LeafLabs Garden » Topic: New Project Guidance & How the Maple Will Work Better

LeafLabs Garden » Topic: New Project Guidance & How the Maple Will Work Better http://forums.leaflabs.com/topic.php?id=857 A place to share, learn, and grow... en-US Fri, 22 Jan 2016 00:07:25 +0000 http://bbpress.org/?v=1.0.2 <![CDATA[Search]]> q http://forums.leaflabs.com/search.php gbulmer on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857&page=2#post-5962 Wed, 10 Aug 2011 15:35:59 +0000 gbulmer 5962@http://forums.leaflabs.com/ galindro - would you give a bit more detail please? Which public key algorithm do you need? How would you do the public key exchange? Would this all be on a private network? What role has the Maple? There are a bunch of Open Source Public Key projects, listed by Google. Some are quite old, and so won't likely require huge machine resources. galindro on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857&page=2#post-5953 Tue, 09 Aug 2011 14:48:27 +0000 galindro 5953@http://forums.leaflabs.com/ robodude666 says: "However, with that said.. There is an official Encryption 128-bit AES/TDES library available from STM/Keil/IAR." AES and TDES are symmetric key algorithm. Are there an implementation of Public Key algorithm for Maple / ARM Cortex? gbulmer on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857&page=2#post-5286 Thu, 16 Jun 2011 19:38:15 +0000 gbulmer 5286@http://forums.leaflabs.com/ becker - an STM32F103RB contains 2 12bit ADC's. Further, the DMA controller can offload the processor from storing the ADC converted values into memory. So two signals can be sampled and converted simultaneously. Also, the ADC+DMA supports an interleave mode, where the same signal can be sampled alternately by two ADC's, giving 2M samples/second at 12 bits. I vaguely remember that the ADC can produce right-aligned samples, and the DMA can read 16 bits, but store only 8 bits. My reading of Section 5.3.18 "12-bit ADC characteristics" of RM0008, specifies an ADC clock frequency is a maximum of 14MHz. I haven't found any mention of higher clock rates and reduced precision conversion, but I haven't attempted to memorise every section :-) The STM32F103RET has three ADC's, but it doesn't support interleaving all the ADC's to get 3M samples/second. fkatzenb on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857&page=2#post-5285 Thu, 16 Jun 2011 19:29:00 +0000 fkatzenb 5285@http://forums.leaflabs.com/ The engine knock was a different application than above. I started it and a friend took it over to code the Arduino with AVR. He is now using a TPIC8101 over SPI since it is designed for detecting engine knock. We are using them to keep from detonating our engines in carnage under heavy boost. becker on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5279 Thu, 16 Jun 2011 17:52:30 +0000 becker 5279@http://forums.leaflabs.com/ The DSOnano device uses a 1Msps rate with the STM32. There is even a hack that does 2M per second, albeit only claiming 8 bit effective resolution. I'm guessing that they base their clock tree around running a single ADC at 14MHz (or 28MHz for the "8 bit", rather than using a chip with 2 ADCs and interleaving). The second generation DSOnano switched to using a high speed external ADC and a gate array. Because no matter what tweaks or slightly improved chip, you won't get a 10x bandwidth improvement. The AVR doesn't need to use much processor time for A/D conversion. But it is awfully slow. And you have to slow it down even more because if the widely spaced clock divisors. Note that you can usually double the speed by looking at the resolution you are actually getting/need, and running the A/D clock faster than the "for maximum resolution" speed. poslathian on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5278 Thu, 16 Jun 2011 17:45:08 +0000 poslathian 5278@http://forums.leaflabs.com/ Interesting project, can you say more about the end application? Last time I heard anyone talk about detecting engine knocks was on a submarine! fkatzenb on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5277 Thu, 16 Jun 2011 17:26:59 +0000 fkatzenb 5277@http://forums.leaflabs.com/ LOL. Thanks guys. I was mostly curious because of how sucky the Arduino was when trying to do enough engine knock samples to actually get a representative sample to call something knock. robodude666 on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5273 Thu, 16 Jun 2011 16:34:44 +0000 robodude666 5273@http://forums.leaflabs.com/ And from what I gathered on the forums so far... "and if you STILL need improvement get an Oak." Imagine implementing AES encryption, ethernet/Wifi, and SD/FAT on a FPGA with VHDL. poslathian on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5264 Thu, 16 Jun 2011 14:13:42 +0000 poslathian 5264@http://forums.leaflabs.com/ Oh and to specifically address your question about SPI - there are two completely hardware spi (with library support) onboard maple, so you can read and write spi without hardly wasting any cycles at all. poslathian on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5263 Thu, 16 Jun 2011 14:12:34 +0000 poslathian 5263@http://forums.leaflabs.com/ My advice to anyone concerned about performance is always the same - if you are uncertain, dont optimize until you need to. On Maple that usually follows this sort of trend: Use Wirish (the Arduino compatibility layer...with all of its SerialUSB.print, and SPI.write() ) if you need improvement, try punting work to the hardware, with hardware spi, hardware i2c, DMA, interrupts, etc. if something still needs improvement drop down to libmaple (the low level libs, like gpio_write_bit, and usb_write_bytes) if you still need improvement drop down to bare C register banging (like GPIOA->BSRR = 1) if you STILL need improvement drop down to assembly. robodude666 on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5246 Thu, 16 Jun 2011 06:37:34 +0000 robodude666 5246@http://forums.leaflabs.com/ The Maple's STM32F103RB's claims 1 million samples per second. Is that fast enough for you? Users haven't reported of fully reaching this performance level, from what I gather, but you can get rather high. The STM32 also has a dual capture mode (or something like that) which effectively doubles performance. In addition to hardware features like DMA, you can capture analog samples without wasting your CPU time on them. This would allow you to continuously capture ADC samples and store them in a buffer, and process them behind the scenes. Just keep in mind that the performance you see from the wirish library, that aims to provide arduino-like functions and compatibility, is not the true performance of the Maple or STM32 (not that I'm saying the wirish library is bad; I'm saying that any APIs that make things look pretty incur overhead). If you access the hardware on a low-C level you'll get splendid results. fkatzenb on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5244 Thu, 16 Jun 2011 03:59:37 +0000 fkatzenb 5244@http://forums.leaflabs.com/ Excellent. I think I will be ordering a Maple here real soon and adding to my range of Arduinos. I am looking forward to having one. So let me ask this. I know the MAPLE is faster, but is it better when it comes to reading things like the A/D? The Ardunio wastes so much time doing trivial things like that. For another project, we ended up grabbing an SPI ADC chip to gain clock cycles back... ALOT of them. Thanks, Frank robodude666 on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5232 Tue, 14 Jun 2011 20:42:26 +0000 robodude666 5232@http://forums.leaflabs.com/ Ah. If you're sending stuff to the server only once per minute, then the Maple will certainly handle that - even if the encryption takes a couple seconds (which is unlikely to be that slow). Communicating with the SD card is not difficult, however you'll need to, of course, encrypt/decrypt the files yourself. Just store them as raw encrypted binary files and you'd be good. fkatzenb on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5231 Tue, 14 Jun 2011 20:12:38 +0000 fkatzenb 5231@http://forums.leaflabs.com/ Thanks for the information. I think it is overkill on this project. Here is a better breakdown.... Read in 3-4 sensor values every 5 seconds Every minute, send the following to the website at that minute - The sensor values at that minute - The 1 and 10 minute running average - The max and min over the last minute Display those values on a simple character LCD Sending entails sending to a PHP based server where the information is logged, graphed, and processed. If the data exceeds certain user parameters, he can be alerted via a SMS message OR his Android application and has the joys of being monitored anywhere in the world. All of this is configured thru the PHP based website. The service on the website won't be free, as web space costs money (which was causing headache over on the Arduino forums), but whatever. However for those who don't want to use the website service, the Android app will poll over local your local area network and will grab that same information and do it's own processing just like the website does but of course has it's limitations. So as you can see, the box is pretty dumb and could use logins without SSL provided a an experienced programmer doesn't leave in any vulnerabilities. The reason for my cryptic nature is that I don't want the commercial world to capitalize on the idea of using the website. Their are products now that do similar, but do so over bluetooth, are limited, and cheap. I want to offer people a way to do something right, not over priced, and give them freedom. Folks will be able to add to it, whether additional things to monitor, utilize some of the outputs for control, etc. I plan on developing an API similar to Munin to allow for that flexibility for those using the website. The open source hardware and software will allow people to expand to their wishes, reduce cost to them, and apply the entire concept to other hobbies. I just want to offer it for this particular hobby first. :) Also my reason about the SD was to provide a place for login, passwords, keys and to possibly allow folks who don't know Arduino/Maple that well to just put the .bin file on to update their unit. Thanks, Frank robodude666 on "New Project Guidance & How the Maple Will Work Better" http://forums.leaflabs.com/topic.php?id=857#post-5228 Tue, 14 Jun 2011 19:03:27 +0000 robodude666 5228@http://forums.leaflabs.com/ What kind of security are you looking for? 128-bit AES is the minimum standard I've seen used for SSL keys, etc. Anything less simply feels like you're false security. I'm very confident that your project would gain great benefit from the BeagleBoard-xM's 1GHz ARM Cortex-A8 processor, 512MB RAM, 10/100 Ethernet, microSD card slot, Real Kernel + OS, etc. While I'm sure you can pull "something" really great off with the Maple - I mean, the Arduino's little 16MHz 8-bit MCU can handle serving web content via the Ethernet Shield/Wifi Shield, read/writing to an SD card, and sampling sensors - it won't be anything compared to what you'd be able to accomplish for your project with a BeagleBoard or equivalent. I love LeafLabs and Maple, but when reading your requirements I simply hear like you need more. A Maple would work, but it would limit what you can do or how quickly you can do it -- unless that's the challenge you're specifically looking for.