[time-nuts] wtd: WWVB info

Jim Lux jimlux at earthlink.net
Thu Aug 6 13:54:06 EDT 2015

On 8/5/15 8:03 PM, Donald wrote:
> On 8/5/2015 6:44 PM, Jim Lux wrote:
>> I'm not sure it would buy you much.. you'd have something running at
>> 240kHz switching the inputs to the detector?
>> It's MUCH easier to just digitize the 60kHz with a high resolution
>> converter.  And have a nice BPF in front of the digitizer.
>> The tayloe/quadrature sampling detector (a key part of the Flexradio
>> original design) is more convenient if you're making a direct
>> conversion receiver that needs to tune up to 10s of MHz, since it
>> allows you to use a slow ADC with lots of bits.  It's basically a I/Q
>> mixer and will take a lot more parts in total than just getting a
>> 192ksps 24 bit converter.
> Thanks for the numbers.
> I was uncertain how to scale the sample rate for the I/Q.

if you want to turn a regular single channel data stream and turn it 
into IQ, an easy way is to do the ++-- approach.

complex sample 1 = real sample 1 + j real sample 2
complex sample 2 = -real sample 3 - j real sample 4
complex sample 3 = real sample 5 + j real sample 6
complex sample 4 = -real sample 7 - j real sample 8
and so forth..

This will give you an I/Q sample stream centered at samplingrate/4.  So 
for 192 ksps (real), fs/4 = 48 kHz, so your 60 kHz signal will be at an 
apparent frequency of +12 kHz (sampled at 96k i/q samples/sec)

As mentioned in a different post, I'd look at a filter/decimate, which 
might be computationally faster.  Say you need 1kHz bandwidth (which I 
think is pretty wide for WWVB)

so you're looking at 192:1 decimation.  with 192 taps, I'll bet you can 
make a fairly decent BPF (and if you're clever, probably bring 60kHz 
down to close to DC at the same time, by combining a synthetic LO with 
the taps)

So you'd be looking at 192 multiplies and adds per 1kHz output point, 
which is a pretty "loafing along" CPU load (e.g. a whole lot of 
microcontrollers could do it)

More information about the time-nuts mailing list