TDC-GPX2 TDC chip
http://ams.com/eng/Products/Precision-Time-Measurement/Time-to-Digital-Converters/TDC-GPX2
Claims to have single-shot accuracy of 10 ps.
"the large print giveth and the small print taketh away"
(Quoted from this great description of modern life:
http://www.tomwaits.com/songs/song/322/Step_Right_Up/ )
I'm not trying to be too critical, but look carefully at the datasheet
for that part:
http://ams.com/eng/content/download/951531/2270299/381482
The "accuracy" can in no way be described as 10 ps. In high
resolution mode:
-
Single shot RMS time resolution: 10 ps typical, 15 ps maximum
-
Integral non-linearity: 20 ps max
-
Differential non-linearity: 5 ps typ
-
Channel to channel isolation: 20 ps typ, 100 ps max
-
Offset error: 200 ps typ
-
Offset error drift vs temperature: 1.5 ps typ, 3 ps max / K
-
FWHM (full width half maximum) histogram: 20-25 ps
So if the temperature changes 6 F (3.3 C), the measurement may drift by
10 ps. The single shot absolute uncertainty due to jitter (see the
histogram) could easily be +/- 25 ps. There is no mention of long-term
drift, since that isn't a typical use of this part. It would be
interesting to put one in a simple oven and see how it drifts over a
few months.
Bill Byrom N5BB
----- Original message -----
From: Christopher Hoover ch@murgatroid.com
To: Discussion of precise time and frequency measurement time-nuts@febo.com
Subject: [time-nuts] TDC-GPX2 TDC chip
Date: Wed, 15 Mar 2017 15:28:20 -0700
http://ams.com/eng/Products/Precision-Time-Measurement/Time-to-Digital-Converters/TDC-GPX2
Claims to have single-shot accuracy of 10 ps.
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.
The datasheet is less than convincing.
The change in resolution/noise with time delay isnt well documented.
Performance may differ when a 1MHz reference clock is used as opposed to the 10MHz clock they appear to have used for some of the plots on the datasheet.
The customer support person appears to know little useful about tdcs.
Have been debating how to establish the actual performance.
A) use random source(s) to produce data for the fill the buckets histogram test.
For statistically useful results an extremely large number of random events is required.
e.g. for a 10MHz reference clock and 10ps bits need around 10,000 hits per bin for 1% rms
statistical variation in bin event totals. i.e. about 1E8 total events.
With a 1MHz reference clock need at least 1E9 events.
B) Use a pair of 10MHz oscillators with a small frequency offset between them one of which is used as the reference clock whilst the other is divided down to produce a suitable event rate. The leading edge of each successive event is swept slowly at a uniform rate across the reference clock cycle in steps of say 10fs or so per event. This requires that the sources have very low spur levels at least 120dB below the carrier and preferably more.
The reference 10MHz can be divided down to 5MHz, 2MHz or 1MHZ for testing with lower reference clock frequencies.
c) As above but use a single 10MHz source that acts both as reference clock for the TDC and as the input source for an extremely low spur offset generator, the output of which is divided down to create a suitable event rate.
D) Compare the TDC measurements with those obtained with a significantly lower noise TDC.
Still (a few of us at least) working on this. Initial crude and somewhat noisy setup suggested better than 10ps rms (limited somewhat by jitter of divider chain used to produce trigger signal). A later version achieved around 3ps rms. However comprehensive testing is still to be done.
Bruce
On 16 March 2017 at 11:28 Christopher Hoover <ch@murgatroid.com> wrote:
http://ams.com/eng/Products/Precision-Time-Measurement/Time-to-Digital-Converters/TDC-GPX2
Claims to have single-shot accuracy of 10 ps.
_______________________________________________
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.
The apparent drift will be determined largely by that of the reference oscillator due to the embedded continuous interpolator calibration. Placing it in an oven will probably degrade the resolution due to the bin size tempco.
All time interval instruments have an offset between channels, This is easily measured and corrected for.
The performance as stated in the datasheet appears comparable if not better than a typical 5370A/B when one takes the various sources of nonlinearity and crosstalk in a 5370A/B into account. It certainly consumes less power even after the power consumption of a good ocxo is added. Its also somewhat cheaper.
However a comprehensive evaluation of its performance beyond that given in the datasheet is required.
A comprehensive test of the performance of a n actual 5370A/B would also be useful.
Bruce
On 16 March 2017 at 15:54 Bill Byrom <time@radio.sent.com> wrote:
"the large print giveth and the small print taketh away"
(Quoted from this great description of modern life:
http://www.tomwaits.com/songs/song/322/Step_Right_Up/ )
I'm not trying to be too critical, but look carefully at the datasheet
for that part:
http://ams.com/eng/content/download/951531/2270299/381482
The "accuracy" can in no way be described as 10 ps. In high
resolution mode:
* Single shot RMS time resolution: 10 ps typical, 15 ps maximum
* Integral non-linearity: 20 ps max
* Differential non-linearity: 5 ps typ
* Channel to channel isolation: 20 ps typ, 100 ps max
* Offset error: 200 ps typ
* Offset error drift vs temperature: 1.5 ps typ, 3 ps max / K
* FWHM (full width half maximum) histogram: 20-25 ps
So if the temperature changes 6 F (3.3 C), the measurement may drift by
10 ps. The single shot absolute uncertainty due to jitter (see the
histogram) could easily be +/- 25 ps. There is no mention of long-term
drift, since that isn't a typical use of this part. It would be
interesting to put one in a simple oven and see how it drifts over a
few months.
--
Bill Byrom N5BB
----- Original message -----
From: Christopher Hoover <ch@murgatroid.com>
To: Discussion of precise time and frequency measurement <time-nuts@febo.com>
Subject: [time-nuts] TDC-GPX2 TDC chip
Date: Wed, 15 Mar 2017 15:28:20 -0700
http://ams.com/eng/Products/Precision-Time-Measurement/Time-to-Digital-Converters/TDC-GPX2
Claims to have single-shot accuracy of 10 ps.
_______________________________________________
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.