VGT Turbo M20 Sleeper ('87 325 Sedan)

[mikey.antonakakis]

oh. much appreciated for the link. and i will definately take you up on the offer if i get stuck

Do you have any recomendation for a CAN reciver module?

Which Arduino board do you have?

[mikey.antonakakis]

Got rid of the manual boost controller and went for a spin this evening to see what boost it’s making on wastegate spring only. ~165kPa absolute, or about 9.5psi “boost”. Note that this would be 9.5psi boost above atmospheric at sea level, about a 1.65 pressure ratio. Because I live ~6000ft above sea level, this is actually a 2.06 pressure ratio (barometric pressure here is about 80kPa vs 100kPa at sea level). So more like 12.5psi boost above atmospheric pressure. But would make roughly the same power as 9.5psi boost at sea level (165kPa MAP).

I’ve got code written to fully integrate the dome pressure PID control into my VGT control (and maybe add a little more flair to the touchscreen OBC), but the PID constants need a little more tuning before I run it. I had it pretty solid at one point but didn’t write down the values, oops. Part of the trickiness is that variation in compressed air supply pressure makes a big difference in the PID tuning, and with a 1-gallon tank and crappy regulator the pressure can fluctuate and drift pretty significantly. I might get into more detail on this later on, but there’s also fluctuation in pressure due to the relatively slow PWM frequency that the solenoids run on. I fixed this with a software low-pass filter, but any filter (analog or digital) will usually introduce a time lag, which can also throw off the stability of the PID loop, so the filter also needs to be reasonably tuned.

To oversimplify a bit, once the dome pressure PID control is tuned, I need to target a dome pressure that gets me to my desired boost, and to do that I need to know the baseline wastegate spring-only boost, hence tonight’s test. Any added dome pressure more-or-less adds to the baseline boost. So if I target 180kPa MAP, and I have 165kPa on wastegate only, I need 15kPa in the wastegate dome to get there.

[tyeler18]

Have you thought about CO2 over compressed air? That would give you a more stable control and not have to worry about your tank fluctuating in pressure as often. A 2lb bottle on this thing would probably last a year or more.

[mikey.antonakakis]

Have you thought about CO2 over compressed air? That would give you a more stable control and not have to worry about your tank fluctuating in pressure as often. A 2lb bottle on this thing would probably last a year or more.

Definitely thought about it. Supply pressure is a little more stable now and PID is getting really close to dialed in. Was messing around with it just now and realized my solenoid open/close thresholds were a bit off (I had changed PID frequency and regulator pressure since my last check and it made a significant difference). Once I recalibrated those thresholds things improved immensely.

[mikey.antonakakis]

Got over a foot of snow yesterday so testing is on pause. Been spending this time doing some thinking on the VGT control algorithm. As-is it works pretty well using only
MAP, boost target, and engine speed as inputs, with only a couple hours of street tuning. As varg and others have pointed out though, it would in theory work better to consider turbine power, which is heavily dependent on mass flow - close vanes too much and you choke the flow, making it counterproductive, and trying to maximize turbine power when I need more
boost would give the best transient response. It’s a little difficult to practically measure and model the relevant parameters, though.

So far I have essentially been iterating my way backwards into this faking this type of control, but it is pretty tough to really get it dialed in with street tuning where I live (there are almost no level or constant-grade stretches of road). My current control method is to have rpm-dependent “most-closed” vane position. I then have a kind of dead band in boost error (target minus actual) before I start opening the vanes, and open them linearly after that (currently 35% I think). For example, if I go to WOT and my boost error is 100kPa, I close the vanes to the allowable minimum, then once my boost has reached 35kPa I allow the vanes to open linearly with the remaining boost error, so vanes reach fully open once I reach my boost target. By tuning this simple parameter of allowable most-closed position I am trying to maximize turbine power iteratively, at least at the start of spooling up. I think my use of a deadband might be backwards though. I probably need to make the vanes immediately start opening as boost rises, and still be partially open once I reach target boost since at lower rpm there’s probably no meaningful difference between say 50% and 100% open.

Better (at least quicker to tune) would be measuring exhaust manifold pressure and temperature and maybe switching to a target turbine expansion ratio (using temperature correction). Since I measure turbine speed I could possibly incorporate it as well, and I think Megasquirt 2 in speed density mode can broadcast calculated mass flow (or at least I can calculate it based on other parameters), maybe I could even get a decent estimate of turbine power. If so, I could pretty quickly tip-in to WOT at various engine speeds to map out how turbine power responds to vane position. Any of this requires me to add exhaust temperature and pressure sensors, though.

In the short term, I can probably get the tuning dialed in better with this method:
-Run with fixed wide-open vanes at a variety of engine speeds, focusing especially on the initial tip-in to WOT to get fuel map tuned decently (should already be pretty close). This gives a baseline AFR.
-Re-enable vanes to close and at a variety of engine speeds, go to WOT and repeat with a variety of allowable most-closed settings at each engine speed. See how AFR and inital turbo shaft acceleration responds to changing most-closed setting. Too much richening indicates loss of mass flow, but shaft acceleration would probably tell the story the best. This should build a very decent map of most-closed position vs. engine speed.
-Once most-closed map is built, focus on mapping the vane ramp rate vs boost with similar method. Difficult to do this repeatably in practice, though. But similar idea, trying to find the edge of noticeable richening of AFR is probably a good starting point, and monitoring shaft acceleration probably tells the best story. Either way I probably need to fix my algorithm to reverse the dead band in vane response, so the vanes start to open a bit as soon as boost starts to rise.

With all of that method though, there’s still a big variable: EGT. This setup is very sensitive, it spools so much harder when the exhaust is good and hot. I really need to get EGT sensor in there.

[mikey.antonakakis]

All that said, I did get a bit of perspective. I’m on a ancient 2.7L with low compression ratio and the “wrong” pistons (fully eta bottom end with 885 head), and a fairly big cam (Z45 turbo, 280/280). My turbo is 60mm in, 60mm out. The new 911 turbo has VGT, a very modern 3.8L, and a 55mm in/61mm out turbo. So, similar on turbine side at least. Makes peak torque from 2500-4000rpm, 640hp at like 6700rpm.

With that in mind, and my still far-from-ideal VGT algorithm/tuning, I am I think doing “not bad” considering that albeit slowly I can make significant boost above 2500rpm, and above 4000rpm I can make lots of boost pretty quickly. Flatshifting and VGT make things really good already from gear-to-gear, even when I am a little slow with the shifts and not truly staying WOT during the shifts:
boost_flatshift_2022-05-19 by Mikey Antonakakis, on Flickr

Notes: rolled on in second gear, running wastegate spring pressure only (160-165kPa). Short-shifted 2-3 at 6300rpm, short-shifted 3-4 at 5000rpm. Lifted to 75% on the 2-3 shift, dropping boost target (worse VGT response). Stayed WOT for 3-4 shift. Shift times (based on spark cut/retard) were about 0.5s. Maintained MAP at 135kPa on the 3-4 shift. Turbo speed dropped from 78k to 64k on the 3-4 shift. VGT opened too early after 3-4 shift, I think.

[mikey.antonakakis]

In other project updates, working on a PCB for the dash display/VGT controller/CAN interface/Boost controller, schematic is 20-30% complete here:

Capture by Mikey Antonakakis, on Flickr

[mikey.antonakakis]

Schematic is “done” for the touchscreen OBC. Contemplating exactly how I want to lay out the PCB. TE Connectivity makes some pretty nice looking enclosures for PCB-mounting bulkhead DTM connectors. I want to do it with dual 12-pin connectors but it is really difficult to source all the parts right now.

Another small job knocked out. Rewired the intercooler heat exchanger fan to run with intercooler pump instead of engine radiator fan. Noticed the intercooler fan had a bad connection, tried to repair it but the motor housing is stamped/spot welded. For $25 I just got a new one, all back together now.

At some point in the not-so-distant future I might upsize the intercooler radiator. I’ve got about 90sq-in of core area, could easily fit a 12”x12” core, might be able to fit a 24”x7” although it could get a little tricky with fittings. 24”x12” also might just barely be doable. It really will only make a difference in heat soak, would need a bigger water-air core for better short-term performance. Going to delete the ABS unit soon, which will clear up a lot more space for a bigger core in the future. Last time I checked, it performs pretty decently for being a cheap, smallest-offered kit. After several pulls on warm days, MAT will get up to ~50F+ambient. Going through a couple gears hardly does anything. Will do more up-to-date testing before I swap any of it out.
Frozenboost.com has their warehouse in town, which is highly convenient, and they have a nice 50% off shelf for items with minor defects.

Untitled by Mikey Antonakakis, on Flickr

[Panici]

I just gave the thread a read. Excellent mix of fabrication and electronics! I've subbed for updates.

Metal work like your turbo inlet tube made to clear the rad hose is ace.

https://www.r3vlimited.com/board/filedata/fetch?id=9892293&d=1576458403

I also love that touch-screen OBC replacement. May have to explore a similar option for my E30 running MS3X (once everything else is done).

[mikey.antonakakis]

I just gave the thread a read. Excellent mix of fabrication and electronics! I've subbed for updates.

Metal work like your turbo inlet tube made to clear the rad hose is ace.

https://www.r3vlimited.com/board/filedata/fetch?id=9892293&d=1576458403

I also love that touch-screen OBC replacement. May have to explore a similar option for my E30 running MS3X (once everything else is done).

Thank you for the kind words! I was checking out your build thread last night, didn’t realize how tight the exhaust side can be on a M52! It’s looking great.

I’m actually working on making a few of the touchscreen OBCs available, will integrate great with MS3X (I just started a thread for that here. The idea is to have a separate enclosure for the brains and I/O of the OBC to make it easier to expand and give flexibility in placement. I will probably order a run of PCBs today, design is done I think:

Screenshot 2022-05-24 182951 by Mikey Antonakakis, on Flickr

[Panici]

Thank you for the kind words! I was checking out your build thread last night, didn’t realize how tight the exhaust side can be on a M52! It’s looking great.

I’m actually working on making a few of the touchscreen OBCs available, will integrate great with MS3X (I just started a thread for that here. The idea is to have a separate enclosure for the brains and I/O of the OBC to make it easier to expand and give flexibility in placement. I will probably order a run of PCBs today, design is done I think:

Thanks for the visit, it's coming together slowly but surely!

Amazing that you're making some of these displays. I commented on your other thread, I am definitely interested in one for my car.
Sure beats bringing a laptop around all the time, and A LOT cleaner then running a tablet.

[mikey.antonakakis]

Thanks for the visit, it's coming together slowly but surely!

Amazing that you're making some of these displays. I commented on your other thread, I am definitely interested in one for my car.
Sure beats bringing a laptop around all the time, and A LOT cleaner then running a tablet.

I don't think is what you meant, but just to be clear I don't have it set up to tune the Megasquirt parameters. I think it would actually be possible to do stuff like flash maps with this unit, but I haven't personally done it, and it doesn't sound fun to do on 2.8" screen lol. But in theory I think you can if you really want to. For purely monitoring though, it definitely beats the hell of out running 52mm gauges everywhere IMO. And having 8 extra analog inputs for MS2 is super handy.

[Panici]

I don't think is what you meant, but just to be clear I don't have it set up to tune the Megasquirt parameters. I think it would actually be possible to do stuff like flash maps with this unit, but I haven't personally done it, and it doesn't sound fun to do on 2.8" screen lol. But in theory I think you can if you really want to. For purely monitoring though, it definitely beats the hell of out running 52mm gauges everywhere IMO. And having 8 extra analog inputs for MS2 is super handy.

Yes, just for purely monitoring once the tune is dialed in.
Right now as ECU output I have a couple warning lights in my cluster and that's it.

Analog inputs would be fantastic. Could use your device as a gauge hub and get some pressure/temperature sensors to compliment my traditional VDO gauge senders.

[mikey.antonakakis]

Yes, just for purely monitoring once the tune is dialed in.
Right now as ECU output I have a couple warning lights in my cluster and that's it.

Analog inputs would be fantastic. Could use your device as a gauge hub and get some pressure/temperature sensors to compliment my traditional VDO gauge senders.

Definitely going to have that feature (I use for oil pressure, fuel pressure, wastegate dome pressure). As long as total 3.3v power usage is less than 250mA, will support almost any analog and digital sensors (e.g. pressure, temperature, accelerometers, GPS, etc.). For reference, analog pressure sensors usually use 5mA or less, and there are up to 14 analog inputs available.

[mikey.antonakakis]

Weather is nice again. I think I figured out a better VGT algorithm and accompanying tuning process. Going to code it up and test hopefully tomorrow.