Posted Mon 20-Aug-2012
Many classic arcade and console video games can be played with a D-pad. While one can use modern controllers with these programs, they lack the elegance of the older designs. However, if one uses one of these older controllers with a USB adapter, one is forced to be tethered to the computer.
Wireless SNES controllers have been done before. However, they usually don't pay close attention to power usage, and none support multiple controllers using a single interface.
The purpose of this project then, is to develop a low latency, ultra-low power device to interface with the original SNES hardware, and connect it to a modern PC. This device must be relatively cheap to manufacture, use components that are currently being manufactured, support multiple devices without interference, and must minimally modify the hardware. Additionally, it should use techniques developed for packet radio to allow for multiple transmitters on the same channel.
It is immediately obvious that one of the most important design decisions is the choice of wireless for use in the controllers. There are several widely available solutions that were considered:
Below is a discussion of the pros and cons of each technology:
38kHz modulated IR light is commonly used for TV remotes. It is well suited for this application: data transmission is one-way, and the transceiver is only energized when actually transmitting. However, the one-way transmission makes it hard to build in support for handling multiple controllers at once.
Bluetooth is an attractive technology, because its widespread use eliminates the need for any kind of specialized receiver hardware. Many PCs have built-in Bluetooth support. Additionally, Bluetooth already has a proven track record in low-power, low-latency devices. The Sony PS3 uses Bluetooth to communicate with its controllers. However, there are some drawbacks to the Bluetooth solution. The Bluetooth HID protocol is simply a wrapper around the USB protocol, and has considerable latency in its inputs. The PS3 requires a custom, low latency hardware stack to meet its deadlines. Also, DIY-friendly Bluetooth solutions are typically rather expensive, and usually are optimized for transmitting data with guaranteed deliveries, something that is unnecessary for the wireless controller.
The pros and cons of ZigBee are very similar to that of Bluetooth, save that a custom receiver would also be required.
These transmitters are extremely cheap, on the order of $2-3 dollars for a receiver/transmitter pair. In fact, this solution has already been used for building wireless SNES Controllers. However, there are several drawbacks to this solution. First of all, the transmitter and receiver are not integrated, and would require two discrete circuits to be included into the SNES Controller, taking up valuable space. Additionally, the extremely low bandwidth (on the order of 1kbps) of the RF link would limit the refresh frequency of the controller, to say nothing of allowing for multiple controllers.
This solution is by far the most promising. This is backed up by real-world evidence. All gaming-quality wireless mice use a custom wireless protocol instead of Bluetooth, as does Microsoft's Xbox 360. 2.4GHz allows for higher bandwidth links with lower power consumption, at the expense of reduced range.
The 2.4GHz solution examined in detail in this document is the Nordic Semiconductor nRF24L01+. This chip is commonly used in industry. For example, it is used in Nike+ Shoes to communicate with iPods, as well as in Microsoft-branded wireless keyboards. In fact, this chip has already been used in some wireless SNES controller designs, however, these designs have not necessarily been executed well.
Unfortunately, the chip is only available in the hard-to-solder QFN package. However, breakout boards are available from 3rd parties such as MDFly for less than $7. This chip has hardware support for Media Access Control and packet checksumming. This makes it extremely easy to transmit data at high data rates, over 1Mbps, without having to worry about handling packet framing in the microcontroller. Additionally, the transceiver supports operating on over 100 different channels in the 2.4GHz spectrum, enabling multiple system to operate in the same environment.
In order to guarantee the successful behavior of the SNES Controller, it is important to make sure that the wireless protocol is robust enough to meet timing deadlines, in the face of interference from multiple transmitters.
The simplest protocol usable that allows from multiple transmitters in the same network is a packetized protocol called Pure ALOHA. In this protocol, all packets are the same length, and all transmitters transmit at the same data rate. This wireless model suits this controller design, as all of the controllers should ideally update at the same rate, and always have the same data to send. Mathematically, we can develop a formula to determine the probability of a successful transmission based on a mathematical analysis of the protocol:
In this equation,
This equation can also be manipulated to solve for any of the variables in the above equation. For instance, assuming a successful transmission rate of 0.95, 4 transmitting controllers, a packet size of 100 bits (The maximum supported by the nRF24L01+), and a bandwidth of 1 Mbps, the maximum refresh rate of the controllers is 64 Hz, which is greater than the 55Hz supported by the original SNES.
There is more to the protocol than just guaranteeing there is enough bandwidth for low latency communication. The protocol also needs to support pairing new controllers. Ideally, the receiver unit would also determine the most noise-free section of the spectrum to transmit on in order to prevent interference from wifi-routers or other controllers. Additionally, it would be nice if metadata, for example, the current charge rate of the controllers, could also be transmitted. These details still need to hashed out.
The hardware design of the SNES Controller has not been completely decided, however, a rough idea has emerged.
A Li-Poly battery will be used to power the SNES controller. The MCP73831T will be used to charge the Li-Poly battery, with the TC54 being used to prevent the battery from being over-discharged. An Atmega328p will be used for the controller brains, due to its ease of use and wide supply range. Also, the TPS79325DBVR will be used as a linear LDO regulator. The choice of the unusual supply voltage for this circuit, (2.5V), is a function of the lithium-polymer battery used in this design. A lithium-polymer battery is considered charged at 4.7 volts, and discharged at 3.0 volts. This straddles the commonly-used voltage of 3.3V. In order to prevent the use of a complex buck/boost converter to stabilize the output at 3.3V, 2.5V was chosen, since it is lower than the lowest discharged voltage of the lithium-polymer device while still being supported by the major ICs (the Atmega328p and the nRF42L01+). A 3.3V LDO regulator could be used, but this would prevent the Lithium-Polymer battery from being discharged past this voltage, which would prevent the entire charge in the battery from being used.
Note that the Receiver design will be similar to the transmitter design, save that it is powered and controlled over USB.
The industrial design of the controller has not been considered much. Ideally, no external switches would be added to the controller, with all wireless control being done through rare button combinations (i.e Start + Select + L + R). However, some external indication will be needed to show that pairing is in progress, or that the controller needs charging. Perhaps this can implemented similarly to the above controllers.