[time-nuts] HP53132A's LSD question

[Magnus Danielson]

On 08/31/2013 05:06 PM, Hui Zhang wrote:
> Dear group:
>
>   I found a strange thing about HP53132A’s output digit, I know the HP53132A is a universal counter that have 12 digits per second performance by read the manual. By use 10s gate time, I thought it will be 13 digits resolution. But I found it can output more digits reads by use GPIB or RS-232 way to collect datas. (This is simple way rather than use offset method). I use RS-232 cable to connected to my counter to observe the frequency of a OCXO, I got these output result ( a part of long sequence):
>
> 9.999,967,286,098,0  MHz
> 9.999,981,814,726,9  MHz
> 9.999,991,573,702,0  MHz
> 9.999,997,482,253,0  MHz
> 9.999,999,945,685  MHz
> 10.000,000,472,030,4 MHz
> 10.000,000,460,991,3 MHz
> 10.000,000,352,570,9 MHz
> 10.000,000,254,108  MHz
> 10.000,000,183,759  MHz
> 10.000,000,136,842  MHz
> ..
> . (Omitted few lines)
> .
> 10.000,000,039,724  MHz
> 10.000,000,006,299  MHz
> 9.999,999,992,692  MHz
> 9.999,999,990,674  MHz
> 9.999,999,989,094  MHz
> 9.999,999,995,662  MHz
> 9.999,999,993,829  MHz
>
> The question is in first four lines the reads is 14 digits, but in line 5, the digits reduce to 13 digits, and then, it’s became 15 digits in line 6(because over 10Mhz). After 3 lines, the reads reduce one digits again to became 14 digits(also over 10Mhz),  again after that few lines the frequency less that 10Mhz, it return to 13 digits like last few line. My question is what determines the HP53132A’s resolution? Why the output is variable digits even the had very close frequency? Any comment will be appreciate.
The HP53132A uses a couple of tricks to achieve high resolution at
relatively high rate. First of all, the counter has a 150 ps single-shot
resolution. Then, it average over a block of measurements. For achieving
frequency resolution, this is a good strategy as it gives a high rate of
high precision readouts.

The read-out format of the HP53132A could have been a bit more normal.
Exactly how they determine the displayed number of digits I don't know,
but I wanted to at least give you  the counter resolution side.

Be aware that this averaging of values does not make it very fit for
ADEV (and friends) calculation, as it provides a pre-filter which makes
the values look unusually good for low ADEV, but for higher taus, the
filtering effect rubs off and more proper ADEV values is achieved. The
bias could be compensated, but that would only amount to remove the
droop, which ends up giving you the properties of the raw values.

Cheers,
Magnus