I might be naïve - high speed data sampling advice.

[Evalon]

Hi,

I may have a need to sample approximately 50 Mbps from 2 channels (1 bit signal) which are stored to e.g. a compact flash card in a format that subsequently allows the stored files to be transferred to a PC for storage.

Since longer recording periods are desired I'm considering e.g. 32 GB compact flash cards for memory.

The basic functionality is "record" and "stop" and "play" (i.e. extract data in exactly the sequence it was entered) , however, in time, fast forward and fast reward, and time counting is desired.

I have thought of an appropriate microcontroller or microprocessor - but, maybe such ready-to-use designs already exist? Also, and I reckon this is important - in case I am to try to do this myself - I have no programming experience (am an engineer, though) and so do not know if this is relatively easy to do - or close to impossible?

Any of you have experience with such a task? Or know of already-made solutions or appropriate IC's?

Suggestions are appreciated

Greetings,

Jesper

 

[Sceadwian]

What is the exact sampling period required and what is the maximum needed recording time? "longer recording times" and "approximatly 50mpbs" are not design criteria that can be worked with =)
Knowledge of the signal you're trying to record would be greatly beneficial as well because even if you need high resolution (50mbs) if the signals don't switch at that frequency compression can dramatically reduce the data you'll have to store.
What micro controller do you have in mind 50msps is HUGE, that's faster than just about any micro controller I know actually runs at let alone can do active data acquisition at.

 

[Evalon]

Hi Sceadwian,

& thanks for your comments & considerations.

As a sort of reply I can say that I'm currently reviewing the possibility of such a recording system and myself have not specified final requirements - simply because I do not know them fully by now. Some aspects are still being considered.

However, to be a bit more specific about what you mention/ask for more information about I can say that:

The signal to be stored will come from a delta-sigma 1 bit A/D converter and as such I guess it will be a challenge to predict the sequences of the data. They must be stored as is, however, I reckon that longer sequences of data - from time to time - can be 1s or 0s only. Don't know if that can be compressed - or if it's straightforward to do so, or necessary ... ?

The recording time is to be one hour. In my calculations that means about 22.5 GB of data/hour for two channels. Which is why I consider a 32 GB compact flash card. Preferably more than one file (i.e. more than one recording session with two channels) should be able to be stored on the card .

The solution I am considering can be microcontroller, microprocessor or ? based, it's not that important, as long as it's accessible from a design point of view. It can also be an existing solution, if so exists.

Functionality basically is "start", "stop", "playback".

Eventually, more than two channels should be daisy chained - probably 8 channels, meaning huge amounts of data. However, for now, it suffice for me to imagine this can be done using some microprocessors and a common clock.

So, since I'm a complete newbie in this area, though with an engineering background - is it likely that it can be done (possibly by me). Since the initial functionality basically is to transfer data to the compact flash card I imagine that programming may be relatively simple ... or ...?

It's also something that I may be willing to pay one of you to do ("within my budgets").

Best regards,

Jesper

 

[Sceadwian]

You're definitely talking about something that requires the use of a micro processor, and some modestly complex ciruitry, an ARM processor might fit the bill but this is most certainly not something you can take on your own. I'm going to guess there aren't many users here that would tackle a project of that level, but it doesn't hurt to ask. You still haven't clued us in as to what the signal is, as that's actually critically important to how you approach the project in the first place. If you could flesh out the ideas of what you're trying to do more than the specs you don't have well defined it'd be easier to suggest the best avenue for you to go down. Electronics is both broad and specialized at the same time. If you're vague you'll get a trillion responces, not one of which might be suitable for YOUR needs, as you don't know your own needs, explain in plain language the entirety of what you're trying to do as you see it. That includes explaining to us what you're trying to do not what you think you need to do to get what you want done.

Also, what's your budget? You gotta ballpark it in some manner. This could be a hundred dollar idea, a thousand dollar idea or a ten thousand dollar idea =) We have NO reference.

 

[Evalon]

Hi Sceadwian,

Thanks for replying. I'm currently considering various information so cannot give a final answer now. Will get back.

'best for your day

Jesper

 

[misterT]

I'm assuming that you are going to use an AD-converter chip with SPI interface.

I would find a development board with a suitable microcontroller (ARM or XMEGA), SPI interface to read the AD-converter and a SD-card interface to store your data.
Maybe one of these: ARM7 Evaluation Boards Comparison
The MCB2140 could be a good choice.

 

[Evalon]

Hello MisterT

Thanks for your suggestions. I took a look at the ARM7 microprocessors, however, aren't they a bit too slow? They operate below a 100 MHz and what I'm trying to acquire is 50 Mbps meaning that there will be only 2 clockcycles for each data acquired ...

I've also looked further into the possibilities and am considering a Korg MR-1000 for a start. It already operates at 5.6 MHz DSD per channel and has all the ancillaries like communication to RAM, harddisk, display etc. in place. It's based on an Analog Devices BlackFin BF532 microprocessor with a 400 MHz core. As far as I can see the ancillaries can operate at higher frequencies (the I/O Bus from the microprocessor operates at 133 MHz), the harddisk appears to sustain 108 Mbps write speed etc.

My question is if it will be accessible to change the coding inside the processor to sample more frequently, such as 5.6 MHz * 4?

Any of you know of this?

Greetings from a very sunny & warm Denmar

Jesper

 

[Oznog]

You should probably start from the beginning... what sort of signal are you reading? What frequency and what sort of resolution are you actually trying to record? Is this audio?? I'm suspecting you're doing an enormous amount of oversampling in place of reading analog levels with greater bit depth, and that may not be ideal.

Note that SD card "Class" specifies minimum guaranteed writing speed and is in terms of 8Mbit/sec per "class". You would probably need Class 7 or above for 50Mb/sec.
And AFAIK, that's for the proprietary SD card interface which you won't be using. You'd be using the SPI interface, which may be much slower but I don't know.

 

[Evalon]

Hi Oznog,

Thank you also for replying. I will, however, need the one bit signal for various reasons and will also need to sample at high sample rates. So my question really is how this can be accomplished?

I have found the Korg MR-1000 to be interesting due to its architecture and a potential for re-programming it. Any of you know if this can be done? Or know of other accessible solutions?

Greetings,

Jesper