Journal98 jun
From Simreal
June 1998
Mon, 1 June 98
Excitement! Yesterday (Sunday) I ran the Leg under PC Brain control (where the Brain software on the big computer sent commands down to the PIC controller, and received feedback from the PIC)... very nice. The pattern I have right now is, upon initialization the leg stretches up and in, then out and down. This morning stretch sets the scale factor for the sensors right up front, so I get good full-scale values out of them. Then, it snaps to a center position under fuzzy-logic positioning control. Very cool. Right now, it overshoots a little bit and has to re-center. Then, I can pull or push on it and, if I strain a lot, I can get it to go off target and it pushes back! Very cool.
Saturday, I woke up feeling unwell -- so I didn't get around to doing anythinguntil noon, and by then my momentum was lost. I tinkered a bit, but nothing big.
Sunday morning was the big time. I spent about 3 hours fiddling with the software and hardware until it worked. And now it is beautiful. Loud. Fast. Precise. I'm still a little bit worried about strength, but I'll be running it through some tests this week.
I'm not using the A/D converter on the PIC anymore... it's slow and not much fun. I have all of my values coming in through the Maxim serial A/D, and I can get a conversion very quickly from it... just a few microseconds, I believe. And it's only about $5 for the chip. I'm beginning to get frustrated with the EPROM on the PIC, too -- the UV erase cycle is a pain. I should go back to the HC11, I suppose, but all my experience is with the PIC now. Atmel sounds good, too. Maybe I'll build boards for each of these later...
I have to finish my board design ASAP... the circuits on my desk are getting totally out of hand, and there are some significant additions I still want to add. The Maxim 3100 UART, for example. And the Dallas NV RAM. That is cool -- I was originally planning a Maxim serial EEPROM (and I may still provide one), but I just found that Dallas has a static RAM with internal lithium battery backup good for 10 years, with good read/write times and 32Kx8 memory. Very sweet. So I'll take over some ports on the PIC to drive that, I think. Well, not take over -- but provide latches and stuff to allow external memory.
This phase of Boris is now complete -- I just barely made my deadline to have the leg working in May. But it is. And it works beautifully. I have still photos already in these pages, and I'll probably add a few more next week -- the real joy is in watching it move, though.
I'm having problems with the pressure sensors. Part of it is that the clips I put on them keep falling off with the abuse the leg gives the frame... this is not a subtle robot. I'll get them working, eventually. Maybe with the new PC board. I hope to have all the parts at least ordered (if not received) this week for the leg computers. I may have to do this twice... errors are bound to show up, especially since I will be designing on specification only (and not experience or prototypes) for the UART and the NVRAM! Scary, but only a $200 committment.
Oh, I also heard from my machinist... he will try to cut my parts this month soon. He wants me to come visit and do it together, but I can't see how I can. It would be fun, though!
A week or so ago, I got a very interesting sensor that might solve everyone's problem with legs -- determining leg angle. I am currently using rotary potentiometers fastened into the joints; messy, but capable. This new Merlin Stretch Sensor is essentially a long, skinny rubber band that changes resistance as it is stretched. It stretches to about 140% of its starting length, which is tight for cylinders and almost perfect for McKibben muscles. I may do a full technical report later, but I wanted to mention it now.
June. I move in a month. In June I must
cut and weld
the six final legs (I will probably use this leg as the weapons arm). I also want to have the leg computers built and running. And I would like to have my fuzzy language ported to the PIC, for local control. And I have to pack to move. And make arrangements in Austin/SanMarcos to live.
AIGH! My head is going to explode! But the excitement I had with the actual, active working leg makes it all worth while. If you've built a robot, you know what I mean. If you don't -- get off your butt and make something!
July. I move, taking a 10 day vacation. Where I am going, I won't have my shop set up -- so I'm stuck with assembly and software for a month or three, depending on how it goes. I still don't have the $4,000 for the pneumatics, and the special order will take two to four weeks to fill.
August. Robot Wars? I don't think so. And I won't be able to make it, anyway, without a miracle. However, there is talk of alternate events (as I've said before) PLUS once RW is up again, they want semi-anual events. So I am now targeting an early winter 1999 event.
Wed, 10 June 98
Well, my head didn't explode -- but that is because I made a decision on Sunday to change my priorities and plans. Since I don't expect to be operational in August 1998, I don't have to make my legs this month. I decided to PACK instead of BUILD in June; if you had as much junk to pack as I do, you would want the whole month to put it into boxes, too.
We had our Going-Away party on Saturday... where I got to show off my robot leg some more! I know, it's just one leg... but it's still fun. I increased its behavior a bit, getting it to wave around some more. By running it all day I got a lot more ideas on how to improve the control and response. Right now, I am only using the position sensor on the leg (I'll pay more attention to the pressure sensors when I get the boards in), and I'm only operating against the absolution position of the leg to determine what signal to send the valves. I am going to take into account the rate of change of the leg's position, and try to match that to the desired velocity chosen by the Brains. I will also make use of the pressure information, but that will wait a bit... getting fine motion and position control on pneumatic cylinders in such an odd and massy arrangement as these legs is... interesting. But definitely doable.
I did a weight test on the leg! I arranged it so that it would push down on a bathroom scale... but to get a reading I had to lean (hard) on the body framework so it wouldn't lift my desk off of the floor. I think I only got about 80 pounds of push on the scale, but I also wasn't able to keep the body from lifting a few inches... considering that the pushing pressure of a cylinder is roughly half of the holding pressure, Boris will be able to stand just fine. Heck, he might be able to jump. Scary. Really really scary.
Today, I sent off my order for Boris Brain Boards from ExpressPCB... this will set me back about $30 a board for 8 boards, and it's going to be worth it. I also sent out an order for parts to populate the board -- heck, the connectors seem to cost more than most of the electronics. Not counting the PIC and a couple of Maxim parts I am getting through samples, the parts order is around $150. So, if I did count the Maxim parts plus the PIC (plus the board), this is about a $200 to $250 controller. More than I expected, actually. Ah well, the price I pay for custom electronics.
While I am waiting for the electronics to come in, I'll be thinking about my airflow requirements again -- on advice from a colleague (Hi Pawan!), I may be able to reduce my actual pump requirements... to take account of peak-load vs. sustained, the pulse modulation I am using, plus the storage capacity of my tanks. Apparently, the pump size (and hence weight) is a function of the cube of the airflow... makes sense... so this could make a big difference in my weight. I'm also looking at my Robot Control language, and I think I'm going to improve and formalize it a bit more. Just a few changes, and it will be far more powerful and easy-to-use.
If I get my boards and parts in Friday (and I should), I will assemble them and test them this weekend (I hope). I also have a lot of packing to do.
One of the software projects I want to take on in my "spare" time is... moving the fuzzy control language down onto the PIC, as an interpreter. Then, I can store the actual "program" onto external serial RAM. Remember last week, when I was talking about the Dallas NVRAM? Well, this week I found what must be the most incredible, awesome, and amazing form of Non-Volatile RAM in the universe, from Ramtron. This stuff is "FRAM", and is a new type of Ram, I guess. It comes in Serial and Parallel form... and has the 10-year holding power of many other types of NVRam. Except this stuff is rated for BILLIONS of writes, and is REALLY REALLY FAST... well, fast for EEPROM; something like 20nS read and write. Yup. Write the same speed as read, plus non-volatile, PLUS they are friendly. You have to buy their stuff through NuHorizons or Future Electronics... but I'm getting a couple of samples for now.
I was able to spend a few more minutes with the Merlin Stretch Sensor ... not bad. I hooked it into a crude Wheatstone bridge and my oscilloscope, and it indeed gives me a decent signal out. With the correct tuning, you could get an excellent position signal out of this sensor. Some drawbacks... if it jiggles too much, you will get vibration in your signal (though not a lot). Also, it is conductive all along the element, so you have to electrically isolate it from the actuator you are measuring. Probably some tape or paint on the underlying metal would be sufficient, though. I am undecided as to whether I want to use them on Boris or not... I'm thinking I like my knee potentiometers for their simplicity, as long as I can find a way to protect them better.
And finally, a heads-up... I and a few dozen other folks are putting together a new organization. The Society of Robotic Combat -- if you got my letter a few days ago, you already know about it. Watch these pages for updates as it matures!
Tue, 23 June 98
Had some exciting e-mail last week... a group of engineers and/or scientists in a Think Tank are interested in Boris and seem serious about sponsoring me. Never a dull moment!
I received the Boris PC boards and electronics, oh, ages ago. I populated one with discretes and sockets, and have tested out about half of the board by now. The two Maxim3100 UARTs are working fairly well (though I need to bit-bang them; they weren't talking to the onboard SPI port. A serious limitation, with Maxim pointing fingers at Microchip).
I've packed all of my workshop and most of my office by now... and there isn't much more that I can do until I get into Texas. I'll probably test more of my PC Board throughout this week, and I would like to hookup the Leg again to show it off to my 9-year old son who is visiting until I move.
I also broke down and purchased a really nice serial Graphics LCD display from Scott Edward's Electronics. Cost me $200, but it does text and graphics, and will be very useful for a local display on the Boris legs for debugging and diagnostics.
The SORC elections are in process right now -- I've been chosen to count the votes, which I'll be doing either tomorrow or later this week depending on the turnout when I check later.
Mostly, until July 3rd when I leave, I will be packing and cleaning, and generally suffering from pre-moving stress.
Oh! And playing with "Sage", my new Congo African Grey parrot. Got her for father's day. Complicates the move a bit (a fiesty bird, two cats, a tarantula, and a guinea pig... not to mention the two kids), but I think she is worth it.
Wish me luck, and I'll start doing Boris and journal updates again on the other side.
