Quote:
Originally Posted by 1023vaughn
This gear is used solely for my educational purposes, development and speeding up my work day. It contains only about 15-20% my own scratch built software, and is used heavily for helping me develop better tools for work-flow, some new 1023 software and tools, and lots of security tools. The base is a 3-node windows Hyper-V fail-over cluster, with 2 redundant flash based v-sans (also windows based) and 2 redundant storage nodes (windows). 2+1 on/off site redundant storage with windows DFS etc. Lots of VMs, some include Windows and Linux Kernel dev, full stack web dev, databases and redundancy, pbx, email, game servers. I built my own email services from on old Hmail fork. Scratch dns, dhcp, http proxy, e2e messaging, "real-time" distributed filesystem replication (early stages), custom win service handlers for managing other software, in win32 and .NET and plenty GCC. Alot more that were based off other builds or libraries but its been a couple years since I needed to touch them. Home automation, security, remote tools, vpn, and probably 3-4 other vms I forget what they do and tons of other vms and software. At one point it was supporting nearly 150 devices and 1g/1g fiber which I dearly miss.
BTW Talking pricing and memory, raw storage in that rack is just over 100TB total, just under 1TB of DDR3 and just under 200 CPU cores just in that image. Living in northern VA at the time its mostly made from retired US government gear.
I have a bunch more modern gear coming in with new nodes to replace the old gear. I had to run 3 20amp circuits to that rack and each one of those R910s could easily pull over 1200VA from the wall, and the power bill was getting to expensive  . Lots of 10g gear and can't wait to get the rack back up and get my whole lab back in order. |
That is a really sexy setup.
It cool that it's built from retired units. I bet that saved a ton of money because to build that from scratch would be prohibitively expensive!
All the stuff in our server rack is now new stuff, replaced and added to over the years as needed. I've debated dumping the Windows install on the 1 server and just migrating it to Linux, but it works well, never crashes, and I just don't feel like messing with it right now. The other 2 full stack Linux servers are great, with one being a redundant mirror of the main Linux server. They never give me a fuss and will run for years without ever needing a reboot. I'm really enjoying learning my way around Linux.
Quote:
Originally Posted by 1023vaughn
I'm sure I have a good 20-30gb of archived data automotive and EEC.  |
Anything you want to throw on a drive and send my way?
Quote:
Originally Posted by 1023vaughn
To ask an on-topic question what was your original process of moving from "yeah that's a table" to "This is base ignition timing" starting out? I'm just curious at a high level how the work-flow started say at "SCT"? (We don't need to talk about disassembly etc)
Did you jump right into hardware "hacking" or was it more organized?
|
Organized?
Not a chance.
So, when I started at Superchips (not SCT... That is something entirely different) they were doing mostly gas tuning and just barely getting started on the Power Stroke. From a Ford aspect, it was basically Mustangs, Cougars, T-Birds, F-Series gassers, and the like. They did GM and Dodge as well, but I didn't do much GM and I didn't do any Dodge.
Anyway, as sad as this may sound, they only thing we were changing in the files in 1997 was timing, open loop fuel, and standard shift points. Rev and Speed limiters as well, if we could find them. But to be honest, that's still all that's really needed to make a decent tune nowadays and I find it interesting how much stuff can be changed unnecessarily with no real improvement in performance.
As I started working on breaking down calibrations even further, we started doing some really... um... "interesting" stuff. We would acquire a vehicle by putting ads on the website offering a free chip for anyone willing to let us test and tune on the dyno. We'd then get the vehicle strapped down and get the emulator plugged into the J3 port. We'd load the file for the vehicle, fire it up and start driving. All of this seems pretty ordinary, right? Yeah, well hold onto your hat. This is where it gets interesting.
While we're running the vehicle, we've got an NGS Scanner and a wideband hooked up to the exhaust. We're doing runs and checking to see how much a change to the timing table makes by flattening the table and then adding or subtracting values to the table. "Oh, look! Every HEX point is 1/2 Deg. of advance." We do the same thing with fuel tables to see how much AFR gets adjusted by adding and subtracting values to those tables. On to Shifts and Limiters. It's all seems pretty sane, until...
We started asking things like, "Where is the Traction Control?" or, "How do we firm up the shifts?" Of course, we don't know where any of that stuff is. We have to find it. Now if you remember, the emulator and software we used had live ROM-Watch so we could see what the PCM is accessing from memory. I'd be sitting in the passenger seat with a laptop running the emulator. Pat Stadjel (one of two other calibration engineers Superchips) would be in the driver's seat with the NGS. So we start going through memory looking for active points and then we'd either "00"them, "FF" them, or just start adding or subtracting values and see what happens.
I mean, what could go wrong, Right? Fortunately, we had an "Undo" so that if we made a change that was unhappy we could revert right back to the previous values. In most cases, the worst that would happen was that the vehicle would stall, we'd drop to neutral, and then just fire it back up. The gas stuff was pretty forgiving like that since the throttles were still cable controlled and the chance of a runaway was basically non-existent. I'd make a change, Pat would hit the gas, and we'd see what the comes up on the dyno. Funzies!
Truth be told, this process, while sounding scary as hell on the surface, really worked quite well. We were able to determine things like global modifiers, PATS switches, WOT Shift RPMs, MAF tables, WOT Multipliers, Injector Flow Rates, and a bunch of other things. To date, I'm not aware of anyone experiencing any damage to their drivetrain except for maybe the loss of a bit of tread on the tires.
Having made great strides in calibration development by this point, Superchips' head engineer Steve Madole decided that I would be the guy that would head up the 7.3L calibration development. He handed me a disk with some base files on it, some very minimal documentation on a couple maps that they thought might be fuel tables, and told me to get on it.
Moving into the Power Stroke realm was something entirely different. Right off the bat, this my our first experience with dealing with a large number of 16 Bit values as most of the gas stuff at that time was still only using 8 Bit for the bulk of the stuff we were modifying. PW Modifier tables, SOI Offset tables, and a number of other things were 16 Bit and dealing with that stupid Intel "Little Endian" crap was a hard thing to wrap my head around. Especially since I was using an 8 Bit editor that didn't translate well into 16 Bit.
Forging ahead with the same "Search and Destroy" techniques I'd used on the gassers, zeroing, adding, or subtracting from values and table to see what changed, I'd occasionally get a runaway situation (basically full throttle) but that was easily remedied with the "Undo" function. Again, I had pretty good success with what we found. Enough so that Superchips "gold chip" quickly became the number one selling chip for the Power Stroke for well over 6 years. It's one of those things that I've always been proud of, even if it was only behind the scenes.
Now keep in mind that when I went into this, I still had NO IDEA what a Normalizer function was. I was going on the assumption that the X and Y axis on a table was static (much like the way GM does their tables. It was quite by accident that I found out when I changed a specific function on a Mustang, the timing curve quit changing. When I went and looked back at the timing table, I saw that the ROM-Watch was no longer jumping around the screen. This revelation completely changed my fundamental understanding of the EEC-IV/V PCMs and changed the way I tuned. I still had to go back into the older stuff (meaning we needed more test vehicles) and figure out which Normalizers went to which axis of which table. This was a pain, to say the least, but it at least gave us the ability to provide a considerable improvement over any earlier tuning we'd created. We were now able to control tables into much higher Load and RPM ranges (say, for a supercharged application) so that we had better resolution where we needed it. It really was quite a leap from where we were before.
Diesel tuning at the time was pretty straightforward. We had no problems making power. Where we had problems was getting rid of those annoying DTCs for stuff like Overboost and MAP Out of Range errors. It took a while to hunt those down, and even then we still didn't have a really good way to eliminate them so Superchips would include a "boost fooler" with every chip. This consisted of a 1/8" brass tube with a hole drilled in it which would be placed inline with the MAP sensor to keep it from seeing accurate boost. Sketchy, I know. But it worked.
This was how much of the tuning I did was handled in the late '90s. It was hard hacked. No decompilers or anything like that. Understanding Binary Shifts, realizing the way Strategy structures broke down into specific groups, understanding Hardware IDs, and even things like figuring out how to checksum the files to prevent P0605 code I had to figure out on my own. It was certainly a challenge, to say the least.
It wasn't until late '98 or early '99 that we got blessed with some actual documentation from a calibration disassembly. This was in the form of a 1992 T-Bird ROM Dump from the compiler.
CCAM0 ROM Listing Sample Page
(click to enlarge)
Binary File Search for Gear Ratio 1
(click to enlarge)
Utilizing a search of existing binary files, we were able to locate something fairly unique in the file within a specific address range, and as such we were able to identify a file which was very close to the one that was related specifically to the documentation. As we didn't have any CCAM0xx based files, the nearest files were CCAQ0xx (X4Z1, X4Z2, etc.) calibrations that very nearly matched the documentation. They were only 20 (0x14) bytes off from the CCAM0xx file, and this shift could be easily identified by browsing through the binary and finding where it diverged from the source. This opened up a whole new world for us as we were now able to better understand the binary shifts, normalizers, modifiers, and many other components of the calibration modeling. Of course, this meant going back and redoing a number of gas calibrations as we now could look for new things based on binary file comparisons between our sample base and our working files. This also helped us to learn more about the naming conventions Ford used for the parameters, maps, and functions. Knowledge which carried us through the later gas and diesel applications.
This is how it went, and I honestly think that the experience helped to make me a better engineer. I was forced to problem solve. I was forced to work in HEX. I was forced to figure it out on my own, and only after I had a pretty solid understanding of what I was looking at did we receive any outside help. I'm grateful for that experience. It was an exciting time to be on the forefront of the tuning industry and being on of the people that was making things happen. It was something I took pride in and it was very personal to me.
I'm sure you realize, how tuning has changed over the years. HEX gave way to 3D Graphs. 8 and 16 Bit binaries progressed into 32 Bit Float values. OBD-II flash replaced the infamous J3 port. However, chips are still a solid market because the Power Stroke 7.3L simply will not die. It has a following that rivals many of the iconic vehicles of yesteryear. Mustangs... Camaros... Corvettes... Stuff like that. It is now 25 years after the first Power Stroke 7.3L hit the street and 17 years after the last one was sold, and they're still highly sought after by fans and enthusiasts.
I've been in this industry 23 years and I can't think of anything else I'd rather be doing. I love the challenges of discovering new things, creating new things, and helping to support an industry that is very near and dear to me.
Anyway, I hope this helps you to understand just how bass-ackwards things were in the beginning. If I hadn't been there, I wouldn't have believed it.
