[time-nuts] notch filter for close in phase noise measurement

Bob Camp kb8tq at n1k.org
Sun Oct 2 19:07:21 UTC 2016


Hi

One of the limitations of the return loss bridge approach can be the loaded Q. Indeed 50 db
would be pretty awful.  You have more control over things like notch depth and bandwidth with
a more “classical” notch filter topology. You can indeed use overtone crystals in that case
which gives you a higher unloaded Q from the resonator. 

If indeed you come up with a 30 Hz wide at 3 db notch filter, you better put it in an OCXO style
enclosure. You also better have a way to tune it to match your signal source. Room ambient
variations will have you chasing it all over the place otherwise. Even a 200 Hz wide filter is going 
to be “twitchy” if that is the 3 db bandwidth. 

Bob

> On Oct 2, 2016, at 12:27 PM, Scott Stobbe <scott.j.stobbe at gmail.com> wrote:
> 
> What's the loaded Q of such a notch filter? 50 dB 100 Hz off of 10 MHz
> sounds like a pretty lousy Q.
> 
> On Sunday, 2 October 2016, Bob Camp <kb8tq at n1k.org> wrote:
> 
>> Hi
>> 
>> The issue is *not* about notch filters and if they are useful. The gotcha
>> is
>> that they are mainly useful far removed from carrier rather than close in.
>> The
>> statement “works 100 Hz off carrier” requires a *lot* of qualifiers to
>> make it
>> apply in real measurements.  Without getting into the limitations, it is
>> very difficult
>> to determine just how close to carrier you can go with a notch. That is not
>> implementation specific it applies to all notches. You *do* need to get
>> into the
>> details.
>> 
>> The stuff we have gone over so far is hardly an exhaustive list. There are
>> many issues.
>> We have yet to get into the amount of power being delivered to the crystal
>> in the
>> notch filter and the behavior of crystals when driven with a lot of power
>> ….
>> 
>> Bob
>> 
>>> On Oct 2, 2016, at 12:10 PM, Adrian Rus <adrian.rus at broadhurst.ro
>> <javascript:;>> wrote:
>>> 
>>> Bob,
>>> There is no issue about using or not using notch filters. We know (?)
>> all of these. It is about a damn simple topology. Yes, this topology has
>> its shortcomings (impedance, variation with freq and the so).
>>> The math is close, but not exact.
>>> I gave up here. If the topology is of no use, this is it. For me it was
>> funy to _discover_ that simple topology doing a notch.
>>> ‎Best,
>>> Adrian
>>> 
>>> Sent from my BlackBerry 10 smartphone on the Orange network.
>>> Original Message
>>> From: Bob Camp
>>> Sent: Sunday, October 2, 2016 18:35
>>> To: Discussion of precise time and frequency measurement
>>> Reply To: Discussion of precise time and frequency measurement
>>> Subject: Re: [time-nuts] notch filter for close in phase noise
>> measurement
>>> 
>>> 
>>> Hi
>>> 
>>> Ok, the next issue with the notch filter technique is the termination of
>> the oscillator
>>> it’s self. The notch may (or more likely not) provide a proper 50 ohm
>> load at the
>>> carrier frequency. Even if it is correct at the carrier, it will go off
>> impedance as it
>>> moves away from carrier.  You either need a pad in series with the
>> oscillator (which
>>> drops sensitivity) or something similar (like an isolator). The gotcha
>> here is that the
>>> phase noise of the device may not be the same when it is incorrectly
>> terminated. The
>>> issue is more significant in minimum stage devices or when the output
>> stage contributes
>>> to the total noise of the device.
>>> 
>>> A bit of math:
>>> 
>>> A good 10 MHz oscillator will be in the -155 to -165 dbc / Hz range at
>> 100 Hz off carrier.
>>> If you have lost 20 db of energy due to the notch width, that is now
>> -175 to -185 dbc / Hz.
>>> If the oscillator is putting out +10 dbm, that would be -165 to -175 dbm
>> / Hz. The lower
>>> number is at the KTB level without any loss in the bridge, a the
>> attenuator, or noise figure in the
>>> post amplifier. The higher number is only 10 db away. If the notch has a
>> bit more loss, things
>>> get even tighter. This is more than just a theoretical issue.
>>> 
>>> After that you do get into the AM + PM thing. The notch is normally
>> proposed for use on
>>> floor measurements. Details are in the FCS paper by Stone back in the
>> 1970’s.  There the argument
>>> is that the noise process *must* be producing equal amounts of AM and PM
>> noise. That makes
>>> the conversion of “what I measured” to phase noise fairly easy. Close
>> in, you can indeed have
>>> processes that produce unequal amounts of AM and PM noise. Without a way
>> to separate the
>>> two, you toss a fairly large bit of doubt into the measurement.
>>> 
>>> Bob
>>> 
>>> 
>>>> On Oct 2, 2016, at 11:03 AM, Adrian Rus <adrian.rus at broadhurst.ro
>> <javascript:;>> wrote:
>>>> 
>>>> Yes. It can be used for offsets starting some 100-200Hz. Plus, the
>> measured noise is PN+AN. Again, the only reason I wanted to share this
>> topology is its outrageous simplicity. All pluses and minuses of notch
>> filer measurement methode, remain.
>>>> Sooner (or later) I shall share with you (after the real life
>> validation) an (again, very simple) interderometric methode.
>>>> Adrian
>>>> 
>>>> Sent from my BlackBerry 10 smartphone on the Orange network.
>>>> Original Message
>>>> From: Bob Camp
>>>> Sent: Sunday, October 2, 2016 17:54
>>>> To: Discussion of precise time and frequency measurement
>>>> Reply To: Discussion of precise time and frequency measurement
>>>> Subject: Re: [time-nuts] notch filter for close in phase noise
>> measurement
>>>> 
>>>> 
>>>> Hi
>>>> 
>>>> The notch is (say) 60 db deep at the carrier frequency. At 100 Hz off
>> the carrier frequency,
>>>> it still has some depth. It might be 50 db deep, it could be 10 db
>> deep. A lot depends on the
>>>> crystal you have. Even if it’s only 10 db deep, the phase noise you
>> measure at 100 Hz off
>>>> carrier will be “off” by 10 db.
>>>> 
>>>> Bob
>>>> 
>>>>> On Oct 2, 2016, at 10:46 AM, Adrian Rus <adrian.rus at broadhurst.ro
>> <javascript:;>> wrote:
>>>>> 
>>>>> Hello,
>>>>> The _generator_ is a reference 10MHz oscillator and the only
>> calibration of the notch is to equal the oscillator freq.
>>>>> The basic idea of the message is its simplicity (as compared to other
>> notch approaches).
>>>>> Best regards,
>>>>> Adrian
>>>>> 
>>>>> Sent from my BlackBerry 10 smartphone on the Orange network.
>>>>> Original Message
>>>>> From: Bob Camp
>>>>> Sent: Sunday, October 2, 2016 17:06
>>>>> To: Discussion of precise time and frequency measurement
>>>>> Reply To: Discussion of precise time and frequency measurement
>>>>> Subject: Re: [time-nuts] notch filter for close in phase noise
>> measurement
>>>>> 
>>>>> 
>>>>> Hi
>>>>> 
>>>>> Getting close to carrier with a notch filter involves a bit of
>> calibration of the notch. It’s not
>>>>> imposible to do, but it is a needed step. The generator you use to do
>> the measurement has
>>>>> to be pretty clean to get adequate data at low offsets.
>>>>> 
>>>>> Bob
>>>>> 
>>>>>> On Oct 2, 2016, at 3:56 AM, Adrian Rus <adrian.rus at broadhurst.ro
>> <javascript:;>> wrote:
>>>>>> 
>>>>>> Hello list,
>>>>>> For those of you interested in phase noise measurement without using
>> fancy/dedicated gear, here you are the way I have got. Disclaimer: as far
>> as I am concerned, all phase noise measurements use a technique to get rid
>> of carrier: quadrature mixing, interferometric [more on that, later] and
>> notch filters.
>>>>>> 
>>>>>> The simplest way use notch filters, and the simplest notch filter can
>> be arranged with just 3 elements:
>>>>>> - one return loss bridge
>>>>>> - one quartz crystal
>>>>>> - one resistor
>>>>>> Hook the crystal on DUT port, the oscillator to be measured on IN
>> port, the SA [spectrum analyzer] on OUT port and the resistor on REF port.
>> The resistor have to be determined by trial and error to equal the series
>> resistence of the crystal at series resonance. From some -50dB up, can hook
>> a potentiometer in parallel to the resistor[s] and fine tune for the
>> deepest notch.
>>>>>> It is easy to get notches as deep as -85-90dB. The filter is useful
>> in close in measurements not closer than 100-200Hz from carrier. Yes,
>> between the notch and SA you should insert a 40-60dB amplifier. The
>> amplifier will not degrade the flicker noise [as there is practical no
>> carrier - see Rubiola papers], but will set the noise floor.
>>>>>> The series resonance freq have to be selected from multiple crystals;
>> I have experienced series resonance in 10MHz crystals ranging from -300Hz
>> to +100Hz against 10MHz sharp, and have selected a crystal resonating at
>> +25Hz at room temperature. For exact fit you can either tune the oscillator
>> @+25Hz, or better, thermostat the crystal; thermostating the crystal will
>> also tune the notch to the desired freq.
>>>>>> My selected crystal was equilibrated by a series resistance of
>> 14.7ohm. Please note, the series resistance of other 11 crystals I have
>> tested range from 14ohm to tens of ohm.
>>>>>> Regards,
>>>>>> Adrian
>>>>>> _______________________________________________
>>>>>> time-nuts mailing list -- time-nuts at febo.com <javascript:;>
>>>>>> 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 at febo.com <javascript:;>
>>>>> 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 at febo.com <javascript:;>
>>>>> 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 at febo.com <javascript:;>
>>>> 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 at febo.com <javascript:;>
>>>> 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 at febo.com <javascript:;>
>>> 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 at febo.com <javascript:;>
>>> 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 at febo.com <javascript:;>
>> 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 at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.




More information about the Time-nuts_lists.febo.com mailing list