All DDC-2/MCP355 (and even DDC 3.2) Pump Owners Please see this !!!!! (Now Updated !)
Updated on July 31st, 2007 - Please see 2/3 of the way down the post.
DDC-2/ MCP 355 Failure Problems
(or The Strange Case of the Stuttering Pump)
(Caveat: Sorry, the following is a little bit technical but may give you a good idea of what is happening with your Liang/Swiftech pump.)
Okay, many of you experienced watercoolers that have been to the forums before or lurk here, know that the Laing/Swiftech MCP 655 pump has a good reputation for reliability and low failure rates. Same for the DDC-1 (also known as the MCP 350). The problem has been the more powerful DDC-2 (also known as the DDC+ or the MCP 355) which puts out 19.5 feet of head, and is no longer being made (the DDC 3.2 has replaced it). I purchased one of these DDC-2 pumps from Petra's Tech Shop (great place BTW !) with their top back in May to use for a series of Showdown articles that I have been releasing here at XS. Surprisingly the pump worked quite well for me for the first one month of testing, even though it was not in continuous daily use.
After about five weeks, I decided that I would put this pump in one of the three computers in my little lab room and everything went fine for about four days until I noticed the E6700 CPU heating up to 83 deg C one morning. Doing a few checks revealed that the DDC-2 was now not running or not running correctly. Well, after looking over all the simple fixes like jimming the molex connectors and making sure the PSU rail was fine, I quickly drained the loop, removed the DDC-2 and replaced it with a MCP655, so that I could take the smaller pump out for inspection.
Two weeks had passed before I had a chance to look at everything carefully and here is what I have now found:
1 – If you take the regular top that comes with the DDC-2 (or Petra's top) and turn it over, you will see the large circular well which is about 8.02 mm deep. (this figure has been corrected from the original 7.75 mm so remember that when viewing Photo #2 below) (sorry, I do not have a micro-caliper at home) If you now look at the naked pump (with the top removed), you will see the impeller head sticks above its base by ~ 7.85 mm. (corrected from the earlier 7.60 mm estimate, so remember that when viewing photo #3 below) This leaves only ~ 0.10 to 0.15 mm difference between the top of the impeller head and the bottom of the screw-down top (or Petra's top, which has the same internal height). Normally, as the impeller spins this 0.15 mm will be filled with water (which at this location almost acts a lubricant). This stack height between the impeller and the 'roof' of the screw-down top is critically important; if it is too large, too much water will hold there, causing cavitation and lost head pressure/flow, if it is too tight the impeller risks bumping into the roof and stopping.
2 – The 'motor' of this pump (and the DDC-1 and MCP 655) is an electromagnet which simply means that it is a looped coil of wire inside the housing that goes around the stator base. When current is applied, the impeller will follow the consequences of the famous 'right-handed rule' of electromagnetics and spin. If it is spinning correctly, the thermal heat generated from the impeller's ride on the ceramic bearing and the electromagnet will pass into the cool water flowing over the impeller. If the impeller is not spinning, the heat will be trapped in the pump and build up (quite quickly, I might add). How accurately, the wire is coiled in this electromagnet will determine if there are any 'gaps' in the electromagnetic field once current is applied whose importantance will be evident in a minute.
3 – As the impeller starts to spin, once current is induced, it will pick up rotational acceleration and physically rise slightly (critically, this rise is determined by how strong the electromagnet in the base is and how much current is passing thru that electromagnet and is described in physics by an ancillary to the Maxwell equations (this takes the form of a triple integral that you would see in multivariable calculus). This rise for these small pumps (and the MCP 655) has as it's largest component, it's angular momentum, which takes place both in the XY planes and the Z axis (along the Z axis is how much the impeller wobbles as it spins around). (A tiny component of this rise is also determined by the water under pressure flooding the ceramic bearing as well) If this 'wobble' is too great, the impeller will careen into the screw-down top (just like a child's spinning top). (Thanks to C-Dale for taking a look at this and bringing everyone's attention to this aspect) Therefore, the distance between the top of the impeller and the bottom of the screw-down tops is critically important. Also, if there are gaps in the electromagnetic field generated by the stator, this 'wobble' will be magnified by some small factor. Again, if there are gaps in the electromagnetic field generated by the stator, and MORE amperage is sent thru the electromagnet especially during startup, then this will further increase the 'wobble' and make it even more likely that the impeller will careen in the bottom of the screw-down top. These effects are most critical during the initial start up phase of the impeller. After start up, angular momentum will be conserved (just as in a gyroscope for example) and the impeller's 'wobble' will be reduced.
4 - The amount of amperage and its time delay onset are determined by the logic pattern laid down on the printed circuit board (PCB) that controls the electromagnet. If the amperage is allowed to start at some minimal value (enough to start the impeller in a multi-block waterloop for example) and then rises to its max, the impeller will start off smoothly without too much 'wobble'. If the amperage is raised too quickly or started at too high a level, then the impeller will wobble too much at the very start and careen into the bottom of the screw-down top.
5 – When Liang put out the stronger version of the DDC-1, which was called the DDC-2 (with the orange impeller), they used a different electromagnet design (according to Petra) and raised the amperage from 0.75 to 1.5 amps. In this design, I believe they failed to take into account that this would now cause the impeller to rise even higher from the base and possibly touch the bottom of the supplied top or Petra's top.
6 – Testing the PCB on the back of the pump with an oscilloscope reveals no problems. The PCB is not the culprit, in so far that there are actual current problems. The logic printed on the PCB is another matter, since I do not have access to Laing's engineering documents.
7 – It takes very little friction (or downward pressue) to keep the impeller from spinning when it first starts up. (Roughly, I gauge this as about a dead weight of 250 grams (notice that grams is not a measurement unit for friction or pressure) – see the experiment in Photo #4 below where I have 170 gram pliers on the impeller to slow down its startup)
8 – This is likely causing the impeller to wobble off and around the ceramic bearing and touch the bottom of the screw down top (which also means you now how a quandary, because if you screw down the top very tight, you lessen this ~0.15 mm headroom and if you leave it looser, you risk a leak). When the top of the impeller touches the bottom of the screw on top, there is now enough friction to prevent the impeller from picking up speed and spinning properly, or to spin more slowly (which is causing the 'sand' sound that many of you are hearing). Many of you will also be able to see this as 'stuttering' (which is the impeller trying to start up then hitting the ceiling, then bouncing back (Newton's third law), then repeating this process) if you look into the pump while you run it dry (just connect the pump to your PSU, but only for 1 ~ 2 seconds, any more and you risk damage to the ceramic bearing and PCB). This is also the reason why Gabe or Alex (Petra) are getting these pumps back and disassembling them only to find out that they work 'correctly' in ninety percent of the cases. This is a consequence of the impeller rising and being stopped before it starts its spin up, and with no top, there is nothing to slow it down. Or, Gabe or Alex then screw on the top again, only though, maybe just a little bit less tight than you had it on there in your home, and the pump seems fine when it starts it up dry. That is why this problem has been so hard to diagnose.
9 – This stoppage of the impeller is mitigated a tiny bit by the orientation that you have the pump in.
10 – Depending upon the viscosity of the coolant you are using, how tight you screwed on the top, what PCB logic design you have, and several other factors, these impeller stoppages could start as soon as you buy the pump or take place months later (or perhaps never occur). That is the ghost in the machine and the reason why it has been so difficult to pinpoint the troubles with this specific pump (the DDC-2). I also increased the viscosity of the water pouring thru the pump with cooking oil (along a known viscosity curve to mimic those using this pump with zerex, hydrix, etc) and found it to be very, very sensitive and unpredictable to viscosity differences.
11 – I suspect, but am not sure (so please do not do this, until we hear back from someone in the know at Laing or Swiftech) that many of these problems would disappear if we were to grind off another 0.25 mm from the circular depression that is on the the bottom of the screw-on tops (or add a gasket to the top of the stator as Jedda suggested below). This would reduce head pressure and flow slightly but it would be an easy fix that those at home could do.
12 – The reason why the failure rate for the DDC-1 pumps is so much smaller is that less amperage is being pumped thru the electromagnet (and a possibly better electromagnet design) (and therefore the spacing of the wires around the electromagnet is also less critical), which causes the impeller to wobble off its base less, (and is within the 0.15 mm safety range (the tops for the DDC-1, DDC-2 and DDC 3.2 are exactly the same). The DDC-1 impeller also spins slower and therefore, if it does occasionally touch the bottom of the screw on top, there is less friction, and less chance that the impeller will be stopped. Additionally, the PCB circuit design for the DDC-1 may allow for a slower ramp up of the amperage (to 0.75 amps) which would mean less wobble at start up before angular momentum is conserved.
How sure do I feel that this is the correct answer to describe most of the problems that you are having or will be having; fairly sure. Although, I am an engineer in the real world, I have only spent about three hours on this problem (in addition to the time I spent setting up the viscosity curve). I do not work for Laing's QC unit and I do not have access to their engineering documents, but I feel the above statements are likely true, although I cannot be 100% sure. I would welcome feedback from any other engineers or technicians who have experience with this pump and can test my results more accurately. There have also been some cases where it looks to me like the heat buildup from the impeller being blocked has led the PCB to be damaged but this is a secondary effect. The critical question is: If the above is true, will the new DDC 3.2 pumps (which are the redesigned slightly less powerful version of the DDC-2) be affected? I cannot answer that, but alot will depend upon how the PCB has been redesigned for the DDC 3.2, how exactly the wire was looped for the DDC 3.2 electromagnet (allowing for no 'gaps'), how much clearance Liang left in the screw down tops (which appears to be the same ~ 8 mm !) and how much amperage they pass thru the electromagnets in those pumps (which I think is 1.5 amps). It is entirely possible that Liang, just looped a bit less wire in the electromagnet for the DDC 3.2 without realizing it may be a on/off friction issue with the impeller head caused by too much amperage or to quickly ramping up the amperage thru the PCB. I have notified Gabe and Petra of this, and I hope this helps out anyone who may own these pumps.
Please keep reading to the Test Section:
Testing Section
Test 1 - Placing a dyed fatty acid on the top of the impeller, shows that the impeller is indeed making contact with the polyoxymethylene (Delrin, the material used in the top) on startup.
Test 2 - The contact between the impeller and the stock top or Petra's top can be mitigated somewhat by how tightly the screws are tightened down. Tightening the screw down on Petra's top till they almost stopped, resulted in about a 90% chance that the impeller would fail at startup. Backing those same screws off by 3 full revolutions decreased this failure rate to about 45%.
Test 3 - Some investigation with the oscilloscope reveals that the amperage is not being being ramped up correctly at impeller start up for the DDC-2 pumps. This is causing the impeller to wobble around more than it should at startup.
Test 4 - Without taking apart the stator and voiding my warranty (which I need, so I can return this pump), I cannot comment on the wiring pattern for the electromagnet.
Test 5 - Likewise, after careful inspection, I cannot comment on the manufacturing of the impeller itself. Many of the problems that we are seeing, would be greatly amplified if the moment of the impeller were off balance (this is most crucial at startup).
Test 6 - The DDC 3.2 pump did exhibit some of these same problems although to a lesser degree (at least for the model I have). The failure rates with the top screwed on all the way are about 40%, and with the screws backed out, it drops to about 5 ~ 10%. After just speaking to Alex (Petra), he has told me that the DDC 3.2 is a complete redesign and that the amperage is raised in a very slow curve allowing for much smoother startup. Although, I have not had a chance yet to examine the PCB of the DDC 3.2, I will look at this in more detail this week.
Therefore, taking all this info in (along with the keen eye of my lab buddy, an experienced electrical engineer and two technicians from our department's machine shop), I would say the most likely culprits behind these strange failures are:
1 - Laing designed the circuit layout incorrectly on the DDC-2 PCB with its more powerful electromagnet, when they made the switch from the DDC-1. They increased the amperage from 0.75 amps to 1.5 amps but did not change the onset time delay in the PCB. Therefore, the impeller starts up too quickly, with too much wobble and careens into the bottom of the screw-down tops. (this is the answer that I personally feel is most likely)
2 - Laing has made a small manufacturing error in how the wire for the electromagnet was looped in the DDC-2 models, causing 'gaps' in the electromagnetic field which induces greater wobble of the impeller at start up.
3 - Laing has made a small manufacturing error in how the impeller itself is weighted (the 'moment' of the impeller is off), which greatly exaggarates any wobble. (some users have said they felt their impeller was defective in some small manner)
Further, the ripple (often called noise) on standard 12 volt direct current PSU rails is not enough to account for the DDC-2's problems. And although, the stack height between the top of the impeller and the bottom of the screw-down tops is critically important, it is not the main culprit. Raising this height, or sanding off material from the bottom of the screw-down tops will simply allow minute cavitation and result in lost head pressure and/or flow. These three problems described above are unfortunately only problems that Laing can correct and not you (at least not easily). Surely, this article will reach Laing at some point and they can then critically review this.
Until that time comes though, here are the home remedies which may help you:
1 - If you have impeller stoppage or hear a 'sand' like grinding sound, back off the bolts of the screw-down top by one to two revolutions. (yes, your head pressure and flow may drop by a very small amount, but this is better than redoing the entire loop and sending the pump back to Laing)
2 - Mount the pump horizontally.
3 - Do not use viscous fluids with the DDC-2. 100% distilled water and biocide will be one of your best bets here.
4 - Although not critical to the failure rate, raising the pump and cooling it with a fan, can only help
After having just spoken to Alex, I will spend some time testing the DDC 3.2 and we will see how she holds up. So, probably some more updates later this week.
I hope this helps out for now and I will notify Laing about this,
Jay
Photo #1 – Taking a look at the DDC-2 to troubleshoot
Photo #2 – The 7.75 depth of the circular depression at the bottom of Petra's screw on top (actually this distance is ~ 8.02 mm, as corrected by Petra)
Photo #3 – The ~ 7.60 mm that the impeller head rises above the base. (actually this figure is about 7.75 ~ 7.78 mm)
Photo #4 – The downward weight (and by calculation, the resultant pressure or friction) needed to stop the impeller is being examined.
http://img253.imageshack.us/img253/3...achmentie1.jpg
http://img253.imageshack.us/img253/5...achmentnb1.jpg
http://img253.imageshack.us/img253/6...achmentva5.jpg
http://img253.imageshack.us/img253/5...achmenttl9.jpg
