xGoat
More tea please.

‘Interesting’ USB key bandwidth step

In about a month’s time, Student Robotics will be shipping robotics kit to several schools in the Southampton and Bristol area. Sixth formers from these schools will be competing their robots against each other some time in April next year. The robots are programmed using USB keys. In the past, we’ve tended to purchase the absolute cheapest USB key we possibly could, and we’ve had quite a lot of them fail.

So this year, we’ll be using some USB keys from a known brand. I bought a couple of Kingston 2GB DataTraveler keys and went about abusing one to try to discover if it’s likely to fail. I hacked together some scripts that continuously rewrote the same 10MB space (of the block device that the key presents) over and over again. Out of curiosity I also logged the time it took for this space to be written to each time.

So I ran my scripts and rewrote the 10MB block 2000 times. Here’s what the bandwidth did:


The step is interesting. It could be something to do with the USB key’s wear-levelling. My bet is that the key uses its flash as a big circular buffer (like JFFS). Initially, all the blocks of flash are erased, so no erases have to happen when writing to them. When the end of the circular buffer is reached, it has to erase each block before it can write to it, which drops the bandwidth. Obviously this is pure postulation…

I’m reasonably convinced that this type of USB key won’t fall apart in the robots. The file the students will be loading onto the robot is of the order of 100k. Also I think it quite unlikely that they’ll reach 2000 programming iterations. So I’m for shipping these keys.

The hacky scripts that I used are in a git repo.

Posted at 8:19 pm on Saturday 11th September 2010

It was never supposed to be this way

At the last Student Robotics doing, Jeremy showed me an exceptionally vile piece of hardware. Here, I share with you the pictures that I took to raise general awareness of this sort of abuse.

USB disgustotron


USB disgustotron


USB disgustotron

Posted at 11:55 pm on Monday 9th November 2009
2 Comments

New bus, new possible destinations

Student Robotics hardware development is heating up. The support for the currently shipped kit has almost reached a point where we don’t need to spend all our time on it, and a bunch of people eager to get involved in developing the new stuff have joined the group.

After a summer of racing to get the old kit into a shippable state, the Student Robotics hardware schedule is finally (ok, almost) in line with the rest of the groups agenda! We get to think about the new hardware design at the beginning of the academic year, giving us the maximum amount of time to both get new people involved and develop the new kit.

One of the major targets for the 2011 kit (uh, the 2010 kit is currently shipping — think of it like magazines) is the I2C bus that connects all the peripheral boards together. It’s getting replaced. From almost day one the bus has been misery and pain. If you’re thinking about using I2C as an interconnect between boards on a robot platform, I encourage you to think in a different direction. Due to the capacitance of the bus in our kit, we’ve had to knock the bus down to something like 20KHz. Furthermore, this bus gets along abysmally with the RFI emitted by the robot’s motors, so many transactions get repeated over and over until the checksums align.

So, what’s going to replace our I2C interconnect? I’ve been thinking about this in the back of my mind for a while, and discussion recently began on the mailing list. Right now, we have two options in front of us: asynchronous serial over RS485 or USB. Both of these are differential buses, and both appear that they could serve us well.

USB

The USB option involves the least explaining. We’d stick a USB-to-serial chip (probably a FT232RL) on each board that’d interface with the board’s microcontroller. All the boards would plug into a USB hub, which would end up plugged into the BeagleBoard.

RS485

RS485 feels like it’s been around since the beginning of time. I expect that somewhere there’s a wire-wrapped array of 555 timers (with the requisite heatsinks…) that are doing something they shouldn’t over RS485, or a central heating control system sitting in some large company’s basement brimming with valves rocking out to RS485. However, the fact it’s still around and is still used extensively in many industrial applications is a good thing. RS485 transceivers are cheap! RS485 is compatible with daisy-chaining, which means that one could just plug another board in without having to get hold of any additional hardware (e.g. a hub).

The downside of RS485 having been around for ages is that most of the cheap transceiver chips are 5V supply only. 5V went out with the dinosaurs. All the cool kids use 3.3V and lower these days. Somewhat surprisingly, it looks like it’s cheaper to get a 5V linear regulator and RS485 transceiver to suit, rather than a 3.3V transceiver. Sounds disgusting, but it’s cheaper and performs just as well.

If we went for a master-slave relationship like we had in I2C, then we’d be OK. However, I’ve always been annoyed that the boards connected to our robot can’t generate interrupts of their own. This simple code:

from sr import *
def main():
    yield io.pin[2] == 1
    print "Beans"

Results in the JointIO board being continuously polled over the I2C bus to see if pin 2 has gone high. This sucks in terms of latency. It gets worse when there’s more than one coroutine running on the robot, and latency goes through the roof. So, I’d really like a multi-master bus. One where boards can shout at the BeagleBoard with an interrupt. That interrupt message can contain information about what the interrupt was as well, so the BeagleBoard doesn’t have to ask it, resulting in a nice short interrupt response time.

However, with multiple-masters come collisions: two boards may try to both write to the bus at the same time. There are a variety of methods of dealing with collisions that we could use, and many of these involve things like random-backoff times and detecting whether the line was driven in the way it was supposed to etc. My current favourite is a token-ring like approach, where we add point-to-point connections between boards that are adjacent in the chain, and get them to act like a big shift register with a single ‘1’ bit passing through it. When a board has the ‘1’, a.k.a. “the token”, it’s the master of the bus. When it’s done with the token, it just passes it on to the next board in the line. This way, there are no collisions to deal with, and each board gets a chance to transmit its data.

Which will win?

I’m not sure which will win at the moment. The RS485 option seems safe, and USB could allow for future exciting expansions. I’ll be doing more reading on USB over the next week or two to try and work out how it’ll perform in noisy situations. A possible option is sticking both the USB and the RS485 transceivers on the PCB design, and only populating one of them when it’s time to ship. Depends on how costly the board space would be.


Image used under CC license from flickr user viriyincy.

Posted at 4:26 am on Wednesday 4th November 2009

Site by Robert Spanton. © 2008 - 2011