The hardware design makes perfect sense. The software leaves a lot to be desired.
I usually try out every product that I tear down. Sometimes I find that the product is useful, and sometimes not. The subject of this month's Tear Down, the Nike SDM Triax Elite training system falls into the latter category. And that's unfortunate because I really wanted to like this product. Let me explain.
If you're not familiar with the Nike SDM Triax Elite, it's a four-piece set that you use/wear while walking or running. As shown below, it consists of a watch, a piece that attaches to your shoe, a piece that straps to your chest, and a piece that connects to your PC. The shoe and chest attachments give real-time responses to the watch through a wireless connection, so you can tell how far you've gone and how fast, what your heart rate is, and so on. Then when you get home, you download that workout information wirelessly to your PC to log the workout.
In theory, that sounds great. The only problem is that I couldn't get it to work properly. The hardware seemed to do it's job flawlessly, but the user interface designed into the software just wouldn't let me do simple operations to track my workouts. I consider myself a fairly astute user, so I assume if I didn't have the patience to figure out how to get it to work, neither would lots of other people.
Now I'll get off my soapbox and describe the hardware design, because from that perspective, it's an excellent design. The product was designed for Nike by Dynastream, which is gaining a great reputation for products of this sort. The SDM Triax Elite was actually the first in a series of products from Nike/Dynastream.
The design starts with a Texas Instruments MSP430 microcontroller on each of the four components. Oddly, it's a different 430 in each instance. That tells me that the designers did their homework and got just the right part for the application.
As you can see from Figure 1, the watch contained an MSP430F135 REV N. I thought it was peculiar that the revisions go all the way up to N, but I was informed that such occurrences are not uncommon. In the case of this MCU, most of those revisions were done internally at TI, so designers likely never knew about them. Only three revisions were made public.
According to Adrian Valenzuela, TI's MSP430 product marketing engineer, “It's not uncommon for a device to release at Rev. J. On these MCUs, the things that generally have the most issues are the serial communications interfaces, where it might have some timing skew, or there might be a bug in the timers. For example, maybe it misses a count or it just doesn't work as expected. In these early 13X and 14X devices, the ADC (analog-to-digital converter) had a number of issues. This is an older device, and that adds to the number of revisions because there's been so much time to find all these bugs. That's typical for the whole industry, not just us.”
The 135 is now considered one of the older parts in the 430 family, but that wasn't the case when the Nike SDM Triax Elite was designed (in 2004). The MCU contains 16-bytes of flash memory, 512 bytes of RAM, two 16-bit timers, and a communications interface that can be configured as a UART or SPI. There's also an on-chip comparator and a 12-bit ADC.
Another component on the watch board is a Microchip serial EEPROM (the 24AA128). It's a 128-kbit part that communicates over I2 C. The wireless communications are handled by a hybrid transceiver, the TX1000, that operates at 916.50 MHz . It was designed by RF Micro Devices (RFMD).
There's a device on the board that I could not determine, as it's covered by a protective plastic insulation. By doing some deduction, I believe the part under the black “goo” is an LCD controller, as that function isn't built into this particular MCU. The reason for the insulation could be so the signals don't get any interference from the wireless signals transmitted by the transceiver. The other possibility is that an unpackaged die was used, which is fairly common in very high-volume applications. The insulation would protect the die and its associated traces.
The piece that attaches to the shoe obviously required some careful design. In addition to two key components, an MCU and a transceiver, it holds a pair of accelerometers to measure the runner's steps. These parts (ADXL321) hail from Analog Devices. With the two accelerometers, a three-axis measurement system can be formed. That's combined with Dynastream's patented algorithms, which interpret the accelerometer signals to derive the runner's speed and distance traveled. ADI claims that up to 97% accuracy can be achieved.
As you can see from Figure 2, the MCU in the shoe attachment is an MSP430F149, Rev. O. The 149 is a higher-end part compared with the 135 discussed earlier. It contains 60 kbytes of flash, 2 kbytes of RAM, and more timers. In addition, it has two serial communication interfaces and a multiplier. The multiplier is likely used for some sort of filtering of the data before it's transmitted to the watch. That transmission is handled by the RFMD TR6000 916.50-MHz transmitter.
The next generation of this product (although I'm not sure if there were iterations in between) doesn't use an accelerometer (see the sidebar, “Next generation cuts cost by a huge margin,” ). It uses a lower-cost hybrid approach to count steps.
|Next generation cuts cost by a huge margin
Although I was impressed with the hardware design of the Nike SDM Triax Elite training system, I felt that it fell short in its software design. But the designers have certainly outdone themselves–both in terms of hardware and software–in future generations. In fact, I purchased the Nike+iPod Sport Kit after using the original product (see a complete Tear Down of the Nike+iPod Sport Kit on EETimes* ). And I was quite impressed with all aspects.
First and foremost, the product is cheap, just $29. Nike gets to that price point by using the processing power of the iPod nano. Yes, it only works with a nano, so factor that into the cost if you don't already own one. I happened to have a nano in my arsenal, so I just had to lay out the addition $29.
If you're familiar with the operation of an iPod, getting the Nike+iPod Sport Kit to work is a breeze. It worked right the first time out of the box. The measurements were initially off by about 8%, but it was easy to calibrate the device to get precise measurements.
So how do they go from $300 to $29? Understand that the functionality in the Nike+iPod Sport Kit is far less than the SDM Triax Elite. However, it did just what I was looking for, calculating distance, time, and pace for my run. And through the iPod, which is always attached to me when I'm running anyway, it gives you voice prompts as to how far you've run, and how far you still have to go.
The biggest cost cutting comes from the elimination of the watch, which is obviously the most expensive piece of the SDM Triax Elite. Another cost cutting comes by eliminating the two accelerometers. An impressive setup was designed as a replacement. The Dynastream designers came up with something that operates like the inverse of a miniature speaker. When your foot hits the floor, it causes a voltage on that speaker, counting it as one step. It's a poor man's accelerometer, but it's pretty clever, and as far as I'm concerned, accurate enough for my purposes.
Another difference between generations is the incorporation of Nordic radios to handle the transmitting and receiving of data. That saved some money as well.
The MCU in the chest monitor is the simplest of the 430 family, designated the 430F1101A. It's pictured in Figure 3. In a 20-pin package, it holds 1 kbyte of flash memory and no dedicated ADC. It has a comparator, a 16-bit timer, and a slope ADC.
Valenzuela comments, “You don't get solid ADC performance, so it's interesting that they used this part for the heart monitor.”
The simplicity comes from the fact that the component simply sends the pulses to the watch, where the processing occurs. The RFMD transmitter on the heart monitor is the T6000. It operates at the same 916.50-MHz frequency.
The final piece of the product, shown in Figure 4, is the part that connects to the PC. It does its thinking with an MSP430F147, a cousin of the 149. The only difference between the two is the flash size, 32 kbytes compared to the 149's 60 kbytes. Otherwise, they're exactly the same. This piece also contains an RFMD TX1000 transceiver (same as in the watch). The serial-to-USB converter is produced by FTDI. This IC provides a dedicated function, taking the serial data from the watch and converting it to a form that can be read by the PC through the USB port. This particular FTDI part is quite common in designs like this one. It handles all the required functionality with very few external components.
Richard Nass is editor in chief of Embedded Systems Design magazine and editorial director of the Embedded Systems Conference. He can be reached at firstname.lastname@example.org.