|
Hopefully people are not getting 'exhausted' with this build. Oh well, I will just reply to myself  So now that the exhaust was sorted, it was time to figure out an intake.  The stock K8 flex coupler and VAF take up a lot of room. This is the straight neck KLZE intake which is the hardest to fit in, but flows best. Most people who swap these engines into MX-3s use a curved neck to clear the battery better. I mocked up a filter, silicone reducer and 90 degree pipe.  Not too bad, but the stock battery was definitely not fitting back in there. Time to go battery shopping at costco.  Top is the stock 323 battery (group 35). Bottom is a group 51R, used in modern Hondas. Height and length are pretty similar. Width gives me enough room for the intake piping. I wasn't happy where the filter was sitting. The stock pickup scoops air from the fender area. I also needed some sort of a support for the filter end; it was too floppy otherwise. I've never welded aluminum before, but figured I might as well put my TIG to good use. I took some scrap pipe and after four or five tries I had this:  Okay, looks decent enough, let's try the real thing.  A little rougher, but not too bad. For whatever reason I needed to use a lot more filler rod on this tubing than my practice piece. It fits in nicely with the new battery. I also welded a support bracket to the underside.  The polished pipe has to go. Nothing else is shiny in the engine bay. I glassbeaded and clearcoated the pipe. It looks a lot more like cast aluminum now. Hard to capture in the pictures.  I also modified and shortened the 323 battery hold down. I had to extend one side, add some new bends and shorten the top piece.  I had a Honda battery box laying around. It also fit well and made everything look more finished and factory. 
|
#
?
Jul 6, 2018 20:03
|
|
|
|
| # ? Jun 3, 2024 16:24 |
|
|
I love this thread. Excellent work on the intake
|
|
#
?
Jul 6, 2018 20:20
|
|
|
That intake looks fantastic. Are you going to build a box for it?
|
|
#
?
Jul 6, 2018 20:49
|
|
|
Well done. Can't wait to hear how it drives!
|
|
#
?
Jul 6, 2018 20:59
|
|
|
Wrar posted:That intake looks fantastic. Are you going to build a box for it? Thanks guys. Not sure yet. I haven't figured out why the factory air pickup tube was in the fender. I don't see a lot of air ducting for it, I might have to think up something for that.
|
|
#
?
Jul 7, 2018 05:16
|
|
|
Sgt Fox posted:Thanks guys. Because it's cooler than the engine bay.
|
|
#
?
Jul 7, 2018 15:09
|
|
|
Larrymer posted:Because it's cooler than the engine bay. This^^^^^ Also, with the bumper and various plastic shielding, it makes it a bit more immune to sucking up water during puddle hopping sessions. With the caveat of course being: don't drive through deep water.
|
|
#
?
Jul 7, 2018 16:32
|
|
|
Yeah, makes sense. The MX-3 even had little ducts from under the lights/fender to the battery box. Given they were scooping air from the front, I assume they were trying to get airflow over the battery. I wonder if this was for cooling or to try to dissipate any hydrogen gas? Battery box wasn't fully closed so gas buildup seems unlikely. I'm tempted to try to retrofit then over.
|
|
#
?
Jul 7, 2018 17:00
|
|
|
Mazda kept that up through at least the MS3 - a duct straight from the grille to the battery box.
|
|
#
?
Jul 7, 2018 20:50
|
|
|
Just catching up on this thread, but holy poo poo, amazing work all the way around, but especially in the reverse-engineering.Sgt Fox posted:Not sure yet. I haven't figured out why the factory air pickup tube was in the fender. I don't see a lot of air ducting for it, I might have to think up something for that. Nthing the "cooler air" part, but it's also to help quiet down intake noises. Every EFI car I've owned had the factory air intake in the fender and/or behind the bumper. It'd be great if you could still keep it pulling cold air from outside, it would really help on hot days. That short ram is going to hurt you on hot days... (but it'll sound a lot better!)
|
|
#
?
Jul 8, 2018 10:10
|
|
|
Let's step back from the mechanical stuff for this post and revisit the ECU maps. If you remember, the MX-3 ECUs (K8) had different dimensioned fuel and timing tables than the KLZE (KL31) Japanese ECUs. For example, here is the high timing map from a K801:  15x17 in size. Here is the KL31:  14x15 in size. Obviously you can't copy the Japanese map into the K8 ECU. This got me thinking. I figured there are two possible reasons for the table size difference. One reason could be that the K801 table has a greater range of values, ie, it has data for higher RPM and higher load. I would have expected the KLZE ecu to have more RPM range, not the little 1.8L ecu. The two tables look very close in shape, it does not look like they added data on the upper end of the ranges. That leaves the other explanation, that the K8 ECU has a higher resolution on load and RPM. Again, not sure why, but it seemed the most likely option. This is obviously all guesswork. If we suppose that the K8 has the same range in the table, but just more data points, that means we have to take the KL31 ECU map and linearly expand it. Think of it like stretching a quilt evenly in size. You have the same pattern, but it occupies more space. How can we do this? Like most of life's problems, this one can be solved with maths. Enter bicubic interpolation. This is a method of calculating data points based on their neighbors in a 2D array. You can get an idea from this (black dot is the interpolated value):  Cubic takes the values of more than it's immediate neighbours into account. Wikipedia walks us through the math:    Yeah, so I'm an Engineer, not a Mathematician and I've conveniently 'lost' my copy of Matlab from university, so let's think of an easier way. What uses bicubic interpolation? Image processing! I bet there is a library for performing this on an array. In fact, Paint.net has this as an option when resizing images..  Great, well I'm sure I can find the source code somewhere. Then I got to thinking. A bitmap image is technically an array of pixels, where a set of values represent each pixel.  For an 8-bit monochrome bitmap image, the image is broken down into an array of 8 bit (or 1 byte) pixels. Each pixel stores the gray level, ie, a value from 0-255. (For a colour image, you will have three bytes per pixel, for red, green and blue levels) Conveniently, the timing value is also 1 byte, or 0-255, 0x00-0xFF.  So, I could easily write a program to create a 8-bit bitmap with the gray level data as the values from my timing array. So, this: code: Which means I can now load it into Paint.net and resize it to 15x17 pixels.  We can then run it back through the bitmap to bit array program we wrote to get: code: You will notice it shares the same shape/magnitude as the smaller KL31 map, just stretched or interpolated over a larger number of points. So, I have no idea if my assumptions are correct, but it at least gives a bit of insight into my insanity. Sgt Fox fucked around with this message at 21:50 on Jul 12, 2018 |
|
#
?
Jul 12, 2018 21:20
|
|
|
Nice abuse of a paint program.
|
|
#
?
Jul 12, 2018 21:24
|
|
|
"An AI Guide to Waterpump Replacement Step 23: Reprogramming your ECU with MS Paint"
|
|
#
?
Jul 12, 2018 22:07
|
|
|
This thread just keeps getting better and better and it's not even up and running yet.
|
|
#
?
Jul 12, 2018 22:30
|
|
|
WHAT THE gently caress  That's really cool. MSPaint can really do anything.
|
|
#
?
Jul 12, 2018 22:32
|
|
|
Haha. That's one way to do things. Any idea if the U.S. KLDE is 15x17? I'm wondering if having a bigger table makes it easier to meet our particular emission requirements?
|
|
#
?
Jul 12, 2018 23:01
|
|
|
Charles posted:Haha. That's one way to do things. KLDE (KL01 KL07, KL48 and KL57 etc) are 14x15 as well. The K8 ECUs (K801, K834) are the only ones which are 15x17.
|
|
#
?
Jul 12, 2018 23:11
|
|
|
angryrobots brought me to this thread, I don't know how I missed it. As far as your intake goes, I fully support the DIY I did with a period Mazda (MX6 GT with F2T). I went to the junkyard and found a Yugo. I grabbed the hood scoop and cut out a "U" outline on the top of the hood, then folded it down over a DIY cone filter setup like yours. That way rain was diverted over the filter, but it still sucked in nice, clean air.
|
|
#
?
Jul 12, 2018 23:15
|
|
EVERYWHERE 
|
Holy gently caress, you mad genius! That's amazing.
|
|
#
?
Jul 13, 2018 00:15
|
|
|
boxen posted:"An AI Guide to Waterpump Replacement Step 23: Reprogramming your ECU with MS Paint" I can't put it better than this.
|
|
#
?
Jul 13, 2018 00:19
|
|
|
*Paint.NET.  MS paint doesn't have the right resampling algorithm. 
|
|
#
?
Jul 13, 2018 00:33
|
|
|
Sgt Fox posted:Bicubic resampling hack As a computer graphics nerd I knew exactly where this was going as soon as I saw the first two diagrams and the problem and I wasn't disappointed in the least. This owns, you own, your car owns, your thread owns.
|
|
#
?
Jul 13, 2018 00:33
|
|
|
Can you just convert the values from hex to see what they are so you can confirm your hypothesis about the RPM and such? e: actually looked at the values and no, that would be too simple.
|
|
#
?
Jul 13, 2018 00:56
|
|
|
Wrar posted:*Paint.NET. At this point, I think Sgt Fox would somehow import the algorithm from some Cold-War-Era Croatian import geiger counter or something by epoxying rare earth magnets to a desk fan and using that as the method of data transmission.
|
|
#
?
Jul 13, 2018 01:34
|
|
|
boxen posted:At this point, I think Sgt Fox would somehow import the algorithm from some Cold-War-Era Croatian import geiger counter or something by epoxying rare earth magnets to a desk fan and using that as the method of data transmission. He'll replace the ECU with nixie tubes.
|
|
#
?
Jul 13, 2018 01:48
|
|
|
JFC you are one hell of a insane genius
|
|
#
?
Jul 13, 2018 02:37
|
|
|
boxen posted:At this point, I think Sgt Fox would somehow import the algorithm from some Cold-War-Era Croatian import geiger counter or something by epoxying rare earth magnets to a desk fan and using that as the method of data transmission. Man, thanks, way to ruin the next part. :| (You were wrong though, its a US cold war geiger counter.) 
|
|
#
?
Jul 13, 2018 03:13
|
|
|
Sgt Fox posted:Man, thanks, way to ruin the next part. :| (You were wrong though, its a US cold war geiger counter.)  Pretty sure we've hit peak AI. It's a shame I only have one 5 to give.
|
|
#
?
Jul 13, 2018 03:33
|
|
|
boxen posted:At this point, I think Sgt Fox would somehow import the algorithm from some Cold-War-Era Croatian import geiger counter or something by epoxying rare earth magnets to a desk fan and using that as the method of data transmission. Sgt Fox posted:Man, thanks, way to ruin the next part. :| (You were wrong though, its a US cold war geiger counter.) I love that there is less than two hours between these posts.
|
|
#
?
Jul 13, 2018 08:03
|
|
|
This kind of thread is why I keep coming back to SA. This is THE BUSINESS.
|
|
#
?
Jul 13, 2018 12:40
|
|
|
Yer a wizard, ‘arry This thread is a continuous source of amazement and wonder
|
|
#
?
Jul 13, 2018 13:48
|
|
|
Local newspaper headline: “Mad Scientist Turns His Efforts to Mazda 323 Swap”
|
|
#
?
Jul 13, 2018 15:35
|
|
|
Sgt Fox, I'm pretty sure I've said it before but it bears repeating, if you're ever in Alberta, beers are on me. Also, I think the mad genius types always end up with rotaries. Am I totally off?
|
|
#
?
Jul 13, 2018 17:40
|
|
|
McTinkerson posted:Sgt Fox, I'm pretty sure I've said it before but it bears repeating, if you're ever in Alberta, beers are on me. He rebuilt a rotary several years ago. He got like the last factory new rotor chambers left in North America.
|
|
#
?
Jul 13, 2018 18:19
|
|
|
wesleywillis posted:He rebuilt a rotary several years ago. He got like the last factory new rotor chambers left in North America.  (OK, I'm also out of good 12A rotors)
|
|
#
?
Jul 13, 2018 18:33
|
|
|
Rotaries have a special place in my shop. I've done a few engines over the years. Eventually on the list is helping to fix up a friend's REPU. In fact, a few months ago the local Mazda dealer called me when a guy was looking for a new 12A. Here is what put the 323 project on hold for a month:  A friend blew his 13B on the first practice event of the year, a month before we were scheduled to go to a race event. He cracked a S4 iron around the rear dowel pin and spread all his oil out on the entire track. Queue a couple weeks of ordering parts and rebuilding and porting. We ended up using Turblown's full stud kit instead of tension bolts. (https://turbosource.com/products/turblown-engineering-13b-rx7-rx8-rotary-engine-stud-kit) This allowed us to go to a higher torque than stock. Interestingly enough, it took over eight passes with a torque wrench to get each exactly equal. Shows there is a fair bit of stretch/load shift on those bolts. Then on the dyno, we found the mazda turbo had a poo poo wastegate. We ended up taking a hole saw and welding an external wastegate up to it. Of course, the week before we were leaving for the race, the poo poo v band clamp broke. Hard thing to find on an island, so I ended up making one out of scrap steel. As for the chassis, its an AE86 coupe. Its name is Rotorolla (Rotary Corolla)  
|
|
#
?
Jul 13, 2018 22:07
|
|
|
Sgt Fox posted:As for the chassis, its an AE86 coupe. Its name is Rotorolla (Rotary Corolla) That's just dirrrrrrty.
|
|
#
?
Jul 13, 2018 22:25
|
|
|
god dammit that loving shift to greyscale trick, take your 5 and go
|
|
#
?
Jul 14, 2018 03:25
|
|
|
Yeah that was pretty sweet. How do you plan on testing whether your initial assumption was correct or not?
|
|
#
?
Jul 16, 2018 09:30
|
|
|
|
| # ? Jun 3, 2024 16:24 |
|
|
In the interest of looking like a bone stock car, I picked up a set of 14" miata steel wheels with Mazda center caps for $40. They were pretty ugly. A quick sandblast and coat of paint later however, they looked alright:  Someone was asking earlier about motor mounts and tight tolerances. These were the mounts I used, from AWB. They really don't have much flex.  Now that the engine was fired up, it was time to take care of some of the little things. One of these was the tachometer. As you remember, the stock cluster was for a 4 cyl tach. I did not want to swap to the MX-3 cluster, it would not have fit well and I also was not a fan of its appearance. There are a lot of different tach converters on the market, but I decided to go with TachMatch (https://www.technoversions.com/TachMatchHome.html). It was adjustable to fit your conversion ratio, had a nicely laid out PCB and looked well made. It was also one of the cheaper ones. I wired it up, intercepting the signal from the coil to the tach.  To intercept the signal, I made a little insulator out of delrin and used a couple ring terminals. This way one ring can touch the flex circuit board, and one can touch the screw, which goes through to the tach itself.  I eagerly turned on the car to try it out. Nothing. Boo. Time to toss an oscilliscope on the signals:  Engine was idling (probably high), TachMatch in 1:1 mode. Top waveform is the trigger input, bottom waveform is the TachMatch output. 10V per division, 10ms. I suspect that the voltage may not be high enough to drive the tach. You can see from the bottom trace that the TachMatch outputs a nice square 12V signal. The top trace is the stock tach signal, which has a high voltage inductive spike. This spike is required to drive the stock Mazda tach.  You can see the voltage spike is quite high, 80+ volts for what the scope can actually capture. Luckily, the company also makes a voltage boost module. It creates the inductive voltage spike required for the Mazda tach.  I added some electronics grade silastic to the inductor to prevent mechanical vibrational stress on the leads. It's a heavy component, only supported via axial leads so adding some mechanical fastening is important. I also sprayed a silicone conformal coating on both boards, to prevent corrosion.  After wiring it into the cluster, I now had a functioning and accurate factory tach! Sgt Fox fucked around with this message at 19:28 on Aug 13, 2018 |
|
#
?
Aug 13, 2018 19:22
|
|