Monday, October 8th 2007 | Ismael Ghalimi
Over the past two weeks, we added many features to our marketing requirements for the Redux Model 1. This gave us a sense for what could be done. Then, we ruthlessly trimmed the feature set down, which allowed us to discover what should be done. Following this process, we researched which alternative chipsets could best support our requirements, and stumbled upon Qualcomm’s upcoming Snapdragon. From the very little information currently available, it looks like we found a perfect match, assuming that we could source the necessary components, and follow an extremely aggressive development schedule.
Qualcomm’s Snapdragon (Cf. datasheet) puts evertyhing we need into a chipset that should be comprised of four chips (assuming that it looks anything like the MSM7600): core CPU and DSP, RF processor, GPS processor, and power manager. The core CPU also includes the video controller, 802.11a/b/g (possibly n as well), and Bluetooth 2.0 + EDR. But what puts the Snapdragon in a category of its own is its performance and direct support for high-resolution displays. From a performance standpoint, its core CPU called Scorpion is clocked at 1GHz and offers 2100 DMips, which is more than twice what Samsung’s S3C6400 offers today. From a display standpoint, it supports XGA resolution directly, and provides 3D graphics acceleration natively. But most importantly, the chips that make the Snapdragon have been designed to work together, therefore integration should be dramatically simplified, compared to the module-based approach described in our Revision Three design candidate.
With that in mind, the biggest challenge in getting there might not be technical but logistical. Indeed, Qualcomm does not seem to work with distributors for its leading edge chipsets, and seems to favor large manufacturers over small ones. Our limited production needs (2,500 for the first batch) might quite simply make it a show stopper. But assuming that we could find a way to work around this problem, here is what a Snapdragon powered device might look like.
At this point, we are pretty much set on using SiMa Systems’ uIP reference design, which supports a patented Multi-Touch/Dual-Force technology, and would allow us to put a touch sensor on both the bottom and top of the device, something that has never been done before. Our Mobile Internet Device (MID) will have a single Home button, and a single DD1 connector for charging, docking, and data exchange. The tablet’s dimensions should be 8.5" x 5.5" (216mm x 140mm), with a thickness of just 0.5" (13mm). The case will be made of machined aluminum, and will be user-serviceable through the use of a single set screw, allowing the latch plate to slide by about 4mm in order to release the case bottom and provide access to internal components.
After much deliberations, we concluded that a standard XGA display (1024 x 768) would be a better option than the W-XGA display (1280 x 768) we considered initially. The reason for this is that most System on a Chip components only support up to 1024 x 1024 resolution, meaning that adding these 256 pixels on the side would require the use of a separate video controller, making the design significantly more complex, possibly reducing overall performances, and certainly increasing power consumption. For a screen that would be about 8" wide, we believe that XGA resolution is good enough, and are currently looking for a suitable LCD module. We also think that a wide screen is less appropriate for supporting dynamic orientation, for it would make the portrait display mode not suitable for very many applications.
One of the most promising innovations incorporated into the Redux Model 1 will be the use of SiMa Systems’ Multi-Touch/Dual-Force touch sensors on both the bottom and top of the device. This will allow the use of ten fingers for supporting innovative gestures, while enabling user-intuitive drag and drop actions. This should also allow us to implement effective ways of supporting copy and paste, which is remarkably absent in Apple’s iPhone. A lot more work is required in this area, but we are fully committed to the technology at this point. Mechanical and electrical details regarding the connection of the touch sensors to the touch controller also remain to be specified.
While we considered getting rid of every single button for our design, pragmatism prevailed, and we decided that we must provide a way to fully turn the device off, while providing a way to turn it back on afterward. This requires the use of one button at the very least. Also, we felt that offering a generic Home button that could be implemented at the operating system level rather than having to be supported by every application at a graphical user interface level would be highly desirable. Therefore, we settled down on the use of a single Home button, to be located next to the screen. The device will also support dynamic screen orientation in order to allow the Home button to be used with either the left or right thumb.
Similarly, we concluded that the device should offer a single connector, to be used for charging, docking, and data exchange. After some further discussions with experts in the field (thank you Dennis), we decided that the use of a magnetic connector would not be appropriate, for several reasons: it is heavy, can create unwanted interferences, and is not practical for safely docking a device of the size and weight we are building. As a result, we decided to adopt a standard connector that could support easy docking, removing the need for expensive research and development at the industrial and mechanical engineering levels. Our choice went to the DD1 connector developed by Japan Aviation Electronics Industry. This connector is used by Apple for the iPod and iPhone, which would allow owners of both iPods, iPhones, and Redux Model 1 devices to reuse existing power adapters and USB connectors. This would also allow us to ship a standard Apple USB Power Adapter for our first production batch, as suggested in this past article. Finally, this would allow use to provide all the ports we need on the docking keyboard, including USB 2.0, audio in, audio out, video out, and serial.
SD Card and SIM Card Slots
Now that our design provides support for cellular connectivity by default, user access to internal components is not required anymore, as it was with a design based on optional and user-replaceable cellular connectivity modules. Therefore, easy access to the SD and SIM cards has become much more desirable. Assuming that manufacturing costs would let us afford the use of aluminum trays for the SD and SIM cards similar to the one used for the iPhone (Cf. iFixit), we will go for such a solution, which would only require drilling of two linear holes into one side of the encasing, and proper placement of connectors on the main PCB. The choice between regular SD and microSD will be dictated by mechanical constraints related to the device’s thickness and the vertical placement of the main PCB inside the enclosure. Tray ejection should be supported by a spring instead of requiring the use of a paper clip as is the case with the iPhone. Should the use of aluminum trays turn out to be prohibitive, the SD and SIM slots should be placed inside the device, without providing easy access to them.
Should we go with the Qualcomm Snapdragon chipset, our main logic board would include just two modules, the touch controller module and the Snapdragon module. It would also embed the accelerometer, ambient light sensor, camera, power manager, backlight LEDs, Home button and DD1 connector (or connectors for cables connected to them), and some of the four antennas required for supporting the device’s wide range of connectivity options (802.11a/b/g, Bluetooth 2.0, cellular, and GPS).
Touch Controller Module
The touch controller will be implemented as a separate module in order to support parallel development with the Snapdragon module and main PCB. The Home button will be used to turn the device on, or to wake it up from sleep mode. As a result, the touch controller will be powered off when the device is turned off or in sleep mode. The touch controller will use its own ultra low power CPU in order to support the embedded processing of simple gestures, thereby offloading the Snapdragon’s CPU.
The Snapdragon Module includes the following components and features:
- Scorpion CPU clocked at 1GHz
- 600MHz low-power, low-leakage DSP
- Video controller supporting up to XGA resolution (1024 x 768)
- Camera controller supporting up to 12 megapixel resolution
- WVGA encoder/decoder
- Digital audio support for MP3, aacPlus and Enhanced aacPlus
- CDMA2000, 1xEV-DO (B, A and 0), HSDPA/HSUPA, WCDMA, GSM/GPRS/EDGE
- 802.11a/b/g (maybe n) and VoIP with VCC
- Bluetooth 2.0 + EDR
According to this article, the Snapdragon’s CPU consumes only 250 to 500 milliwatts.
The Snapdragon supports Qualcomm’s gpsOne hybrid position-location technology, which functions in four different modes of operation. Chosen automatically or specified by software, the four modes are Standalone GPS, Mobile Station (MS)-based, MS-assisted, and MS-assisted/Hybrid. In the A-GPS modes, gpsOne technology utilizes assistance data from a location server in the wireless network in combination with A-GPS circuitry
and software in the wireless device.
Based on the design of existing handheld devices using a Qualcomm chipset (MSM7200) to support the exact same range of connectivity options as well as the gpsOne technology, it should be possible to embed all the device’s antennas into the enclosure itself, without requiring the use of any external antennas. This would not offer the best possible quality of connection, but it would make for a cleaner industrial design.
In order to support dynamic screen orientation, the device must embed one or two 3D accelerometers to be placed on the main CPB. For these, we could use the LIS302DL from STMicroelectronics, the same as the one used for the iPhone.
Ambient Light Sensor
In order to dynamically control the display’s brightness for power management purposes, the device could embed an ambient light sensor like the Intersil EL7900. This component should be considered optional though, for it could be replaced by a simple brightness control widget directly accessible from the Home screen.
As discussed in this previous article, the device will support audio input and output through Bluetooth 2.0 and the DD1 connector only. The Bluetooth 2.0 interface will support EDR for wireless connectivity to Bluetooth mono and stereo headsets. The device will also provide audio support for MP3, AAC, aacPlus, Enhanced aacPlus, AMR-WB/+, Windows Media Audio, and RealNetworks Audio.
In order to support web conferencing, the device might support an embedded camera located next to the screen, opposite to the Home button. This component should be considered optional though, for it could be integrated into the docking keyboard.
The power manager should support the use of multiple Lithium-ion polymer batteries, one embedded into the MID, and up to three provided by the docking keyboard. All major components being integrated into the Snapdragon chipset, the power management chip should be provided by the chipset itself, and will most likely be a variation of Qualcomm’s PM7500. The power manager should support charging through a standard Apple USB Power Adapter.
Since the use of an integrated chipset would significantly reduce the use of real estate by active components on the PCB, off-the-shelf Lithium-ion polymer batteries enclosed in a rigid package should be preferred over batteries enclosed in a soft package. Based on the device’s dimension, it should be possible to embed a battery similar to the one used for the HTC Shift, which offers 2700mAh. Since the device will support instant on operations, the use of LEDs for battery charge indication is not necessary, and can be replaced by graphical indications on the Home screen. If the battery is to be placed between the PCB and the backlight diffuser, custom moldings on the backlight diffuser could be used to secure the battery in place.
The software stack will be based on the Linux operating system (more details there).
At this point, all we need to move forward with this design is the ability to source the Snapdragon chipset from Qualcomm, and access to some RF engineers who could help us with the design of the four antennas that are required. If you know the right people at Qualcomm, and/or know some good engineers, please drop us a line.
Entry filed under: Office 2.0