First Week of January 1999
Happy New Year to all our readers. This last week has been a bit traumatic. The new computer card is complete but have been having a few problems with the new software. It all works fine except when walking, where about half way through the walk cycle the the program gets into a loop, which stops it walking, and ignores incoming commands. The watch dog should restart the program, but this is not happening and the program can only be restarted through a hard reset. Despite running the program on a simulator for at least two walk cycles the problem does not reveal itself.
Anyway investigation into why the software is failing proceeds, but on returning to continue the testing, the disc drive on the computer on which we have been developing the software stopped responding. There was then a very frantic few hours in trying to recover the disc drive. We did not have a back up of any of the "Jim Struts" software. The disc drive problem was finally traced to a loose connector. We have spent the last two days catching up on backing up our data. We could not recommend more strongly that you should back up your data...
Needless to say the software has still not been fixed.
The software problem has not been solved and we can not replicate the fault on the simulator so have started looking to see if the problem is in the hardware. During these investigations we have found that the switch mode power supply output has more ripple on its output compared to the original linear power supply. This is causing dither on the Analogue to Digital convertors, this would not crash the program but is not helpful. With the linear PSU the dither was near nil, with the switch mode it is about plus minus 1 bit. Still considering how to solve this as adding various decoupling capacitors dose not seem to cure this!
We have solved the ripple problem by upping the switch mode PSU volts and feeding it into a linear PSU, this has reduced the dither. This arrangement still gives a reasonable efficiency. The card now only draws 100 mA, a good reduction from the original 1 amp. This now gives the possibility to run the hydraulics and computer card on three packs of racing car Nicads, this would give a weight saving of about 800 gms. During the reworking of the power supply we had a minor explosion where a electrolytic capacitor was installed the wrong way round, there was a big bang and a three foot stream of gas squirting out the top of the case for a few seconds.
Finally got the card to run the walk sequence without crashing. It was caused by too smaller resistor between the CPU and the MOSFETS. Must of been getting spikes or volts below the CPU ground coming back from the MOSFETS, locking up the CPU. So the problem was not the software, but the hunt to find problems in the software did reveal some minor bugs.
There is now a big sigh of relief as we should now be able to see if the walking sequence for Miss Struts can be made to work before we embark on her construction.
We have received our contracts from "Mentorn" for series three, a bit belated due to local problems at the post office. Have ordered some interesting materials to construct some legs.
Now that we have a working CPU card we have been trying some synchronised leg walking. Err.. walking backwards when asked to walk forward does not seem correct. It has not been a success, if we can not get this to work there will be no point in building 'Miss Struts'. We added some code to move the legs to their centre positions. When this code is run the legs are not consistent in where they park by up to 10 cm, several inches. Ummmm...