Well, his excuse was that the only way to remove power from the controller was to open the main power contactor, which also removed power from the motors. So the only time there was a problem was that 0.1 seconds between when the contactor was de-energized and it finally opened, and that never actually happened in normal operation because the contactor was opened with the PWM off.
Never mind what might happen if the main power contactor welded shut...
Yeah, you're right. Some design tradeoffs were made. One was that things were crammed in really tight into this forklift, and there wasn't room to put the cap bank closer to the motor controllers.
In later years we were working on a new controller that had a row of capacitors right across the bus bars in the controller, and it wouldn't have needed that external cap bank.
In 1994 when the original system was designed they couldn't find caps small enough to fit in the space under the top cover that could still handle the ripple current, so an external bank was used.
That reminds me of another thing. Way back when I started at the forklift company in 1994, the engineering department was a bit lacking in equipment. So for a while, we had to use an old WYSE terminal for debugging. The forklift had a serial port on the controller board and you could plug in a terminal or computer with terminal program, and look at things like all the sensor readings, examine memory contents, etc.
When we would run the lift motor at about half PWM, which resulted in maybe 250 amps or so, the image on the CRT of that old WYSE terminal would shake up and down about 1/4 inch, even though it was on a table a couple feet away.
Magnetic-field generation and wiring-loop inductance go hand in hand. I'm surprised more attention wasn't paid to the current path, wires beside each other, twisted, etc. If twisting is not practical, often the case with fat cables, one can drop the loop inductance (and magnetic field) dramatically by using a bundle of four wires instead of two, with each paralleled pair arranged on opposite corners of the bundle. Reducing the wiring and load inductance is usually beneficial to the controller in some way.
What's better -- I've once visited an aluminium foundry. In the hall with all the electrolysis cells you could stack half a dozen (iron) coins on top of each other standing straight up on their edges. When you moved in the hall this "stack" would lean over on one or the other side, depending on which direction the field was...
In a way, he had a legitimate grip. The digital scopes advertise that they run at high sampling rates and digitally filter the data thereby avoiding aliasing. But actually as you turn the timebase knob, at high ratios of input frequency-to-scan-rate, the antialias filter fails and all mayhem breaks loose. This is as true for the $20k Agilent Infinium as it is for the digital Tektronix.
These wires were all thick enough that twisting wouldn't be practical. You might get one twist per foot or so.
Interesting. That sounds perfectly logical, but I never would have thought of it.
Had that been suggested, I suspect the management would have immediately shot it down due to the extra wire cost. That's the way things went there.
Here's another story. Not high current related, but it is from the same engineering lab. One of the engineers there had the idea that he knew everything, and then wondered why nobody liked him. He was trying to debug a controller board one day, claiming that the microprocessor clock was running at something like 700 Hz. The micro being a static design would happily run at that speed, but should take forever to do anything since it normally ran at 20 MHz. Anyway, since he "knew everything" nobody was volunteering to help him out with the problem.
He was looking at the clock signal with a digital oscilloscope. It was a Tektronix TDS420. Digital scopes are great as long as you understand the differences in how they behave compared to an analog scope. The scope was showing a nice sine wave on the screen, and he even had a measurement set on screen that also verified the frequency to be about 700 Hz.
After an hour or so with him tinkering around at the work bench, he lifted the board up while it was running. Mounted on the back side of the controller was the display board, which was a dot matrix LED display. The display was multiplexed and scanned in software run in the timer interrupt on that very same microcontroller that supposedly was running with a 700 Hz clock. But I saw that the display was scanning normally.
At that point, I realized that the clock circuit was just fine. He just had the digital scope timebase set way to slow, and it was aliasing. I walked up to the bench, said "Here's your problem" and pressed the "Autoset" button on the scope. It automatically reset the timebase and showed him the 20MHz clock. As I walked away I heard him muttering something about "damn digital scopes."
Mentioning coins in a thread where I've been talking about forklifts reminded me of this one.
While working at the forklift company I learned how to pick quarters up with a forklift. :-)
You place the quarter on the floor a bit in front of one of the forks. Tilt the mast forward as far as it will go and lift it up so that the tip of the fork is just off the floor. Drive the forklift forward so that the tip of the fork is over the quarter, then slowly lower the mast a bit until some of the weight of the fork is resting on the quarter. Then slowly back up the forklift. When the fork tip slides off the top of the quarter, the quarter will flip up and land on the tip of the fork.
Hmm, we are talking the same time frame. That must have been a rather small compartment. We managed the riple current with seven 35mmx35mm snap-ins. The end product was rather smaller than the SCR systems it replaced.
Of course one of the adavantages is that the duty cycle is such that you don't spend much time at the worst case ripple and can treat those as surge cases. If you had to rate for full current and 50% PWM continuously you would need a lot more caps.
Seen that too. It sounds like we had similar experiences at roughly the same time. Any chance your lift truck was a reach? :)
Speaking of current limits. The controller was designed, over my objections, to use the NMI to signal an over-current. The hardware would cut out the PWM for one cycle and leave it up to the SW to correct the problem more permanently. Under the right circumstances, which of course were not too hard to produce, you would get a cascade of NMIs as the SW tried to react to one NMI and failed to do so before the next occurred. The end result was either a small hesitation which was not really noticable unless you were looking for it, or overloading the processor so much that it either blew the stack or triggered the watchdog resetting the controller in the process or worst of all loading the processor down such that it was effectively stuck for a short period of time. I managed to eliminate the last in SW. The board was eventually redesigned to put a one-shot on the NMI. Ugly, but workable.
Most of the systems we went into didn't have a line contactor, only a manual emergency disconnect. They relied on that and the direction contactors for interlocks. Needless to say, we couldn't relay on timing there.
One of the first application note I did was for adding a line contactor. It could be wired in for two purposes, 1) to prevent the spark from plugging a battery into a bank of 21+mF work of caps and/or 2) keep a battery of reverse polarity from being plugged in.
Reversing a 600AH 48V battery across the controller doesn't do it a whole lot of good.
I am going to run out of stories eventually. And being a lot younger than most of the regulars on this group, I haven't had as much time to accumulate stories.
One day I was sitting in our lab area when the power went off. It wasn't a clean quick power loss like when a breaker trips. This time the lights went dim and kind of flickered for a few seconds before going out. It was clearly a "something really bad just happened in the factory" kind of power loss.
It turns out they had a problem with one of the forklifts they were building where the bearings in the mast rails were binding up so that the forks wouldn't lower all the way. The width of the channel in the mast rail probably wasn't to spec, but even so it's easier to just get out the grinder and grind off a little in the offending area than to tear it all apart. So that's what they tried to do. The problem was that this mast was an optional mast that was unusually tall. They couldn't extend it up enough inside the building to get access to the area on the rail that needed adjustment. So they drove it outside to do the job, and neglected to notice the power lines coming into the factory right above the area they decided to work in.
So the power loss was when they lifted the mast right into the power lines. I don't remember the exact voltage of those lines, but I believe the safety guy later told me they were several thousand volts.
Fortunately neither of the guys working on the forklift were hurt. One was standing in the forklift (these were stand up models) and was lifting the mast at the time of contact with the power lines. But the controls of the forklift were all plastic, and apparently well enough insulated that he didn't even get a shock. He had enough sense to let go of the controls and hop out of the back of the forklift in one jump. I don't know if the other guy, with the grinder, was touching anything on the forklift or not. Probably not, since they were still lifting the mast at the time.
When we moved the forklift the conductive rubber static strap that used to be underneath was gone and there was a sooty blast mark where it used to be. And there was a 1 inch hole in the concrete below that point where all that power burned through the concrete looking for a path to ground.
The power lines had big notches melted out of them at the point of contact. The electric company's fix for that was to cut the line right at that point, clean up the ends, and push them into some sort of splice connector that locked the wires in because of their own tension. Sort of like an 18 inch long steel chinese finger cuff for power lines.
I was somewhat proud of the fact that the electronics that our group designed survived the incident intact. They replace the mast bearings just in case of any pitting caused by the current, and replaced all the electronics just in case the reliability had been affected.
So we took all the motor controllers and system controller box and put them on a shop forklift just to see if they would work reliably. That forklift drove around the shop, used by the production workers, for years with no problems.
Once I managed to get the NMI down to something like three instructions it worked rather well. It all depends on latency, how close you are to the end of the pwm cycle when you hit current limit and the motor inductance. It worked for all but the most degenerate cases. Nonetheless capping the maximum repeat rate was a definite improvement.
Saw a controller (I did the production test benches for it) that used that for an inner current loop. Quite effective.
I did software current control on multiple bridges in a single micro to do similar control. Once it's tuned it's very effective. You do have to be careful while tuning though.
We did run into an issue with plugging. It turns out that under certain circumstances once you start plugging it will self sustain and the only way to get it to stop braking is to open the direction contactors. The motors we first tested on actually braked very nicely in that mode.
Our boards all had the overcurrent implemented totally in hardware so such things couldn't happen. But there was a logic signal back to the controller so that it could tell when the overcurrent was tripping.
The overcurrent was controlled by a one-shot. So if you continually pushed the forklift into an overcurrent condition, it would make a ratchet like sound as the overcurrent re-tripped on the next PWM cycle after the one-shot. The problem is that while that is happening the average power was far less than what the controller could do right at the edge of current limit, so the only way out of it would be to back off on the control handle and ease into it again.
Later we were designing a new controller that did cycle by cycle current limiting. That worked so much better there was no comparison. It would just take the controller to maximum power and sit there. You could even slam the PWM on to 100% and just watch each cycle current limit a little later as the motor sped up. But that design never made it to production, as it was in progress at the time of the Great Downsizing that left the company with 1 employee.
Schaeff Incorporated. They eventually somewhat came back and are still around.
What happened was a combination of a company that was always on the edge of financial trouble, and the bad economy around 2001 and 2002, and a merger with another company.
Normally most companies have blown their capital equipment budgets before December rolls around, so they always had low sales in December. Traditionally they would have a temporary shutdown for a couple weeks in December, to let some orders build back up. But normally us engineers weren't included in that unpaid shutdown because we always had plenty of work to do.
But in December 2004 I think their financial troubles were worse than usual, because their shutdown was supposed to be for a month and included us engineers.
Now the Schaeff Inc. forklift company in Sioux City, IA, was a division of Schaeff Inc. in Germany. And in December 2001, Schaeff Inc in Germany was bought out by Terex.
This part is conjecture on my part, but I think Terex just wanted the construction and mining equipment that Schaeff made in Germany, and didn't care about the relatively tiny forklift manufacturing business in Iowa. So they just let them try to continue on their own without any further financial help, and they didn't have the money to bring us all back, so the temporary layoff turned permanent about 6 weeks later. Actually at that point they still had more than 1 employee. But they had gone from something like 120 people down to maybe 10.
Then Terex moved the whole operation to someplace near Chicago. Only one employee from the Sioux City plant went to Chicago. Terex later sold the whole operation to someone else.
Their web site is at
formatting link
The products I worked on were the W-Series and Schaeff ECHO forklifts.
That reminds me of another one. We sort of accidently discovered an interesting thing about our forklifts. If you got them up to full speed, and while continuing to hold the control handle forward, reached across with your left hand and hit the battery disconnect, the momentum of the forklift turning the traction motors while coasting to a stop would generate enough power to continue to power all the electronics of the forklift. The power would feed backwards right through the armature mosfets and feed back into the B+ bus. It provided enough current to sustain the field winding current and keep the generator effect going, and even power the electric steering pump motor.
If you let go of the control handle, the armature mosfets would turn off and cut the current flow to the B+ and everything would shut down pretty much instantly. And on one model that had electric brakes, the electric brakes would close and the forklift would screech to a halt. :-)
So, of course, the next "unofficial" experiment was to push one forklift with another, then have the one being pushed disconnect the battery. This worked well enough that the one being pushed could run all the functions including lifting and lowering the mast. As long as you held the control handle forward so the controller would keep the armature mosfets on, it would continue to work as a generator.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.