Discussion:
10MHz standard for comms receivers
Add Reply
Bob Betts
2018-10-01 16:43:48 UTC
Reply
Permalink
Hi All. Over the years, I have experimented with WWVB and GPS and Rubidium timing to establish a 10MHz Standard for LO sync in communications receivers. My career has kept me away from the hobby for some time and, frankly, I am not very well versed in some of the timing techniques that I read here...actually I'm just a hacker. Anyhow the thought occurred to me that with some of these newer computer clock timing systems, that a (for example) a 3GHz clock could be divided down by 300 to obtain a fairly tight 10MHz reference. It does have an effect on SDR pgms, but I'm curious about analog systems as well. It would seem (to me) that whatever error there may be at 3GHz, that it would also decrease to a tighter tolerance during the "divide by" process.
Okay, so I've been an RF and Audio products designer for 55 years and really (probably) shouldn't venture into territory unfamiliar to me. But this part is a hobby...so how's a guy supposed to learn.

Will someone please explain the errors in my thinking? it would be appreciated.
Bob, N1KPR http://www.bobsamerica.com  http://www.youtube.com/n1kpr

Engineering: Where Enigma meets Paradox
_______________________________________________
time-nuts mailing list -- time-***@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions the
jimlux
2018-10-01 17:36:26 UTC
Reply
Permalink
Post by Bob Betts
Hi All. Over the years, I have experimented with WWVB and GPS and Rubidium timing to establish a 10MHz Standard for LO sync in communications receivers. My career has kept me away from the hobby for some time and, frankly, I am not very well versed in some of the timing techniques that I read here...actually I'm just a hacker. Anyhow the thought occurred to me that with some of these newer computer clock timing systems, that a (for example) a 3GHz clock could be divided down by 300 to obtain a fairly tight 10MHz reference. It does have an effect on SDR pgms, but I'm curious about analog systems as well. It would seem (to me) that whatever error there may be at 3GHz, that it would also decrease to a tighter tolerance during the "divide by" process.
ppm scales with frequency - a 1 GHz oscillator with 1ppm will divide by
100 to 10 MHz with 1ppm.


Where it gets interesting is with phase noise. Assuming no other
sources of noise (not generally valid, but a good starting place):

If you multiply or divide a source by N, the phase noise (at a given
offset from the carrier) changes by 20 log N. So if phase noise at 1
GHz is -80dBc at 10kHz away, if you divide by 100, the phase noise will
be -120dBc at 10kHz away.

In general, the quietest sources (in terms of phase noise) are crystal
oscillators in the 5-10 MHz range - to a first approximation, it's
because the physical "rock" is big.

So, it's more likely you'd get a quieter 1 GHz source by multiplying
10MHz up, than the other way around.

That's for close in noise...(within some kHz of the carrier) For far
out noise, that's not necessarily the case. You might find that the
microwave oscillator has better noise at 1 Mhz out than the 10 MHz
multiplied up (and degraded by 40 dB). Maybe, maybe not - you need to
look at your particular oscillator.

In general, the wider the tuning range, the worse the noise.


I work a lot at 8.4 GHz - taking something like a HMC506, the phase
noise is -50dBc at 1kHz, -105dBc at 100khz and -125 dBc at 1 MHz

If I lock that to a crystal at 10 MHz.. let's take an inexpensive tiny
OCXO with good but not special noise..
-160 dBc at 1kHz, -165 at 10kHz, and probably the same out to 1 MHz.

I'm going to need to multiply up by 845. so 20logN is 58 dB

At 1kHz, the multiplied up crystal is -102dBc, which is a lot better
than my VCO at -50.
At 100kHz, though, my crystal is -107dBc, and my VCO is at -105 -so
that's pretty much a wash.

At 1 MHz out, my crystal is still -107, but my VCO is down at -125, so
it's quieter.

If I were building a synthesizer, I'd probably make the loop bandwidth
right around 100kHz.
Post by Bob Betts
Okay, so I've been an RF and Audio products designer for 55 years and really (probably) shouldn't venture into territory unfamiliar to me. But this part is a hobby...so how's a guy supposed to learn.
Will someone please explain the errors in my thinking? it would be appreciated.
Bob, N1KPR http://www.bobsamerica.com  http://www.youtube.com/n1kpr
Engineering: Where Enigma meets Paradox
_______________________________________________
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- time-***@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructi
Dana Whitlow
2018-10-01 17:44:09 UTC
Reply
Permalink
Isn't it the case that computer clocks these days are subject to two
influences that make them worthless for timing?

1. Deliberate random FM to spectrally spread RFI leakage.

and

2, Wild variations of clock speed according to usage needs of the moment,
in order to reduce average power consumption and thermal loading.

Dana
Post by Bob Betts
Post by Bob Betts
Hi All. Over the years, I have experimented with WWVB and GPS and
Rubidium timing to establish a 10MHz Standard for LO sync in communications
receivers. My career has kept me away from the hobby for some time and,
frankly, I am not very well versed in some of the timing techniques that I
read here...actually I'm just a hacker. Anyhow the thought occurred to me
that with some of these newer computer clock timing systems, that a (for
example) a 3GHz clock could be divided down by 300 to obtain a fairly tight
10MHz reference. It does have an effect on SDR pgms, but I'm curious about
analog systems as well. It would seem (to me) that whatever error there may
be at 3GHz, that it would also decrease to a tighter tolerance during the
"divide by" process.
ppm scales with frequency - a 1 GHz oscillator with 1ppm will divide by
100 to 10 MHz with 1ppm.
Where it gets interesting is with phase noise. Assuming no other
If you multiply or divide a source by N, the phase noise (at a given
offset from the carrier) changes by 20 log N. So if phase noise at 1
GHz is -80dBc at 10kHz away, if you divide by 100, the phase noise will
be -120dBc at 10kHz away.
In general, the quietest sources (in terms of phase noise) are crystal
oscillators in the 5-10 MHz range - to a first approximation, it's
because the physical "rock" is big.
So, it's more likely you'd get a quieter 1 GHz source by multiplying
10MHz up, than the other way around.
That's for close in noise...(within some kHz of the carrier) For far
out noise, that's not necessarily the case. You might find that the
microwave oscillator has better noise at 1 Mhz out than the 10 MHz
multiplied up (and degraded by 40 dB). Maybe, maybe not - you need to
look at your particular oscillator.
In general, the wider the tuning range, the worse the noise.
I work a lot at 8.4 GHz - taking something like a HMC506, the phase
noise is -50dBc at 1kHz, -105dBc at 100khz and -125 dBc at 1 MHz
If I lock that to a crystal at 10 MHz.. let's take an inexpensive tiny
OCXO with good but not special noise..
-160 dBc at 1kHz, -165 at 10kHz, and probably the same out to 1 MHz.
I'm going to need to multiply up by 845. so 20logN is 58 dB
At 1kHz, the multiplied up crystal is -102dBc, which is a lot better
than my VCO at -50.
At 100kHz, though, my crystal is -107dBc, and my VCO is at -105 -so
that's pretty much a wash.
At 1 MHz out, my crystal is still -107, but my VCO is down at -125, so
it's quieter.
If I were building a synthesizer, I'd probably make the loop bandwidth
right around 100kHz.
Post by Bob Betts
Okay, so I've been an RF and Audio products designer for 55 years and
really (probably) shouldn't venture into territory unfamiliar to me. But
this part is a hobby...so how's a guy supposed to learn.
Post by Bob Betts
Will someone please explain the errors in my thinking? it would be
appreciated.
Post by Bob Betts
Bob, N1KPR http://www.bobsamerica.com http://www.youtube.com/n1kpr
Engineering: Where Enigma meets Paradox
_______________________________________________
To unsubscribe, go to
http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
Post by Bob Betts
and follow the instructions there.
_______________________________________________
To unsubscribe, go to
http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- time-***@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
jimlux
2018-10-01 20:54:43 UTC
Reply
Permalink
Post by Dana Whitlow
Isn't it the case that computer clocks these days are subject to two
influences that make them worthless for timing?
1. Deliberate random FM to spectrally spread RFI leakage.
and
2, Wild variations of clock speed according to usage needs of the moment,
in order to reduce average power consumption and thermal loading.
yes and no..

microcontrollers and things with microcontrollers do use spectrum
dithering, it's less common in a PC.

While the "instruction rate" might vary with the needs of the moment and
die temperature, there's usually some clock and corresponding counter
that runs at a relatively constant rate so the CPU knows what time it is.

That said, a 10 ppm (or even 50 ppm) oscillator in a PC would be "high
performance"

My macbook air shows 56.652 in ntp.drift


_______________________________________________
time-nuts mailing list -- time-***@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
Bob kb8tq
2018-10-01 21:24:31 UTC
Reply
Permalink
Hi

At least on the parts I’ve used, spread spectrum is a software controlled feature.
You use it or not depending on what you are trying to do. There are a lot of systems
out there that have fairly tight timing needs (though not time nut level stuff).

Yes, this all *assumes* you are writing code from scratch for the micro. That may or
may not be the case……

The main reason for spread spectrum is to make EMI requirements easier to pass.
Once you get to a small enough part, it really doesn’t generate all that much EMI internally.
Again, a lot of assumptions get into that.

Bob
Post by jimlux
Post by Dana Whitlow
Isn't it the case that computer clocks these days are subject to two
influences that make them worthless for timing?
1. Deliberate random FM to spectrally spread RFI leakage.
and
2, Wild variations of clock speed according to usage needs of the moment,
in order to reduce average power consumption and thermal loading.
yes and no..
microcontrollers and things with microcontrollers do use spectrum dithering, it's less common in a PC.
While the "instruction rate" might vary with the needs of the moment and die temperature, there's usually some clock and corresponding counter that runs at a relatively constant rate so the CPU knows what time it is.
That said, a 10 ppm (or even 50 ppm) oscillator in a PC would be "high performance"
My macbook air shows 56.652 in ntp.drift
_______________________________________________
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- time-***@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-n
Loading...