PSN-L Email List Message

Subject: Re: Hekla volcano geophone planned
From: ian ian@...........
Date: Sat, 05 Jan 2008 22:08:22 +0000


Hi,

I was waiting nervously for your reply and an array of astute questions; 
here they are!  :-)   I'll try to answer as best as possible.

My A/D is a National Instruments AT-MIO-16XE-50 which plugs directly 
into the motherboard, so no communication issues.  The data sheet is at 
http://www.ni.com/pdf/products/us/2mhw254-255e.pdf and the manual is at 
http://www.ni.com/pdf/manuals/370507a.pdf

 From the manual, sampling rate is only a consideration when switching 
gain between channels (which I am, see later), otherwise full tilt is 
fine.  The board does have a micro with which it can perform various tricks.

Higher bit systems give both higher resolution and higher dynamic 
range.  I think the main concern for seismometry, amongst cash limited 
amateurs, is the dynamic range to try to avoid saturation when a strong 
signal comes in.  So, with my 16 bit A/D, I feed the signal into 3 
separate channels and sample each of them at 3 different gains, +/-0.1V, 
+/-1.0V and +/-10V.  The software then chooses the reading with the 
greatest gain which hasn't saturated.  So that gives it an "effective" 
bit rating of 20V/3.05e-6V or just over 22 bits for dynamic range but 
still 16 bits for resolution.

As said above, changing gain between channels does have settling 
issues.  To eliminate this effect, I should increase the intra-channel 
sampling time.  However, this would increase the skew between the 3 
samples, which would introduce another source of noise.  I should really 
measure/calculate these competing effects and find the optimum point.

For this application, though, I am content.  I used to use a 12 bit 
system, so my present system is much better even with these features.

I'm squirting the data across my own intranet to the graphing computer, 
so am getting most of the available 100 Mb/s bandwidth with minimal 
latency.  For those using the internet, there shouldn't be a problem 
provided the data are timestamped at source.  It then doesn't matter how 
long the data takes to arrive or whether the times between samples 
varies, the data can be properly reassembled using the individual 
timestamps.  Things though are more limited with 56K modems.  I'd be 
interested to hear how there is a 12 Km "gap" in your system. 

Cheers

Ian
http://www.iasmith.com/
http://www.festivalpreviews.com/


ChrisAtUpw@....... wrote:
> In a message dated 05/01/2008, ian@........... writes:
>
>     Hi,
>     actually, I don't think I'm near any limits.  The A/D I use can
>     handle 20K samples/sec. 
>
> Hi Ian,
>  
>     What ADC are you using? 20 K SPS is 50 micro sec / sample. My ADC 
> takes 20 muS/S.
>     How does it's accuracy depend on it's sample rate?
>     Does it have an onboard processor chip to take and average 
> multiple samples?
>     What data rate does the ADC board to computer link support?
>
>     I only have 2 instruments (though I sample each on 3 A/D channels
>     to get the resolution up to 22 bits), so that's only 6 x 50, or
>     300 samples/sec.  
>
>     You need to average four samples to get 1 additional bit of 
> accuracy, 16 samples to get two bits extra, etc.
>
>     So I could connect up all of the A/D's 16 input channels and still
>     not stress it.
>
>     This is likely to be ~40 bytes total with the overheads at 50 SPS, 
> say 16 K bits/ sec. If you are using a 24 bit ADC, it is likely double 
> this. Then it largely depends on whether you are sending datapackets, 
> or individual bytes.
>     If you are sending asynchronous bytes, you have to wait for the 
> signal to be transmitted, the receiving server to respond and the ACK 
> to be received. Transmission delays can be significant. 
>     I am 12 km from the phone terminal, so the delay would be well 
> over 80 micro sec per byte. Coupled to a a 56 K modem, I certainly 
> could not transmit this much data.
>
>     The data server does burn up 80% of the PC's CPU but it's only an
>     old 800 MHz machine and wouldn't cost much to replace with one
>     twice the speed.  Data across the network connection is only 32
>     characters x 50 or 1600 bytes/sec.  Less than a 500th of the 100
>     Mb/s network bandwidth
>     (being generous).
>
>     So, do you know what minimum speed you can actually get for 
> asynchronous transmissions? The broadband data rates quoted by the 
> service providers are usually maximums in the best possible 
> conditions, not the average and certinly not guaranteed. They may not 
> allow for transmission delays. Reality may be only a small fraction of 
> the specs advertised! A recent BB survey in the UK suggested a far 
> lower preformance, nearer to that of the 56 K modems.
>  
>     Regards,
>  
>     Chris Chapman

-- 


__________________________________________________________

Public Seismic Network Mailing List (PSN-L)


[ Top ] [ Back ] [ Home Page ]