Ultrasonic System

(click photo for full-size image)

The ultrasonic system is a composed of four Polaroid ultrasonic transducers/ranging modules, an 8-bit microcontroller which supports the I²C serial bus, a few 7405 open-collector inverters, and lots of LEDs. The original goal was to distribute the ultrasonic sensing to multiple controller board so that up to 32 transducers at a time could be added to the system while not bogging down the robot's main CPU. On each controller board, a sequence of DIP switches is used to select the I²C bus address of each board. The ultrasonic transducers/ranging module are the standard off-the-shelf versions available from Polaroid, modified to replace the thin 9-conductor cable with a longer and more standard 2x5 header cable (only 9 pins are used; the tenth hole is plugged to make sure the cables can't be inserted incorrectly). The LEDs give some visual conformation that the sensors are working, and the 7405s are needed since the logic of Polaroid ranging module's ECHO output was apparently designed "upside down" (see the discussion on ECHO logic below).

The controller board drives the ranging module of each board in a sequential fashion. It does this in three stages. First, the module is activated by asserting the INIT and BLNK inputs. This initiates the transmission of a 49KHz pulse by the module. Second, the BLNK input is deasserted to enable the detection circuitry on the module. Normally this circuitry is automatically disabled by the module until 2.38ms after the pulse is transmitted to avoid false ECHO detection, but this also means that objects closer that approximately 15 inches cannot be detected. We've shortened this delay to about 0.75ms, giving us reliable detection down to around 4 inches. Third, the INIT input is deasserted to turn off the module. An external interrupt, caused by a ECHO signal from the active ranging module, is used to determine the time-of-flight of the returning sound pulse. The board stores the most recent time-of-flight readings until the data is requested by the robot. This information is then transmitted over the I²C serial bus.

The original ultrasonic controller board was breadboarded in summer of 1995 for one ultrasonic module at a time. It was re-breadboarded in fall of 1996 to support four modules at a time and was used on Brak through the summer of 1997. These designs used a Philips Semiconductor 8XC751 microcontroller and an RS-232 serial interface to communicate with Brak since there were no other I²C bus boards built at the time. The 8XC751 does not have a built-in USART as do most other 8051 derivatives (the I²C serial bus replaces it) so serial I/O is done in software by bit-banging, giving a transmission speed of about 9600 bps. One wirewrapped version of the circuit was briefly built in the summer of 1997 before I realize it would be cheaper and faster to have the printed circuit boards manufactured. These board were built later that summer and one was used to replaced the breadboarded version.

The version of the board now in use on Brak uses the I²C serial bus for communications. However, difficulty in locating 8XC751 microcontrollers for development caused us to switch to the Microchip Technology PIC16C63 microcontroller in summer of 1998. The devices unfortunately are not pin-compatible with the 8XC751, so some nasty retrofitting has been done on the board to accept the new microcontroller. The PIC16C63 has proved to be easy to work with and rewriting the software to drive the ultrasonics and handle the I²C bus communication was pretty easy.

The final board uses surface mount SOP devices, which are cheaper (the 74HC05 is 20 cents in SOP and 53 cents in DIP). Since 2/3 of the inverters are used to drive the LEDs you can save a few dollars by not using the LEDs and eliminate one 74HC05. But people think the blinking lights are cool, so I'd leave 'em. The SOPs are also loads of fun to solder (yeah, uh-huh). The board's only problem, which is somewhat typical of ultrasonic ranging modules, is the modules require quite a bit of current to drive them and the board occasionally hangs when the CPU's voltage is pulled low by the surging current draws of a firing transducer. This however is really more of a power supply problem than a board problem, but you may have to deal with it.

Polaroid used to sell evaluation kits for their ultrasonics for $99. It's been many years since I bought them so I have no idea what the actual price is now or even if they still sell them. They also would separately sell the transducers and ranging modules in bulk (packs of 10) for something like $150 and $250 respectively, so since Brak will eventually use twelve sensors it's a better deal if you can buy them this way.

ECHO illogic (or perhaps ill logic)

I can't for the life of me understand why the ECHO logic on the board is designed the way it is. At least I can't understand why it wasn't designed the way I want to use it. To explain my frustration, let me first explain some about the internal circuitry of the ECHO output. It is an open-collector output, so if you know what that means you can skip the next paragraph.

An open-collector output is an output which can generate a true logic 0 but not a true logic 1. Most logic devices output logic 0 and logic 1 by driving one of two internal transistors, one which gives the logic 0 and the other the logic 1. One of these transistors is always on, but never both. The logic 1 transistor is removed in an open-collector output, which means that when device tries to output a logic 1 the output pin can do nothing. There is no electric circuit without the logic 1 transistor and with the logic 0 transistor off. This logic level is sometimes referred to as "logic Z". To complete the circuit, an external pull-up resistor to +5VDC is needed to give a logic 1. What is the point of all this, you ask? Well, open-collector outputs can be used to build a "wire-OR" connection by connecting many open-collector outputs together with one pull-up resistor to form a single output. If all devices output a logic Z, the pull-up resistor makes the single output a true logic 1; if any one device outputs a true logic 0, the single output is a true logic 0. If you try doing the same thing with a "normal" output, the combination of logic 0 and logic 1 outputs would cause a small short-circuit.

Anyway, the problem with the ranging module is that each module outputs a logic 0 when it is inactive or waiting to detect an echo; it only outputs a logic Z when it detects an echo. This means the ECHO outputs from each module cannot be wire-ORed together to give a single ECHO signal since at least one module is always inactive. Remember, the wire-OR works when all device are normally outputting logic Z. The ranging modules are just the opposite. Hence the 7405 devices, which contains six open-collector inverters. The ECHO output from each ranging module is inverted to give the correct behavior for a wire-OR design.

Now, this might cause you to ask two other questions: (1) doesn't each ECHO output need a pull-up then before going into the 7405?, and (2) why are you wire-ORing things in the first place? The answer to (1) is "No", since these logic devices will treat an unconnected input as though it were a logic 1. So logic 0 in gives logic 1 out, and logic Z in gives logic 0 out. The answer to (2) could be "Well, the wire-OR seemed like such a neat idea I just needed to try it", but it really goes back to the original design for the ultrasonic controller which supported six ranging modules on a 24-pin Philips 8xC751 microcontroller. The available I/O pins at that point were very precious and without the wire-OR I would need six INIT outputs plus six ECHO inputs plus one BLNK output plus one POWER output, so I was already at 14 I/O pins used. In addition, the ECHO outputs needed to cause an interrupt on the microcontroller and the 8xC751 didn't have six interrupt input pins. This alone was the major motivation for using a wire-OR.

Parts List

Part number	Description	Qty	Price
Polaroid
N/A	Polaroid ultrasonic evaluation kit	2	$99.00
Digikey
PIC16C63-04/SP-ND	Microcontroller, I²C interface	1	$7.73
SN74HC05	Inverter, hex O/C, SOP-14	2	$0.20
Mouser
163-5008	DC power jack, 1.3mm	1	$0.73
154-UL6641	PCB modular jack, RJ11, 6 position/4 contact	4	$0.87
73-XT49S400-20	Crystal, 4MHz, HC-49/US	1	$0.73
571-23825713	IC Socket, DIP-28 0.300" thru-hole	1	$0.19
Jameco Electronics
67811	Header, 5x2 0.100" shrouded	4	$0.29
101856	DIP switch, 3 position	1	$0.25
97851	Resistor pack, 330, one common, 10 pins	1	$0.27
24660	Resistor pack, 4.7K, one common, 10 pins	1	$0.27
28628	Transistor, 2N2222 NPN, TO-92	2	$0.10
34761	LED, green diffused	4	$0.09
34796	LED, red diffused	4	$0.08

Board designs

V1.x Schematic in Adobe PDF (21Kb)
V1.x Schematic in PostScript (115Kb)
OrCAD Schematic files and libraries
Gerber 274-X format

Software (includes source code, project files, and Intel HEX-format files)

Version 1.09 (98/July/28) (18Kb)

Tools (programs you may need)

Microchip Technology MPLAB-IDE for Windows (V3.40 or later)