Discussion:
Allan Deviation question
(too old to reply)
Bill Dailey
2012-07-04 20:23:30 UTC
Permalink
Raw Message
I am measuring a 10MHz OCXO. I am wondering about my methods and their
effect of Allan Deviation.

SETUP:

I am running the signal into my radio (gps disciplined) and taking the
resulting audio and running it to a sound card (oscillator gps disciplined)
then looking at the result in Spectrum lab. I am using a sample rate of
48kHz and an FFT length of 32,768 and decimating the signal by 48 to give
me decent resolution. The resulting FFT window is 32 seconds. Then
looking at the peak frequency or phase every second (can measure both
instantaneously).

QUESTION:

Does that window time affect my Adev at short time intervals of say 1s to
30s or can these numbers be trusted?

Thanks for your time. Please save esoteric and playful discussions, just
interested in the ability to trust my measurements.

Doc
KX0O
--
Doc

Bill Dailey
KXØO
Azelio Boriani
2012-07-04 23:20:38 UTC
Permalink
Raw Message
I think that no one can say anything on that setup. Anyway, try this: run
the GPSDO 10MHz into the radio and take the measurements. Then run your
signal and let us know the difference. Usually running the reference
against itself (if possible) gives the noise floor of the measurement
setup... if you can't see any difference then maybe the noise is too high.
Post by Bill Dailey
I am measuring a 10MHz OCXO. I am wondering about my methods and their
effect of Allan Deviation.
I am running the signal into my radio (gps disciplined) and taking the
resulting audio and running it to a sound card (oscillator gps disciplined)
then looking at the result in Spectrum lab. I am using a sample rate of
48kHz and an FFT length of 32,768 and decimating the signal by 48 to give
me decent resolution. The resulting FFT window is 32 seconds. Then
looking at the peak frequency or phase every second (can measure both
instantaneously).
Does that window time affect my Adev at short time intervals of say 1s to
30s or can these numbers be trusted?
Thanks for your time. Please save esoteric and playful discussions, just
interested in the ability to trust my measurements.
Doc
KX0O
--
Doc
Bill Dailey
KXØO
_______________________________________________
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Magnus Danielson
2012-07-04 23:22:45 UTC
Permalink
Raw Message
Hi Bill,
Post by Bill Dailey
I am measuring a 10MHz OCXO. I am wondering about my methods and their
effect of Allan Deviation.
I am running the signal into my radio (gps disciplined) and taking the
resulting audio and running it to a sound card (oscillator gps disciplined)
then looking at the result in Spectrum lab. I am using a sample rate of
48kHz and an FFT length of 32,768 and decimating the signal by 48 to give
me decent resolution. The resulting FFT window is 32 seconds.
At this point you should have 1 kHz sampling rate, and the FFT window
will be 32,768 s long. So far my only question is what your radio is
dialled into, as it will give the beat frequency on the audio side and
also the scaling factor for the phase-deviations.
Post by Bill Dailey
Then looking at the peak frequency or phase every second (can measure both
instantaneously).
How do you do that? If you have a 32,7 s long FFT window, you only get
results at that rate. Are they interleaved? 32 interleaved will give you
measures every 1,024 s.
Post by Bill Dailey
Does that window time affect my Adev at short time intervals of say 1s to
30s or can these numbers be trusted?
If you taking the readings out of the top bin on your spectrum plot,
then the FFT window time will scale the "instruments" hardware
bandwidth, as to be expected from the Nyquist theorem. This has proven
not to be all the truth, as there will be a droop (bias error) for the
short-tau ADEV measures, which wears off and is essentially gone after a
decade or so. Hence, using long FFT windows will reduce the rate of
samples and cause short tau measure of interest to be biased and untrusted.

The recommended practice is to use a higher sampling rate, and in your
case shorter FFT window, maybe skip the decimation, such that you can
afford to remove low tau0 multiples such that what you see is what you
can trust.

Please have a look at the Allan Deviation wikipedia page, I think you
will find some useful stuff there. I'd happy to adjust it if needed.
Post by Bill Dailey
Thanks for your time. Please save esoteric and playful discussions, just
interested in the ability to trust my measurements.
It's a fair question to ask. It's also a good question to ask.

It would be good if you could calibrate your setup by using known
sources with known phase-noises. While I like the tool I have, I will
sure that I do my calibrations one way or another.

Cheers,
Magnus
Stewart Bryant
2012-07-06 14:43:51 UTC
Permalink
Raw Message
... and running it to a sound card (oscillator gps disciplined)

How did you achieve this?

Thanks

Stewart
Azelio Boriani
2012-07-06 14:51:02 UTC
Permalink
Raw Message
Yes, I'm also interested in how-to. At the moment I think it is a hack:
there is no sound card AFAIK that accepts a reference input. I have
recently bought an Acqiris/Agilent DP105/U1067A 150MHz 500Ms/s digitizer
PCI card that accepts an external 10MHz as a reference for the sampling
process.
Post by Stewart Bryant
... and running it to a sound card (oscillator gps disciplined)
How did you achieve this?
Thanks
Stewart
______________________________**_________________
To unsubscribe, go to https://www.febo.com/cgi-bin/**
mailman/listinfo/time-nuts<https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts>
and follow the instructions there.
John Ackermann N8UR
2012-07-06 14:58:17 UTC
Permalink
Raw Message
I think a number of higher-end sound cards accept a "word clock" or
"world clock" (I've seen it both ways) that's intended to allow syncing
to an external source. The challenge I've seen is that the frequency
(either in the 12 or 24 MHz range) is one that's not simple to
synthesize precisely (i.e., zero offset from nominal) and with low
jitter/phase noise from a 5 or 10 MHz reference.

I think it would be a great project to come up with a synthesizer block
that could do that. It's been on my list for a long time, but hasn't
yet risen near the top.

John
----
Post by Azelio Boriani
there is no sound card AFAIK that accepts a reference input. I have
recently bought an Acqiris/Agilent DP105/U1067A 150MHz 500Ms/s digitizer
PCI card that accepts an external 10MHz as a reference for the sampling
process.
Post by Stewart Bryant
... and running it to a sound card (oscillator gps disciplined)
How did you achieve this?
Thanks
Stewart
______________________________**_________________
To unsubscribe, go to https://www.febo.com/cgi-bin/**
mailman/listinfo/time-nuts<https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts>
and follow the instructions there.
_______________________________________________
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Heinzmann, Stefan (ALC NetworX GmbH)
2012-07-06 15:52:08 UTC
Permalink
Raw Message
A few cards accept a reference clock in the MHz range (usually under the term "Superclock"), but most wordclock inputs are made for the wordclock frequency itself, for example 48 kHz. An internal PLL (of varying quality) multiplies that up to a multiple of it for driving the converter and shift clocks.

Those PLLs often have a rather wide lock range, so you'd be able to feed 50 kHz (or 100 kHz or 200 kHz) into it, which can be divided down from 10 MHz quite easily.

Not sure whether that's what you want, however.

Cheers
Stefan
-----Ursprüngliche Nachricht-----
Auftrag von John Ackermann N8UR
Gesendet: Freitag, 6. Juli 2012 16:58
An: Discussion of precise time and frequency measurement
Betreff: Re: [time-nuts] Allan Deviation question
I think a number of higher-end sound cards accept a "word clock" or
"world clock" (I've seen it both ways) that's intended to allow syncing
to an external source. The challenge I've seen is that the frequency
(either in the 12 or 24 MHz range) is one that's not simple to
synthesize precisely (i.e., zero offset from nominal) and with low
jitter/phase noise from a 5 or 10 MHz reference.
I think it would be a great project to come up with a synthesizer block
that could do that. It's been on my list for a long time, but hasn't
yet risen near the top.
John
----
Post by Azelio Boriani
Yes, I'm also interested in how-to. At the moment I think it is a
there is no sound card AFAIK that accepts a reference input. I have
recently bought an Acqiris/Agilent DP105/U1067A 150MHz 500Ms/s
digitizer PCI card that accepts an external 10MHz as a reference for
the sampling process.
Post by Stewart Bryant
... and running it to a sound card (oscillator gps disciplined)
How did you achieve this?
Thanks
Stewart
______________________________**_________________
https://www.febo.com/cgi-bin/**
mailman/listinfo/time-nuts<https://www.febo.com/cgi-
bin/mailman/listi
Post by Azelio Boriani
Post by Stewart Bryant
nfo/time-nuts>
and follow the instructions there.
_______________________________________________
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Magnus Danielson
2012-07-06 23:57:38 UTC
Permalink
Raw Message
Post by John Ackermann N8UR
I think a number of higher-end sound cards accept a "word clock" or
"world clock" (I've seen it both ways) that's intended to allow syncing
to an external source. The challenge I've seen is that the frequency
(either in the 12 or 24 MHz range) is one that's not simple to
synthesize precisely (i.e., zero offset from nominal) and with low
jitter/phase noise from a 5 or 10 MHz reference.
I think it would be a great project to come up with a synthesizer block
that could do that. It's been on my list for a long time, but hasn't yet
risen near the top.
A correct word clock should run at sample rate, i.e. 48 kHz or 96 kHz,
which isn't stellar in synthesis difficulty.

For description of Word Clock, please see Appex B in AES 11.

There are system clocks also being used, typically x128 of the sampling
rate, so 6.144 MHz, 12.288 MHz and 24.576 MHz hence is popular. Again
not stellar to lock.

Cheers,
Magnus
Jim Lux
2012-07-06 15:08:36 UTC
Permalink
Raw Message
Post by Azelio Boriani
there is no sound card AFAIK that accepts a reference input. I have
recently bought an Acqiris/Agilent DP105/U1067A 150MHz 500Ms/s digitizer
PCI card that accepts an external 10MHz as a reference for the sampling
process.
while there are no "sound cards" (in the sense that they physically plug
into the PC bus) with external clock inputs, there are quite a few sound
interfaces that take a sample clock input (e.g. at 44.1, 48,96, 192) of
some sort.

A lot of them use 1394/Firewire (for historical reasons)

what you're looking for is "word clock in" or similar things
Post by Azelio Boriani
Post by Stewart Bryant
... and running it to a sound card (oscillator gps disciplined)
How did you achieve this?
Thanks
Stewart
______________________________**_________________
To unsubscribe, go to https://www.febo.com/cgi-bin/**
mailman/listinfo/time-nuts<https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts>
and follow the instructions there.
_______________________________________________
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Jim Lux
2012-07-06 15:09:30 UTC
Permalink
Raw Message
Post by Azelio Boriani
there is no sound card AFAIK that accepts a reference input. I have
recently bought an Acqiris/Agilent DP105/U1067A 150MHz 500Ms/s digitizer
PCI card that accepts an external 10MHz as a reference for the sampling
process.
oops I was wrong.. there ARE people making PCI multichannel audio
interfaces with external clock.. RME in Germany makes more than one.
Azelio Boriani
2012-07-06 15:24:35 UTC
Permalink
Raw Message
OK, found it: the RME HDSPe RayDAT PCIe audio card has this reference but
with the optional extension card (of course). It is $950 for the card + the
expansion card for the world clock...
Post by Jim Lux
Post by Azelio Boriani
there is no sound card AFAIK that accepts a reference input. I have
recently bought an Acqiris/Agilent DP105/U1067A 150MHz 500Ms/s digitizer
PCI card that accepts an external 10MHz as a reference for the sampling
process.
oops I was wrong.. there ARE people making PCI multichannel audio
interfaces with external clock.. RME in Germany makes more than one.
______________________________**_________________
To unsubscribe, go to https://www.febo.com/cgi-bin/**
mailman/listinfo/time-nuts<https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts>
and follow the instructions there.
John Ackermann N8UR
2012-07-06 15:30:29 UTC
Permalink
Raw Message
I don't know if they've been discontinued, but a number of the M-Audio
cards had word clock inputs as well. They are/were pretty widely
available on eBay.

John
Post by Azelio Boriani
OK, found it: the RME HDSPe RayDAT PCIe audio card has this reference but
with the optional extension card (of course). It is $950 for the card + the
expansion card for the world clock...
Post by Jim Lux
Post by Azelio Boriani
there is no sound card AFAIK that accepts a reference input. I have
recently bought an Acqiris/Agilent DP105/U1067A 150MHz 500Ms/s digitizer
PCI card that accepts an external 10MHz as a reference for the sampling
process.
oops I was wrong.. there ARE people making PCI multichannel audio
interfaces with external clock.. RME in Germany makes more than one.
______________________________**_________________
To unsubscribe, go to https://www.febo.com/cgi-bin/**
mailman/listinfo/time-nuts<https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts>
and follow the instructions there.
_______________________________________________
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Mark Spencer
2012-07-06 15:27:14 UTC
Permalink
Raw Message
Word clock generators appear to exist that will accept standard external reference frequencies. One of the vendors also sells a stand alone 10 MHz rubidium reference for driving their word clock generator.

Sent from my iPod
Post by Azelio Boriani
there is no sound card AFAIK that accepts a reference input. I have
recently bought an Acqiris/Agilent DP105/U1067A 150MHz 500Ms/s digitizer
PCI card that accepts an external 10MHz as a reference for the sampling
process.
oops I was wrong.. there ARE people making PCI multichannel audio interfaces with external clock.. RME in Germany makes more than one.
_______________________________________________
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Magnus Danielson
2012-07-07 00:01:58 UTC
Permalink
Raw Message
Post by Jim Lux
Post by Azelio Boriani
there is no sound card AFAIK that accepts a reference input. I have
recently bought an Acqiris/Agilent DP105/U1067A 150MHz 500Ms/s digitizer
PCI card that accepts an external 10MHz as a reference for the sampling
process.
oops I was wrong.. there ARE people making PCI multichannel audio
interfaces with external clock.. RME in Germany makes more than one.
digital audio => need to lock clocks.

Long ago existed audio boards which sample rate converted (badly)
digital audio, this proved to be a killer for pro apps. With digital
audio you need to control your clocking and have things tied to a
master, especially when having multiple sampling boxes. etc.

The analogue world was so much easier in this regard...

Cheers,
Magnus
Bill Dailey
2012-07-06 15:59:54 UTC
Permalink
Raw Message
Removed the oscillator and send it the required freq (24.576MHz) that is disciplined to 10MHz. Put an sma connector next to the card in the little cover on the pci slot. Currently using my ds345 but am getting a Valon Synthesizer board with a divider. Tried ve1alq but the dividers aren't right and can't seem to fix that. Have a vcxo but no way to discipline it. Just going to break down and go the Valon route. Will also use it for 125MHz for my qs1r.

Have done this setup with 2 sound cards. One is a SB xfi surround USB and the other is fata1ity extreme gamer. Works like a charm for both. 20uHz plus minus through my radio (with 10MHz signal) Should be much better with audio but can test because my ds345 is tied up. I would bet I will have substantially sub-uHz precision is the audio only domain.

Doc
KX0O


Sent from my iPhone
Post by Stewart Bryant
... and running it to a sound card (oscillator gps disciplined)
How did you achieve this?
Thanks
Stewart
Azelio Boriani
2012-07-06 16:09:24 UTC
Permalink
Raw Message
OK, I'll take a look at my cheap audio cards searching for the oscillator.
Post by Bill Dailey
Removed the oscillator and send it the required freq (24.576MHz) that is
disciplined to 10MHz. Put an sma connector next to the card in the little
cover on the pci slot. Currently using my ds345 but am getting a Valon
Synthesizer board with a divider. Tried ve1alq but the dividers aren't
right and can't seem to fix that. Have a vcxo but no way to discipline it.
Just going to break down and go the Valon route. Will also use it for
125MHz for my qs1r.
Have done this setup with 2 sound cards. One is a SB xfi surround USB and
the other is fata1ity extreme gamer. Works like a charm for both. 20uHz
plus minus through my radio (with 10MHz signal) Should be much better with
audio but can test because my ds345 is tied up. I would bet I will have
substantially sub-uHz precision is the audio only domain.
Doc
KX0O
Sent from my iPhone
Post by Stewart Bryant
... and running it to a sound card (oscillator gps disciplined)
How did you achieve this?
Thanks
Stewart
_______________________________________________
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hal Murray
2012-07-06 15:27:52 UTC
Permalink
Raw Message
I think a number of higher-end sound cards accept a "word clock" or "world
clock" (I've seen it both ways) that's intended to allow syncing to an
external source. The challenge I've seen is that the frequency (either in
the 12 or 24 MHz range) is one that's not simple to synthesize precisely
(i.e., zero offset from nominal) and with low jitter/phase noise from a 5
or 10 MHz reference.
What happens if you just feed them 10 MHz when it expects 12?

I'd expect it's just the clock to the ADC and DSP. If so, everything will
scale.
--
These are my opinions. I hate spam.
Hal Murray
2012-09-30 17:03:34 UTC
Permalink
Raw Message
My recollection was a bit off -- we saw about 22ns on the two-port 58535a
and about 15ns on the 8-port 58517A. I would guess the 4-port unit would be
similar.
Interesting that it's so far off from the 40 ns in the data sheet.
--
These are my opinions. I hate spam.
Bill Dailey
2012-09-30 17:35:49 UTC
Permalink
Raw Message
I think the usually quote worse case.

Sent from my iPhone and Hunter Lambert is my hero!
Post by Hal Murray
My recollection was a bit off -- we saw about 22ns on the two-port 58535a
and about 15ns on the 8-port 58517A. I would guess the 4-port unit would be
similar.
Interesting that it's so far off from the 40 ns in the data sheet.
--
These are my opinions. I hate spam.
_______________________________________________
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hal Murray
2015-02-08 20:06:18 UTC
Permalink
Raw Message
OK, so I had an HP 58535A two-port GPS splitter handy and put it on the
VNA. It clearly has a filter of some sort, as shown by the S21 frequency
response. The delay at the center of the passband is about 21ns, and it
increases to about 26ns at the edges.
Thanks.

It seems a bit strange that HP didn't mention the delay. That part seems
likely to be used with their GPS gear that is setup to compensate for cable
delays.
--
These are my opinions. I hate spam.



_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hal Murray
2015-02-15 21:28:18 UTC
Permalink
Raw Message
For each reading I clear the counter, then poll using SRQ until data is
ready. The measurement cycle seems to always take between 1.5 and 2
seconds, even if the start/stop signals arrive almost immediately. The GPIB
timeouts are short enough that I don't think they are causing the delay.
I've seen something similar. It may be a delay built into the 5334 before it
starts running a command. That might make sense if you wanted to send
another command to modify the setup.
--
These are my opinions. I hate spam.



_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hal Murray
2015-03-04 20:03:16 UTC
Permalink
Raw Message
Per Tom Clark, who came up with the idea, they are *not* intended to
provide a near-end line termination to 50 ohms, but are simply there to
protect the paralleled devices if the gates have slightly different delays
(in which case one gate could end up sinking the other two).
Is that a real problem? How far off can the prop delay be for 2 gates on the
same chip?

I seem to remember reading something saying it was OK to just wire them up in
parallel. It could have been an app note, or it could have been a rumor on
usenet.
--
These are my opinions. I hate spam.



_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Charles Steinmetz
2015-03-05 00:44:29 UTC
Permalink
Raw Message
Is [one gate sourcing or sinking current into paralleled gates that
don't switch at
exactly the same time] a real problem? How far off can the prop
delay be for 2
gates on the same chip?
I seem to remember reading something saying it was OK to just wire them up in
parallel. It could have been an app note, or it could have been a rumor on
usenet.
Most manufacturers sanction direct paralleling of gates on the same
chip (and forbid any paralleling, ballasted or otherwise, of outputs
from different chips). However, there are other considerations
(short-circuit protection, maximum output current rating per gate,
maximum supply and/or ground current rating per package, maximum
power dissipation, etc.). Good design practice is to put series
resistors on each gate output that will prevent any of these ratings
from being exceeded into a short circuit to ground or to Vcc. For
HC, AC, NC7NZ, and NC7SZ, the per-gate output current rating is +/-
24 (or 25) mA, which suggests that each gate should have a 200 ohm
series resistor -- but you need to check all of the other ratings
mentioned above for the chip you use, to make sure they won't be violated.

Unfortunately, if the load is 50 ohms to ground, pulling it to even
4.5v from a 5v logic supply requires a source resistance of only 5.6
ohms, or 36 parallel gates each with 200 ohms in series (this assumes
that the gates can pull all the way to 5v while delivering rated
current, which they can't -- so the reality is even worse). You can
make things a little better by terminating the output into 50 ohms to
1/2 Vcc (i.e, the center of a voltage divider with 100 ohms to Vcc
and 100 ohms to ground), but (i) it only gets a little better, and
(ii) now it won't pull all the way to ground.

All that said, why do you want to generate a high-current square
wave, anyway? If you're distributing a frequency standard, it is
much better to distribute a sine wave (you don't have to worry about
the harmonics being skewed, and there is exponentially less of a
problem with it radiating and getting into every radio and other
sensitive electronic device in your house and the other houses on the block).

Best regards,

Charles



_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Dan Kemppainen
2015-03-04 20:57:56 UTC
Permalink
Raw Message
Hi,

Correct me if I'm wrong but being that AC series gates are MOS devices,
isn't there inherent current limiting in the MOS junction itself? I
would think that for the few nanoseconds of skew across gates the tens
of ohms of junction resistance would make 'shoot through' negligible in
terms of heating and gate damage.

Of course, a TTL device would be a completely different story and I
would fully expect summing or balancing resistor would be needed there.

Does anyone have further input regarding paralleling MOS logic devices?

Dan
One comment on the parallel AC gate approach. It may not be directly
applicable to Martyn's issue, but there is a common confusion about the
value of the summing resistors.
Per Tom Clark, who came up with the idea, they are*not* intended to
provide a near-end line termination to 50 ohms, but are simply there to
protect the paralleled devices if the gates have slightly different
delays (in which case one gate could end up sinking the other two).
So, the commonly used 47 ohm value isn't magic. You can use a lower
value, and thus get more voltage at the far end. I haven't experimented
to see how far you can take that idea before destroying gates.
John
_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hal Murray
2015-09-01 08:21:58 UTC
Permalink
Raw Message
Attached is a frequency plot of my 5065A ...
The label at the top says "HP 5334A".

Your graph has a resolution of roughly 1E-12. Maybe twice as good. How do
you get that from a 5334? I can only get 11 digits. I'm guessing it's
tangled up with the "tau0 = 3600", but I don't know what that means.
--
These are my opinions. I hate spam.



_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Bob Camp
2015-09-01 10:50:18 UTC
Permalink
Raw Message
Hi

If you hook up a pps into your 5334 from the GPSDO as the “start” and a pulse
from the Rb as the “stop” you will get about 2 ns resolution.

Tau zero is the sample time used. 3,600 sec sounds a lot like a sample per hour.
If you have a tag each hour your resolution is 2x10^-9 / 3600 = ~6 x10^-13.

======

Given what we now know about the sensitivity of the C field, things like humidity
may indeed be part of the long term drift ….

Bob
Post by Hal Murray
Attached is a frequency plot of my 5065A ...
The label at the top says "HP 5334A".
Your graph has a resolution of roughly 1E-12. Maybe twice as good. How do
you get that from a 5334? I can only get 11 digits. I'm guessing it's
tangled up with the "tau0 = 3600", but I don't know what that means.
--
These are my opinions. I hate spam.
_______________________________________________
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hal Murray
2017-10-23 20:52:30 UTC
Permalink
Raw Message
But if any of the EXT REF are low-z inputs, that won't work so well.
Unless there is only one. Then you can use it as the last one and you don't
need the explicit terminator.
--
These are my opinions. I hate spam.



_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
John Ackermann N8UR
2017-10-23 21:05:20 UTC
Permalink
Raw Message
Post by Hal Murray
But if any of the EXT REF are low-z inputs, that won't work so well.
Unless there is only one. Then you can use it as the last one and you don't
need the explicit terminator.
Good point.
_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hal Murray
2018-01-25 21:41:49 UTC
Permalink
Raw Message
The challenge is that the chip is a 7x7 mm 44-QFN package and really wants
to be put on a six-layer circuit board. That's doable, but challenging,
for home assembly.
Can anybody comment on the toaster oven approach?

Is it practical for things like this? How much does a solder mask cost? How
much other stuff do I need? Does the solder paste need to be refrigerated
and other quirks like that?

What are the chances of a newbie getting a 44-QFN right on the first try?
--
These are my opinions. I hate spam.



_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Bob kb8tq
2018-01-25 21:53:04 UTC
Permalink
Raw Message
Hi

You can mod toaster ovens, they work ok, EEVB has *lots* of info on that. When you buy
your PCB you can get solder stencils at the same time. Bought as a package they are in the
$10 or so range. Without buying the boards with them, I’m sure the price goes up a bit. The
metal ones are a bit more than plastic ones.

Solder paste should be refrigerated if you want it to last. How long it will do sitting on the bench
is a “that depends” sort of thing. Amazon will send you a (small) tube for $10 or so.

The whole “can you do it” depends a bit on how good your microscope is and how steady your
hands are.

Again, EEVB is your friend if you really want to get into this.

Bob
Post by Hal Murray
The challenge is that the chip is a 7x7 mm 44-QFN package and really wants
to be put on a six-layer circuit board. That's doable, but challenging,
for home assembly.
Can anybody comment on the toaster oven approach?
Is it practical for things like this? How much does a solder mask cost? How
much other stuff do I need? Does the solder paste need to be refrigerated
and other quirks like that?
What are the chances of a newbie getting a 44-QFN right on the first try?
--
These are my opinions. I hate spam.
_______________________________________________
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Mark Goldberg
2018-01-25 21:55:23 UTC
Permalink
Raw Message
I have had very good luck with a converted toaster oven, GC-10 solder
paste, and OSH Stencils metal stencils. Basically, if the temperature
profile is good and you have good solder paste application, decent
placement, good solder mask and correct pad sizes, everything solders
itself. I have built 70 of my boards with zero solder defects. I use a 4
pin castellated part.

https://sites.google.com/site/markstcxo/
https://sites.google.com/site/markscontroleo2build/
https://drive.google.com/open?id=0B2gd5QRdoS7BVTRSNzZFTTB6RlU

In this case though, I would probably opt for the eval board at $150. I
have spent way too much time on my small board project and there are lots
of little details to getting a clean oscillator.

73,

Mark
W7MLG
Post by Hal Murray
The challenge is that the chip is a 7x7 mm 44-QFN package and really
wants
to be put on a six-layer circuit board. That's doable, but challenging,
for home assembly.
Can anybody comment on the toaster oven approach?
Is it practical for things like this? How much does a solder mask cost?
How
much other stuff do I need? Does the solder paste need to be refrigerated
and other quirks like that?
What are the chances of a newbie getting a 44-QFN right on the first try?
--
These are my opinions. I hate spam.
_______________________________________________
To unsubscribe, go to https://www.febo.com/cgi-bin/
mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
J. Grizzard
2018-01-26 00:54:18 UTC
Permalink
Raw Message
Post by Hal Murray
Can anybody comment on the toaster oven approach?
Is it practical for things like this? How much does a solder mask cost? How
much other stuff do I need? Does the solder paste need to be refrigerated
and other quirks like that?
What are the chances of a newbie getting a 44-QFN right on the first try?
If you have a toaster that you have modified to retain heat better and
has an actual reflow controller, the toaster approach works /really/
well. The toaster I built follows the JEDEC-standard reflow profile
pretty much to the letter, other than taking longer to cool down than it
should (due to not having any sort of fan to move air). So as far as the
solder and components are concerned, it's basically no different than a
decent commercial reflow oven.

If you (or anyone) goes the toaster route, I really recommend getting a
conversion kit -- I used the Controleo2 (from http://whizoo.com/) --
they have a v3 now with a graphical touchscreen. You want the kit
(rather than just the controller) because a normal toaster oven just
leaks too much heat, and won't heat up fast enough, so you need to do a
bunch of other things to one to actually be able to hit your curves.
(They have their whole build guide at http://whizoo.com/reflowoven, so
you can see what you have to do. I built mine over a weekend. I believe
their current kit comes with everything you need but the toaster (their
previous kit required the addition of sealant, a tray, and the
insulation blanket)).

It's basically fire-and-forget -- populate board, stick in oven, press
'start', come back in a bit...

Stencils are pretty cheap (and fast!). I don't remember the exact price
off the top of my head, but I did a not-small board recently (60x110mm),
and the stencil (via OSHStencils) was less than $10, and got to me in
two days, and they're pretty much the same stencils any assembly house
would use.

The main problem with solder paste is that the flux degrades over time
(and you /really/ want your flux). Refrigeration slows that down, but
there's limits. You can, though, buy your solder paste in small
quantities (I think a 15g syringe, which will do a pretty decent number
of boards, is ~$15), so having it eventually go bad isn't that big a
deal. People have reported getting ok results with years-old paste,
though -- I suspect the results will be at least partially dependent on
the details of your board design (how fine-pitch the footprints are, and
such).

Odds of getting a 44-QFN right on the first try are pretty good. Neither
the chip nor the solder paste need be perfectly aligned for things to
work, at least if you're using a PCB with a proper soldermask. At that
pitch, you could probably even just smear some solder paste over the
pads, place the chip, and have success (though I still suggest a
stencil). The biggest problem source (for me, at least) is getting /too
much/ solder paste down (and ending up with bridge pins because there's
just no other place for the solder to go), thus the use of a stencil...

-j
_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Forrest Christian (List Account)
2018-01-26 00:54:53 UTC
Permalink
Raw Message
So, I happen to have a full low-volume SMD assembly line here... for our
own products (although I did have similar thoughts to John about spinning a
couple of carrier boards for these type of parts but designed so they are
suitable for assembly on our line).

Our take on QFN's is that they're not as bad as one would imagine,
especially if one extends the pads outside of the QFN footprint so you can
have a chance of reworking them, since most re-work issues are an issue of
a bit of flux and heating the solder up to melting point. I still try to
avoid them if at all possible, because I hate something you can't easily
inspect, and QFN's are pretty much in that category. The problem with
inspection being that you have to generally have expensive inspection
equipment, the most common being an x-ray machine, in order to really tell
if the part is soldered correctly. If you don't have this, you pretty much
have to rely on a functional test which can be problematic since there are
a lot of solder defects which result in boards which test fine, yet are not
truly soldered correctly - which fail in the field.

But, everytime that I've relented and used them, they've been remarkably
trouble-free, often easier to deal with than an equivalent pitch TQFP since
bridges/soldering seem to happen less often than on a leaded part, and
usually these issues clear just by applying some flux and reheating the
joints. Assuming you find the bridge/defect in the first place.

The especially troublesome QFN's are the ones with 'interior pads' since
there is no easy way to see how well they soldered, and reworking those
joints are a challenge. The single ground mostly-thermal pad ones aren't
too bad (such as the part we're talking about), with the caveat that you
have to put some thought into how to handle the ground vias so they don't
suck the all the solder from the pad into the via holes. This usually
means plugged vias (aka small enough that the plating fills 100% of the
hole). In addition there is a lot of discussion about how much voiding
(unsoldered area) is acceptable on that center pad, and the answer
generally is that "it depends". But, when soldering, with a reasonable
stencil design, you're going to typically get more than enough fill to not
have any problems.

The QFN's with multiple interior pads, I've tried successfully to stay away
from, since it seems that defects are much more likely on these. Although
there's a voltage regulator wih this pad style that I've got my eye on that
I'm seriously considering. But that one is unique in that the center pads
share vias with perimeter pads, so you can just run a single pad all the
way from the edge under the unit, so it would still be possible to reheat
the joint.

I understand that some people have had luck hand-soldering QFN's with the
center pads by adding good sized vias where heat and possibly solder can be
applied *through* the board. I'm not sure I would trust this for
production, though.

As far as doing this at home in a toaster oven, I wouldn't be surprised if
it was not only possible but worked well, assuming everything else was
fine. With modern pastes and components, the soldering process is
remarkably insensitive to variations.

One hint: If you do want to experiment, there are 'dummy components'
available which could help with the verification process and cost less than
the real chips. If you search for "QFN44 dummy component" you'll find
topline and maybe another vendor or two. These are definitely less
expensive than expensive parts, but in most cases, I've also discovered
that I have been able to find some other very low cost 'real' component in
the same package.

NXP has a app note at
https://www.nxp.com/docs/en/application-note/AN1902.pdf which covers the
basics of what I described above.

One final note to mention: Many/most QFN's are moisture sensitive. This
pretty much means that once you open the package you have a limited amount
of time to either mount them, or put them in a dry box. (or re-package
them in a moisture proof container with an appropriate dessicant pack).
If this doesn't happen correctly, then the part absorbs enough moisture
from the air that when you bake it, the part cracks as it turns to steam.
This is sometimes visible, sometimes not. Either way, is causes
reliabilty issues.

The data I have access to indicates that the SI5340A is currently rated at
MSL2, which means that the 'open time' is 1 Year (assuming normal humidity
levels). BUT... you never know until you get the package, and even then
you should double check with the manufacturer based on the exact
manufacturing date and factory.
Post by Hal Murray
The challenge is that the chip is a 7x7 mm 44-QFN package and really
wants
to be put on a six-layer circuit board. That's doable, but challenging,
for home assembly.
Can anybody comment on the toaster oven approach?
Is it practical for things like this? How much does a solder mask cost?
How
much other stuff do I need? Does the solder paste need to be refrigerated
and other quirks like that?
What are the chances of a newbie getting a 44-QFN right on the first try?
--
These are my opinions. I hate spam.
_______________________________________________
To unsubscribe, go to https://www.febo.com/cgi-bin/
mailman/listinfo/time-nuts
and follow the instructions there.
--
*Forrest Christian* *CEO**, PacketFlux Technologies, Inc.*
Tel: 406-449-3345 | Address: 3577 Countryside Road, Helena, MT 59602
***@imach.com | http://www.packetflux.com
<http://www.linkedin.com/in/fwchristian> <http://facebook.com/packetflux>
<http://twitter.com/@packetflux>
_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Mark Sims
2018-01-25 22:10:39 UTC
Permalink
Raw Message
OSHSTENCILS.COM sells stencils by the square inch. They have 4 mil stainless, 3 mil Kapton, and 5 mil Kapton... I usually use the stainless ones. They also sell syringes of solder paste. I keep mine in the fridge. I have year old paste that works fine. There are now pastes that do not require refrigeration... I don't know how good they are.

QFN44 pads are spaced 0.5mm. You don't have to get the chip set down perfectly. Surface tension will align it when the solder melts. I've hand placed chip scale packages with like .25mm pads.

I use a modified toaster oven. My temperature control PID is derived from Lady Heather's temperature control algorithm. There are a LOT of Arduino, etc reflow oven controllers out there.

---------------------
Post by Hal Murray
Can anybody comment on the toaster oven approach?
Is it practical for things like this? How much does a solder mask cost? How
much other stuff do I need? Does the solder paste need to be refrigerated
and other quirks like that?

What are the chances of a newbie getting a 44-QFN right on the first try?
_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
David fav
2018-01-25 22:22:33 UTC
Permalink
Raw Message
I used to do SMT If the solder paste does not absorb moisture when
stored it should be OK

People have had success with an iron frying pan on the stove .
I have used a hot air paint stripper gun with a metal funnel to re flow the
paste.
There cheap SMT re work stations on Alibaba now <$100
Post by Mark Sims
OSHSTENCILS.COM sells stencils by the square inch. They have 4 mil
stainless, 3 mil Kapton, and 5 mil Kapton... I usually use the stainless
ones. They also sell syringes of solder paste. I keep mine in the
fridge. I have year old paste that works fine. There are now pastes
that do not require refrigeration... I don't know how good they are.
QFN44 pads are spaced 0.5mm. You don't have to get the chip set down
perfectly. Surface tension will align it when the solder melts. I've
hand placed chip scale packages with like .25mm pads.
I use a modified toaster oven. My temperature control PID is derived
from Lady Heather's temperature control algorithm. There are a LOT of
Arduino, etc reflow oven controllers out there.
---------------------
Post by Hal Murray
Can anybody comment on the toaster oven approach?
Is it practical for things like this? How much does a solder mask cost?
How
much other stuff do I need? Does the solder paste need to be refrigerated
and other quirks like that?
What are the chances of a newbie getting a 44-QFN right on the first try?
_______________________________________________
To unsubscribe, go to https://www.febo.com/cgi-bin/
mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Mark Sims
2018-01-26 02:13:34 UTC
Permalink
Raw Message
I'm using a toaster oven I got from Walmart. It said it was 1500 watts on the box, but is actually 1800 watts. It has two sets if heating elements. I didn't insulate it at all and am only using one set of elements... 900 watts. It hits my PbSn curve quite well. When the heat cycle is done, it beeps and you open the door. The cool down is close to the official recommendations. My controller is set to keep the oven pre-heated to 100C.

The controller is based on a ATmega 2561 with a LCD touchscreen. The temp control PID is based upon the one in Lady Heather. I've been meaning to put a more powerful fan in it and/or a servo to open the door and/or try using the second set of heating elements and/or try some insulation / reflective film, but my lazy bastarditis keeps getting in the way.

--------------------
You want the kit (rather than just the controller) because a normal toaster oven just
leaks too much heat, and won't heat up fast enough, so you need to do a
bunch of other things to one to actually be able to hit your curves.
_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hal Murray
2018-03-29 21:57:24 UTC
Permalink
Raw Message
From info for a 332A, YMMV, etc...
(a) is the 6P6C modular cable
connecting the receiver and decoder straight, or reversed
Receiver/decoder cable RJ11-4/6 wired pin 1 to pin 1 - 500
feet maximum length
(b) what are the four DIP switches on the back of the 333 used for?
Option switches - 4 position dip switch provides installed
option selection as follows:
1 Battery backup - ON is enabled
2 Report time each second - ON is enabled
3,4 Reserved

Backup Battery Enable
Upon power up, place Option Switch 1 in the ON
position to enable RTC battery backup. During ship-
ment and periods of no use, set this switch to OFF to
avoid battery cell reversal.

Reset
The decoder may be reset by removing power and
setting Option Switch 1 to OFF for a few seconds.
--
These are my opinions. I hate spam.



_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Hal Murray
2018-04-13 22:09:29 UTC
Permalink
Raw Message
The modular cable connecting the receiver to the decoder is wired straight
through, not reversed as most telephone cables are. My fear is that
someone (like me) might at one point have used a reverse cable and thus put
reverse polarity on the board; I don't see any reverse power protection.
Many thanks for that comment. I hadn't checked that before.

My cable was reversed. A new/correct one arrived today. The meter is
working. With a bit of practice, I'll bet I could read the data pattern off
the wiggles.

But after an hour, the LCD still shows time-since-power-up rather than date
and time.
--
These are my opinions. I hate spam.



_______________________________________________
time-nuts mailing list -- time-***@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Loading...