Archive for the ‘components’ Category

Group import LED matrices?

Friday, May 20th, 2011

If you love LEDs as much as I do, you might find this blog post interesting. From the comments:

The P4 (the 64*32 indoor one, 4mm pitch made of 8*8 elements) costs around 160 dollars and the p6 (32*16 outdoor one, 6mm pitch made of SMD leds) around 40 , and it is just for fun! Note that I’m also able to get 32*32 6mm pitch SMD led matrices for around 70

Oooooh, seriously tempting :D I’ve been wanting these things for years, but always stayed away due to the high pricetag. But this thing changes that. Here’s a quick comparison. You might have heard of those 8×8 LEDsee RGB modules for 7 euro (or 10 dollar or so). Those are the cheapest RGB modules I’ve run across available to mere mortals. Now imagine a 128×64 matrix. That’s 128 modules aka 896 euro! With the group import you’d need 4 panels to get the same resolution. Smaller dimensions, but at roughly 460 euro. And a crapload less work because the factory has already done all the LED soldering for you :D


Thursday, May 12th, 2011

Whoa, looking forward to giving these big ass LEDs a testdrive. 10mm diffused RGB and the first testcase for the LED drivers. The 3mm red one and 5mm strawhat are a quick size comparison to give you an idea how big 10mm actually is (personally I never saw 10mm ones before irl).  Gotta love Chinese manufacturers :D

TLC5940 driver boards

Wednesday, May 11th, 2011

So with all those LED drivers lying around it’s time to do something fun with them. Given the amount of chips a PCB certainly wouldn’t be overkill so I whipped one up. It’s the max size Eagle freeware allows, but everything fits nicely. Of course using SMT would enable a smaller PCB, but since the drivers are already present, that’s not an option. The TLC5940 comes in 3 packages; SPDIP, SSOP with PowerPAD and QFN.  Soldering PowerPAD packages is a pain because you need to put solder underneat and adapt the PCB to lead away heat. QFN packages have no leads, which is also a pain for a hobbyist. SPDIP is the only one easily used.

Each board has 3 TLC5040 chips and data is clocked in as 8x8x8 (aka RR…RGG…GBB…B). The board is designed to be Daisy chained, including a return path. The return path may offer possibilities to automatically detect the size of the array via the status information bits. It would be nice to have RGB…RGB, but 16 channels per chip isn’t really dividable by 3. Same problem arose when I thought up smaller PCBs, e.g. use 1 driver per PCB.

A potential problem with these big boards is that they’re harder to build into a system which has small dots. The jury is still out on what to use as a light source. Options are (large) regular LED’s, LED strips cut up into smaller pieces, etc. The TLC5940 can drive up to 120 mA, so that allows for a range of options. Using separate LEDs requires a huge amount of hole drilling, so for the time being I much prefer using LED strips. By cutting it into pieces holding 3 LEDs, oval shapes (in milk coloured acrylic) should be easy to accomplish.

One thing I greatly underestimated was the cost of connectors. Usually only a couple are needed, but in this case I’ve got enough drivers to make 15 boards. Each board has 16 RGB channels, which makes a grand total of 240 connectors. The initial idea was to use polarised connectors, but even low cost connectors from Thailand or China drove up the price too much for my taste. To give you an idea, the male part alone costs $140 at Mouser’s! So I settled on using standard angled pin headers and cheap Molex C-grid like connectors. Should work fine, just have to make sure that during hookup the connection is not reversed.

To be continued….

What to do with these?

Sunday, May 8th, 2011

Imagine you’ve got a bunch of  TLC5940 ICs lying in front of you and you’re thinking “What can I possibly make with these….”.

Ooooooh the possibilities >:-)

US export regulations

Wednesday, April 20th, 2011

Going to play around with a Coolrunner-II CPLD so I thought it would be a good idea to get Xilinx’ free ISE WebPACK. Guess what, that was a bad idea. The US wants my home address and phone number:

US export regulations require that your First Name, Last Name, Company Name and Shipping Address be verified before Xilinx can fulfill your request. Please provide accurate and complete information for immediate processing. Sorry, addresses with Post Office Boxes and names/addresses with Non-Roman Characters with accents such as grave, tilde or colon are not supported by US export compliance systems.

Fucking nazis…. No way I’ll give them the requested info. Hell they already monitor my bank transactions (do they really think any half decent terrorist uses the ‘system’ and risk a big fat red flag behind their name!?). Filling in bogus info does the trick just fine ;)

On a related matter, Mouser does not sell Xilinx products. Looks like an alternative CPLD is needed for any future development.

Oh crap…TI acquires National Semiconductor

Tuesday, April 5th, 2011

Another mega company and less market competition, meaning we as developers are going to get screwed anytime now… bah!

Btw, the mail they send contained some pretty ‘funny’ lines:

I want to reinforce TI’s commitment to you, our customer,  as we merge our two companies.

Sure, keep believing that. Your commitment is to your own financial benefit, or you wouldn’t have merged in the first place. You know it, we know it, so just be honest and say management is greedy. Just stop feeding us this corporate bullshit.

National’s 12,000  products plus TI’s 30,000 means more performance, power and packaging options when selecting the right ICs for your application.

Euhm how does that chance anything? By simply adding the other’s productline to your own does not magically bring new more powerful products into existence. Would be a neat trick though.

Our combined sales and applications team of 2,500 will be larger than any in the industry so we can provide more customers with greater face-to-face support than ever before.

HAHAHAHAHA, in which parallel universe!? Give me 1 example of a big company getting even bigger where customer experience an actual improvement.

TI’s fabs and National’s available capacity can enable higher production levels.

Again, how does this acquisition increase something? You could move products from one productionline to one at another facility, but that will not increase you production level. Also I must point out that I love the EvalBot stunt they pulled, but it took 3 months to be produced. Suffice it to say I have my doubts when it comes to production levels (this is not a TI/National specific issue, just head over to Mouser, Digikey, etc and look at some of the lead times)

Ok, enough for today ;)

Using a SPI DataFlash

Sunday, December 19th, 2010

Memory chips used to be a pain with all their address and data lines, but no more. Today I hooked up the 32 Mbit DataFlash (an Atmel AT25DF321A in case you skipped my previous post) to the Bus Pirate and I can barely believe how easy it was. Here’s a shot of the entire setup:

All you need to do it hook up the SPI lines and apply 3,3V power. As a first test requesting it’s manufacturer and device ID seemed logical:

SPI>[0x9f r:4]
READ: 0x1F 0x47 0x01 0x00

Only 4 bytes are returned, but it indicates the chip is working as it should be. The first byte (0x1F) means Atmel is the manufacturer. The 2nd and 3rd byte hold device information and exactly match the datasheet (AT25DF series with 32-Mbit density). The last byte is optional according the JEDEC standard and appears to be unused. Well this confirmed things were working, so next would be reading and writing data.


Upgrading musicplayer’s storage capabilities

Saturday, December 18th, 2010

Something that’s been bothering me is the limited storage capabilities of SD cards (from a FLAC player’s perspective). According the SDHC spec 32GB is the max. Beyond that SDXC must be used which, as far as I know, is not supported by currently available microcontrollers such as the STM32. Offcourse 2 SDHC cards could be used to double the capacity, but I would like to try a different route. Also there’s another reason why I’m not adding more card slots, but we’ll get to that another time (offloading FAT handling to a different chip).

Using a harddisk was out of the question due to the noise it makes (aka the moving parts), although the massive storage capabilities are nice. For a while I’ve been drooling over SSD’s, but didn’t pursue that option because prices needed to come down first. Recently prices dropped under €100 for an OCZ 64GB SSD, so time to explore that option again (I consider that price the max I want to spend on this and the capacity a minimum for FLAC storage). Funny thing is that SD cards are actually more expensive than SSD looking at cost per GB and on top of that SSD is much faster and can, also be used in your desktop pc/server/seedbox/etc if you ever want to upgrade to a bigger drive.
Anyway, since IDE is pretty much dead these days, SATA is left as the only option. Directly interfacing SATA devices is a bit of a pain, so the 2nd best option seems to be using an USB adapter. Last month I already looked at a cheap USB SATA bridge. These adapters are easy to obtain and use so I’m going to ignore their use in this post.

Sooo this leaves us with the problem of interfacing USB. Thankfully there are a lot more viable options for this. We’ve got various controllers with an integrated USB interface and there are various separate solutions. The latter is what I’m going to experiment with.

If you’re reading this, chances are you’re familiar with FTDI Chip. Until now I’ve only used their USB device devices, but that’s about to change.  On my desk lies a V2DIP2-32 module. This module has FTDI’s Vinculum II chip and offers USB host functionality. Nice extra is that the toolchain offers bulk mass storage with FAT32 support (see page 324 of the Vinculum II User Guide).

What does this mean for the musicplayer? Well, simply put the idea is to replace the SD card using SDIO with this module and connect it to the STM32 using SPI or 8-bit FIFO. Speed-wise it both interfaces should work, although actual numbers are a bit sketchy since FTDI only provides numbers for the 1st generation devices transferring data to a monitor port. We’ll see how this plays out.

Another addition is an Atmel 32Mbit DataFlash. It’ll mainly be used for storing trackinfo. Given the large file storage and poor embedded FAT32 performance this should be a useful addition. Will do some quick testing soon.

Don’t order components while sleeping…

Thursday, December 16th, 2010

… or the right one is what you get, while you were expecting something sized like the left one. LOL

STM32 design contest

Sunday, November 21st, 2010

STMicroelectroics distributor EBV Elektronic is hosting a STM32 design contest and sends you a STM32 Discovery Kit for free if you register to participate. Mine arrived yesterday and is a welcome dev board for testing purposes :)

Offcourse I entered the contest as well. The project is called FLACie and I intend to get the official FLAC decoder running on the Cortex-M3 core. There is a port done for the Rockbox project of the FFmpeg decoder, but it’s old (5+ years and from before the reference decoder was optimised) and only works on ARM7TDMI controllers. I’m talking about a port based on the reference decoder and optimised to run on Cortex-M3 controllers. General idea is that if it runs on ARM7TDMI cores it must definitely be possible on Cortex-M3 cores.

If this succeeds the VS1053B decoder can be taken out of the musicplayer design. And since that component is the most expensive one it should have a nice impact on the total component cost. Besides it would be a kicker if this actually works!

To be continued…

On a sidenote; remember my rant about shipping costs? The dev board was send through Deutsche Post from Germany using a normal postage stamp. Thankfully no UPS, DHL or any other expensive customer unfriendly courier. Props to EBV for using the postal system as it’s supposed to be used!