There is a lot going on right now in the world of solid state disks. Prices are falling, while capacities and IO speeds are constantly climbing. It’s now possible to pick from a good variety of SSDs available from mainstream vendors. I decided to pop for a new 30GB model from OCZ: the 1SLD30G. It is new enough that I couldn’t find any reviews for it. Here is what you need to know.

This drive is capable of belting out some impressive read speeds. Sequential writes seem to be about on par with a mechanical disk. The image below compares the 1SLD30G (left) to a Samsung 320GB mechanical disk (right). The first row in the benchmark shows sequential reading and writing. The next two rows show the performance reading and writing 512kB and 4kB blocks in random locations.

Applications blast open now. Even apps like Photoshop, which used to take 10 seconds or so to start, snap open in just a couple seconds. The machine boots up and shuts down much faster, though I didn’t time it. And it’s kind of weird that it does all of that in silence, without the typical grinding noise of a mechanical disk. Additionally, access time was confirmed at 0.2 ms and CPU utilization was reported as low as 3.2%.

That’s all well and good, but my curiosity had been piqued. I had to pop the drive open and see what control chip was running the show. Low and behold, it is the JMicron JMF602, which is used in many other MLC SSDs. Actually, the OCZ Core V2 drives use the same chip. So it appears that the Solid series is at least very closely related to the Core V2 series. What’s so bad about that?

The problem is that writing a lot of small files can bring the JMF602 to its knees. All it takes is unzipping a file. During an unzipping operation, the computer slows down by a huge amount. I think this has been the reason for so much of the negative hype surrounding the Core V2, and other manufacturer’s drives that use the same chip. Perhaps OCZ is just looking to move more drives by capitalizing on the fact that the average user doesn’t know the Core V2 and Solid series drives use the same controller chip.

Also, keep in mind that motherboard chipsets can have a major impact on SSD speed. My SATA controller is based on the Intel ICH9R. It seemed to handle the SSD like a champ, but many laptop chipsets cripple the read and write speeds of SSDs. Find out what southbridge you have and do some research before buying an SSD.

So what is my final verdict? For my use, this drive is fast most of the time. Yes, it slows down during unzipping and some other activities. And sure, there faster drives, but at a measely $70, the 1SLD30G seems like a good deal. Something to remember is that OCZ has just announced their new Vertex series of SSDs, which although still based on MLC chips, are supposed to have 200 MB/s read and 160 MB/s write capabilities. And the MSRP for the 30G Vertex is $129. So if you have to have an SSD right this minute (I did, heh) the Solid series might be for you. Otherwise I would hold off for the Vertex.

A while back I wrote about a project called OpenServo which basically replaces the guts of normal RC servos with higher performance electronics. I thought this project was really cool, mainly because you could pump a series of position values to the servo and it would fit a Bezier curve to the points for you. The main problem was that the project was setup for analog feedback only - encoders were not supported - which means it would be hard to adapt this project to normal servo motors. It turns out that adding encoder support would be hard because the microcontroller (the ubiquitous ATmega168) was basically tapped in terms of processing power.

I really wanted to add encoder support so that decent motion control could be added to projects for not much cash (hopefully less than $25 per board). So I have taken a first stab at a solution. I’ve picked the LS7366 to handle encoder reading duties. This chip decodes the two encoder phases and updates a 32 bit position value, which is then pumped out over an SPI bus to the ATmega168. This should be a fairly straightforward code change, but I still need to calculate or experiment to see if the SPI comms are as fast or faster than the analog reading they are replacing.

The original OpenServo uses a 50 milliohm shunt to measure motor current. I have replaced that shunt with the ACS712 hall effect current sensor. The reason for this change was two fold. The first reason is to allow currents greater than 2 amps to be measured (the power rating of the 50 mOhm resistor means only 2 amps can safely pass through the resistor). The second reason is just to improve the efficiency of the board, even if only slightly, by not burning any power in a shunt.

So anyway… above you can see my version of the OpenServo. It is almost ready to be sent for manufacturing. Updates will follow. Oh by the way, shipping was $13 from US Digital for 3 LS7366 chips… ouch.

If you are anything like me, you probably subscribe to about fifty sites that have RSS feeds, and therefore you get several hundred news items per day. The ability to get news and updates on exactly the stuff that you care about is great, but there is so much room for improvement. There have been efforts before to algorithmically pick which news items might interest you, but what I have in mind is different.

I would like to see an RSS reader that is smart enough to group news items that are repeated across multiple feeds. There are just too many stories that get picked up by several sites, so it makes sense to collapse the duplicates down and just show one of them. Of course, you could expand the duplicate list if you felt like you needed more detail than what one of the items provided. This could be really difficult to handle though. For instance, the RSS reader would need to be smart enough to distinguish between a site picking up a story much later than everyone else, and a site posting a story that is somewhat different than a story being picked up by many others.

Also, I would love to have the ability to block news items that contain certain phrases. I can’t tell you how tired I got of hype for the 3G iPhone… I don’t own one, and I’m not going to own one. I could really give a damn if PwnageTool works or not or how many people are lined up outside the Apple Store! Similarly, I want the ability to truncate or block really long news items. TechCrunch is horrible about posting insanely long, in depth analyses of sites or people that I don’t care about and just skip over anyway.

So uh, someone make an intelligent RSS reader and give me a cut of the profits.

Here is a project I have been following over at AVRfreaks. A guy known only as Brad has developed an entirely AVR based game console in a surprisingly short amount of time. The console uses one microcontroller for the low-level video stuff, and another micro for sound generation. The “cartridge” can consist of any microcontroller of your choosing so long as it has enough IO lines to interface with the console. A unique RAM sharing setup makes the whole system very fast. I’m thinking about designing PCBs for this project when he releases his schematics.

I must have been living in a cave. I-Wei Huang has been building steam powered RC vehicles for a couple of years now. They are expensive, complicated, impractical, and basically just all around ridiculous. And he’s built lots of them. But there is just something cool about seeing and hearing these contraptions chugging along. Check out some of the videos on his site… this is one thing I definitely won’t be attempting to recreate.

Not quite a year ago, Spark Fun started carrying the first 5 Hz GPS module I can recall seeing. At $99, the price really isn’t that bad considering all of the cool stuff you could do with the faster position updates. I was blown away to find a company called Ohararp that is selling the exact same E-Tek EB-85A GPS module for only $55! It’s not very often you find that big of a swing in prices.

A guy known only as ChaN has got some serious engineering skills. His talents include hardware, software, firmware, DSP, etc, but the thing that always strikes me is his protoyping abilities. He hardly ever makes real PCBs, instead he prefers to use tiny enamel coated wire to make connections between itty-bitty surface mount parts. His latest project is no exception. I think the concept of moving the actual device around to control it is kind of goofy (it’s an MP3 player, not a Wii controller), but the implementation is something to behold. Check out the rest of his site, he’s done tons of cool stuff.

A few weeks ago, I stumbled across this page and signed up for a free MatchPort sample made by Lantronix. This sweet little device has an internal web server and allows you to send RS232 data over B/G wifi networks at up to 921 kB/s. My sample part showed up last Friday and I can imagine all kinds of cool stuff to do with it. I’ll have to finish up my other projects before I do anything with it though…

So I finished up the PCB design, ordered boards, and populated one board. The LM3404 based current regulators work fine, but I haven’t scoped the critical signals yet to make sure there are no serious problems. I need to get an IR receiver (the missing 3 pin device on the bottom right corner of the board) and write the firmware. I think the pushbutton will be used to select different fade effects. I also need to buy some RGB LEDs, currently I am just using white 3 watt Cree LEDs. They are INSANELY bright. Once the firmware is finished up, I will post the Eagle files, gerbers, and firmware for all to enjoy.

This is pretty wild, and I won’t pretend like I completely understand all of what’s going on here, but a company called Flylogic has been tearing down CMOS devices and posting some awesome images of their discoveries. Their latest effort is based on one of my favorite chips, the Atmel ATmega88. They have analyzed many other devices, highlighting possible security flaws, weird design features, and other anomalies. Pretty incredible what you find in a $3.00 chip!