J06 HP-59992A time interval calibrator for HP-531xx counters
I think the way the fine cal works by checking the the intervals between four different edges that a lot of asymmetries in the signals are nulled out in the software.
How good are 1:2 180 degree phase shifters at exactly shifting by 180 degrees? At what cost?
Also coax and RF relays cost a lot. Pretty soon your BOM cost is over what a 59992A will run... assuming you can find one.
So far my design is tending towards: 10MHz ref input -> Minicircuits doubler -> Wenzel squarer -> 74AC74 divider -> 74AC04 buffer -> level shifter. The doubler/divider might not be needed, but I think it will give a more symmetric output. I might include a space for a 10 MHz TTL oscillator for non time-nut users... hopefully it might be stable enough over the short time interval for a cal measurement cycle.
For the gain cal an 5V LM4040 (or other) reference. Routing signals to the output BNC's via 2P4T slide switches. The slide switches might be a weak link... but they are less than 50 cents each. Whatever, it should beat the pants off doing just the "quick TI cal"
The reason I started looking into this is that I want to upgrade the old firmware to a much newer version that allows the comma separators in the decimal digits. Upgrading the firmware requires a new calibration.
On Sun, Jul 9, 2017 at 2:15 AM, Mark Sims holrum@hotmail.com wrote:
So far my design is tending towards: 10MHz ref input -> Minicircuits doubler -> Wenzel squarer -> 74AC74 divider -> 74AC04 buffer -> level shifter. The doubler/divider might not be needed, but I think it will give a more symmetric output. I might include a space for a 10 MHz TTL oscillator for non time-nut users... hopefully it might be stable enough over the short time interval for a cal measurement cycle.
once upon a time, I was experimenting with digital signals to derive
stable RF transverter
LOs. I've "found" that feeding a XOR gate with a signal and his
replica delayed by 2 inverters
did result in a crude frequency doubler (well it's rather an edge
detector). Since I was going to use the double frequency just to drive
a divider by two, the actual duty cycle out of the doubler didn't
matter.
HTH
Frank
With a sinewave input LF feedback from the 74AC04 output to the shaper inputs could be used to regulate the output duty cycle by adjusting the switching threshold. However ensuring that the input amplitude is sufficiently large to override potential LF oscillation. The LF feedback will also compensate for delay asymmetry in the 74AC04 as well.
The output transition times of ACMOS is relatively slow compared to modern LVCMOS parts.
Bruce
On 09 July 2017 at 12:15 Mark Sims holrum@hotmail.com wrote:
I think the way the fine cal works by checking the the intervals between four different edges that a lot of asymmetries in the signals are nulled out in the software.
How good are 1:2 180 degree phase shifters at exactly shifting by 180 degrees? At what cost?
Also coax and RF relays cost a lot. Pretty soon your BOM cost is over what a 59992A will run... assuming you can find one.
So far my design is tending towards: 10MHz ref input -> Minicircuits doubler -> Wenzel squarer -> 74AC74 divider -> 74AC04 buffer -> level shifter. The doubler/divider might not be needed, but I think it will give a more symmetric output. I might include a space for a 10 MHz TTL oscillator for non time-nut users... hopefully it might be stable enough over the short time interval for a cal measurement cycle.
For the gain cal an 5V LM4040 (or other) reference. Routing signals to the output BNC's via 2P4T slide switches. The slide switches might be a weak link... but they are less than 50 cents each. Whatever, it should beat the pants off doing just the "quick TI cal"
The reason I started looking into this is that I want to upgrade the old firmware to a much newer version that allows the comma separators in the decimal digits. Upgrading the firmware requires a new calibration.
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.
Hi
On Jul 9, 2017, at 9:35 AM, Francesco Messineo francesco.messineo@gmail.com wrote:
On Sun, Jul 9, 2017 at 2:15 AM, Mark Sims holrum@hotmail.com wrote:
So far my design is tending towards: 10MHz ref input -> Minicircuits doubler -> Wenzel squarer -> 74AC74 divider -> 74AC04 buffer -> level shifter. The doubler/divider might not be needed, but I think it will give a more symmetric output. I might include a space for a 10 MHz TTL oscillator for non time-nut users... hopefully it might be stable enough over the short time interval for a cal measurement cycle.
once upon a time, I was experimenting with digital signals to derive
stable RF transverter
LOs. I've "found" that feeding a XOR gate with a signal and his
replica delayed by 2 inverters
did result in a crude frequency doubler (well it's rather an edge
detector).
You can do the same thing with a filter on one input of the XOR. Tuning the filter for best
symmetry is fairly easy. Watching the result on a spectrum analyzer usually is quickest.
Bob
Since I was going to use the double frequency just to drive
a divider by two, the actual duty cycle out of the doubler didn't
matter.
HTH
Frank
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.
On Sat, Jul 8, 2017 at 5:15 PM, Mark Sims holrum@hotmail.com wrote:
I think the way the fine cal works by checking the the intervals between
four different edges that a lot of asymmetries in the signals are nulled
out in the software.
How good are 1:2 180 degree phase shifters at exactly shifting by 180
degrees? At what cost?
The original parts were nothing special! PSC-2-1 is spec'd at 4 degrees
max phase unbalance - though typical is a lot less - and seems to be about
$14.
https://www.minicircuits.com/pdfs/PSC-2-1.pdf
https://www.minicircuits.com/pdfs/PSCJ-2-1+.pdf
Hi
People have been known to sort MiniCircuits parts and use the “extras” in
something else. A certain major oscillator manufacturer once bought a bunch
of RPD-1’s , sorted them for the “one in a hundred” examples, and then returned
the rest for credit…… Somehow I doubt HP didi it quite that way.
4 degrees at 10 MHz. A LVCMOS inverter has a delay of about 2.5 ns. Delay time
on cheap flip flops is no longer broken out in enough detail to really see how they
are doing.
Bob
On Jul 9, 2017, at 1:14 PM, Orin Eman orin.eman@gmail.com wrote:
On Sat, Jul 8, 2017 at 5:15 PM, Mark Sims holrum@hotmail.com wrote:
I think the way the fine cal works by checking the the intervals between
four different edges that a lot of asymmetries in the signals are nulled
out in the software.
How good are 1:2 180 degree phase shifters at exactly shifting by 180
degrees? At what cost?
The original parts were nothing special! PSC-2-1 is spec'd at 4 degrees
max phase unbalance - though typical is a lot less - and seems to be about
$14.
https://www.minicircuits.com/pdfs/PSC-2-1.pdf
https://www.minicircuits.com/pdfs/PSCJ-2-1+.pdf
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.
On 7/9/17 10:27 AM, Bob kb8tq wrote:
Hi
People have been known to sort MiniCircuits parts and use the “extras” in
something else. A certain major oscillator manufacturer once bought a bunch
of RPD-1’s , sorted them for the “one in a hundred” examples, and then returned
the rest for credit…… Somehow I doubt HP didi it quite that way.
Some manufacturers will cherry pick for you. Maybe the datasheet says
gain is 14 to 18 dB, and you order a batch that you want to be 15.5-16.5
dB.. They'll pick those out, but of course, now the remainder have a
hole in the distribution. Particularly when you're buying high-rel or
space grade, where they have to 100% test anyway.
On Sun, Jul 9, 2017 at 10:55 AM, jimlux jimlux@earthlink.net wrote:
On 7/9/17 10:27 AM, Bob kb8tq wrote:
Hi
People have been known to sort MiniCircuits parts and use the “extras”
in
something else. A certain major oscillator manufacturer once bought a
bunch
of RPD-1’s , sorted them for the “one in a hundred” examples, and then
returned
the rest for credit…… Somehow I doubt HP didi it quite that way.
Some manufacturers will cherry pick for you. Maybe the datasheet says
gain is 14 to 18 dB, and you order a batch that you want to be 15.5-16.5
dB.. They'll pick those out, but of course, now the remainder have a hole
in the distribution. Particularly when you're buying high-rel or space
grade, where they have to 100% test anyway.
Right, but there is nothing to indicate there is anything special going on
in this case. The theory of operation section of the manual states that
the splitters provide "nearly" in-phase and "nearly" out-of-phase signals
without defining "nearly". It does not sound like it's critical and to me,
it sounds like they are acknowledging that there are phase mismatches
through the splitters.
Then going on to the other link I posted:
It accounts for the individual delays through the splitters/cables and
shows how they are eliminated by doing multiple measurements and solving
the resulting simultaneous equations.
Indeed, the programs in the J06's manual mention that 'cable' mismatches
(presumable phase mismatches) are included in the calibration constants and
the same cables should be used for doing actual measurements: (the
formatting did not survive and I fixed some obvious OCR errors)
40 ! THIS PROGRAM EXECUTES THE CALIBRATION ALGORITHM DESCRIBED IN D. CHU'S
50 ! PAPER "CALIBRATION OF SYSTEMATIC ERRORS IN PRECISION TIME-INTERVAL
60 ! COUNTERS", INTERNATIONAL TEST CONFERENCE, PHILADELPHIA, 1985;
70 !
80 ! SET-UP PROCEDURE
90 ! 1) CONNECT A PULSE SOURCE TO THE INPUT OF CALIBRATOR, 3 DB LARGER
100 ! THAN THE DESIRED SIGNAL TO BE MEASURED AND APPROXIMATELY THE
110 ! THE SAME RISE/FALL TIMES, -50% DUTY-CYCLE, STABLE 1 TO 100 MHz.
120 ! 2) CONNECT A PAIR OF CABLES FROM CALIBRATOR OUTPUTS A & B
130 ! TO COUNTER START & STOP INPUTS RESPECTIVELY
140 ! (NOTE: CABLE MISMATCHES ARE INCLUDED IN THE CALIBRATION CONSTANTS,
150 ! AND SAME CABLES SHOULD BE USED LATER FOR DOING MEASUREMENTS;
160 ! ALSO "CABLES" INCLUDE LINEAR, PASSIVE OR ACTIVE PROBES)
170 ! 3) SET COUNTER TO SEPERATE: DC/50 ohms/Xl/PRESET to BOTH CHANNELS
180 ! 4) HPIB ADDRESSES: COUNTER-707, CALIBRATOR-705
190 ! 5) CONNECT A PRECISE OFFSET VOLTAGE SOURCE TO THE OFFSET INPUT: ENTER
200 ! THE EXACT SAME VALUE WHEN PROMPTED. DEFAULT IS 0.00 VOLT
In message DM5PR1101MB21057D2E3C55E34307A0FDE2CEAB0@DM5PR1101MB2105.namprd11.prod.outlook.com, Mark Sims writes:
Also coax and RF relays cost a lot. Pretty soon your BOM cost
is over what a 59992A will run... assuming you can find one.
The crucial feature of the 59992A was support for DC-bias, and
that seems to have dictated its design.
If you have no need for DC-bias, an 8-pin microcontroller with a
stable crystal and suitable resistor networks on the outputs will
do fine.
TI measurements on top of DC-bias was important in development and
manufacturing of disk drives: Measurement of jitter of mechanical
origin must happen on the analog side of the differential read
amplifier, which usually balances a couple of volts up for cost
reasons.
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
In message 8fdafefb-3563-94be-e68d-63fe00164612@earthlink.net, jimlux writes:
On 7/9/17 10:27 AM, Bob kb8tq wrote:
Hi
People have been known to sort MiniCircuits parts and use the “extras” in
something else. A certain major oscillator manufacturer once bought a bunch
of RPD-1’s , sorted them for the “one in a hundred” examples, and then returned
the rest for credit…… Somehow I doubt HP didi it quite that way.
Some manufacturers will cherry pick for you. Maybe the datasheet says
gain is 14 to 18 dB, and you order a batch that you want to be 15.5-16.5
dB.. They'll pick those out, but of course, now the remainder have a
hole in the distribution. Particularly when you're buying high-rel or
space grade, where they have to 100% test anyway.
I've been told by somebody from HPs manufacturing in Europe that
there were one or two "magic" components where the manufacturer
would send their entire production to HP.
HP would test, measure and sort into four bins: "Failed", "Good",
"Really good" and "HP".
The first three bins got shipped back to the manufacturer.
He suspected that back at the manufacturer, the "Really good"
got a "MIL-GRADE" stamp :-)
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.