Invertabot Construction Log


May 13, 2003: Log begins

This is the new robot I'm bulding for Dragon*Con Robot Battles 2003.It will fight in the 12 pound class. It currently has no name and is incomplete. Missing are the batteries, controller, covers and antiwedge skirts. The bot is fully invertable. Everything works the same both upright and inverted. While similar to my previous Dragon*Con bot T-Zero , this bot corrects several design defects in its predecessor. The drive wheels are set wide to provide much better turning torque and the speed has been almost doubled. Antiwedge skirts will help fend off attacks from the rear.

The bot is constructed mainly of .050 6061-T6 Aluminum sheet and is powered by two custom wound RC car motors. The motors were rewound with 150 turns of #30 wire to run off 24 volts. They are coupled to the four wheels with a custom made 20:1 worm gear and 1.5:1 pully system. The 4 inch wheels were made for RC cars.

The main weapon is a pneumatic flipper powered by 150 PSI CO2 from a 12g BB gun CO2 cartridge. A paint ball regulator provides the 750 to 150 PSI pressure drop. I get about 12 "flips" per cartridge.

The Lexan faced boxes on the left and right of the flipper are infrared emitters. The slot in the center of the flipper is an infrared sensor. Hopefully this will be the heart of the semi-autonomous opponent tracker. I intend to mix the left-right targeting signal from this system with the manual joystick control to help steer directly into the opponent. This is the same basic system used on my undefeated autonomous mini-sumo bot Delta Force.

May 17 2003 Progress Report

Several more tasks have been completed. The battery holder tray and Electronics tray have been built and installed. The 20 AA NIMH cells have been glued together forming the 24 volt battery pack. I still need to solder all series connecting straps to the batteries.

At this point the total weight of the bot was becoming a problem. I needed to save about 8 oz. somewhere. Also, the new high flow rate valve controlling the flipper only functioned properly up to about 160 psi. Flipping force seeming about the same as T-Zero at 200 psi because the rate of flow was much faster. Still, I wanted more power in the flipper. I began to wonder if unregulated 750 PSI from the CO2 cartridge could be used directly without any regulator. This would increase the flipper power and reduce the weight of the bot by at least 9.5 oz (weight of regulator, gage and plumbing). Calculations showed that a 0.75 inch bore actuator would provide 330 pounds of force at 750 PSI. Divided by 6 (flipper lever ratio) gives 55 pounds at the flipper tip. This is more than double what I got with the low pressure system. I had a couple of 0.75 inch bore air cylinders. Would they survive 750 PSI? Time to pull out the hydrostatic tester. After filling the tester and the cylinder with water I pumped up to the max on the dial of 1000 PSI. Nothing failed. In fact, nothing at all happend except the actuator rod extended. This was encouraging. Now I needed an electric valve that could handle 750 PSI. I'd previously modified the valve used in T-Zero for greater flow rate by building a new valve body with a larger orifice. Calculations indicated that reducing the orifice by a factor of two should allow the valve to work at 4 times the pressure. eg: 200 PSI to 800 PSI. I was not worried about the valve exploding because it was all machined steel. I proceeded to machine yet another valve body with a 5/64 inch orifice from 303 stainless.

Time to test my creation. Will it leak or fail to open? Lets find out. I hooked it up to a CO2 cylinder and slowly opened the cylinder valve. The plastic hose stiffened up indicating pressure. There were no signs of leaks. With the valve firmly clamped in a bench mounted vise and a shield over my face, I applied 24 VDC to the solenoid coil. SSSSSSSSSSSSSS!!!! The sound of success. It actually worked!

Ultimate test

Using hardware store brass fittings I rigged up connections between the valve and the actuator then installed the assembly in the flipper. I placed the flipper on the floor and placed an old dot matrix printer (weighs 12 pounds) on top. With my face shield on I connected up the CO2 bottle and opened the valve. Keeping a safe distance I touched the solenoid valve wires to a pair of 12 volt batteries. POW! The old printer was tossed into the air and landed upside down! Excellent! Of course the flipper stayed open because there was no way for the gas to escape but I'd fix that later when I made the final version of the plumbing from 2024 aluminum.

Final version of the lower flipper assembly. Between the valve and the actuator is my homemade coupler and bleed orifice. It provides a slow leak to drain all the CO2 from the actuator within 1 second after firing.

Complete flipper in open position

Here's what it looks like with the electronics tray installed and the regulator removed.

So far the flipper seems to be working well. The power is more than enough even without a buffer tank. Nothing has broken yet. I have not done testing to see how many flips per CO2 cartridge I'll get. My guess is 6 to 9. A total of 10 ounces was saved while doubling the flippers power. A win win if there ever was one. The bot still has not been named. One posibility is "ICU" for "Invertable Combat Unit" or "I See You" (reference to it's crude vision system) and it may put your bot in the ICU, Intensive Care Unit. :-)

May 18 2003 Progress Report

I've finally come up with a name I really like. Invertabot. Pretty much says it all.

May 26 2003 - Major milestone

Cover plates have been made and the controller has been wired and mostly debugged. Several wiring errors were found and corrected. Fortunatly I didn't smoke anything during the debug process (came close though!). Since the controller is almost identical to the one in T-Zero I was able to use the same firmware code for initial tests. The bot responds to commands from the 900 mhz radio and drives well. I need to add code to the firmware that does the IR target sensing and possibly debug hardware if problems are found. Several other firmware changes need to be made to optimize it for this motor/voltage combination. I also need to figure out where to put an antenna. Perhaps a "Slot Antenna" would work. This would involve cutting a 1/2 wave length slot in the skin somewhere and connecting the receiver coax to opposite sides of the slot in its center. Unfortunatly the slot might weaken the structure.

Invertabot is now 6 oz overweight! Correcting this will no doubt involve many hours making custom parts from aluminum and drilling lightning holes in other parts.

invertabot with cover on

Invertabot with cover on

Closeup of controller

Homemade controller

May 27

Strange motor control behavior was observed after connecting up shielded cables that go to the flipper mounted IR sensors. The right motor would run in reverse but not forward. In fact, the software/electronic current limiter was being activated to save my mosfets. A dead short was present only in forward mode! I discovered that the right motor had a short between one input connection and its frame. Of course this only started to be a problem when the bots chassis was grounded by the shielded wires going to the IR sensors! I took it apart and found that 4 screws went all the way through the insulated brush assembly and one was contacting the floating ring that allows adjustment of the motor timing! I got out the dremel tool and ground down all 4 screws. I guess I need to do the same to the left motor as well.

Severe radio interference problems discovered! The control range was very short due to RFI from the motors. Putting a clip-on ferrite noise suppressor did nothing. The spectrum analyzer showed wide bandwidth noise going up to over 1 gigahertz. The motors only have a single capacitor connected across the brushes. I added 2 more. I connected a pair of 1500 pf caps from each brush connection to the frame (ground). While I still see lots on broadband noise on the spectrum analyzer, the 900 mhz onboard radio receiver hears a lot less. Range in greatly increased to at least 65 feet. More test are needed but this may be good enough. It's looking like these cheap hand rewound $15 RC car motors kinda suck.

Note: All commands to the robot are in 9600 baud data packets with a 16 bit CRC code on the end. The software in the packet receiver computes the CRC value for the received packet and rejects it if it doesn't match the code on the end of the packet that was computed prior to transmission. If more that 50 milliseconds go by without any good packets received the motors are stopped and the flipper deactivated. This keeps the bot from running amok due to weak signal or interference.

May 28

According to my stop watch Invertabot runs at 5 feet/second. This is a little less than the 6 feet/second I wanted but not too bad. It's 66% faster than T-Zeros 3 feet/second speed. Pushing power seems to be less than T-Zero. The wheels do spin when Invertabot encounters an immoveable object but not very fast. Also the current limiter kicks in at 10 amps and the battery is quickly drained. This is what happens when it's geared for speed, not torque. The flipper weapon should reduce the need for raw pushing power anyway :-).

May 29

I'm now testing the IR opponent detector. I copied the C code from my mini sumo bot Delta Force. The IR LED target illuminator driver code and electronics work fine. The IR LEDs are pulsed at about 2 amps for 100 microseconds every 10 milliseconds (100 times/sec). 2 amps may sound like alot but it's only a 1% duty cycle so the average current is 20 milliamps. The LEDs don't even get warm.

According to the waveforms on my scope the PIN photodiode and amplifier are working correctly. I'm noting one problem - the IR illuminators do not cover the area directly in front of the bot. I hooked up my trusty monochrome TV camera and directly observed the infrared output. Turns out this problem is caused by the metal housing around the LED assembly being too narrow and restricting the horizontal beam spread. I need to rebuild it 1 inch wider. More weight! Arrrg! I'll use .032 aluminum instead of .050. Maybe it will actually be lighter than the original.



Naked IR Illuminator


Enclosed IR Illuminator


IR waveforms. Upper trace: Left then right pulse when target is offset to the right. Note the pulse to the right of the center is higher. Bot should turn right until both pulses are equal.


TV image of bot showing how the IR illuminator shows up

May 31

New wider IR illuminator housings have been made from .032 6061-T6 aluminum and weigh 0.4 ounces less than the originals. This saves 0.8 ounces total and the beam spread is almost enough. I may still need to reposition some of the IR LEDs to get it exactly right.

I finished up the rest of the code to do autonomous steering and it works pretty well. I have 4 buttons to control the auto-steer functions. The left button makes the bot spin counter clockwise until it centers on the target. The right button does the same except spins clockwise. The top button runs the bot forward at max speed until a target is seen to the left or right. The motor speeds are adjusted to direct the bot towards the target. The bottom button runs it in reverse while keeping the nose pointed at the target.

I made some videos of my testing session today. They show Invertabot flipping 4#, 8# and 12# objects, autonomous driving and manual driving. Both RealPlayer and Quicktime formats are available.

Invertabot tests, Quicktime 2.2 meg

Invertabot tests, RealPlayer 3.8 meg

June 1-4: Invertabot goes on a low-iron diet

Invertabot was 6 ounces over the 12 pound limit and was still in need of several anti-wedge skirts and some other minor things. A diet plan was developed. I spent many hours doing things such as machining bolts from aluminum, drilling holes in bronze gears, replacing bronze bearings with UHMW bearings and replacing solid shafts with tubing. So far I've completed the weight reduction in the power transmission units and saved about 5 ounces. Next I need to build an aluminum air cylinder and also rebuild the valve body from aluminum.


steel bolts

Original hardware store steel bolts

aluminum bolts

Homemade 6061-T6 bolts save 0.85 oz

lightned gear

This gear weighs 0.5 oz less due to the 8 holes I drilled


Completed transmission weighs about 2.5 oz less

June 7: Dale builds an air cylinder


In the continuing effort to reduce weight I spent all day Saturday machining an air cylinder from 2024 Aluminum. It's bore is 0.75 inch and the stroke is 1 inch. The actuator shaft is 1/4 inch drill rod that I heat treated to increase hardness. The shaft rides on two bronze bearings in the end cap which is mounted to the body with 8 6-32 screws. The wall thickness is 1/8th inch. Early tests indicate it opens much faster than the commercial air cylinder due to the gas inlet hole being much larger. Overall I saved 2.85 ounces over the commercial version.
Commercial air cylinder

Commercial Air Cylinder

Homemade air cylinder

Homemade air cylinder not only weighs less but is stronger and looks better

Turning on lathe

Turning cylinder body on lathe

Drilling bolt circle

Drilling bolt circle with rotary table

June 9: Buffer tank/coupler

I felt I needed a small CO2 buffer tank and a pressure gage so I set about building a new integrated tank coupler/buffer-tank with gage port from 2024 aluminum. This was not as much of a challenge as the air cylinder but it took 3 trys to get it right. This assembly added an ounce due the the addition of the pressure gage. The 0-1200 psi gage was purchased at PaintBall Atlanta for $15.

Original CO2 tank coupler

Original CO2 tank coupler

Custom tank coupler

New integrated coupler and buffer tank with gage

June 10: Valve body and coil

To save more weight I rebuilt the electric valve body from 2024 aluminum. It took 2 trys to get it right.

Steel valve body

Steel valve body

Aluminum valve body

2024 Aluminum version

Original solenoid coil The solenoid coil assembly on the valve weighed over 7 ounces! This seemed excessive so I looked through my junk and found another coil that looked promising. Unfortunatly it didn't have the proper iron assembly on it. I built my own from 1 inch water pipe and some steel bar stock. The new coil assembly works as good as the original but weighs 3.25 ounces less.
coil parts

New solenoid coil parts

New coil

Fully assembled coil

The weightloss program is now complete and Invertabot weighs in at 11 lbs, 8 ounces. This is a full 14 ounces less than before!

June 14: Paint job

Today Invertabot got a paint job and 2 cool blue LEDs. The paint is "powder coat" wrinkle black. I used the powder coat system from Harbor Freight. This stuff is great. Just spray on the dry powder and bake in an oven at 400 degrees for 10 minutes and it's fully cured. It's also very tough.

The top cover has a Lexan window attached with solid rivets. There's also a small finger sized slot that allows access the the power switch.

I also made new IR LED perf boards taking care to aim of the LEDs to enhance horizontal coverage. I also put in bright blue LEDs for effect.

Painted Invertabot

June 16: Raising and lowering the flipper leading edge

The Robot Battles combat surface consists of four raised carpeted stage platforms pushed together with the seams taped up. The seams constitute a hazard for wedges. The leading edge can get stuck at the seam location. Invertabot would normally have to pass over the seams backwards if not for this little innovation. I used a small RC servo to raise the flippers leading edge when I push a button on the transmitter. Oh yeah, it also works when inverted. This feature was designed in from the start but I finally got the hardware and software working correctly today.

Flipper in normal down position

Flipper in normal position, resting on floor

Flipper in up position

Flipper is raised to pass over seams

June 19: Secret weapon installed and tested


secret Today I installed and tested a new defensive weapon. I've never seen this on any other robot. Due to the advantage it will provide in the Dragon*Con Robot Battles environment, simplicity and ease of duplication, I'm going to delay discussing or showing it here until after the event.
Spur Brake
Edited 9/7/03
Now that Invertabot has won Dragon*Con robot Battles 2003 I'll show you the secret device. It's a "Spur Brake" designed to prevent the bot from driving off the stage. Here is video of the first prototype being tested on T-Zero. Another video shows Invertabot being tested with the Spur Brake attached.
Spur Brake activated

June 26: Voltage Booster

Voltage booster switching power supply Invertabot runs on 20 AA NiMH cells which provide about 24 volts. The motors can turn about 10,000 rpm at no load at this voltage. At 36 volts, in theory, they would turn at 15,000 rpm. Adding 10 more AA NiMH batteries would increase the bots weight by about 8 ounces and over the 12 pound limit.

Fortunatly switching power supply technology has advanced to the point of single chip controllers. With a Linear Technology LT1270 chip and a few other parts I built a 24 to 36 volt boost mode power supply. It weighs 2.8 oz. It's good for about 7 amps. Since the motors only draw high current when under heavy load the microcontroller turns the boost OFF if the load exceeds 6 amps and back on when it drops to less than 4 amps. The motors draw little current when cruising at top speed. It works wonderfully and is transparent to the driver. Invertabots top speed is now 7 feet per second, 41% faster than before. This works out to 4.8 mph.

June 28: Hobby Town USA Club Open House

ARFC table at Hobby Town I took Invertabot to the Club Open House event at Hobby Town USA. Both the Atlanta Hobby Robot Club and Atlanta Robot Fight Club attended. A Model Warship Combat club was also there. They have battles with air powered BB cannons and attempt to sink each other!

I let a couple of other people "borrow" the controller to see how Invertabot drives. Pretty soon TestBot got in on the action and a small robot battle broke out in the back corner of Hobby Town.

Not having the controller in hand, all I could do at that point was watch the action. Invertabot did well considering the drivers had no experience. The anti-skid spikes on the front and rear worked mostly as planned but a couple of times TestBot managed to get Invertabot on top and carry him around. Most of the time the spikes dug into the carpet bringing TestBots pushing to a sudden halt. Invertabot got under TestBot several times and pushed him all the way across the room. I didn't charge the flipper with CO2 for the safety of the many spectators. TestBot eventually had trouble with a loose wheel and the battle was over.

Invertabots motors were pretty hot but the batteries and power mosfets were barely warm. The only damage was a slightly bent rear anti-skid spike. This was a good shake down test.

I picked up some good information from the Model Warship Combat guys. There is a place in Marietta to buy pneumatic components called PHA. It looks like a good source of valves, cylinders, hoses, fittings and other components.If you want to see how a remote controlled BB cannon works click here

August 4 2004: Improvments for upcoming 2004 Dragon*Con Robot Battles

This log entry combines all the modifications I've been making for the past couple of months.

New Motor Controller

New motor controller

First of all I wanted to redesign Invertabots controller using the same technology as Thrasher, my 1 pound ant bot. I wanted to do it on a real PC board so I could quickly make duplicates for future bots. The new controller I designed uses 50 amp, 60 volt MOSFETs. With the small heatsinks I'm currently using it has no problem controlling a pair of 14.4 volt 30,000 rpm size 550 monster RC truck motors.

In addition to controlling motor speed, it also has several other features.

  • High current driver for solenoid valve
  • Servo driver with 5V at 1.6 amps to power it
  • Socket for Linx 900 mhz data receiver (RXM-900-HP3)
  • Connector for optional telemetry transmitter
  • Socket for sensor daughter board which has:
    • IR opponent sensor circuit
    • Inversion detector
    • Battery voltage monitor
    • Connector for optional gyro

Software features include:

  • Exponential speed and steering with dead zone (single stick control)
  • Current limiter
  • Locked anti-phase PWM (28.8 khz)
  • Signed current feedback for constant speed with varying loads
  • Battery voltage compensation for constant speed with varying voltage
  • Several infrared opponent targeting methods selected from remote control
  • Low battery indicator

Note: This controller doesn't work with conventional RC gear. I use my custom 900 mhz transmitter and hacked PS2 controller.
What you see in the photo is the first prototype. There were several ground loop issues and the wrong hole size file was sent to PCB Express. I fixed all the known problems and await rev 2 which should be here on Aug 5th.
Schematics will be made available eventually as "open source".

New Motors and Gears


gears Motor

I got tired of rewinding motors for 24 volts so opted for "stock" 550 size RC monster truck motors. The new controller effortlessly drives these motors. The new motors turn 30,000 RPM at 14 volts. Since the previous rewound size 500 motors turned only about 13 or 14 thousand RPM this new configuration was way too fast for Dragon*Con with the 30:1 transmission and 4 inch tires. I estimate 10 mph! To get a sane speed I had to ditch the 1.5:1 timing pully and replace it with 2.666:1 gearing. There was no room for a larger pully. The overall ratio is now 53:1 and the top speed is very fast, about 9.8 feet/second or 6.6 mph max. Actual running speed is probably somewhere between 5.5 and 6 mph.

New Batteries


Bottom view batteries

The photos above show the new batteries on the bottom side. They are 3300 mAh NiMH 6 cell packs for RC cars. They are connected in series for 14.4 volts and weigh 14.2 oz each. Both battery packs and the CO2 cartrage can be changed out quickly by removing the bottom cover (not shown).

Flipper Solenoid

Switching from 24 to 14 volts to use stock motors created another problem. The flipper solenoid valve operated on 24 volts. Arrrg. The coil I had was totally encapsulated so I had to estimate the wire size and number of turns so I could wind a new one that had the same magentic flux at l2 volts or less. After measuring the inductance, DC resistance and physical size I figured it was about 800 turns of #26 wire. What counts here is ampere turns, ie: amps * turns. I needed the same ampere turns at half the voltage. The 24 volt coil pulled 2.6 amps. 800 * 2.6 = 2080 ampere turns. Turns out all I needed to do was use wire with twice the circular area and wind it into the same space. Since I didn't have any #24 magnet wire I used 2 parallel strands of #26 which equated to #23... close enough. The new coil fires the solenoid at 10 volts or higher and pulls 8 amps at 10 volts. #24 wire would have been closer to my goal but this is good enough. The excessive current is only drawn for 30 milliseconds anyway. One more problem solved.

New IR Opponent Detectors


New IR assy New ir assy outside

The new batteries and motors totaled about 10 ounces more than the old ones. Inverabot had to loose some weight! Part of the weight reduction was achieved with the lighter mounting system shared by both the batteries and controller board and removing the "*Spur Brakes". The rest I got by totally redesigning the infrared "head lights". The originals used 6 IR LEDs pointed into a metal reflector to concentrate the light into a pattern that was about 40 degrees in the horizontal dimension and only about 5 degrees vertical. The assembly was large, heavy and subject to puncture wounds inflected by spinner bot "Storm".

To reduce the size and weight I replaced the reflector with a homemade Lexan lens. I put a 1 inch diameter Lexan rod in the lathe and polished it to the point it was transparent. I cut it in half with a band saw which produced two "D" shaped lenses. I cleaned up the flat surface on the mill and then used the dremel tool to polish it to transparency. The IR LEDs are mounted inside the frame now and .050 6061-T6 aluminum guards are bolted to the outside to protect the lenses. I folded some clear plastic I salvaged from some of that consumer product packaging you can't open without a chain saw into the slots of the lens guards for scratch protection. I saved 4 ounces, increased robustness, and improved the performance a little.

I'm currently having fun playing around with software algorithms for auto-steering into opponents.

* The spur brakes didn't work at Dragon*Con last year because there is a 1 5/8 in steel border around the carpeted surface. They can't dig into steel! A major redesign would have been required to correct the problem. Since Inverabot has outsanding brakes and precise control and I needed to cut weight I decided to remove them.