Horizon Sensor:
Someone forgot that we could not hook up a horizon sensor to the real satellite
until we were ready to launch it. We needed a test fixture which would act like the
sensor and interface to real flight electronics without damaging them. They gave me
6 weeks, a block diagram and access to all the CMOS 4000 series I wanted. Finished
the design in 5 weeks including building the unit and tested it the 6th. They were
very happy with the results, the sensor worked as designed and firmware responded
properly.
How do you model a turret with electronics?
I received a call to design a test fixture for a turret. Ok what's a
turret and why do you need a test fixture. Well we need something to emulate the
turret, this one hangs on the bottom of an AC130 aircraft with optics on it. It
weights about 800 pounds, do you think you can do this, by the way we need it about 3 to 4
months and we want to have a PC interface to it. Can you do it?
I asked and received one tech, one software coder (mathmatics expert) and a
budget of $10K for parts. For that I needed to get may parts from the local used
part electronic store.
The control system guy who was working on this design hands me a 11x17 sheet
block diagram of the control system, it seems to have an outer loop working at 4 Hz and an
inner loop working at 30 Hz. I decide to build the inner loop as analog control
system. The step response is kind of hard to meet with all digital, did I tell you I
have a 10MHz 286 as my computer.
The tech points out that the driver is expecting a big inductive load, I wasn't
ignoring that fact just wanted to make sure we could do the hard stuff. I send the
tech off to wire some inductors for me with a couple of 0.1 ohm resisters to us as current
sensors. I need to know how much current the thing is driving to know how to
respond.
After three, yes three board revs on the analog inner loop I finally get it
working. By the way I didn't have real boards for the analog section the last was
copper plate with the sockets ground out and solder posts for resisters and
capacitors. All in all it was ready after 6 months of effort, firmware for the
driver was ready when the real turret arrived and other then a few operational modes which
we couldn't anticipate (you control folks know about modes) it worked.
We bought this great Array Processor but it can't do...
I was brought into this project by another engineer who I admire a lot and had
worked with on two other projects. Our task was to design a board to get data into a
high speed, DSP laden array processor. Software engineers had designed this system
(just a getting us back for all the system that hardware engineers design for them).
We needed to reformat the data as it was coming form four different streams in a very
strange order. I don't ask I do! As we are developing this little card that
goes into the array processor and the little reformat box, the software gurus discover
that they don't have the overhead to perform the gain/offset correction of the image data.
We step up and say we can do that for you what are the coefficients? Well
it turns out that it is really non-linear gain and offset correction and they need to
download the info to us. We have to now build a bi-directional interface to the
array processor and the non-linear gain and offset. No problem give us a little more
schedule and we have it.
To make matters worst (better in my mind) we have a new Mentor Graphics system
which I helped procure and I am proving to management it can make your life better.
I have everything simulating and we are going to go to PCB from the schematic and skip
that pesky wire-wrap board phase.
About a month passes by and we have the non-linear gain/offset working in
simulation and the interface board in the array processor is coming along well. We
have some tricky clock problems because of all the pipelining but have them figured out
also. Those pesky software guys now tell us they can't do the special windowing
functions they need in the array processor and could we tack that onto our board
also? Well this board is now a 3-tiered VME 270 cm board, damn big as boards go and
I just found out that I need 4 of these hummers in the chassis, I am keep in the dark
about many aspects of this design.
We commit to building another board with all this cool functionality on
board. Management is getting nervous and want a project review. We are shown
as the long pole in the tent, hell 6 months earlier we weren't even in the tent.
Talk about pressure. Management decided that the PCB approach is to risky and want a
wire-wrap board made, I push back and tell them the new one could be wire-wrap but the
first is going out to PCB fabrication; they give me that.
The PCB comes back, we stuff, we find some bad sockets, we get it into the
chassis, I get my first look at the chassis (man is it large), we test and in about
2-weeks of playing with code the thing works. Ok, I did make 3 changes in the PALs;
did I tell you about the 30 (yes thirty) PALs I designed for just this board (it would
have been nice to have FPGAs back then, just a year to late). No wiring changes at
all, damn I was proud.
The other board has a micro-sequencer on it from AMD and I have more engineers
helping me than I can stand. We finish the first iteration of this board in 6 weeks
which fills about 65% of the board space. Two weeks in the lab and back at the
workstation and it is finally working. Just a couple of more software items pushed
into hardware and we are done. Total project time 1 year, my involvement 7 months, 2
PCB's, 1 wire-wrap board, 1 custom chassis, 45 PALs and Sequencer and total success.
All the units were sold.
Clients Success