ISM band maximum bandwidth

[mariocampeau]

Is there a maximum bandwidth tolerated by the FCC if I use a single-channel in the 900 MHz ISM band?

Thanks,

 

[zappedagain]

The 900 MHz band is pretty wide open on what you can and can't do; that's why it is called a trash band. See FCC Part 15, sections 15.231 and 15.247 for your BW options with different modulation types.

 

[mariocampeau]

Thanks.

In 15.231, there is a paragraph that reads:

"... For devices operating above 900 MHz, the emission shall be no wider than 0.5% of the central frequency. Bandwidth is determined at the points 20dB down from the modulated carrier."

This means that at 915 MHz, my maximum bandwidth (20 dB) should be 4.57 MHz, right? Does this means that I can't use a larger bandwidth (i.e. 20 MHz) in order to obtain a sharpest rising time of my baseband signal?

 

[VisiGoth]

mariocampeau
Why do you have a sharp risetime baseband signal? Could you encode your information into a narrow band modulation?

All,
What about this UWB stuff? From early reading on it there was talk about ignoring interference in the interest of economy and allowing very wide band RF. Is there a special band for unlicensed UWB?

 

[mariocampeau]

Because I am trying to do some RF localisation by reading a phase change in a signal. The sharpest the transition is at the phase change the more precise I will be in my localisation.

 

[zappedagain]

Mario,

You can go wider under 15.247 with some additional power and modulation constraints (read past the frequency hopping section to the 'digital modulation techniques' section). If you are trying to optimize an existing system you might be on the wrong track though, because there are a couple of things to consider:

First, your transmitter probably has a limited slew rate that it can change frequencies at. For example, say it will slew 5 MHz in 5 uS or 20 MHz in 20 uS; either way the transition is still 1 MHz/uS. Unless you are considering a completely new transmitter design you aren't going to buy yourself anything.

Second, you don't want to modulate at a bandwidth wider than your receiver bandwidth. Any time your signal is outside of the receiver bandwidth you'll lose amplitude and start to amplitude modulate the signal. AM is a killer for most of the phase detectors that I have worked with (differential phase error).

You can eliminte the AM effect by widening your receiver filter; unfortunately this brings up your noise floor because noise power is proportional to bandwidth. Your phase accuracy is very dependent on signal to noise ratio, so that will clobber you again. Your transmitter and receiver need matched bandwidths for optimum accuracy.

 

[mariocampeau]

It is a brand new design from A-to-Z so I have full control of Tx and Rx. And I don't think there is much other solutions than using the widest bandwidth possible :-(

In article 15.247, you are talking about the paragraph stating 'a minimum 6 dB bandwidth of at least 500 kHz'? Well that is good news

Thanks a lot for your guidance.

 

[zappedagain]

And the paragraph in 15.247 that states to keep the power under 8 dBm in a 3 KHz bandwidth. This section was written for M-ary modulation techniques so it expects a flat passband. If your signal is only 2-ary you'll need to make sure the peaks meet this limit. I haven't seen any FCC certifications showing 2-ary in 15.247 yet but people are claiming they can do it (it has been 12+ months since I last checked and this change to 15.247 is fairly new).

 

[VisiGoth]

zappedagain
Thank you. I didn't know that info about wider bandwidths.

mariocampeau
You can get significant performance gains with a slower transition signal by processing gain. If you intentionally slow up your transitions with a filter (say a Gaussian filter as in the GSM phone system) you can integrate the correlated signal across a long time and get a very accurate time measurement. You can look under Cramer-Rao lower bound (CRB or CRLB) to get the equation that shows you what the variance of your time measurement will be as a function of SNR, bandwidth, and your integration time. By not integrating you may be giving up an awful lot of performance.

 

[mariocampeau]

Thanks VisiGoth, I'll take a look at that.

 

[VEBill]

"And I don't think there is much other solutions than using the widest bandwidth possible..."

Hmmm...

My gut reaction tells me that Shannon's Law (or something similar) probably applies to location detection. At its common sense root, location info is not a 'high bandwidth' data set. Most location data can be represented by a few hundred bits of data, updated perhaps once per second (YMMV). In most cases, it isn't a lot of data.

So unless the subject is flying around at a very high rate of speed with constantly changing acceleration (like an out-of-control, rocket powered bumble-bee), you shouldn't need a huge bandwidth.

(Of course, path loss enters the equation as well...)

Just as an example, you could integrate innumerable low bandwidth pings or returns to provide the same information as one (less often) very high bandwidth ping or return. Also, look up Barker Codes (a recent thread). There are many techniques other than raw bandwidth.

This 'bandwidth versus info' tradeoff is a whole science in itself in the radar field.

 

[VisiGoth]

The Cramer-rao bound shows the improvement in time estimate with increased bandwidth and increased SNR. You must have BW, and it is the best way to get accuracy. You would want to fill your BW allowed as much as possible within complexity/cost restraints. Integration (processing gain) would improve your SNR.

VE1BLL mentioned motion and Barker codes.

Motion could cause problems. If you do not ave any motion and you do not have osc. drift problems, then you do not have to worry about your coherence time.

Barker codes are one way of ameliorating the ambiguity problem. If you have BW, but poor sidelobe suppression on the corelated signal, you could get mixed up on exactly where your sinals correlate. Barker codes do this, as well as many other codes. GSM uses a set of color codes that have good corelation and anti correlation poperties and have a well defined modulation and control over BW. I would recommend looking into the GSM definitions and scale the numbers to fit your requirements.