PicOx - Open Source Motor Controller
For Light Electric Vehicle PMDC Motors -
Part 3 ...
March 2008
Disclaimer
- This project is in a preliminary state of development, every effort is being
put forth to share it's progress, and note issues as they arise... Though
theworkshop.ca and parties noted in the development of the PicOx Open Source
Controller are held blameless and without liability with regard to how this
information is used.
Section
2 closed with a schematic and a proto-typed circuit that was tested to
16Amps at 24V and had a working current limit circuit that kept the motor within
it's operating range of 24V @ 6Amps...
The
PCB layout to the right is the translation of the schematic listed in the
previous installment.
With
the addition of a second IRFB4110, a third 20200 Diode pair and three 470uF
capacitors in place of the single 820uF.
The
red lines denote jumper wires that link circuit elements.
The
layout process for this board took essentially a full day spread out over 2
(two). The Allegro ACS75x sensors still have not arrived, so rather then wait,
the "Home Brew" current sensor that was hacked out of a Hall throttle
has graduated to Ver 1.02.
There
was a bit of putzing about to get the image out of ExpressPCB, mostly related to
having to clean the Laser Printer to get a clean image.
As an
aside the transfer paper that I used was just a page out of "Nuts &
Volts" magazine that had no colour ads on it and was a code listing for a
project that I know that I'll never have any interest in building... I've had
great success since dropping commercial transfer papers and just using pages cut
out of magazines that have a light weight paper and glossy finish.
The
listing below is of images of the various steps that I take to make a PCB and
should be applicable to any such project... Certainly each person has the option
to use or disregard the advice listed but it seems to work ok for me...
Key
points, Scrub board with an abrasive agent such as Ajax or similar powder
cleaner and water.
Wipe
dry with clean cloth or paper towel...
Avoid
touching the copper surface, handle PCB by edges.
Just
prior to applying the transfer, wipe the board with isopropyl Alcohol, or
equivalent that will dry quickly with no residue.
Set
the iron to "High", but ensure that it has no water in it, as you
don't want the transfer to swell or start wrinkling before it has a chance to
adhere to the PCB.
As
soon as you can handle the board drop it into a pan of COLD
water.
The
toner is a plastic particle that melts with the heat of the iron, and the cold
water is like a thermal fixer of sorts.
The
brd can sit for 5 or 10 minutes to soften-up the transfer paper.
I use
either a course steel scour pad like the one shown to the right or a softer
plastic pad that looks similar.
The
pad is applied very lightly just to rough-up the paper and start reducing it to
a pulp.
This
image is the last of the scour pad.
The
pulp is essentially rubbed off by your finger tips, again with very light
pressure...
As
tedious as this all sounds, I'm talking about 15 minutes tops from the time the
brd hits the water till here..
This
is the step I always forget, and typically makes life much harder down the road
when I do...
For
fast dry-times use nail-polish for touching-up the PCB traces, or you can
use latex or oil based paints, but it will add a couple of hours at minimum of
drying time...
This
board had 6 or seven spots that needed minimal touch-up and I opted to add my
initials in a blank part of the brd.
The
lettering turned out nice and crisp...
Hopefully
it comes out in the copper as well.
Personally
I feel that since we can't legislate common sense, similarly we should not be
able to litigate on the premise of stupidity...
To
that end; Please read all labels, warnings, MSDS material that relates to your
chemicals and handle per their directions...
After
15 minutes...
After
30 minutes...
After
45 minutes...
Once
you think that it's done as much as it's going to... Wash thoroughly with cold
water.
I
used to scrub the toner off the copper at this point...
...but
have learned the virtue of "Patience" and have been repaid with
superior solder joints as the copper is not corroding in the air and from
excessive handling.
Each
hole is lightly punched to create a dimple that the drill-bit can start into
without skating an 1/8 of an inch before breaking off at the chuck.
Initially
I like to drill every hole with a 0.020" or 0.025" drill bit and then
scale-up to what ever is needed for larger parts and hardware mounting.
This
is only the second time that I've opted to transfer a parts mask to the a PCB...
I
found that the first time saved some hassles and questions about parts
placement, so I figured I'd try it again.
It's
easier to apply the transfer after the holes are drilled, and it will last
longer as it won't be abused by the drilling process.
If I
was a real keener I could have added the part#'s as they correspond to the
schematic to really save time...
Over-all
though it looks good, it's obvious where the IC's go, and the like...
Any
holes that had to be enlarged were...
And
all over-sized parts are test fitted to the PCB, and/or fabricated as needed.
The
heatsink is just a small slab of 1/4" aluminum that is cut to size, drilled
and tapped.
Now
you can scrub down the toner and see that shiny fresh copper for the first
time...
I
have all my parts and misc supplies ready to start soldering immediately after
the copper is exposed...
This
board was stuffed and soldered in about 3 hrs...
Bear
in mind that I have very poor eyesight, so 45 minutes wouldn't be unrealistic if
this wasn't your first time...
The
bottom line is that it should be done as soon as possible after the copper is
exposed...
Soldering
half on a Sunday night and the other half a week later is a really bad idea (I
know...)
Assembled...
The
Power connectors were fabricated out of a small length of used household 1/2"
copper pipe.
It's
not that an energy apocalypse or economic Armageddon are on the horizon (which
they are...) but rather that I couldn't find a suitable source locally or via
the web that could have the finished part in my hands in under 15 minutes at
near zero cost.
These
are a couple of 30A thermal breakers that I picked up from a supply house...
Actually
I bought all that they had (the hording has already started...)
The
mechanism is a concave tab that pops up at a predetermined temperature.
These
are a TruFlex B150R material that has impressive specs available on-line.
After
the success of a "Home-Brew" pre-charge circuit for the BFx
controller, I built this unit exactly the same way and added the 30A breaker
at the top of the circuit closest to the Battery V+ lead.
This
resolves a minor issue that I didn't want to mention on the previous page as I
hoped that this would do the trick.
The
pre-charge of the caps also keeps the motor from doing a quick jump as the
PicAxe power's up and the Fet's conduct for a split second.
Essentially
a soft-start for the controller electronics.
Basic
operation is hold the RED button down for 5 Sec and flick the "Mains"
switch up...
If
some one has a hard time with instructions as simple those, perhaps they
shouldn't be driving an electric vehicle (though you may want to keep that
safety helmet on them...)
Finally
I have the gear assembled with fresh AWG#8 wiring that is crimped and soldered, equipment
has been moved from the office to the shop and set-up on a bench.
This
the most nerve wracking part, powering up for the first time...
All
goes well, the initial waveforms from Section 2 all scope out properly prior to
attaching the MY68 motor.
Even
with the motor attached it spins-up and runs under no load to full speed at 24V
just like in the office...
With
a locked rotor the current limiting function is not working as before... I
anticipated this as the sensor was moved from below the IRFB4110 between the
Source leg and ground, up to being in direct line with the Motor.
The
copper coils on the sensor are 4 turns of AWG#18 wire times 2.
The
wire was doubled-up to handle more current, while the number of turns was
reduced to get a broader scale of amperage readings.
So
the CurrADC will have to be re-calibrated in the software.
I
started to do just than when I found that the current readings were pretty
screwy...
The
Motor had an analog meter wired in-line that would read 3, 5, 8, 10 amps in a
linear fashion as the throttle was advanced, but the pc was returning 2, 6, 0, 1
amps or any combination randomly of similar numbers.
With
a 5V/div voltage scale, there is a pronounced negative going 20V spike on the
ADC output of the UGN3503 that is feeding the PicAxe processor...
Surprisingly
the picaxe did not heat-up, but just offered consistently random readings...
The
simplest solution was to see if there would be enough Magnetic field off the
single conductor to the motor ground to provide a CurrADC value.
In
reality I don't need to have even single amp precision, 5 Amp resolution would
be good if it was reliably consistent.
The
torroid is hot glued to the Hall sensor and the power terminal to avoid
mechanical vibrations from disrupting the ADC readings.
Unfortunately
I can't report that this is a functional configuration...
Since
the current limiting routine was non-functional, it was commented out of the
code and the PicAxe updated, so that I could try and determine the new CurrADC
scale. In the course of making numerous measurements I was able to run the
current on the Analog amp meter up to 20amps at full throttle.
Initially
I would try brief periods of 5 secs, 10 secs, 15 Secs with smooth ramp-up and
down of the current... certainly the tiny MY68 was getting hot and was reading
45C on it's outer shell, but otherwise was holding up.
As I
was letting it cool down, I tried ramping up the current in short fast snaps of
the throttle from 0 to full. when something strange happened.
The
PicAxe locked-up...
I
knew immediately that it was locked as the terminal that I use as a data log
from the controller displayed garbled ASCII as it's last entry.
Power-down,
back-up and all is well...
After
a few tries it locks again.
If I
hold the throttle pinned on with a slow ramp-up after about 20 sec the processor
locks.
This
is a real problem, especially considering that twice the picaxe locked at full
throttle...
The implications
of not resolving this problem at this point would be akin to playing Russian Roulette.
There
are several days of work to this point and I'm pretty bummed out... After
disconnecting the CurrADC line from the PicAxe and numerous other trouble
shooting procedures I turned to the Endless Sphere EV Forums for some ideas...
Over
the course of the evening a few folks replied, and the response pointed toward
ground noise.
I had
considered that there could be a ground problem, but it seemed impossible that certain portions of the ground plane were noisier than others. Up to this
point I thought that all points along a conductive path would be at the same
electrical potential as every other point along the same plane.
But
that is for an ideal conductor under ideal circumstances, and unfortunately this
is reality...
This
the source of the noise, a negative going spike that measures approx 5 Volts
directly off the Source leg of the FET that is closest to the section of the
ground plane that services 7 or 8 logic level components in the control
electronics section.
After
looking at the PCB layout, and literally having it pointed out to me, I'd
created a long serpentine path for the noise to service virtually every part
possible BEFORE being clamped by the bank of capacitors.
The
short term solution, or more accurately "Proof" was to add this 820uf
capacitor with heavy AWG#14 solid copper legs to clamp the noise as close as I
could at the source, which ironically is the "Source" lead from the
MosFet...
The
Noise has been tamed to a semi-manageable level of 1.5V at the Grd connection of
the LM317 regulator...
There
is no amount of throttle snapping that can lock the PicAxe now, similarly the
motor has sustained 20Amps @ 24 V for 30 Secs with no ill effects beyond the
predictable heat rise...
So in
closing this section, it would appear that I've accomplished very little beyond
getting an education in PCB design and the associate problems with noise.
Since
the next logical step is to either add a larger motor to draw more current or
bump the voltage up to 36V, it's obvious that more and more capacitors would be
need to be added to mitigate quite possibly a text-book example of the worst
noise management design of the circuit at hand...
So
over the next few days the PCB will be completely revised to place the Large
Capacitor Bank as close to the FET ground plane as possible with the logic level
circuitry on the other side.
Support
theworkshop.ca