Correct sampling frequency for Ham remote receiver 472-479 KHz?

Retired physician and long time Ham.  DSP NOOB.

I want to put a remote receiver for the proposed 472-479 KHz Ham band in a
quiet site about 10 miles away line of sight.  Already have an
experimental tkt.  Will be solar/off grid transceiver...digital conversion
to VHF digital link to shack....data stream to pretty standard baseband
such as the Softrocks use.  Slow digital data only.

Looked at many possibilities.  Settled on an AD7764 with tight RF filter
on input.  Take advantage of the 256 X decimation to lower the data rate. 
Dump into a small Arduino or such board to buffer the data stream and into
a simple RF transmitter for the link.  Almost no need for a very stable LO
as I can send a weak signal to the RX of known frequency for calibration
on the RX end.

I've been through Lyon's book where he covers the choice of Fs in detail,
but I'm lost in the details:

Is the BW I use the somewhat greater than 7 KHz band of interest or a much
narrower CW or such bandwidth with slow keying or frequency shift?

The internal clock of the device, FICLK, is spec'd from 250 KHz to 20 MHz.
 Do I want to use a lower clock and end up with a narrow output bandwidth
or a near max clock and do something with the signal either before the
link or in the shack?

Guessing the output data should be a 3 to 10 KHz range to fit an audio
bandwidth and stay away from the DC component and RF filter slopes.

Or should I just punt and hang a softrock out there, send AF data, and
play like a softrock with stuff in the middle.

Sorry for the long question, Lyon's book and lots of stuff on the WEB well
explain what is going on but I'm in the stage of learning as in 55 years
ago in Calc 1 where the text says "It is obvious that....."  

Thanks,  N0UU


---------------------------------------
Posted through http://www.DSPRelated.com

On Tue, 14 Jul 2015 07:08:22 -0500, N0UU wrote:

> Retired physician and long time Ham.  DSP NOOB.
> 
> I want to put a remote receiver for the proposed 472-479 KHz Ham band in
> a quiet site about 10 miles away line of sight.  Already have an
> experimental tkt.  Will be solar/off grid transceiver...digital
> conversion to VHF digital link to shack....data stream to pretty
> standard baseband such as the Softrocks use.  Slow digital data only.
> 
> Looked at many possibilities.  Settled on an AD7764 with tight RF filter
> on input.  Take advantage of the 256 X decimation to lower the data
> rate. Dump into a small Arduino or such board to buffer the data stream
> and into a simple RF transmitter for the link.  Almost no need for a
> very stable LO as I can send a weak signal to the RX of known frequency
> for calibration on the RX end.
> 
> I've been through Lyon's book where he covers the choice of Fs in
> detail, but I'm lost in the details:
> 
> Is the BW I use the somewhat greater than 7 KHz band of interest or a
> much narrower CW or such bandwidth with slow keying or frequency shift?
> 
> The internal clock of the device, FICLK, is spec'd from 250 KHz to 20
> MHz.
>  Do I want to use a lower clock and end up with a narrow output
>  bandwidth
> or a near max clock and do something with the signal either before the
> link or in the shack?
> 
> Guessing the output data should be a 3 to 10 KHz range to fit an audio
> bandwidth and stay away from the DC component and RF filter slopes.
> 
> Or should I just punt and hang a softrock out there, send AF data, and
> play like a softrock with stuff in the middle.
> 
> Sorry for the long question, Lyon's book and lots of stuff on the WEB
> well explain what is going on but I'm in the stage of learning as in 55
> years ago in Calc 1 where the text says "It is obvious that....."

This may help: http://wescottdesign.com/articles/Sampling/sampling.pdf

Basically, you want to sample at a high enough rate so that after 
aliasing, the off-frequency stuff that comes through the filter and gets 
aliased into your passband is significantly smaller than the expected 
atmospheric noise.

Of course, this depends on what you expect to be out there on the skirts 
of your filter, so there's some b'guess and b'gosh, but that's a typical 
problem for radio.

You also want to sample at a rate that puts the center of your filter 
passband roughly at 1/4 of your sampling rate.

Beyond that, sampling as fast as you can will let you average out the ADC 
noise (it's more or less getting you coding gain).

-- 
www.wescottdesign.com

Thanks for the link.  Takes care of the high/low sampling question.  Will
use Lyon's example to pick a high Fs that gets the desired signal to where
I can transmit it and go from there (with the band limited stuff above
that basically ignored for later processing).  N0UU
---------------------------------------
Posted through http://www.DSPRelated.com

"N0UU" <107156@DSPRelated> writes:

> Retired physician and long time Ham.  DSP NOOB.
>
> I want to put a remote receiver for the proposed 472-479 KHz Ham band in a
> quiet site about 10 miles away line of sight.  Already have an
> experimental tkt.  Will be solar/off grid transceiver...digital conversion
> to VHF digital link to shack....data stream to pretty standard baseband
> such as the Softrocks use.  Slow digital data only.
>
> Looked at many possibilities.  Settled on an AD7764 with tight RF filter
> on input.  Take advantage of the 256 X decimation to lower the data rate. 
> Dump into a small Arduino or such board to buffer the data stream and into
> a simple RF transmitter for the link.  Almost no need for a very stable LO
> as I can send a weak signal to the RX of known frequency for calibration
> on the RX end.

Hi N0UU,

Perhaps I missed it, but are you planning on doing a direct conversion
receiver (with an analog mixer at the front-end)?  
-- 
Randy Yates
Digital Signal Labs
http://www.digitalsignallabs.com

Tim Wescott <tim@seemywebsite.com> writes:

> On Tue, 14 Jul 2015 07:08:22 -0500, N0UU wrote:
>
>> Retired physician and long time Ham.  DSP NOOB.
>> 
>> I want to put a remote receiver for the proposed 472-479 KHz Ham band in
>> a quiet site about 10 miles away line of sight.  Already have an
>> experimental tkt.  Will be solar/off grid transceiver...digital
>> conversion to VHF digital link to shack....data stream to pretty
>> standard baseband such as the Softrocks use.  Slow digital data only.
>> 
>> Looked at many possibilities.  Settled on an AD7764 with tight RF filter
>> on input.  Take advantage of the 256 X decimation to lower the data
>> rate. Dump into a small Arduino or such board to buffer the data stream
>> and into a simple RF transmitter for the link.  Almost no need for a
>> very stable LO as I can send a weak signal to the RX of known frequency
>> for calibration on the RX end.
>> 
>> I've been through Lyon's book where he covers the choice of Fs in
>> detail, but I'm lost in the details:
>> 
>> Is the BW I use the somewhat greater than 7 KHz band of interest or a
>> much narrower CW or such bandwidth with slow keying or frequency shift?
>> 
>> The internal clock of the device, FICLK, is spec'd from 250 KHz to 20
>> MHz.
>>  Do I want to use a lower clock and end up with a narrow output
>>  bandwidth
>> or a near max clock and do something with the signal either before the
>> link or in the shack?
>> 
>> Guessing the output data should be a 3 to 10 KHz range to fit an audio
>> bandwidth and stay away from the DC component and RF filter slopes.
>> 
>> Or should I just punt and hang a softrock out there, send AF data, and
>> play like a softrock with stuff in the middle.
>> 
>> Sorry for the long question, Lyon's book and lots of stuff on the WEB
>> well explain what is going on but I'm in the stage of learning as in 55
>> years ago in Calc 1 where the text says "It is obvious that....."
>
> This may help: http://wescottdesign.com/articles/Sampling/sampling.pdf
>
> Basically, you want to sample at a high enough rate so that after 
> aliasing, the off-frequency stuff that comes through the filter and gets 
> aliased into your passband is significantly smaller than the expected 
> atmospheric noise.
>
> Of course, this depends on what you expect to be out there on the skirts 
> of your filter, so there's some b'guess and b'gosh, but that's a typical 
> problem for radio.
>
> You also want to sample at a rate that puts the center of your filter 
> passband roughly at 1/4 of your sampling rate.
>
> Beyond that, sampling as fast as you can will let you average out the ADC 
> noise (it's more or less getting you coding gain).

Hi Tim,

I guess you're assuming he wants to do direct conversion using
bandpass sampling?

Unfortunately I don't think the AD7764 would support this. On p. 6 of
the datasheet, it is stated that the differential amplifier's bandwidth
"for 0.1 dB flatness" is 125 kHz. I don't think you can assume that such
a front-end will be good at almost two octaves (479 kHz) higher.

Or am I missing something?
-- 
Randy Yates
Digital Signal Labs
http://www.digitalsignallabs.com

Planned to AD with the AD7764 right from the antenna and bandpass filter,
dumping the serial string right into some simple board, throw away the
last 6 non-signal bits and send the resulting 24 bit data stream over RF
to the shack.  Choices are to sample fast (roughly Fs 20 MHz) and then do
a quick filter and decimation at the remote location before sending the
roughly 7 KHz wide band data I want to watch OR to use a slower Fs in the
range of 5 MHz so as to basically down convert the 472-479 KHz signals to
the (say) 3-10 KHz range for processing at the shack with whatever AF like
programs I want such as Spectran, the Softrock stuff, etc.  The down
conversion is in the AD sample process.  Thanks,
---------------------------------------
Posted through http://www.DSPRelated.com

On Tue, 14 Jul 2015 17:43:28 -0400, Randy Yates wrote:

> Tim Wescott <tim@seemywebsite.com> writes:
> 
>> On Tue, 14 Jul 2015 07:08:22 -0500, N0UU wrote:
>>
>>> Retired physician and long time Ham.  DSP NOOB.
>>> 
>>> I want to put a remote receiver for the proposed 472-479 KHz Ham band
>>> in a quiet site about 10 miles away line of sight.  Already have an
>>> experimental tkt.  Will be solar/off grid transceiver...digital
>>> conversion to VHF digital link to shack....data stream to pretty
>>> standard baseband such as the Softrocks use.  Slow digital data only.
>>> 
>>> Looked at many possibilities.  Settled on an AD7764 with tight RF
>>> filter on input.  Take advantage of the 256 X decimation to lower the
>>> data rate. Dump into a small Arduino or such board to buffer the data
>>> stream and into a simple RF transmitter for the link.  Almost no need
>>> for a very stable LO as I can send a weak signal to the RX of known
>>> frequency for calibration on the RX end.
>>> 
>>> I've been through Lyon's book where he covers the choice of Fs in
>>> detail, but I'm lost in the details:
>>> 
>>> Is the BW I use the somewhat greater than 7 KHz band of interest or a
>>> much narrower CW or such bandwidth with slow keying or frequency
>>> shift?
>>> 
>>> The internal clock of the device, FICLK, is spec'd from 250 KHz to 20
>>> MHz.
>>>  Do I want to use a lower clock and end up with a narrow output
>>>  bandwidth
>>> or a near max clock and do something with the signal either before the
>>> link or in the shack?
>>> 
>>> Guessing the output data should be a 3 to 10 KHz range to fit an audio
>>> bandwidth and stay away from the DC component and RF filter slopes.
>>> 
>>> Or should I just punt and hang a softrock out there, send AF data, and
>>> play like a softrock with stuff in the middle.
>>> 
>>> Sorry for the long question, Lyon's book and lots of stuff on the WEB
>>> well explain what is going on but I'm in the stage of learning as in
>>> 55 years ago in Calc 1 where the text says "It is obvious that....."
>>
>> This may help: http://wescottdesign.com/articles/Sampling/sampling.pdf
>>
>> Basically, you want to sample at a high enough rate so that after
>> aliasing, the off-frequency stuff that comes through the filter and
>> gets aliased into your passband is significantly smaller than the
>> expected atmospheric noise.
>>
>> Of course, this depends on what you expect to be out there on the
>> skirts of your filter, so there's some b'guess and b'gosh, but that's a
>> typical problem for radio.
>>
>> You also want to sample at a rate that puts the center of your filter
>> passband roughly at 1/4 of your sampling rate.
>>
>> Beyond that, sampling as fast as you can will let you average out the
>> ADC noise (it's more or less getting you coding gain).
> 
> Hi Tim,
> 
> I guess you're assuming he wants to do direct conversion using bandpass
> sampling?
> 
> Unfortunately I don't think the AD7764 would support this. On p. 6 of
> the datasheet, it is stated that the differential amplifier's bandwidth
> "for 0.1 dB flatness" is 125 kHz. I don't think you can assume that such
> a front-end will be good at almost two octaves (479 kHz) higher.
> 
> Or am I missing something?

No, I didn't even bother to look at the data sheet.  Bad me.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

On Tue, 14 Jul 2015 16:48:47 -0500, N0UU wrote:

> Planned to AD with the AD7764 right from the antenna and bandpass
> filter, dumping the serial string right into some simple board, throw
> away the last 6 non-signal bits and send the resulting 24 bit data
> stream over RF to the shack.  Choices are to sample fast (roughly Fs 20
> MHz) and then do a quick filter and decimation at the remote location
> before sending the roughly 7 KHz wide band data I want to watch OR to
> use a slower Fs in the range of 5 MHz so as to basically down convert
> the 472-479 KHz signals to the (say) 3-10 KHz range for processing at
> the shack with whatever AF like programs I want such as Spectran, the
> Softrock stuff, etc.  The down conversion is in the AD sample process. 

See Randy's comment about the AD7764's front end -- there may be useful 
signal at 475kHz, but it'll be down, and the noise will be unabated.

So you're probably looking at an external sampler, or at mixing the 
signal down.  If you look at the schematics for things that digital guys 
call "samplers", you'll see that they're what radio guys call "mixers" -- 
so all the difficulties of dynamic range and impedance matching and 
whatnot that apply to mixers also apply to samplers.  Fortunately, Zeus 
(or whoever the God of Electrostatic Discharge is) provides lots of 
atmospheric noise in that frequency, so you can make excuses at the low 
end -- but that just eliminates excuses at the high end when you 
attenuate the signal down so that the atmospheric noise just barely 
overwhelms the front-end noise.

I alluded to this before, but since I've mentioned the ADC front-end 
noise, my policy whenever I care about wringing the most performance from 
an ADC is to sample it as fast as I can, then average the results.  Most  
ADCs that are near the top of their class in bandwidth / bit count have 
well over 1LSB of front-end noise -- the faster you sample, the more you 
can average this noise out.

-- 

Tim Wescott
Formerly KG7LI, formerly N1FUW
Wescott Design Services
http://www.wescottdesign.com

Thanks all.  Gave me some insight on the process.  I hadn't planned on
using the preamp for dynamic range reasons and also missed the 1 dB
bandwidth.  I'm gone for three weeks and will see how this goes after
that.  Will much later respond with results.  As I said before, there are
many standard ways to do this, but looking at the chip I thought maybe a
simple chip for the RX, another for the data handling and an RF link using
only a bit of power might be interesting.  And 24 bits rather than 16 used
for the HF stuff.  We'll see.  Thanks again for the help.  N0UU, Ham call 
WG2XUX, LF experimental tkt.

---------------------------------------
Posted through http://www.DSPRelated.com

Tim Wescott <seemywebsite@myfooter.really> wrote:
> On Tue, 14 Jul 2015 16:48:47 -0500, N0UU wrote:
 
>> Planned to AD with the AD7764 right from the antenna and bandpass
>> filter, dumping the serial string right into some simple board, throw
>> away the last 6 non-signal bits and send the resulting 24 bit data
>> stream over RF to the shack.  

(snip)
> So you're probably looking at an external sampler, or at mixing the 
> signal down.  If you look at the schematics for things that digital guys 
> call "samplers", you'll see that they're what radio guys call "mixers" -- 
> so all the difficulties of dynamic range and impedance matching and 
> whatnot that apply to mixers also apply to samplers.  Fortunately, Zeus 
> (or whoever the God of Electrostatic Discharge is) provides lots of 
> atmospheric noise in that frequency, so you can make excuses at the low 
> end -- but that just eliminates excuses at the high end when you 
> attenuate the signal down so that the atmospheric noise just barely 
> overwhelms the front-end noise.

There is a lot of both natural and manmade noise at lower frequencies.

Reminds me years ago if plans for a 20kHz receiver and antenna
for the digital form of WWV.  The antenna is a shielded loop.
You make a loop out of electrical conduit, connect it to a box
with an insulated coupling on one end, and run some not too large
number of turns of wire around the loop (inside the conduit).

One way to look at it is that the loop picks up the magnetic 
part of the EM wave, instead of the electric part that most
antennas pick up. It seems that makes a big difference at the
lower frequencies.

> I alluded to this before, but since I've mentioned the ADC front-end 
> noise, my policy whenever I care about wringing the most performance from 
> an ADC is to sample it as fast as I can, then average the results.  Most  
> ADCs that are near the top of their class in bandwidth / bit count have 
> well over 1LSB of front-end noise -- the faster you sample, the more you 
> can average this noise out.
 
-- glen