the ejector has to be very precise in order to work, and you gain about 10% efficiency average and sometimes better by eliminating most of the throttling losses you get from a TXV or cap tube. I thought of getting one out of a junked Prius for my GSHP. Keep us posted on your success!

If you don't need potable water rated pumps, check out the Grundfos Alpha and Alpha2-self adjusting to maintain a given amount of head regardless of changing circumstances, and about $145 for iron and about $250 for stainless. They take from 5 to 45 watts.

Laing/Bell & Gossett make potable and non-potable ECM pumps-they come in several types that aren't PWM controllable but very efficient from 11-60 watts.

All of these are much more efficient than the usual Taco/B&G induction motor pumps. The Grundfos might already do what you want out of the box (unless the fun is in the designing, of course).

I installed OpenFOAM on my PC to try to simulate the ejector, but there sure is quite a learning curve! (I'm definitely going to ask some friends who have studied chemical or mechanical engineering.)

The pump does have to work with potable water, as does the heat exchanger. I just looked up the ECM pumps and they're not that much more expensive than the conventional ones, probably break even or better after considering the cost of an inverter. What I'm not sure about is whether or not they're easily hackable to allow speed control from a microcontroller. Looks like my best bet is to get one of the solar pumps and then emulate a solar panel.

I'm also not sure exactly what variant of ECM technology they use, since from my experience with computer fans, it makes a pretty big difference. My old Nidec ECM fan, once considered the most advanced fan available, is completely blown away by a new Delta. Nidec's original ECM technology is just a 5 level drive (very easy to implement in an 8 bit microcontroller) and makes a very distinctive buzz even at low speeds. Delta's Cindy Wu sensorless FOC and Nidec's new ECM technology (the two work in a very similar manner, both use 32 bit DSPs) drive sine waves to the motor at low speed and trapezoidal waves at high speed, as well as measure the voltages and currents coming back from the motor in order to get the phase just right.

For your information, I got curious and decided to scope a Delta fan with Cindy Wu sensorless FOC. Top trace is the phase to ground voltage, bottom trace is the neutral to ground voltage. (Unfortunately, I wasn't able to take it apart far enough to see the control logic. The real magic is going to be in the code inside the microcontroller anyways...) Modern ECM motors sure are complex! (As for a Delta fan being Y wired, that's just ironic...)
http://i42.tinypic.com/2ijsx9t.png

Here's a Delta fan someone else took apart, but it's much different from the one I have. Also there are the pictures inside a Sanyo fan and a (newer) Nidec. The interesting part is that the Sanyo has a resolver whereas the Deltas and Nidecs don't.
Pictures of A Dead Delta Fan and It's Internal Parts █ NO 56K-TONS OF PICS █ - Overclockers Forums

Laing D4 (aka mpc650) pcb mod
http://www.honigtopf.mynetcologne.de/D5internal.JPG
It looks like the Laing pumps have some sort of DSP drive. In the detailed picture, the VPP, DAT, and CLK testpoints means that it is most likely a 16 bit dsPIC. Not as powerful as the 32 bit C2000 and ARM processors in the Delta and Nidec fans, but still plenty of power to do good motor control. It looks like it should be pretty easy to tap a wire to control it from a microcontroller, but I'm not sure what kind of control range to expect. (Sensorless drives generally have difficulty at very low speeds.)

Assuming 5200BTU/hr of heat dissipation to the water (stock rating of A/C, would expect more after mods), 70F inlet water, and 140F outlet water (dishwashing mode), it translates into about 0.15 gallon/minute if I did my calculations right. The high pressure variable version of the Laing D5 seems to vary down to about 1 gallon/minute so it looks like the real problem might be the pump not being able to throttle low enough. I can easily workaround that by adding a bypass valve. There will be some efficiency loss but it would be tiny compared to the rest of the machine. I could start out with a bigger compressor but it seems wasteful when I would expect a 5200BTU/hr unit to easily keep up with all of my needs for hot water during the summer.

So are you going to build a batch-type water heater device or an on-demand type device? The intended use eludes me at this point. With (only) 5000 Btu of capacity, it seems you would need either a large HX with a long, somewhat restrictive counterflow water path to sufficiently extract enough heat or some kind of storage tank to use as a buffer.

It's going to use the existing water heater as a storage tank. There will be a wireless sensor for checking the water heater temperature. (Think of it as similar to the Airtap, but located distant from the water heater.)

I'll probably go with the higher powered version of the D5 so there'll be plenty of headroom for highly restrictive plumbing. The fact that the D5 series uses something very similar to the Cindy Wu algorithm (or that might be exactly what it uses) means it should be possible to throttle it down quite a lot. (The Cindy Wu algorithm is tuned for the quirks of small fan motors, but a centrifugal pump is very similar to a fan as far as the torque/speed curves go.)

Be sure and post lots of photos, it's very helpful to anyone wanting to do something similar.

-AC

I would definitely buy, beg, scrap or steal a prius evap assembly for this one. Ejector assemblies of any type are a research scientist's dream. They are a hacker's nightmare. Even firemen and sailors don't like them because they are finicky. And those designs are time tested.

How's this project coming? It shows lots of promise.

-AC

Just got the donor window A/C, a $100 Walmart special GE. The compressor is a Rechi, which is actually a pretty good unit. I went with a R410a unit since the compressors are designed for higher operating pressures, which means it won't be overstressed during the 140F "dishwashing" mode.

In it's stock configuration, it uses about 500W on high cool. In high fan, it uses 50W, which means the fan accounts for 10% of the total power usage! (That A/C is just begging for Cindy Wu or some other ECM drive...)

I ended up getting a Topsflo TS5 pump. Интернет-магазин "Ваш Солнечный Дом" - автономные системы энергоснабжения +7-499-7489064, +7-499-7489072 (That's probably not the exact one, just a similar one.) It appears to use some rather sophisticated ECM drive with a DSP. I'll do some testing once i get it.

Still deciding on the heat exchanger. The plate exchangers are surprisingly cost competitive compared to making some from copper tubing. I'll definitely need a filter to keep everything clean, I'll probably go with a whole house filter available from Home Depot.

Denso Nippondenso A C Condenser 477 0575 | eBay
Is this what your looking for?
http://www.ebay.com/itm/Denso-A-C-Ev...eaedd0&vxp=mtr

I'm going to be making the ejector from copper tubing. There will also be a phase separator to allow the gas (boiled off in the secondary evaporator) to flow directly to the compressor, currently looking at parts for making that. (Trickiest part is the float, since ES22a is very light. I'll probably have to use a spring to assist in closing the valve.) The primary expansion valve is most likely going to be a TXV with a PWM driven resistor thermally coupled to the sensing bulb to electronically adjust the superheat. (It's hard to design a linear solenoid for a high pressure difference that's still energy efficient.) The secondary expansion valve only has to deal with a small pressure drop (on the order of 10-20 PSI) and is relatively easy to design.

I'll also have to get an EPA certification (don't legally need it for what I'm doing but the parts stores in my area give a hard time to anyone who doesn't have it...) and in the meantime, I'm working on the mechanicals as well as the control electronics. The wireless sensor turns out to be surprisingly complex if I want it to be robust. (Throw in a whole bunch of ECC, add CRC and other data integrity checks, revert to using internal sensors if remote sensors fail...) I'm sure my friend Tiffany Yep could give me a solution, except she'll come up with some FPGA design that would be massive overkill.

It's a big project but there's a lot of good learning in it.

I'm confused now. I reread the thread and it still didn't help. So I have some questions.

1. Are you planning on this heat exchanger? The link you provided was dead:

B3-12A 40 Plate Universal Beer Wort Chiller [HX1240BWC] | DudaDiesel Biodiesel Supplies

This bphe isn't rated for the pressures you intend to put through it. Design pressure is 145 psi. Way lower than r410a pressures and half r22 pressures. However, it just might work at r12/r134a/butane pressures. Check out surplus city liquidators dot com, they have turbotec turbo-flow coax coil exchangers for dirt cheap.

2. Are you planning on using r410a or an r22 equivalent? I see you bought a r410a unit. Running it at r22 pressures would bring you down to refrigerator-capacity range. This would amplify your pumping and fan energy usage compared to overall capacity. Add in an electronic control system, and the "pwm resistor/bulb heater", and sensors, and valves, and...whatever else. Even if you went with hyper-efficient individual components, they would still most likely add up to more than the original, low-tech unit fan.

3. What refrigerant does the prius use? The ejector-driven plans I've seen use mainly carbon dioxide. An example:

http://www.sciencedirect.com/science...40700710002720

I understand how the inventors have achieved a huge level of efficiency gain using CO2 as a refrigerant. The pressure levels and compression ratios are orders of magnitude higher than hydro(chloro/fluoro)carbon-based systems, so there is a definite advantage by splitting the low-side pressure flows. But with the lower compression ratios associated with r12/r22/r410a systems, there is not as much potential energy to be split in the low side. The potential for energy savings is still there, it's just not so large.

Don't get me wrong, I'm not trying to poke holes in your project. In fact, I find it very amazing and captivating. A lot of this design research is above my level: although I can understand the theory behind the basic design, when valves and controls and such are thrown in I get lost in the details. I hope you can bring your system to life and look forward to witnessing you progress in that direction.

The heat exchanger I plan to use is this:
B3-12A 40 Plate Heat Exchanger with M5-.08 Mounting Studs [HX1240] | DudaDiesel Biodiesel Supplies

AC Hacker used a similar but smaller unit. I'm oversizing since that improves efficiency and part of the exchanger will work as the subcooler. (I initially planned on making a separate subcooler but then I realized how much easier the plumbing would be if I made part of the condenser the subcooler.)

It is going to be running a R22 substitute, but the capacity loss is not as much as you think. A R410a compressor being used with R22 or an equivalent gets 2/3 the capacity, or about 3400BTU/hr in my case. But that assumes a standard Rankine cycle. Oversized heat exchangers (using the original condenser as a second evaporator) gain some capacity back and the ejector gains even more capacity back. I expect similar to original capacity or a little less. An additional benefit is that a R410a compressor is designed to work with higher low side pressures, so the evaporator fans going full blast are not going to overload it. (That is one of the reasons why I decided to go with Cindy Wu drive Deltas instead of the plain BLDC Panasonics I initially looked at.)

You also really underestimate the efficiency of the electronics. A 120mm HHE series Delta fan with Cindy Wu drive uses 8W at full power, and there will be two of them, so that's 16W total or about 1/3 the draw of the original fan. They're some older ones out of an old HP workstation, so not as powerful as the GHE series monster in my PC. But they should be plenty for this application. At minimum speed, it only uses 0.75W! Keep in mind that Cindy Wu sensorless FOC technology was designed with mobile computing as one of the end uses, so it's not surprising just how efficient it is. One thing I have noticed is that even when commanding minimum speed, the fan "kicks" on startup as the DSP applies pulses of current to the motor coils until it sees the back EMF and starts doing FOC. But the compressor (which I'm not planning to inverter drive, at least not yet) would kick far more, so there should be no real increase in startup noise.

The pump is rated at 15W, but I'm pretty sure actual draw would be less than that under typical operating conditions. (It's potted which is great for waterproofing, but I can't nondestructively analyze it apart from measuring the current and sensing the magnetic fields.) The electronics would adjust it to maintain the correct outlet temperature, so many factors (pressure drop, inlet temperature, commanded outlet temperature, etc.) would affect the actual power draw.

The TXV bulb heater is small, less than 1W. If you ever touched a 2W power resistor after it has been running at any significant amount of its rating, you'll understand that it takes very little power to heat up a small object. (I'll probably have to put something between the suction line and bulb to prevent the line from drawing away all the heat yet still allow the valve to regulate properly.)

As for the microcontroller/microprocessor, the fact that they're often used in battery operated products says a lot about how energy efficient they are...

The Prius uses R134a. I didn't get to take a look at one of those evaporator assemblies, but I'm starting to think that the main evaporator might also work as a phase separator.

photos...
 

If you can get a HX that has braze fittings (AKA: 'sweat fittings') on the refrigerant side it will make your life happier.

Don't forget to post the photos as you go, it is extremely useful to other people who may want to try a similar project.

-AC

I have no idea what this means, but it sounds like something sheldon would say and offend the whole room.

I assume this is a new uber-efficient fan for overclockers' super-servers or sumthin. With less than a KW of heat to deal with, two will probably do the job at 4000 rpm. If not, just make a fan sandwich.

I have another question: Which system are you going off of? I see two:

1.easier "ejecs"

2.more complicated "ejecs II"

Preview article of the systems' descriptions

Looks eerily like a 1-compressor workaround for the hallowell low-temp heat pump. Yikes.

It's sort of a mix of those two designs. It's going to be similar to the first one with an extra evaporator between the phase separator and compressor, along with a bypass so gas can go directly to the compressor. The air first passes over that evaporator, then goes on to be cooled by the other evaporator. Two stage cooling with two different pressure zones increases efficiency and enhances dehumidification. It's more "costly" to pull down to lower temperatures, so that design allows the conventional zone to do a lot of the work before the reduced temperature/pressure zone does even more cooling. The reduced temperature/pressure zone then goes to a lower temperature for the same input power, thus improving dehumidification.

Pretty much all sensorless FOC drives do this "kick" on startup, not just Cindy Wu. For those who don't know, a BLDC or ECM motor is just a 3 phase (usually, some very small ones are 2 phase) permanent magnet synchronous motor (PMSM) with an integrated inverter. Traditionally, the inverters used resolvers to detect rotor position, but that has disadvantages in cost, reliability, as well as phase difficulties at high speed. The more modern approach, made increasingly practical for small motors by cheaper DSPs, is to detect position by measuring the voltages and currents the motor is feeding back. The problem is that at startup, a stationary motor doesn't generate any back EMF so the DSP can't tell its position. What is generally done is to apply some low frequency pulses while monitoring for back EMF. Half the time, the pulses start the motor moving backwards for a moment before the DSP detects the position and starts accelerating it in the correct direction. The sudden acceleration is what makes it "kick". (It also does that if the rotor starts moving in the correct direction, but to a much lesser degree.)

Where things get interesting is at high speed. Traditional resolver based drives have issues with "phase angle" due to the combination of the phase delay of the resolver filters and the inductive nature of the motor. While it's possible to compensate for that electronically, the complexity it adds is almost as much as going to sensorless FOC. Sensorless FOC drives advance the timing (almost analogous to the timing of an engine) so that the peak magnetic field is developed while the rotor is at the optimum angle. Cindy Wu drive technology (sensorless FOC for small fan motors) has been available since 2008 or so and is very common in Dell and HP computers. (Interestingly, Intel bundled coolers are still stuck with Sanyo fans and their comparatively primitive resolver based drives, along with numerous complaints from overclockers!)

I believe the upwind evaporator in the ejecs 2 system is what you describe. It looks like the OEM built the separation into the upper tank of the evap unit. It sprays the ejector fog towards a baffle or screen. The liquid then drops thru parallel channels at the far side of the evap to the bottom tank. The liquid then flows into the remaining channels, boils off, and dries any remaining fog on its way out. They also built a liquid receiver into the condensor side of the system, which leads me to believe they had problems with starving the evap at some point.

The article states that one can simply retrofit one of their ejecs 2 evaporator assemblies into a conventional a/c system and start enjoying more capacity and reduced energy without changing other parts of the system. So I guess this can be done, given enough time and effort. Good luck and godspeed.

Please take lots of pics. This will be interesting to say the least.

Got a link to this information?

-AC

The "preview" article from post #17, Section 3.2.1 says things like "perfectly compatible" and "easily modified by replacement" like it's a clip-on tie or sumthin japanese and generic. Soon they will be everywhere....

http://www.atzonline.com/Article/104...g-Systems.html

They do have a lot of low prices on parts, but they don't calculate the shipping until after you have placed the order! Are they yet another "jack up the shipping costs so we can list attractive prices" scam?

For the coils, how much efficiency would I lose if I use them in an orientation other than horizontal? I'm trying to keep all parts self contained in the original A/C case if I can. And how would the BTSSN-09 (for example) compare to the 40 plate in terms of efficiency?

This is not an equal comparison. The bphe is rated for up to 17000 btu at whatever flow rate v.temp, whereas the coax he is rated for half that capacity under normal heat pump operating conditions. The plate he is $100 plus shipping, I'm guessing it would do roughly 3 times more heat transfer under the same head pressure (based roughly on price). Surplus city has the next size up coax exchanger, rated for 1.7 tons (20kbtu) for $43 plus shipping. Compared to this coax exchanger, the bphe would still probably be more efficient. But plate exchangers have more maintenance needs than coax exchangers and efficiency would eventually drop to below the coax exchanger due to fouling. In contrast, coax exchangers take very little maintenance because they don't have little nooks and crannies for crud to settle or grow in.

In my dealings with Surplus city, they have never charged an outrageous amount for shipping. They just don't calculate shipping on the website until your order is completed because they usually ship all the smaller items in one box. Also, they sell many large, heavy items that can only be shipped by freight. If in doubt, just call them. They can supply you with white papers and such for your items as well, you just have to ask for them.

Dremd just ordered one of the 1.7 ton coax exchangers for his bayou cabin cooler project. It was around $80 shipped. He posted his surprise at the unexpected "heavy-dutyness" of the exchanger in his thread.

Yes, the shipping seamed high, until the coax coil arrived, now it seems about right.
I have to wait for the heat to break so I can pull a window unit out of service to start experimenting, can't wait though.
Subscribed:

Take Lots of Photos!
 

I guess in Louisiana, the supply of good used A/C units this time of year is pretty tight. Up here in Western Oregon, you gotta look where you step so you don't trip over one!

dremde, I wish you the best on this Heat Pump Water Heater project. Living in Louisiana, your COP should be quite gratifying, even through the winter months.

As I'm sure you already know, your project parallels THIS ONE. Having a well-documented and well-illustrated project to follow make things much easier.

And I should add, take lots of photos as you go along so we can all learn from your example!

Best,

-AC_Hacker

Any updates on this project Mike?

I have enough done to get it running, but it needs the controls to be usable.

Any pictures to share?

http://i43.tinypic.com/29fp3qu.jpg
The condenser didn't fit in the original case, so I put it in an old 2U rack server case and mounted the A/C on top of it. There are two extra TXVs (dirt cheap surplus ones that were under $1 each), one for the low pressure evaporator and one for the phase separator.
http://i41.tinypic.com/2utteg2.jpg
I added a line filter so the PLC receiver would operate better, along with a surge protector and homemade hard start kit.

So, how is this project running?

Do you have any performance data yet?

-AC

I'm working on the control software. At this point, I can "run" it by manually adjusting the water pump and fan speeds, good enough to prove that it works but not very practical. Can't get any real performance data (apart from that it draws only 300W or so with the fans maxed out and the suction pressure very high for a ES22a/R433b system) without the tank sensors and in any case, I would have to wait until next year to get meaningful results.

There's a dsPIC for the low level control and an OpenWRT (embedded Linux) platform doing the web UI stuff. There's a whole lot to learn. I'm much more familiar with hardware than software.
http://i39.tinypic.com/2i8zrxe.png
The interlink between the processors is a high speed UART link. I originally planned to use SPI, then realized it was already used by the SD card... The in-band framing is a little more difficult to program for, but I have learned a little about how to do it. (Reserve two values - a frame delimiter and an escape code, then XOR conflicting data bytes and denote that with the escape code.) The link can run up to 460kbaud (with a very clean signal according to my scope) so even realtime streaming of the voltage and current measurements (i.e. to detect compressor speed with FFT or to implement a "soft scope" in the web UI) should be easy to handle.

BTW, if you think all that is overkill for HVAC control, it is. I'm using it as a learning experience to get more familiar with software development.

Overkill? Yes, of course it is.

Seems that if you put it together with duct tape and chewing gum and one syllable words, it would already be doing useful work for you... and would only get better, as your refinements were being applied.

At least, that would be my approach.

I learned a tremendous amount when I let my prototype heat pump run off of water from the backyard loop for a couple of months. I learned things that totally came in from left field, things that I didn't know were there to be learned, things like, "rain is a heat event". I never would have guessed that such was the case.


Best,

-AC

It's pretty cold (by Texas standards anyways) at the moment so I have a lot of time to work on the software. Once warmer weather is back, I'll be able to do a proper test and tune all the control parameters.

I'm nerding out over everything Linux_HPWH. You are logged in as root, lots of control there. :D That's a cool interface, you are winning the golden cocoon award for sure.

I rooted every Android device I have owned within a day of receiving them, often running a custom ROM if there's a good one out there. (The Jessica Simpson board runs OpenWRT instead of Android, but both are Linux...) That's what real engineers do...

Ok, I believe this is the machine you spoke of in the other thread. Am I correct?

If so, I still can't see how you rigged the unit to do its job. Did you finish, or is this still a work in progress?

I'm working on the firmware. The dsPIC is running FreeRTOS and does all the low level control. At this point, it is able to run, but the PID loops need tuning. Also, some features still have to be implemented.

BTW, I'm new to programming with RTOS so that's why it's taking a while. Plenty of learning, though!

OK, so is it in place or still on a bench?

I'm curious as to the plumbing, flow, and control scheme you have come up with and how it is working.

I remember discussing the design phase with you last year, but I never got the goods on how you did an ejector without a manufactured ejector. It looks from the pic u posted the unit has more txv in it than usual. Did u end up using a txv or 2 for the ejector nozzle(s)?

You never completely explained your refrigerant circuit, either. A sketch would be nice and would explain almost better than words. The last thing I heard was that it's a prius/acadia hybrid. A pricada, if you will, rhymes with cicada.

Hopefully it lives longer than a cicada.

Did you get inspiration from Rube, or are you working on a dissertation in modern refrigeration architecture? This rig looks very complex and confusing. And that only covers the mechanical side. The fact that you are still (1 YR LATER) tweaking and twiddling the control logic makes me not even want to ask about it. A simple diagram or 2 (like the mini-manuals hidden inside clothes dryers) along with some general installation specs and operation impressions would go a long way.

Here's the refrigerant flow diagram:
http://i57.tinypic.com/vqoevq.png
The ejector is a MIG welding tip brazed into a tee. The extra TXVs are for the reduced pressure zone and vapor return. The ejector creates the reduced pressure zone, making good use of energy that is otherwise wasted. The two stage cooling enhances dehumidification. Since the refrigerant exiting the ejector has a lot of vapor, a phase separator and third TXV are used to allow some vapor to bypass the first stage evaporator, reducing the pressure drop across it.

Ok, so I drew some arrows to try to make sense of this drawing.

http://ecorenovator.org/forum/member...4-mikeunit.png

Did I get the flows right? This looks like an ejecs II system, with a TXV instead of a cap tube driving the downwind evaporator, and TXV 3 added in.

original ejecs 2:
http://www.globaldenso.com/en/newsre...090519-01b.jpg

Mike's rig:
http://ecorenovator.org/forum/member...235-ejecs2.jpg

If so, I have a question: Why does TXV 3 have its sensing bulb tied to the inlet of EVAP 1? I get that it is meant to digest extra gas and send it to the suction line, but what if? Such as what if EVAP 1 frosted up, causing the phase separator to suddenly burp? This would vent fog/burp juice straight into the compressor. Only after it burped sufficiently would the temperature drop enough to close the valve. Then it's too late, the compressor has to deal with the vomit from TXV 3.

The other looming question is: How does this piggyback to the faucet supply? Draw from cold, pump into hot, or vice versa, or with a third "suction" line? And what controls the water pump?