Frequency Planning

On Oct 23, 9:32=A0am, jim <fiscalconsoc...@gmail.com> wrote:
> On Oct 23, 8:59=A0am, Vladimir Vassilevsky <nos...@nowhere.com> wrote:
>
>
>
>
>
> > Tim Wescott wrote:
> > > On 10/23/2010 06:59 AM, Vladimir Vassilevsky wrote:
>
> > >> Brian wrote:
>
> > >>> Dear list,
>
> > >>> Once I know I will be in the 900 MHz ISM band and the bandwidth of =
my
> > >>> signal will be about 10 MHz, how then do I go about selecting RF
> > >>> center frequency, crystal frequency, intermediate frequency, and
> > >>> sample rate? Is it good enough to say:
>
> > >>> RF: 900 MHz
>
> > >> Nobody would allow that, especially with 10 MHz spectrum.
>
> > >>> Crystal: 100 MHz
>
> > >> No such crystal.
>
> > >>> IF: 25 MHz
>
> > >> Bad choice. Besides, there is no such IF filter.
>
> > > Not to criticize the rest of your critique, but at a 10MHz signal
> > > bandwidth, a 25MHz IF bandwidth can be realized fairly easily with a
> > > home-brew LC filter. =A0Depending again on which of the other unreali=
stic
> > > goals aren't, it could maybe even be made to sacrifice skirt performa=
nce
> > > but use off the shelf parts.
>
> > > Now, if it needed exceptionally sharp skirts or something, then it'd =
be
> > > a pain.
>
> > A filter that needs tuning is a pain in production and a potential
> > problem with reliability. But, this is an academic exersize with no
> > sense or value whatsoever.
>
> > There used to be a legend in our school about the successfully defended
> > master thesis which had the following phrase somewhere in the middle:
> > "As nobody is going to read this anyway, we will make the conductors ou=
t
> > of wood".
>
> > VLV- Hide quoted text -
>
> > - Show quoted text -
>
> If his analysis bandwidth is truly 10 MHz, and he=92s looking at
> anything around 70 dB of dynamic range he=92s pretty much forced into at
> least a 14 bit sub sampling system for an A/D. =A0Something like a AD
> 9245 series would do the job. =A0They're only 12 or so bucks in 100 lot
> sizes. =A0You can add dither to really get a clean spectrum.
>
> I assuming, but sounds like his ref of 100MHz is to a system clk.
> Plenty of choices for that. =A0The question is what=92s his spec ..
> that=92ll be the cost driver.
>
> Somebody has to tune the filters. =A0If there bought or done in house.
> He=92ll have a requirement for pre-selection as well =85 buy=92em.
>
> Then again, he could always do wha V suggest ... pay somebody who
> knows what their doing.- Hide quoted text -
>
> - Show quoted text -


Unless he's after a purely academic exercise he's not going to get
through Tims books and get this job done, at least based on his
opening questions.

Wes's book is highly recommended ..

I didn't see if Tim listed the Analog Devices page ... but it'll be a
good resource for back end info.

On Oct 23, 9:39=A0am, jim <fiscalconsoc...@gmail.com> wrote:
> On Oct 23, 9:32=A0am, jim <fiscalconsoc...@gmail.com> wrote:
>
>
>
>
>
> > On Oct 23, 8:59=A0am, Vladimir Vassilevsky <nos...@nowhere.com> wrote:
>
> > > Tim Wescott wrote:
> > > > On 10/23/2010 06:59 AM, Vladimir Vassilevsky wrote:
>
> > > >> Brian wrote:
>
> > > >>> Dear list,
>
> > > >>> Once I know I will be in the 900 MHz ISM band and the bandwidth o=
f my
> > > >>> signal will be about 10 MHz, how then do I go about selecting RF
> > > >>> center frequency, crystal frequency, intermediate frequency, and
> > > >>> sample rate? Is it good enough to say:
>
> > > >>> RF: 900 MHz
>
> > > >> Nobody would allow that, especially with 10 MHz spectrum.
>
> > > >>> Crystal: 100 MHz
>
> > > >> No such crystal.
>
> > > >>> IF: 25 MHz
>
> > > >> Bad choice. Besides, there is no such IF filter.
>
> > > > Not to criticize the rest of your critique, but at a 10MHz signal
> > > > bandwidth, a 25MHz IF bandwidth can be realized fairly easily with =
a
> > > > home-brew LC filter. =A0Depending again on which of the other unrea=
listic
> > > > goals aren't, it could maybe even be made to sacrifice skirt perfor=
mance
> > > > but use off the shelf parts.
>
> > > > Now, if it needed exceptionally sharp skirts or something, then it'=
d be
> > > > a pain.
>
> > > A filter that needs tuning is a pain in production and a potential
> > > problem with reliability. But, this is an academic exersize with no
> > > sense or value whatsoever.
>
> > > There used to be a legend in our school about the successfully defend=
ed
> > > master thesis which had the following phrase somewhere in the middle:
> > > "As nobody is going to read this anyway, we will make the conductors =
out
> > > of wood".
>
> > > VLV- Hide quoted text -
>
> > > - Show quoted text -
>
> > If his analysis bandwidth is truly 10 MHz, and he=92s looking at
> > anything around 70 dB of dynamic range he=92s pretty much forced into a=
t
> > least a 14 bit sub sampling system for an A/D. =A0Something like a AD
> > 9245 series would do the job. =A0They're only 12 or so bucks in 100 lot
> > sizes. =A0You can add dither to really get a clean spectrum.
>
> > I assuming, but sounds like his ref of 100MHz is to a system clk.
> > Plenty of choices for that. =A0The question is what=92s his spec ..
> > that=92ll be the cost driver.
>
> > Somebody has to tune the filters. =A0If there bought or done in house.
> > He=92ll have a requirement for pre-selection as well =85 buy=92em.
>
> > Then again, he could always do wha V suggest ... pay somebody who
> > knows what their doing.- Hide quoted text -
>
> > - Show quoted text -
>
> Unless he's after a purely academic exercise he's not going to get
> through Tims books and get this job done, at least based on his
> opening questions.
>
> Wes's book is highly recommended ..
>
> I didn't see if Tim listed the Analog Devices page ... but it'll be a
> good resource for back end info.- Hide quoted text -
>
> - Show quoted text -


One other but important point =85 you=92re looking at several hundred K
worth of equipment to do this job, assuming that you buy brand new.

Now you could go gather this stuff up on Ebay =85but you need to know
what to look for.  I suspect if you knew that you wouldn=92t be asking
this question on this news group.

I=92m not saying this to discourage you =85 in fact I hope that you will
pursue your interest if that=92s what this is about.

On the other hand, if your trying to get a project started do consider
the cost of lab equipment.  Renting is certainly an option =85

Hope this helps.

On Sat, 23 Oct 2010 06:36:31 -0700 (PDT), Brian
<heilig.brian@gmail.com> wrote:

>Dear list,
>
>Once I know I will be in the 900 MHz ISM band and the bandwidth of my
>signal will be about 10 MHz, how then do I go about selecting RF
>center frequency, crystal frequency, intermediate frequency, and
>sample rate? Is it good enough to say:
>
>RF: 900 MHz
>Crystal: 100 MHz
>IF: 25 MHz
>Sample rate: 100 MHz
>
>Am I missing something?

As others have suggested, there's a lot to this sort of planning.
There are lots of ways to do this that involve all sorts of tradeoffs
that differ depending on whether the system needs to go into
production or is just a test system, is just a receiver or just
transmits, what type of signal is used, etc., etc., etc.

So don't expect any highly relevant answers in a simple form, because
the answers will depend an awful lot on details that you haven't
given.

FWIW, you could build the system just as you've described, but, as
mentioned, you might have a few issues to overcome along the way.

You might consider alternate architectures, like a direct conversion
system, which is easily done from 900MHz these days.

Be aware that FCC regs (and most international regulating bodies)
limit the either or both of the total transmit power or the
transmitted power spectral density.   As signals get wide, like you
10MHz case, your range may be limited by the transmit power limits.
The good news is that 1W is allowed in that band, so you're probably
not in bad shape.

Since the 900MHz ISM band is only 26MHz wide, you'll have to be very
careful about controlling out-of-band emissions.

Finding a 100MHz oscillator and system clock is not at all difficult,
but why that number?   IMHO 100MHz sounds pretty high for a 10MHz
signal bandwidth unless you have a lot of other processing that needs
to be done, like a video codec or something.

So, again, you can probably make those numbers work, but they may not
represent good tradeoffs for whatever it is you're trying to
accomplish.  The issues are very complex and likely not something
you're going to be able to sort out over the internet for a real
system of any significance.


Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com

I am building a transmitter and receiver. It is not a master's thesis.
I've already graduated. I am a professional DSP developer. I was
performing some system simulation in Matlab and ran into snag. See
this post:

http://groups.google.com/group/comp.dsp/browse_thread/thread/278de8e682a350e7#

I didn't get an answer so I thought I would ask it a different way. I
admit I hadn't done any of the RF system design yet. For example I
have read the ITU spec (a long while ago) but had forgot that 900 MHz
isn't actually in the region 2 band. Silly me.

It is a CDMA system with a chipping rate of 10.23 Mcps in the 900 MHz
band. I figure a 100 mW transmitter will suffice. It's a single
channel design.

I'm in the middle of writing the front end Verilog code for the
receiver correlator (integrate and dump filter at 100 uSec). I went to
test it with an IF of 25 MHz (I just picked it because it is 1/4 the
sampling rate) and the test failed. My test vector did not include any
IF filtering. Turns out the error came from my choice of IF and I
don't know why.

Oh by the way my initial receiver will use the USRP2 for a prototype
which has a 100 MHz VCXO and LTC2284 14 bit ADC at 105 Msps.

Brian  <heilig.brian@gmail.com> wrote:

>I am building a transmitter and receiver. It is not a master's thesis.
>I've already graduated. I am a professional DSP developer. I was
>performing some system simulation in Matlab and ran into snag. See
>this post:
>
>http://groups.google.com/group/comp.dsp/browse_thread/thread/278de8e682a350e7#
>
>I didn't get an answer so I thought I would ask it a different way. I
>admit I hadn't done any of the RF system design yet. For example I
>have read the ITU spec (a long while ago) but had forgot that 900 MHz
>isn't actually in the region 2 band. Silly me.
>
>It is a CDMA system with a chipping rate of 10.23 Mcps in the 900 MHz
>band. I figure a 100 mW transmitter will suffice. It's a single
>channel design.
>
>I'm in the middle of writing the front end Verilog code for the
>receiver correlator (integrate and dump filter at 100 uSec). I went to
>test it with an IF of 25 MHz (I just picked it because it is 1/4 the
>sampling rate) and the test failed. My test vector did not include any
>IF filtering. Turns out the error came from my choice of IF and I
>don't know why.

Seems like it would be more typical to code this in C++ or Matlab
and verify the algorithms (especially including the frequency plan) 
before proceeding to code it in Verilog.

But, whatever works for you.

Steve

> Seems like it would be more typical to code this in C++ or Matlab
> and verify the algorithms (especially including the frequency plan)
> before proceeding to code it in Verilog.

Yes, I coded it in Matlab and was able to produce the same error as my
hardware correlator. When I changed the IF frequency to a different
value (I tried 21 MHz, 20 MHz, 20.46 MHz) the error went away.

Tim Wescott  <tim@seemywebsite.com> wrote:

>On 10/23/2010 06:36 AM, Brian wrote:

>> RF: 900 MHz

>Can you legally transmit a 10MHz wide signal centered on 900MHz?

People commonly say "900 MHz" when what they mean is "An unlicensed
band near 900 MHz" (which will be a different band in different
regions).

Steve

On Sat, 23 Oct 2010 21:21:35 +0000 (UTC), spope33@speedymail.org
(Steve Pope) wrote:

>Tim Wescott  <tim@seemywebsite.com> wrote:
>
>>On 10/23/2010 06:36 AM, Brian wrote:
>
>>> RF: 900 MHz
>
>>Can you legally transmit a 10MHz wide signal centered on 900MHz?
>
>People commonly say "900 MHz" when what they mean is "An unlicensed
>band near 900 MHz" (which will be a different band in different
>regions).
>
>Steve

And the OP specifically said the 900 MHz ISM band, which is 902-928
MHz.  I don't think there's any problem transmitting a 10MHz wide
signal there as long as the peak power, psd, and out-of-band emission
limits are observed.


Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com

On Sat, 23 Oct 2010 14:13:20 -0700 (PDT), Brian
<heilig.brian@gmail.com> wrote:

>> Seems like it would be more typical to code this in C++ or Matlab
>> and verify the algorithms (especially including the frequency plan)
>> before proceeding to code it in Verilog.
>
>Yes, I coded it in Matlab and was able to produce the same error as my
>hardware correlator. When I changed the IF frequency to a different
>value (I tried 21 MHz, 20 MHz, 20.46 MHz) the error went away.

There's nothing that should prevent you from using an IF at fs/4.
That's a fairly common thing to do since it's a convenient way to
build an architecture.


Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com

> There's nothing that should prevent you from using an IF at fs/4.
> That's a fairly common thing to do since it's a convenient way to
> build an architecture.

Thanks, that's what I thought.