rambler roadster

closed-loop cooling system

part 1: hardware

the stock water pump, installed "new" around 2007, failed on the road this past fall, plus the OEM style brass radiator, always a PITA due to it's age and status (brass radiator parts and repair a dead field) leaked yet again (requiring a field repair) so i decided to replace the whole problematic system with modern stuff -- aluminum radiator and electric pump. since those require electronics and software control now was the time to install the new (V2) computer i'd worked on all through 2015. i added code for the water pump and installed it all. took many weeks to pull all the trivia together and arrange it into a system i could trust on the road.

there doesn't seem to be a lot of rodders using electric pump setups, and most of those that do are simply replacing the stock (aka "chevy") mechanical with an electric motor. i have no idea what they use for a temperature controller, maybe the whole hobby really is devoted to quarter-mile use and doesn't care. but mine is a road car.

i chose the Davies-Craig (Australia) EWP-80, purchased from Pegasus Racing in Wisconsin. this is a very small pump, recommended for small import four cylinders, but as i suspected, it is far more than adequate -- in actual use it rarely runs more than 1/3rd capacity and even that only in bursts. more on system behavior later.

physical install was pleasant enough. clearly no one makes anything for this engine so everything was fabricated. the major subtasks were radiator install, water pump block-off and block inlet, the radiator outlet to pump adapter that contains the drain and second temperature sensor, and the head outlet to radiator inlet coupling.

the two 90-degree plastic hose adapters are Davies-Craig parts, fiber-filled nylon. quite good quality. the silicone hoses are from Speedway and Pegasus, in stock lengths. the top hose is an aluminum section plus a 1.5" 90-degree hose; the pump-to-block is a shortened 90-degree 1.25" hose. the pump couples to the lower radiator outlet via the fabricated adapter/drain/temp sensor, made from a 1.75" to 1.25" aluminum adapter/drain that i drilled and tapped to take a Stewart-Warner aftermarket temperature sensor. that required a short piece of 1.75" silicone hose plus a leftover piece of the 1.25" to couple to the pump.

all of the hose clamps (from Pegasus) have an inner extension of the smooth stainless strap that prevents the worm-drive grooves from cutting into the silicone hoses. required!

the tubes and fins of my old brass radiators always got bunged up, from bugs up front and me dropping wrenches in back, and with aluminum being so much softer, i sandwiched it -- both sides -- between honeycomb nomex. the race parts suppliers want $20 for each piece precut, but i got a whole 48" square of it from my local Industrial Metal Supply. the front piece i painted red. both set snugly between the tanks and ends, and are held in place by a screw-down tubing clamp cut down into a fork held in place with a #8 sheet metal screw into the top and botton supports.
i did a lot of reading about user experiences with aluminum radiators. most problems seem to stem from mounting issues. vibration, fractures from welded brackets, and what i suspects harms a lot of radiators but goes unconsidered, chassis/body torsion twisting the radiator, slightly but constantly. i can't weld aluminum, and definitely did not want to use a rigid mount on a delicate critical object. the solution was simple -- i made a sheet steel saddle that mimiced the OEM radiator mount, and sat the aluminum radiator into it, spaced out with blocks of silicone foam rubber. the stuff is very expensive, but i scrounged mine surplus, a big box of pre-cut 1" squares 1/2" thick. the stuff is good for 400F. held in place with Permatex Gray (high-strength) silicone. the radiator is completely isolated, no metal-to-metal contact anywhere. it's a friction fit into the saddle, with a very small clip at the top that keeps it aligned.
the trickiest fab job was actually the head to radiator inlet part. the radiator inlet is actually most of an inch lower than the thermostat pod on the head. the adapter is milled at an angle to point it downward. i suppose it's possible for a bit of air to trap in there, 10's of ccs? but normal expansion handling should deal with it over time. to be clear, there is no thermostat under there.
the pump block off was straightforward, and at the last minute i realized that the OEM pump has an important side effect; it throws water to each side of the first cylinder, not straight at it. so i made a flat plate that's does the impeller's implicit, secondary task. other than that it is exactly what it looks like, a flat plate. the lower radiator-to-pump adapter is a black anodized machined aluminum thing that i inserted a stainless steel mesh filter and a second temperature sensor in. the pump impeller is plastic and very tight clearances, a small chunk of metal could presumably break it (oh woe is me, out in the desert). i extremely thoroughly cleaned and flushed the system, and have been running pure water for the initial couple weeks of testing. this weekend i'll flush and drain and fill with Zerex G-05, an HOAT coolant designed for mixed aluminum and iron systems.
the cylinder head temperature sender is mounted in a hole i drilled in the head back when i rebuilt it (2007?). the OEM sensor up in the thermostat pod might work, but it's slower to respond since it's nearly embedded in the metal. this one, a Stewart-Warner aftermarket job, dangles fully into the water. the other sensor is installed in the lower hose adapter, and is much slower to respond.
mounting the pump itself, was easy, but involved a couple hours of juggling and various visualizations to work out orientation. the trick was to be able to use stock straight and angle hose selections from my suppliers (primarily Speedway and Pegasus). as you can see it's at an odd angle -- this orientation can't trap air in the pump (the primary install warning from the manufacturer), the wire lead vent opens downward, and there is no torque force on any hose. not visible in the pics is the floating anti-vibration mount under the pump, gently supporting the pump upward from one of the old generator block mounting holes, clamped between a 2" x 4" chunk of aircraft-surplus red silicone foam rubber. the pump weighs essentially nothing and so this just keeps it from vibrating/oscillating, seeing how it's mounted by the silicone coolant hoses mainly.
here's the OEM boatanchor. this pump weighs about as much as all of the new cooling system components together.
various views of the complete system. the engine compartment is super clean now. no fan and one belt to the alternator. the two 10" electric pusher fans were well placed over the old brass radiator (wider but shorter) but are too far apart here; but i'll soon move one of them to the very center of the radiator and leave the other off. the hood latch support is currently in the way but it's an easy thing to fabricate. the front view shows the tremendous surface area of the new radiator. visible to the left is the box containing the motor driver MOSFETs, located away from underhood temperatures but still close to the motors. a (reasonable) alternative would have been to run a pair of large (#10) wires from the controller/computer which is on the inside firewall. as it is, a small cable delivers the digital PWM signals to the MOSFETs. LEDs on the front show operation.

system performance

now i knew ahead of time that the software needed to properly control the cooling system wasn't simple, but i had no idea how difficult it was going to be to get right. even Davies-Craig sells only a simplistic controller that seems utterly useless for road cars, seemingly aimed at drag racers.

traditional cooling systems are disgusting things; the horribly inefficient pump, that ineffeciency absolutely required by design, simply churns water in place when the thermostat is closed/mostly closed. the thermostat responds slowly, and water temperature actually varies up and down, with the temperature guage intentionally slow-response to average it out. the pump does the most work, to least effect, at highway speeds, and all energy provided to it from the drive belt is wasted.

i have no hard data from the original system, and wish i did for comparison. but the new system has a two sensors, one in the head and one in the radiator outlet. with the smallest/least expensive crossflow radiator from Speedway ($139) i routinely see 80 - 110F temperature drop inlet to outlet (not a typo). it's is far, far too much cooling for this application. here in Los Angeles, on an admittedly warm February morning, 67F ambient, at 65 mph on Route 2 with 185F water flowing into the radiator, it was 65F -- 75F at the outlet.

the antique (truly) OEM mechanical system works surprisingly well, and is extremely simple, at a cost of 3 - 5 HP. the electronic system uses far less energy, and does a much better job, but the controller software is very complex. it requires a mathematical model of the engine's gross thermodynamic characteristics, which i wrote, using PID (Proportional Integral Derivative) algorithms, one for the pump and one for the fan. they interact at low (< 25 mph) speeds, but above approximately 30 mph, there is sufficient airflow that the fan never turns on.

through a fortunate accident, back when i built this motor i drilled the head for two additional Stewart-Warner aftermarket sensors, between cylinders 2 and 3 and 4 and 5. the casting there is relatively shallow so the sensor portion is fully immersed in the water, meaning that it is fairly fast to respond. immaterial for a traditional guage, for the PID algorithm very fast control is needed -- currently three times a second, and it will probably change to ten (eg. 100mS loop time) if the sensor is fast enough.

both pump and fan speeds are controlled with PWM (pulse-width modulation) rather than on/off. speeds are also smoothly ramped up and down as necessary for lowest noise and wear. the thermodynamic modelling is required; in a nutshell, as combustion heats up metal, the metal heats the water. when the pump pushes "cold" water into the hot head, it remains cold for one to a few seconds; and the metal "upstream" of the sensor is already cooling. the cooling effect of turning the pump on isn't instantaneous; it takes time; therefore temperature undershoot and overshoot are inevitable without rate-of-change prediction. also it's not really possible to truly "control" either heat-from-combustion (controlled by the idiot in the seat with foot on gaspedal) nor "cold" water (at the mercy of the weather and chassis speed, idle crawl to freeway gale).

at the moment (late Feb 2016) my controller is holding water temp at the head to nominal (185F) plus/minus 5 degrees, ranging over a 5 or 6 second time period. given the mass of the head and block this is essentially constant. i'm working at holding it tighter and faster.

the controller throws in for free nicely controlled post-turn-off cool-down cycle work (extracting heat from the block) etc. nothing runs unless it's needed, so the engine is overall quieter.

the software itself is discussed in the code section above.