[time-nuts] Measuring GPS noise floor (sort of)
mspencer12345 at yahoo.ca
Mon Dec 3 13:20:54 UTC 2012
Sorry I meant to say I'm confident the measurements are above the noise floor after approx 100 seconds.
On Mon, 3 Dec, 2012 4:56 AM PST Mark Spencer wrote:
>Thanks for the comments and suggestions.
>With regards to the equipment used to collect the data, I am using two HP 5370B's. The 1 pps output of the PRS10 is connected to the start input of each counter, the 10Mhz output of each of the GPSDO's is connected to the stop input of one of the counters. The 10 Mhz output from the PRS10 is also routed via a distribution amp to the reference input of both counters. I'm confident that the measurements are below the noise floor of the counters beyond approx 100 seconds. The trigger settings on the start input of each counter are identical so I'm hopeful the measurements are being taken at more or less the same time.
>With regards to the common mode issues, yes this crossed my mind, and I expect what I will end up measuring will likely be a function of the longer term performance of the lesser of the two receivers rather than the true GPS noise floor. I should probably have made the "sort of" disclaimer in the title of the thread stronger (: Still I expect this would give me an indication of the likely long term performance of the two GPS receivers in question which from my end user perspective is still useful.
>The two receivers do however receive different numbers of satellites so I'm thinking some GPS system bias may work it's way into the measurements over the long term. I am curious if at some point I will see the computed Adev between the two GPSDO's flatten out and trend upwards.
>The comment regarding the differences in the performance of the GPSDO's vs the GPS system itself also makes sense.
>> Message: 5
>> Date: Mon, 03 Dec 2012 00:17:47 +0100
>> From: Magnus Danielson <magnus at rubidium.dyndns.org>
>> To: time-nuts at febo.com
>> Subject: Re: [time-nuts] Measuring GPS noise floor (sort
>> Message-ID: <50BBE19B.3040800 at rubidium.dyndns.org>
>> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>> On 12/02/2012 11:37 PM, Mark Spencer wrote:
>> > I'm wondering how useful it is to compare two different
>> GPS receivers to a single Rb to get a sense of the GPS noise
>> floor. I've been collecting some data for
>> the last week or so seem be making some headway with this.
>> > Not having a cesium or an Hmaser I elected to compare
>> two different GPSDO's to a single Rb and then assuming the
>> data seemed sane, I thought I would then compute the Adev
>> between the two GPSDO's. My assumption is
>> that by comparing the two GPSDO's to the Rb it should be
>> possible to detect when one GPSDO is mis behaving or
>> possibly when the Rb mis behaves. So far other than a
>> bit of a glitch at the start of the process things seem to
>> be proceeding as I expected. I hope to
>> collect another two weeks or so of data if all goes
>> according to plan.
>> > I'm really only interested in the longer term numbers
>> and I wouldn't put to much trust in the shorter term
>> > Comments would be welcome especially if there is a
>> fundamental flaw with this process.
>> > Looking at the data gave me something to do this
>> afternoon during some down time on a business trip (:
>> I reazlize this approach is probably not how the pros would
>> do this.
>> You do realize that several issues with the GPSDOs will be
>> common mode,
>> such as the errors of the ionspheric model and actual
>> ionspheric effect
>> and that of multipath. The rubidium will to some degree aid
>> illustrate these shifts, but it will be accounted for on the
>> deviation from the common. If you look at the CRIT paper we
>> yesterday, they did exactly what you proposes. For most
>> stuff the 5
>> recievers they used didn't show much differences. If you
>> increase the
>> resolution you might get out more. You might have use for
>> local references, in order to separate out other effects.
>> For instance,
>> with three rubidiums measured, you could single out which of
>> them has an
>> error, and any common mode effects of the GPSDOs could be
>> fairly well
>> separated out from the local performance of your rubidiums.
More information about the time-nuts