 |
|
 |
| Parallax Forums > Public Forums > Propeller Chip > Secure digital read/write code released | Forum Quick Jump
|
| 
rokicki
Registered Member
      
Date Joined Feb 2006
Total Posts : 876
| Posted 12/26/2006 11:02 PM (GMT -8) |   | {{
fsrw 1.0 by Radical Eye Software
This is secure digital and FAT16 read/write code.
For connectivity, see "sdspifasm.spin". You only need four
Propeller lines and six pullup resistors and some sort of SD or
microSD or miniSD socket.
We only support the root directory at the moment, and only one
file open at a time. The file can be open for read, write, or
append. We also support delete (open a file in mode "d"), as well
as a traversal of the root directory.
No long file name support. No (real) date support yet. No
formatting utility yet.
I've tested it on 10 different secure digital cards, and it works
on all of them. There is some variance in the speed. For reads,
I see speeds of between 271 to 304 KBytes/sec. For writes, I see
speeds of between 114 to 228 KBytes/sec, depending on the card.
This is with full file system overhead, and the file system code
is completely SPIN. If you use only the block I/O layer you can do
better than this; if you write small files or flush frequently you
will see worse than this.
I am supplying three different block level implementations. You
should start with the "sdspi" code which is slow and completely in
spin. If that works, move up to "sdspiasm" and then "sdspifasm"
which increases the frequencies the card is driven at. All cards
should support the fastest mode, but you have to make sure your
wires are not terribly long and stuff like that. You change the
block layer used in the object section of the "fsrw.spin" source.
I've tested this code on cards in their virgin (out of the bubble
pack) state, and when formatted with XP, and when formatted with
Win2K, and when formatted with the Canon XTi camera, and when
formatted with the Canon SD400 camera. No problems in any case.
This code will *only* read FAT16, not FAT12 and not FAT32. This means
the number of clusters on your volume must be greater than 4096 and
less than 65530 (or thereabouts). Clusters can range from 512 bytes to
32K (they must be a power of two in size), so this means we can support
cards from 4M up to 2G.
When you format your card, you can use XP to do it, but you have to pick
a cluster size that will make the card end up in FAT16 (which means the
cluster count must satisfy the constraints above). For performance, the
larger the cluster size, the better (since fewer metadata writes and
reads are required). This does introduce some additional space loss due
to fragmentation, but with the large cards available now, this should not
be a problem.
The following table shows the range of cluster sizes you can select in
XP when formatting in order to guarantee a FAT16 volume. The number
of clusters will be the size of the card divided by the cluster size
(approximately).
min max
4M 512 512
8M 512 1K
16M 512 2K
32M 512 4K
64M 1K 8K
128M 2K 16K
256M 4K 32K
512M 8K 32K
1G 16K 32K
2G 32K 32K
If you do not select the right cluster size, your volume may be formatted
in FAT12 or FAT32, neither of which is supported by this code.
FAT16 does not support volumes greater than 4G. If you must use a 4G
card, you will have to ensure the first volume is 2G or less, and this
code will only mount that first volume. In reality this is not much of
an issue, because this code only supports the SD protocol, and cards
of 4G or larger require the new SDHC protocol. (I have no way at this
time to know if an SDHC card will default to the SD protocol for the
first portion of the card.)
}}
OBJ
fsrw: "fsrw" ' the main FAT16 reading and writing code
sdspi: "sdspi" ' low-level SD block I/O in spin (slow)
sdspiasm: "sdspiasm" ' low-level SD block I/O in simple asm
sdspifasm: "sdspifasm" ' low-level SD block I/O in faster asm
fsrw_speed: "fsrw_speed" ' a test of speed
sdrw_test: "sdrw_test" ' a simple example of use
I have removed the file attachment, and instead am pointing everyone
to use this URL:
http://tomas.rokicki.com/fat16work/
which will always have the latest and greatest as well as status.Post Edited (rokicki) : 12/29/2006 6:40:36 PM GMT | | Back to Top | | |
 |
Mike Cook
Registered Member
Date Joined Jul 2004
Total Posts : 678
| Posted 12/27/2006 12:19 AM (GMT -8) | | |
FYI,
If you like to roll you own, I used this SD socket from www.mouser.com
Mounted this socket on a little perf board with an 11 pin header so it will plug into a bread board, attached is a couple of photos and the data sheet for the SD socket from mouser.
' Assumes SD card is interfaced using four consecutive Propeller
' pins, as follows (assuming the base pin is pin 0):
' 3.3v
'
' 20k SD CARD PIN-OUT
' p0 ──────┻─┼─┼─┼─┼─┼────── do (PIN-7) DO
' p1 ────────┻─┼─┼─┼─┼────── clk (PIN-5) CLK
' p2 ──────────┻─┼─┼─┼────── di (PIN-2) DI
' p3 ────────────┻─┼─┼────── cs (dat3) (PIN-1) CS
' └─┼────── irq (dat1) (PIN-8) DAT1
' └────── p9 (dat2) (PIN-9) DAT2
'
' (PIN-6) GND
' (PIN-3) GND
' (PIN-4) +3.3
SD CARD Socket Pin-out:
PIN SD CARD Propeller
-----------------------------------------------------
1 (NC)
2 (PIN-9) DAT2 Pull-up
3 (PIN-1) CS Pull-up P3
4 (PIN-2) DI Pull-up P2
5 (PIN-3) GND GND
6 (PIN-4) +3.3 VCC
7 (PIN-5) CLK PUll-up P1
8 (PIN-6) GND GND
9 (PIN-7) DO Pull-up P0
10 (PIN-8) DAT1 Pull-up
11 (CD SW) GREAT WORK ROKICKI!!
Mike Post Edited (Mike Cook) : 12/27/2006 11:36:12 AM GMT Image Attachment :
000_0673.jpg 244KB (image/pjpeg)This image has been viewed 641 time(s). | Image Attachment :
000_0674.jpg 126KB (image/pjpeg)This image has been viewed 649 time(s). | Image Attachment :
100_1106.jpg 1.12Mb (image/pjpeg)This image has been viewed 984 time(s). | Image Attachment :
SD-DATA.JPG 50KB (image/pjpeg)This image has been viewed 904 time(s). |
File Attachment : SCDA1A0901.pdf 113KB (application/pdf) This file has been downloaded 387 time(s). | | Back to Top | | |
|
rokicki
Registered Member
Date Joined Feb 2006
Total Posts : 876
| Posted 12/27/2006 8:54 AM (GMT -8) | | Thanks! I should definitely mention that Mike Cook provided a lot of help in testing the driver and making
suggestions for its improvement (not all of which I've implemented yet).
There are some sneaky counter tricks in the assembly code if anyone is interested. (I use PHSB as a
shift register to shift data out with FRQB=0 in NCO mode, and I use counter A to generate the clock
in several different ways.)
I can make the assembly code faster if anyone needs more speed; I have some preliminary versions
but I'm not yet happy with them so I'm holding them back for the moment.
Here's a picture of the secure digital cards I've tested:
I tend to buy the cheapest cards; it's quite possible more expensive cards would also be faster.
I will also note that I have been trying to wear out a card (so far unsuccessfully) by doing millions of
random writes (random block writes should be the worst case for one of these flash cards). | | Back to Top | | |
|
rokicki
Registered Member
Date Joined Feb 2006
Total Posts : 876
| Posted 12/27/2006 10:40 AM (GMT -8) | | My reading of the Sandisk specification (referenced from the Wikipedia article) is that it is the host controller's responsibility
to pull up all lines that might be in an input state with a resistor between 10K and 100K. Floating inputs on CMOS chips can
cause all sorts of nastiness.
(The propeller itself seems to deal with unconnected inputs reasonably well; anyone know how? Are there tiny pullups or
pulldowns we don't know about? Or is it possible we need to pull up all the unused propeller lines too, or set them to output?)
I would strongly recommend spending the $15 and getting the breakout board from SparkFun; the time it will save you will
more than pay for itself. (I just soldered on a male header so I can plug the breakout board directly into a breadboard.)
For my main project, using the Propeller protoboard which isn't out yet, I will probably go with a microSD socket breakout
board since it is physically smaller and microSD cards are pretty cheap.
SD cards are so commoditized now; the smallest I can find is 512MB, and they go for about $10 to $15 a pop. I could add
MMC support, but I don't want to make the code any larger when SD cards are all available so cheaply. This is the same
reason I only support FAT16; since we can format a card in FAT16, there's no sense wasting code space on FAT12 or FAT32. | | Back to Top | | |
|
David B
Registered Member
Date Joined Jul 2004
Total Posts : 337
| Posted 12/27/2006 11:08 AM (GMT -8) | | Looks great, Rokicki. Much cleaner than my WAV player SD code.
Have you tried the 4 bit mode IO? I'm wondering if that's another way to get more speed out of these cards.
David | | Back to Top | | |
|
Mike Cook
Registered Member
Date Joined Jul 2004
Total Posts : 678
| Posted 12/27/2006 12:39 PM (GMT -8) | | |
Originator,
In Rokicki's archive there is a simple example, you might want to start with that.
This is what I've been doing with this and I'm not proud of it, but I'll post it anyway. This archive is a 'work in progress' for testing several ideas, with the SD Routines.
What it is, is a data logger, it uses:
1. DS1307 to get time
2. DS1820 to get temp
3. Then writes this mess to a file
Right now it's configured to run for 10 hours. WARNING I have slightly modified TV_TEXT.spin to work similar to the FullDuplexSerial.spin, this way I can easily switch back and forth from the TV and a PC running a terminal program.
If it's too confusing, please delete it! 
Mike Post Edited (Mike Cook) : 12/27/2006 8:43:14 PM GMT
File Attachment :
_MJC_sdrw_test_TimeAndTempStamp - Archive [Date 2006.12.27 Time 14.38].zip 33KB (application/x-zip-compressed)This file has been downloaded 439 time(s). | | Back to Top | | |
|
Mike Cook
Registered Member
Date Joined Jul 2004
Total Posts : 678
| Posted 12/27/2006 1:16 PM (GMT -8) | | |
I could probably write a module to take the recorded data (10 hours), read it in and average it or do some other calculation. But what I'll be doing with the data file is removing the SD card from the Propeller and then analyze the recorded data in Excel.
Some ideas I've had were:
Back in the olden day's before there was an IBM Home PC, I had a Commodore 64 and a TRS-80 Model 12. A buddy and I wrote a BBS (Bulletin Board Software) kind of like the old WildCat BBS software, but this was before WildCat was written. We used the data storage to feed menus back to the BBS user. Since our software was written in BASIC we did not want to hard code all the User Menus for the BBS, so we made all menus a *.txt file, and just passed them to the user.
Actually If Mike Green can integrate this into his Propeller OS that would be pretty awesome. All sorts of stuff can be done with these routines, let your mind run wild!
<EDIT>
originator: Another Idea! I did a quick search on some of your posts, it looks like your a CNC junkie. I am too, still going through CNC-101 on my home built machine (www.allsurplus.net/CNC). Granted a Full blown PC running a CNC is THE way to go, however you could use these SD routines to read in G-Code to drive a small CNC run from a propeller chip.
</EDIT>
Mike Post Edited (Mike Cook) : 12/27/2006 9:49:24 PM GMT | | Back to Top | | |
|
rokicki
Registered Member
Date Joined Feb 2006
Total Posts : 876
| Posted 12/27/2006 2:05 PM (GMT -8) | | On using four data lines: this would require operating the SD card in SD mode, not SPI mode.
I think this is possible, but it may involve licensing fees. I'm not entirely clear on this.
There are other good ways to get the speed up; moving a bit more into assembly should be
able to double the speeds I posted above, and then using multiblock mode would probably
get you even higher. On the other hand, the speeds I posted above are pretty good. I think
you should have no trouble playing back 44KHz 16-bit audio (assuming you use sufficient
buffering; I suspect as little as 2K buffering would be enough). Recording 44KHz 16-bit
audio would be much trickier and would require some attention to detail (and some cards
give significant delays---I've seen up to 150ms---on some block writes) so more buffering
would be needed, but I think it's possible.
What I want to build with the propeller is a fully self-contained microcomputer with a built-in
BASIC-like interpreter. I want it to drive a TV, have built-in mass storage (the SD card),
use a cheap PC keyboard and mouse, and be buildable for next to nothing. So far the
propellor and the protoboard, plus a few odds and ends, seems to be the best solution.
The goal here is accessible introduction to programming for kids, no PC required. In a
perfect world I'd like to build a few hundred of these and give them away for free. I'm not
sure it's a perfect world, but I'm gonna see how close I can get.
A SD card has so much storage I can have all the documentation right on the device itself
and even provide built-in help. Also, I can swap out the interpreter/compiler (if I choose to
separate the two) fairly quickly, if needed; I can even swap chunks of code out and in as
needed. And I can package the thing with hundreds of small sample programs.
Other than this, though, clearly the SD card can be used for a lot of different things, like
sound storage, image storage, video storage, code storage. You can even put the boot
code in the EEPROM that boots from the SD card, and bingo, you swap out SD cards to
change the functionality of your embedded device without requiring its physical removal
or attachment to a PC. SD cards are cheap enough where you can just send them out by
mail as "software updates" for people using whatever product you have.
I remember when the TRS-80 color computer floppies were only something like 150K, yet
we stored a lot of programs on them. These days a $10 512MB SD card has more than
3,000 times the storage space. Surely we can figure out more things to do with this
than we did with the old 150K storage on the 895KHz Coco.
-tom | | Back to Top | | |
|
rokicki
Registered Member
Date Joined Feb 2006
Total Posts : 876
| Posted 12/27/2006 2:38 PM (GMT -8) | | Sure, the above code will read SD cards no problem, and you can just comment out the write routines to save space if you want.
Note that the secure digital card can only physically be attached to one thing at a time---either the PC or the propeller. Either one can
read or write at any time (you can move the card back and forth) arbitrarily. So you can format the card on the PC, add files,
move it to the propeller, the propeller can read those files, add its own, and back and forth and so on. It's just like a huge floppy disk. | | Back to Top | | |
|
rokicki
Registered Member
Date Joined Feb 2006
Total Posts : 876
| Posted 12/27/2006 11:50 PM (GMT -8) | | Hey, Bill! Thanks for the comments! Yeah, I spent some good time with the scope figuring out the cycle timings of the
counters when you modify the various registers, and also spent some time checking for glitches. It's very important
to avoid the glitches. Luckily I was able to drive everything from just the counters which simplified some things.
I've got a new version of the block driver coming along, and I am thinking of using the memory space that the
assembly occupies in the object as the sector buffers too, so I can make the overall memory consumption smaller at
the same time I make the whole thing much faster. But doing those sorts of stupid memory tricks can get a bit uglier
than I'd like. | | Back to Top | | |
|
rokicki
Registered Member
Date Joined Feb 2006
Total Posts : 876
| Posted 12/27/2006 11:57 PM (GMT -8) | | Bill's got a great trick, as reported here:
http://mikronauts.com/2006/12/28/cheap-sd-socket/
He just buys a miniSD or microSD which comes with an SD adapter card, and solders directly to the pads on
the adapter card. That's a pretty cool way to go! I don't know how well the plastic will hold up, and I
think at least the microSD isn't *guaranteed* to support SPI (but I bet they all do), but it's a neat idea.
Bill, do share how it goes! | | Back to Top | | |
|
Mike Cook
Registered Member
Date Joined Jul 2004
Total Posts : 678
| Posted 12/28/2006 7:10 AM (GMT -8) | | |
I'm starting a routine to add current file date and time:
In Tom's fswr.spin, function: pri pdate
pri pdate
{{
' Get the current date and time, as a long, in the format required
' by FAT16. Right now it"s hardwired to return the date this
' software was created on (December 26, 2006). You can change this
' to return a valid date/time if you have access to this data in
' your setup.
}}
return constant(((2006-1980) << 25) + (12 << 21) + (28 << 16) + (12 << 11))
I added the following comments and assignments: (just to make it more clear, on how to assign values)
pri pdate
{{
' Get the current date and time, as a long, in the format required
' by FAT16. Right now it"s hardwired to return the date this
' software was created on (December 26, 2006). You can change this
' to return a valid date/time if you have access to this data in
' your setup.
}}
' YEAR MONTH DAY HOUR MINUTE SECOND**
return constant(((2006-1980) << 25) + (12 << 21) + (28 << 16) + (23 << 11) + (59 << 5) + (29))
'
' ** Seconds are counted with 2 seconds interval, so a value of 29 in this field gives 58 seconds.
Next job is to get the date/time from the DS1307 in there!
Having a ball playing with this! This is like going from cassette to a floppy disk back in the late 70's!
Mike Post Edited (Mike Cook) : 12/28/2006 3:44:38 PM GMT
Image Attachment :
_FileDate.JPG 30KB (image/pjpeg)This image has been viewed 404 time(s). | Image Attachment :
_FileDateDos.JPG 24KB (image/pjpeg)This image has been viewed 396 time(s). | | |
| | Back to Top | | |
|
bassmaster
Registered Member
Date Joined Dec 2006
Total Posts : 181
| Posted 12/28/2006 12:00 PM (GMT -8) | | | does anyone know of a windows utility to partition a sd card?, I have one that has a "bios" setting that say kodak 512 and fat12, but logically formatted as 64mb fat16, I think I need to delete all partitions and create a fat16 partition. I am currently dl knoppix to attempt this on a live cd of linux. | | Back to Top | | |
|
rokicki
Registered Member
Date Joined Feb 2006
Total Posts : 876
| Posted 12/28/2006 12:34 PM (GMT -8) | | For windows 2K and XP, just go to control panel -> administrative tools -> computer management -> disk management.
This has all the formatting and partitioning tools you need. You can right-click on the "removable disk" and ask it to
reformat, and when you reformat, pick the cluster size in order to guarantee it's FAT16 as documented in the first
message in this thread.
A 64MB card can be formatted as FAT12, FAT16, or FAT32 depending on the sectors per cluster. I suspect your camera
may have formatted it to FAT12 with 32K clusters (since this gives the best write performance, and this is what cameras
care most about). If we want it to be FAT16, we need to use a maximum cluster size of 8K (which is still pretty reasonable
and will give pretty good performance). So make sure in the disk management tool that you pick this cluster size.
Otherwise it will default to some value that may not be what we want.
Note that frequently windows will format a card without a partition table (that is, the entire card is a filesystem partition
and there is no MBR with partition information). The code I posted above handles that case too.
If you can get me the first dozen sectors on the card, I can let you know how it's organized. | | Back to Top | | |
|
rokicki
Registered Member
Date Joined Feb 2006
Total Posts : 876
| Posted 12/28/2006 8:49 PM (GMT -8) | | I've put an updated fsrw.spin (v1.1) on http://tomas.rokicki.com/fat16work/fsrw.spin. This fixes a bug where the
creation time was not being set correctly. | | Back to Top | | |
|
Mike Cook
Registered Member
Date Joined Jul 2004
Total Posts : 678
| Posted 12/29/2006 9:30 AM (GMT -8) | | |
FYI - Just some notes on using the file DATE/TIME stamp in the SD routines: (See Tom's link for an updated fsrw.spin file)
The ONLY reason the time stamp was implemented was to eliminate the following message from a Windows 'DOS' box:
"If you open a command prompt in Windows (W2K or XP) and do a directory (DIR<ENTER>) you will get "The Parameter is incorrect", before the 'DIR' command displays the file name of each file. Windows is wining about the file date/time not being present and or all ZEROs. If you are only accessing the files written from the Windows GUI then you'll never see this message."
Some tips on using 'time stamps' on your files. Note this assumes that you have access to the current date and time via a DS1307, GPS receiver, or whatever other device that can provide time:
1. To time stamp a file, you will need to assign variables to the pri pdate routine in fsrw.spin BEFORE you open a file for write. By doing this, when reading the directory of the SD on a Windows box you will see the date and time the file was created.
2. If you are appending to a file, the time stamp will be when the file was originally created. The time stamp on the file will not be updated when you are writing in append mode.
I'm speaking for Tom here, so Tom come in and 'slap me' if I'm 'off base' here. I believe Tom's original intent with the SD routines was to be able to add mass storage capability to the propeller with the smallest/fastest code possible.
In all reality if your logging Date/Time stamped data to a file, this data (Date/Time) should be part of each 'record' that is written to the contents of the log file.
Go out and get some SD cards and a socket! I guarantee you'll have fun with this!
| | Back to Top | | |
|
Gavin
Registered Member
Date Joined Aug 2006
Total Posts : 134
| Posted 12/29/2006 6:26 PM (GMT -8) | | Got some SD cards the day before this code showed up:)
Today got a USB card reader and a microSD card.
Is seems the Micro SD cards come packaged with a SD adapter, get two cards and you have a spare adapter, so you could solder one up.
Got a long weekend to have fun with this, well done Rokicki.
Gavin | | Back to Top | | |
|
Bill Henning
Registered Member
Date Joined Sep 2006
Total Posts : 1283
| Posted 12/29/2006 10:01 PM (GMT -8) | | | Here's a quick and dirty test program for Rokicki's SD routines - they work great !
I also posted it to my blog.
Best,
Bill
www.mikronauts.com - a new blog about microcontrollers File Attachment : testsd.spin 1KB (application/octet-stream) This file has been downloaded 323 time(s). | | Back to Top | | |
|
iam7805
Registered Member
Date Joined Jun 2006
Total Posts : 14
| Posted 1/2/2007 5:17 PM (GMT -8) | | I can't download this right now, just want to ask a quick question.
Are the hardware hookups for this version the same as your read-only version you posted earlier?
Regards | | Back to Top | | |
| 89 posts in this thread.
Viewing Page : 1 2 3 4 |
| Forum Information | Currently it is Monday, February 08, 2010 6:10 PM (GMT -8)
There are a total of 415,872 posts in 57,629 threads.
In the last 3 days there were 78 new threads and 858 reply posts. View Active Threads
| | Who's Online | This forum has 18513 registered members. Please welcome our newest member, SEL.
42 Guest(s), 12 Registered Member(s) are currently online. Details
Peter Jakacki, CannibalRobotics, W9GFO, Bob Lawrence (VE1RLL), Computer Geek 101, tonyp12, Harley, Jimm2, ciamo, Bane, Peter KG6LSE, atlstang |
Forum powered by dotNetBB v2.42EC SP2.02
dotNetBB © 2000-2010 |
|
|