JP
Jim Palfreyman
Thu, Jun 19, 2008 7:08 AM
Hi Folks,
Well my 5370B purchased off ebay arrived from the US in perfect
condition. It passed all the operator checks (all well within spec)
and the only issue being that the internal 10811 oscillator was
running a tad slow. A quarter of a turn with the screwdriver fixed
that.
Having played with it for a day or so I have a few questions. I was
wondering if there was an FAQ anywhere?
My first issue was powering it up. It said on the socket on the back
110-240V and I nearly just plugged it in out of habit (as you can do
with modern power supplies). I decided to read the manual, however and
was so glad I did. 240V into it without flipping the little board over
and changing the fuse would have been a disaster! I didn't have a 1.25
amp fuse but a 1 amp worked ok.
Second issue was opening the cover! It wasn't in the manual and I
couldn't find it anywhere on the net. I figured it out in the end
(undo the screw at top rear and remove the top two feet and slide the
top back). My top didn't slide easily and so it wasn't obvious.
My third issue I'm not sure how it happened and I think it was a bug.
After lots of prior fiddling I started taking measurements between two
1 PPS readings from two different GPSDOs (Thunderbolt and Z3815A) and
it was telling me the difference was 980 nsec. Checking the waveforms
on an oscilloscope said it should have been of the order of 3 nsec.
This had me stumped for ages until I clicked the DSP REF button and it
showed -977! This was obviously from previous fiddling but I thought
the SET REF button would be lit if REFERENCE was non zero. I haven't
been able to repeat this. Any ideas? Motto here is always click CLR
REF to be sure.
Another thing I don't understand is to do with frequency measurements.
Gating with 0.01, 0.1 or 1 second gives good results for a reference
10 MHz, but why does measuring the same frequency with the 1 PERIOD
set and say 1000 samples not give good results?
Thanks in advance,
Jim
Hi Folks,
Well my 5370B purchased off ebay arrived from the US in perfect
condition. It passed all the operator checks (all well within spec)
and the only issue being that the internal 10811 oscillator was
running a tad slow. A quarter of a turn with the screwdriver fixed
that.
Having played with it for a day or so I have a few questions. I was
wondering if there was an FAQ anywhere?
My first issue was powering it up. It said on the socket on the back
110-240V and I nearly just plugged it in out of habit (as you can do
with modern power supplies). I decided to read the manual, however and
was so glad I did. 240V into it without flipping the little board over
and changing the fuse would have been a disaster! I didn't have a 1.25
amp fuse but a 1 amp worked ok.
Second issue was opening the cover! It wasn't in the manual and I
couldn't find it anywhere on the net. I figured it out in the end
(undo the screw at top rear and remove the top two feet and slide the
top back). My top didn't slide easily and so it wasn't obvious.
My third issue I'm not sure how it happened and I think it was a bug.
After lots of prior fiddling I started taking measurements between two
1 PPS readings from two different GPSDOs (Thunderbolt and Z3815A) and
it was telling me the difference was 980 nsec. Checking the waveforms
on an oscilloscope said it should have been of the order of 3 nsec.
This had me stumped for ages until I clicked the DSP REF button and it
showed -977! This was obviously from previous fiddling but I thought
the SET REF button would be lit if REFERENCE was non zero. I haven't
been able to repeat this. Any ideas? Motto here is always click CLR
REF to be sure.
Another thing I don't understand is to do with frequency measurements.
Gating with 0.01, 0.1 or 1 second gives good results for a reference
10 MHz, but why does measuring the same frequency with the 1 PERIOD
set and say 1000 samples not give good results?
Thanks in advance,
Jim
W
wje
Thu, Jun 19, 2008 11:53 AM
Congratulations! It's a wonderful instrument, isn't it?
I can answer a few of your questions.
Set/clear ref - after power-on or a reset, the ref value is (should be)
zero.
When set-ref is pressed, the current time interval average value is
stored. This remains set until a power-off/reset or you clear it. The
buttons don't light up.
Freq vs time interval - simplistic answer: when you're using
time-gating, you're looking at more samples than in averaged one-period
mode. For 10Mhz, a gate of 0.1 secs is 1 million periods. In period
mode, you're averaging a maximum of 100,000 periods. Accuracy is
proportional to the square root of the number of samples.
The 'specifications' section of the manual gives the formulas for
determining both resolution and accuracy (which are NOT the same) for
the various modes.
I would suggest you go through the detailed alignment procedures even
though the unit passes the operational checks. These units almost always
have drifted out of alignment in the analog front-end unless you were
lucky enough to get a freshly-calibrated one. You can usually
significantly improve channel-to-channel trigger consistency and lower
the jitter by doing so. It's a tedious process, and some of the pots are
virtually impossible to reach, as are some of the test points, but it's
worth the pain. The theoretical one-sample measurement accuracy is 20ps,
but a non-adjusted unit can add many tens of ps of jitter. HP states a
typical jitter of 35ps; with very careful tweaking, you can get better
than that; mine reports 21-23 ps averaged over 1000 samples. (I'm not
entirely sure I believe that, though)
Bill Ezell
They said 'Windows or better'
so I used Linux.
Jim Palfreyman wrote:
Hi Folks,
Well my 5370B purchased off ebay arrived from the US in perfect
condition. It passed all the operator checks (all well within spec)
and the only issue being that the internal 10811 oscillator was
running a tad slow. A quarter of a turn with the screwdriver fixed
that.
Having played with it for a day or so I have a few questions. I was
wondering if there was an FAQ anywhere?
My first issue was powering it up. It said on the socket on the back
110-240V and I nearly just plugged it in out of habit (as you can do
with modern power supplies). I decided to read the manual, however and
was so glad I did. 240V into it without flipping the little board over
and changing the fuse would have been a disaster! I didn't have a 1.25
amp fuse but a 1 amp worked ok.
Second issue was opening the cover! It wasn't in the manual and I
couldn't find it anywhere on the net. I figured it out in the end
(undo the screw at top rear and remove the top two feet and slide the
top back). My top didn't slide easily and so it wasn't obvious.
My third issue I'm not sure how it happened and I think it was a bug.
After lots of prior fiddling I started taking measurements between two
1 PPS readings from two different GPSDOs (Thunderbolt and Z3815A) and
it was telling me the difference was 980 nsec. Checking the waveforms
on an oscilloscope said it should have been of the order of 3 nsec.
This had me stumped for ages until I clicked the DSP REF button and it
showed -977! This was obviously from previous fiddling but I thought
the SET REF button would be lit if REFERENCE was non zero. I haven't
been able to repeat this. Any ideas? Motto here is always click CLR
REF to be sure.
Another thing I don't understand is to do with frequency measurements.
Gating with 0.01, 0.1 or 1 second gives good results for a reference
10 MHz, but why does measuring the same frequency with the 1 PERIOD
set and say 1000 samples not give good results?
Thanks in advance,
Jim
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Congratulations! It's a wonderful instrument, isn't it?
I can answer a few of your questions.
Set/clear ref - after power-on or a reset, the ref value is (should be)
zero.
When set-ref is pressed, the current time interval average value is
stored. This remains set until a power-off/reset or you clear it. The
buttons don't light up.
Freq vs time interval - simplistic answer: when you're using
time-gating, you're looking at more samples than in averaged one-period
mode. For 10Mhz, a gate of 0.1 secs is 1 million periods. In period
mode, you're averaging a maximum of 100,000 periods. Accuracy is
proportional to the square root of the number of samples.
The 'specifications' section of the manual gives the formulas for
determining both resolution and accuracy (which are NOT the same) for
the various modes.
I would suggest you go through the detailed alignment procedures even
though the unit passes the operational checks. These units almost always
have drifted out of alignment in the analog front-end unless you were
lucky enough to get a freshly-calibrated one. You can usually
significantly improve channel-to-channel trigger consistency and lower
the jitter by doing so. It's a tedious process, and some of the pots are
virtually impossible to reach, as are some of the test points, but it's
worth the pain. The theoretical one-sample measurement accuracy is 20ps,
but a non-adjusted unit can add many tens of ps of jitter. HP states a
typical jitter of 35ps; with very careful tweaking, you can get better
than that; mine reports 21-23 ps averaged over 1000 samples. (I'm not
entirely sure I believe that, though)
Bill Ezell
----------
They said 'Windows or better'
so I used Linux.
Jim Palfreyman wrote:
> Hi Folks,
>
> Well my 5370B purchased off ebay arrived from the US in perfect
> condition. It passed all the operator checks (all well within spec)
> and the only issue being that the internal 10811 oscillator was
> running a tad slow. A quarter of a turn with the screwdriver fixed
> that.
>
> Having played with it for a day or so I have a few questions. I was
> wondering if there was an FAQ anywhere?
>
> My first issue was powering it up. It said on the socket on the back
> 110-240V and I nearly just plugged it in out of habit (as you can do
> with modern power supplies). I decided to read the manual, however and
> was so glad I did. 240V into it without flipping the little board over
> and changing the fuse would have been a disaster! I didn't have a 1.25
> amp fuse but a 1 amp worked ok.
>
> Second issue was opening the cover! It wasn't in the manual and I
> couldn't find it anywhere on the net. I figured it out in the end
> (undo the screw at top rear and remove the top two feet and slide the
> top back). My top didn't slide easily and so it wasn't obvious.
>
> My third issue I'm not sure how it happened and I think it was a bug.
> After lots of prior fiddling I started taking measurements between two
> 1 PPS readings from two different GPSDOs (Thunderbolt and Z3815A) and
> it was telling me the difference was 980 nsec. Checking the waveforms
> on an oscilloscope said it should have been of the order of 3 nsec.
> This had me stumped for ages until I clicked the DSP REF button and it
> showed -977! This was obviously from previous fiddling but I thought
> the SET REF button would be lit if REFERENCE was non zero. I haven't
> been able to repeat this. Any ideas? Motto here is always click CLR
> REF to be sure.
>
> Another thing I don't understand is to do with frequency measurements.
> Gating with 0.01, 0.1 or 1 second gives good results for a reference
> 10 MHz, but why does measuring the same frequency with the 1 PERIOD
> set and say 1000 samples not give good results?
>
> Thanks in advance,
>
> Jim
>
> _______________________________________________
> time-nuts mailing list -- time-nuts@febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
>
>
>
CH
Chuck Harris
Thu, Jun 19, 2008 12:18 PM
Freq vs time interval - simplistic answer: when you're using
time-gating, you're looking at more samples than in averaged one-period
mode. For 10Mhz, a gate of 0.1 secs is 1 million periods. In period
mode, you're averaging a maximum of 100,000 periods. Accuracy is
proportional to the square root of the number of samples.
Also, the trigger point jitter when in one-period mode affects every
sample, but when in gate mode it only affects the first and last
cycle.
The 'specifications' section of the manual gives the formulas for
determining both resolution and accuracy (which are NOT the same) for
the various modes.
I would suggest you go through the detailed alignment procedures even
though the unit passes the operational checks. These units almost always
have drifted out of alignment in the analog front-end unless you were
lucky enough to get a freshly-calibrated one. You can usually
significantly improve channel-to-channel trigger consistency and lower
the jitter by doing so.
What does one do where the tests call for an HP8082A pulse generator?
The silly things are so rare that they are priced in the stratosphere!
-Chuck Harris
wje wrote:
> Freq vs time interval - simplistic answer: when you're using
> time-gating, you're looking at more samples than in averaged one-period
> mode. For 10Mhz, a gate of 0.1 secs is 1 million periods. In period
> mode, you're averaging a maximum of 100,000 periods. Accuracy is
> proportional to the square root of the number of samples.
Also, the trigger point jitter when in one-period mode affects every
sample, but when in gate mode it only affects the first and last
cycle.
>
> The 'specifications' section of the manual gives the formulas for
> determining both resolution and accuracy (which are NOT the same) for
> the various modes.
>
> I would suggest you go through the detailed alignment procedures even
> though the unit passes the operational checks. These units almost always
> have drifted out of alignment in the analog front-end unless you were
> lucky enough to get a freshly-calibrated one. You can usually
> significantly improve channel-to-channel trigger consistency and lower
> the jitter by doing so.
What does one do where the tests call for an HP8082A pulse generator?
The silly things are so rare that they are priced in the stratosphere!
-Chuck Harris
W
wje
Thu, Jun 19, 2008 1:01 PM
Why, you use your 8082A! I got one for a steal a couple of years ago on
EBay.
However, all you really need is a fast rise-time pulse generator that
has external triggering and a variable trigger delay, and can generate
a +1/-1 volt swing. The frequency the tests use aren't that critical,
anything 1Mhz or higher should work. The only semi-critical parameter
is the rise time; the faster it is, the less trigger uncertainty you'll
have. 1ns is good. :) Also, unless the delay time setting is rather
fine, it's extremely difficult to actually do the anti-coincidence
alignment.
What I'm missing is a spectrum analyzer, so I haven't done the
frequency multiplier alignment on either of my 5370s. Any tricks for
doing that without one?
Bill Ezell
They said 'Windows or better'
so I used Linux.
Chuck Harris wrote:
wje wrote:
Freq vs time interval - simplistic answer: when you're using
time-gating, you're looking at more samples than in averaged one-period
mode. For 10Mhz, a gate of 0.1 secs is 1 million periods. In period
mode, you're averaging a maximum of 100,000 periods. Accuracy is
proportional to the square root of the number of samples.
Also, the trigger point jitter when in one-period mode affects every
sample, but when in gate mode it only affects the first and last
cycle.
The 'specifications' section of the manual gives the formulas for
determining both resolution and accuracy (which are NOT the same) for
the various modes.
I would suggest you go through the detailed alignment procedures even
though the unit passes the operational checks. These units almost always
have drifted out of alignment in the analog front-end unless you were
lucky enough to get a freshly-calibrated one. You can usually
significantly improve channel-to-channel trigger consistency and lower
the jitter by doing so.
What does one do where the tests call for an HP8082A pulse generator?
The silly things are so rare that they are priced in the stratosphere!
-Chuck Harris
time-nuts mailing list -- [1]time-nuts@febo.com
To unsubscribe, go to [2]https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
References
- mailto:time-nuts@febo.com
-
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
Why, you use your 8082A! I got one for a steal a couple of years ago on
EBay.
However, all you really need is a fast rise-time pulse generator that
has external triggering and a variable trigger delay, and can generate
a +1/-1 volt swing. The frequency the tests use aren't that critical,
anything 1Mhz or higher should work. The only semi-critical parameter
is the rise time; the faster it is, the less trigger uncertainty you'll
have. 1ns is good. :) Also, unless the delay time setting is rather
fine, it's extremely difficult to actually do the anti-coincidence
alignment.
What I'm missing is a spectrum analyzer, so I haven't done the
frequency multiplier alignment on either of my 5370s. Any tricks for
doing that without one?
Bill Ezell
----------
They said 'Windows or better'
so I used Linux.
Chuck Harris wrote:
wje wrote:
Freq vs time interval - simplistic answer: when you're using
time-gating, you're looking at more samples than in averaged one-period
mode. For 10Mhz, a gate of 0.1 secs is 1 million periods. In period
mode, you're averaging a maximum of 100,000 periods. Accuracy is
proportional to the square root of the number of samples.
Also, the trigger point jitter when in one-period mode affects every
sample, but when in gate mode it only affects the first and last
cycle.
The 'specifications' section of the manual gives the formulas for
determining both resolution and accuracy (which are NOT the same) for
the various modes.
I would suggest you go through the detailed alignment procedures even
though the unit passes the operational checks. These units almost always
have drifted out of alignment in the analog front-end unless you were
lucky enough to get a freshly-calibrated one. You can usually
significantly improve channel-to-channel trigger consistency and lower
the jitter by doing so.
What does one do where the tests call for an HP8082A pulse generator?
The silly things are so rare that they are priced in the stratosphere!
-Chuck Harris
_______________________________________________
time-nuts mailing list -- [1]time-nuts@febo.com
To unsubscribe, go to [2]https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
References
1. mailto:time-nuts@febo.com
2. https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
CH
Chuck Harris
Thu, Jun 19, 2008 1:33 PM
Why, you use your 8082A! I got one for a steal a couple of years ago on
EBay.
I've been looking for a year now, and I haven't found any 8082A's below $400.
Well, there were a couple, but they were a real mess.
5370A's, which are equivalent to the 5370B in all respects have been
going for less than $100 on the 'bay.
[The only significant difference I have been able to find between the "A" and
"B" is the CPU card and memory have been consolidated. Specification wise, they
are identical in all respects. HPIB wise they are also identical. As are they
in measurement rate, and HPIB throughput.]
However, all you really need is a fast rise-time pulse generator that
has external triggering and a variable trigger delay, and can generate
a +1/-1 volt swing. The frequency the tests use aren't that critical,
anything 1Mhz or higher should work. The only semi-critical parameter
is the rise time; the faster it is, the less trigger uncertainty you'll
have. 1ns is good. :) Also, unless the delay time setting is rather
fine, it's extremely difficult to actually do the anti-coincidence
alignment.
From reading the alignment procedure, the 8082A barely meets the needs.
I haven't found anything else that comes close in one package.
What I'm missing is a spectrum analyzer, so I haven't done the
frequency multiplier alignment on either of my 5370s. Any tricks for
doing that without one?
141T systems are available for much less than the cost of an 8082A, and
are much more useful generally.
-Chuck Harris
wje wrote:
> Why, you use your 8082A! I got one for a steal a couple of years ago on
> EBay.
I've been looking for a year now, and I haven't found any 8082A's below $400.
Well, there were a couple, but they were a real mess.
5370A's, which are equivalent to the 5370B in all respects have been
going for less than $100 on the 'bay.
[The only significant difference I have been able to find between the "A" and
"B" is the CPU card and memory have been consolidated. Specification wise, they
are identical in all respects. HPIB wise they are also identical. As are they
in measurement rate, and HPIB throughput.]
> However, all you really need is a fast rise-time pulse generator that
> has external triggering and a variable trigger delay, and can generate
> a +1/-1 volt swing. The frequency the tests use aren't that critical,
> anything 1Mhz or higher should work. The only semi-critical parameter
> is the rise time; the faster it is, the less trigger uncertainty you'll
> have. 1ns is good. :) Also, unless the delay time setting is rather
> fine, it's extremely difficult to actually do the anti-coincidence
> alignment.
From reading the alignment procedure, the 8082A barely meets the needs.
I haven't found anything else that comes close in one package.
> What I'm missing is a spectrum analyzer, so I haven't done the
> frequency multiplier alignment on either of my 5370s. Any tricks for
> doing that without one?
141T systems are available for much less than the cost of an 8082A, and
are much more useful generally.
-Chuck Harris
P
phil
Thu, Jun 19, 2008 1:48 PM
Why an 8082A, that's a 250MHz unit. Would not an 8015A 50MHz work. I think
it's the same only with a lower bandwidth.
----- Original Message -----
From: "wje" wje@quackers.net
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Thursday, June 19, 2008 9:01 AM
Subject: Re: [time-nuts] 5370B arrived - any FAQ
Why, you use your 8082A! I got one for a steal a couple of years ago on
EBay.
However, all you really need is a fast rise-time pulse generator that
has external triggering and a variable trigger delay, and can generate
a +1/-1 volt swing. The frequency the tests use aren't that critical,
anything 1Mhz or higher should work. The only semi-critical parameter
is the rise time; the faster it is, the less trigger uncertainty you'll
have. 1ns is good. :) Also, unless the delay time setting is rather
fine, it's extremely difficult to actually do the anti-coincidence
alignment.
What I'm missing is a spectrum analyzer, so I haven't done the
frequency multiplier alignment on either of my 5370s. Any tricks for
doing that without one?
Bill Ezell
They said 'Windows or better'
so I used Linux.
Chuck Harris wrote:
wje wrote:
Freq vs time interval - simplistic answer: when you're using
time-gating, you're looking at more samples than in averaged one-period
mode. For 10Mhz, a gate of 0.1 secs is 1 million periods. In period
mode, you're averaging a maximum of 100,000 periods. Accuracy is
proportional to the square root of the number of samples.
Also, the trigger point jitter when in one-period mode affects every
sample, but when in gate mode it only affects the first and last
cycle.
The 'specifications' section of the manual gives the formulas for
determining both resolution and accuracy (which are NOT the same) for
the various modes.
I would suggest you go through the detailed alignment procedures even
though the unit passes the operational checks. These units almost always
have drifted out of alignment in the analog front-end unless you were
lucky enough to get a freshly-calibrated one. You can usually
significantly improve channel-to-channel trigger consistency and lower
the jitter by doing so.
What does one do where the tests call for an HP8082A pulse generator?
The silly things are so rare that they are priced in the stratosphere!
-Chuck Harris
time-nuts mailing list -- [1]time-nuts@febo.com
To unsubscribe, go to
[2]https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
References
- mailto:time-nuts@febo.com
-
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Why an 8082A, that's a 250MHz unit. Would not an 8015A 50MHz work. I think
it's the same only with a lower bandwidth.
----- Original Message -----
From: "wje" <wje@quackers.net>
To: "Discussion of precise time and frequency measurement"
<time-nuts@febo.com>
Sent: Thursday, June 19, 2008 9:01 AM
Subject: Re: [time-nuts] 5370B arrived - any FAQ
> Why, you use your 8082A! I got one for a steal a couple of years ago on
> EBay.
> However, all you really need is a fast rise-time pulse generator that
> has external triggering and a variable trigger delay, and can generate
> a +1/-1 volt swing. The frequency the tests use aren't that critical,
> anything 1Mhz or higher should work. The only semi-critical parameter
> is the rise time; the faster it is, the less trigger uncertainty you'll
> have. 1ns is good. :) Also, unless the delay time setting is rather
> fine, it's extremely difficult to actually do the anti-coincidence
> alignment.
> What I'm missing is a spectrum analyzer, so I haven't done the
> frequency multiplier alignment on either of my 5370s. Any tricks for
> doing that without one?
> Bill Ezell
> ----------
> They said 'Windows or better'
> so I used Linux.
>
> Chuck Harris wrote:
>
> wje wrote:
>
>
>
> Freq vs time interval - simplistic answer: when you're using
> time-gating, you're looking at more samples than in averaged one-period
> mode. For 10Mhz, a gate of 0.1 secs is 1 million periods. In period
> mode, you're averaging a maximum of 100,000 periods. Accuracy is
> proportional to the square root of the number of samples.
>
>
> Also, the trigger point jitter when in one-period mode affects every
> sample, but when in gate mode it only affects the first and last
> cycle.
>
>
> The 'specifications' section of the manual gives the formulas for
> determining both resolution and accuracy (which are NOT the same) for
> the various modes.
>
> I would suggest you go through the detailed alignment procedures even
> though the unit passes the operational checks. These units almost always
> have drifted out of alignment in the analog front-end unless you were
> lucky enough to get a freshly-calibrated one. You can usually
> significantly improve channel-to-channel trigger consistency and lower
> the jitter by doing so.
>
>
> What does one do where the tests call for an HP8082A pulse generator?
> The silly things are so rare that they are priced in the stratosphere!
>
> -Chuck Harris
>
> _______________________________________________
> time-nuts mailing list -- [1]time-nuts@febo.com
> To unsubscribe, go to
> [2]https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
>
> References
>
> 1. mailto:time-nuts@febo.com
> 2. https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> _______________________________________________
> time-nuts mailing list -- time-nuts@febo.com
> To unsubscribe, go to
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
P
phil
Thu, Jun 19, 2008 1:58 PM
I Didn't look at pulse bandwidth. 2.4ns compared to 10 ns
Other than that about the same unit.
----- Original Message -----
From: "phil" fortime@bellsouth.net
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Thursday, June 19, 2008 9:48 AM
Subject: Re: [time-nuts] 5370B arrived - any FAQ
Why an 8082A, that's a 250MHz unit. Would not an 8015A 50MHz work. I think
it's the same only with a lower bandwidth.
----- Original Message -----
From: "wje" wje@quackers.net
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Thursday, June 19, 2008 9:01 AM
Subject: Re: [time-nuts] 5370B arrived - any FAQ
Why, you use your 8082A! I got one for a steal a couple of years ago on
EBay.
However, all you really need is a fast rise-time pulse generator that
has external triggering and a variable trigger delay, and can generate
a +1/-1 volt swing. The frequency the tests use aren't that critical,
anything 1Mhz or higher should work. The only semi-critical parameter
is the rise time; the faster it is, the less trigger uncertainty you'll
have. 1ns is good. :) Also, unless the delay time setting is rather
fine, it's extremely difficult to actually do the anti-coincidence
alignment.
What I'm missing is a spectrum analyzer, so I haven't done the
frequency multiplier alignment on either of my 5370s. Any tricks for
doing that without one?
Bill Ezell
They said 'Windows or better'
so I used Linux.
Chuck Harris wrote:
wje wrote:
Freq vs time interval - simplistic answer: when you're using
time-gating, you're looking at more samples than in averaged one-period
mode. For 10Mhz, a gate of 0.1 secs is 1 million periods. In period
mode, you're averaging a maximum of 100,000 periods. Accuracy is
proportional to the square root of the number of samples.
Also, the trigger point jitter when in one-period mode affects every
sample, but when in gate mode it only affects the first and last
cycle.
The 'specifications' section of the manual gives the formulas for
determining both resolution and accuracy (which are NOT the same) for
the various modes.
I would suggest you go through the detailed alignment procedures even
though the unit passes the operational checks. These units almost always
have drifted out of alignment in the analog front-end unless you were
lucky enough to get a freshly-calibrated one. You can usually
significantly improve channel-to-channel trigger consistency and lower
the jitter by doing so.
What does one do where the tests call for an HP8082A pulse generator?
The silly things are so rare that they are priced in the stratosphere!
-Chuck Harris
time-nuts mailing list -- [1]time-nuts@febo.com
To unsubscribe, go to
[2]https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
References
- mailto:time-nuts@febo.com
-
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
I Didn't look at pulse bandwidth. 2.4ns compared to 10 ns
Other than that about the same unit.
----- Original Message -----
From: "phil" <fortime@bellsouth.net>
To: "Discussion of precise time and frequency measurement"
<time-nuts@febo.com>
Sent: Thursday, June 19, 2008 9:48 AM
Subject: Re: [time-nuts] 5370B arrived - any FAQ
> Why an 8082A, that's a 250MHz unit. Would not an 8015A 50MHz work. I think
> it's the same only with a lower bandwidth.
>
> ----- Original Message -----
> From: "wje" <wje@quackers.net>
> To: "Discussion of precise time and frequency measurement"
> <time-nuts@febo.com>
> Sent: Thursday, June 19, 2008 9:01 AM
> Subject: Re: [time-nuts] 5370B arrived - any FAQ
>
>
>> Why, you use your 8082A! I got one for a steal a couple of years ago on
>> EBay.
>> However, all you really need is a fast rise-time pulse generator that
>> has external triggering and a variable trigger delay, and can generate
>> a +1/-1 volt swing. The frequency the tests use aren't that critical,
>> anything 1Mhz or higher should work. The only semi-critical parameter
>> is the rise time; the faster it is, the less trigger uncertainty you'll
>> have. 1ns is good. :) Also, unless the delay time setting is rather
>> fine, it's extremely difficult to actually do the anti-coincidence
>> alignment.
>> What I'm missing is a spectrum analyzer, so I haven't done the
>> frequency multiplier alignment on either of my 5370s. Any tricks for
>> doing that without one?
>> Bill Ezell
>> ----------
>> They said 'Windows or better'
>> so I used Linux.
>>
>> Chuck Harris wrote:
>>
>> wje wrote:
>>
>>
>>
>> Freq vs time interval - simplistic answer: when you're using
>> time-gating, you're looking at more samples than in averaged one-period
>> mode. For 10Mhz, a gate of 0.1 secs is 1 million periods. In period
>> mode, you're averaging a maximum of 100,000 periods. Accuracy is
>> proportional to the square root of the number of samples.
>>
>>
>> Also, the trigger point jitter when in one-period mode affects every
>> sample, but when in gate mode it only affects the first and last
>> cycle.
>>
>>
>> The 'specifications' section of the manual gives the formulas for
>> determining both resolution and accuracy (which are NOT the same) for
>> the various modes.
>>
>> I would suggest you go through the detailed alignment procedures even
>> though the unit passes the operational checks. These units almost always
>> have drifted out of alignment in the analog front-end unless you were
>> lucky enough to get a freshly-calibrated one. You can usually
>> significantly improve channel-to-channel trigger consistency and lower
>> the jitter by doing so.
>>
>>
>> What does one do where the tests call for an HP8082A pulse generator?
>> The silly things are so rare that they are priced in the stratosphere!
>>
>> -Chuck Harris
>>
>> _______________________________________________
>> time-nuts mailing list -- [1]time-nuts@febo.com
>> To unsubscribe, go to
>> [2]https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> and follow the instructions there.
>>
>> References
>>
>> 1. mailto:time-nuts@febo.com
>> 2. https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> _______________________________________________
>> time-nuts mailing list -- time-nuts@febo.com
>> To unsubscribe, go to
>> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> and follow the instructions there.
>
>
> _______________________________________________
> time-nuts mailing list -- time-nuts@febo.com
> To unsubscribe, go to
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
P
Pete
Thu, Jun 19, 2008 2:17 PM
As to the frequency measurement based on 1 period
timing; I think you'll find that what's actually measured
is the period, whose reciprocal is displayed as frequency.
This means that the channel A trigger point set the start
& the channel B trigger point set the stop of the period
measurement. A very small mismatch of those 2 points
will result in an apparently large error in the computed
frequency.
As was pointed out earlier, this technique also
doubles the trigger uncertainty of the measurement
vs. the more traditional single channel method.
Pete Rawson
As to the frequency measurement based on 1 period
timing; I think you'll find that what's actually measured
is the period, whose reciprocal is displayed as frequency.
This means that the channel A trigger point set the start
& the channel B trigger point set the stop of the period
measurement. A very small mismatch of those 2 points
will result in an apparently large error in the computed
frequency.
As was pointed out earlier, this technique also
doubles the trigger uncertainty of the measurement
vs. the more traditional single channel method.
Pete Rawson
CH
Chuck Harris
Thu, Jun 19, 2008 2:27 PM
Because you need something that has a finely variable delay in the
range of 2 to 5ns, a pulse width adjustable from 5 to 50ns, and a minimum
transition time of less than 1ns. And, it has to be able to drive 50 ohms
to +/-1V.
The 8015A's shortest delay is 20ns, and its minimum pulse width is 10ns, and
its minimum transition is 6ns.
Basically, it isn't even close to being good enough to calibrate a 5370[AB].
Looking through the available pulse generators, I find some have one needed
characteristic, but none of the others. The 8082A seems to be exactly what
is needed (but just barely).
-Chuck Harris
phil wrote:
Why an 8082A, that's a 250MHz unit. Would not an 8015A 50MHz work. I think
it's the same only with a lower bandwidth.
Because you need something that has a finely variable delay in the
range of 2 to 5ns, a pulse width adjustable from 5 to 50ns, and a minimum
transition time of less than 1ns. And, it has to be able to drive 50 ohms
to +/-1V.
The 8015A's shortest delay is 20ns, and its minimum pulse width is 10ns, and
its minimum transition is 6ns.
Basically, it isn't even close to being good enough to calibrate a 5370[AB].
Looking through the available pulse generators, I find some have one needed
characteristic, but none of the others. The 8082A seems to be exactly what
is needed (but just barely).
-Chuck Harris
phil wrote:
> Why an 8082A, that's a 250MHz unit. Would not an 8015A 50MHz work. I think
> it's the same only with a lower bandwidth.
JM
John Miles
Thu, Jun 19, 2008 6:26 PM
I would suggest you go through the detailed alignment procedures even
though the unit passes the operational checks. These units almost always
have drifted out of alignment in the analog front-end unless you were
lucky enough to get a freshly-calibrated one. You can usually
significantly improve channel-to-channel trigger consistency and lower
the jitter by doing so. It's a tedious process, and some of the pots are
virtually impossible to reach, as are some of the test points, but it's
worth the pain. The theoretical one-sample measurement accuracy is 20ps,
but a non-adjusted unit can add many tens of ps of jitter. HP states a
typical jitter of 35ps; with very careful tweaking, you can get better
than that; mine reports 21-23 ps averaged over 1000 samples. (I'm not
entirely sure I believe that, though)
I actually think you're better off not touching anything except the trigger
level adjustments unless there's a specific problem, such as self-test
jitter above 20 to 30 ps. Aligning the interpolators properly requires an
8082A pulse generator and a very fast scope.
The front end can be challenging to align but it does need to be checked
carefully with respect to the operation of all the slide switches. With
luck, nothing needs to be done to it.
-- john, KE5FX
> I would suggest you go through the detailed alignment procedures even
> though the unit passes the operational checks. These units almost always
> have drifted out of alignment in the analog front-end unless you were
> lucky enough to get a freshly-calibrated one. You can usually
> significantly improve channel-to-channel trigger consistency and lower
> the jitter by doing so. It's a tedious process, and some of the pots are
> virtually impossible to reach, as are some of the test points, but it's
> worth the pain. The theoretical one-sample measurement accuracy is 20ps,
> but a non-adjusted unit can add many tens of ps of jitter. HP states a
> typical jitter of 35ps; with very careful tweaking, you can get better
> than that; mine reports 21-23 ps averaged over 1000 samples. (I'm not
> entirely sure I believe that, though)
I actually think you're better off not touching anything except the trigger
level adjustments unless there's a specific problem, such as self-test
jitter above 20 to 30 ps. Aligning the interpolators properly requires an
8082A pulse generator and a very fast scope.
The front end can be challenging to align but it does need to be checked
carefully with respect to the operation of all the slide switches. With
luck, nothing needs to be done to it.
-- john, KE5FX
