Re: [time-nuts] Close-in phase noise measurements
From: Richard (Rick) Karlquist [EMAIL PROTECTED] Subject: Re: [time-nuts] Close-in phase noise measurements Date: Sun, 30 Mar 2008 21:05:50 -0700 Message-ID: [EMAIL PROTECTED] The relationship between phase noise and Allan variance is a complex one and was described in papers at FCS in 1976 and 1978 by my previous manager Mike Fischer (then of HP). I think these papers come closest to answering your question. Add to that the added complexity that you would like to have a window function before you DFT/FFT. Turning a FFT into Allan deviation or the reverse is among the lost causes. Also, there is only a small overlap of interest mostly, so I think they are best handled separatly, except when analysing spikes in the Allan deviation. Cheers, Magnus ___ 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.
Re: [time-nuts] Close-in phase noise measurements
The relationship between phase noise and Allan variance is a complex one and was described in papers at FCS in 1976 and 1978 by my previous manager Mike Fischer (then of HP). I think these papers come closest to answering your question. Rick Karlquist Jeff Mock wrote: This is a half-baked idea I've thinking about for awhile. I wonder if it might be possible to create a single measurement to combine allan variance and phase noise in the same plot. Allan variance usually plots tau in seconds on the x-axis. Instead, you might plot 1/s or frequency on the x-axis. This way, allan variance looks more like very close-in phase noise. For example, a point where tau=1000s becomes the phase noise at 1mHz (milli-hertz) from the carrier. Combining this with more typical phase noise measurements, you can create a single log-log graph covering micro-hertz to hundreds of kilo-hertz. The advantage of combining the measurements into a single entity is that you get most of the characterization parameters for a timebase in a single graph. Would this work? Half-baked, I know... jeff Shane wrote: Do you know much about the RS FSUP50? http://www2.rohde-schwarz.com/en/products/test_and_measurement/product_categ ories/spectrum_analysis/FSUP-%7C-Key_Facts-%7C-4-%7C-966.html -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Bruce Griffiths Sent: Tuesday, March 25, 2008 8:22 PM To: Discussion of precise time and frequency measurement Subject: Re: [time-nuts] Close-in phase noise measurements Shane wrote: Wenzel has a setup you can purchase at low cost. http://www.wenzel.com/pdffiles1/PNTS%201000/BP-1000-SC.pdf Phase noise test sets can be pricey... $200K Shane Their calibration method is somewhat problematic at the low frequency end where the effect of the PLL and the audio amplifier low frequency cutoff may be significant. The NIST calibration technique: http://tf.nist.gov/timefreq/general/pdf/1000.pdf is far superior. Bruce ___ 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.
Re: [time-nuts] Close-in phase noise measurements
Richard (Rick) Karlquist wrote: The relationship between phase noise and Allan variance is a complex one and was described in papers at FCS in 1976 and 1978 by my previous manager Mike Fischer (then of HP). I think these papers come closest to answering your question. Rick Karlquist Essentially the mapping from phase noise to Allan variance isnt 1 to 1. AVAR formula Whilst it is possible to calculate the Allan variance from the phase noise spectrum using the above formula or a variant thereof, the reverse isnt an unambiguous process as it is possible for different phase noise spectra to have the same Allan variance. In practice the Allan Variance low pass filter frequency response isnt necessarily rectangular as assumed in the above formula. Bruce ___ 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.
Re: [time-nuts] Close-in phase noise measurements
Missing formula attached bruce inline: AVARCalc___ 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.
Re: [time-nuts] Close-in phase noise measurements
This is a half-baked idea I've thinking about for awhile. I wonder if it might be possible to create a single measurement to combine allan variance and phase noise in the same plot. Allan variance usually plots tau in seconds on the x-axis. Instead, you might plot 1/s or frequency on the x-axis. This way, allan variance looks more like very close-in phase noise. For example, a point where tau=1000s becomes the phase noise at 1mHz (milli-hertz) from the carrier. Combining this with more typical phase noise measurements, you can create a single log-log graph covering micro-hertz to hundreds of kilo-hertz. The advantage of combining the measurements into a single entity is that you get most of the characterization parameters for a timebase in a single graph. Would this work? Half-baked, I know... jeff Shane wrote: Do you know much about the RS FSUP50? http://www2.rohde-schwarz.com/en/products/test_and_measurement/product_categ ories/spectrum_analysis/FSUP-%7C-Key_Facts-%7C-4-%7C-966.html -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Bruce Griffiths Sent: Tuesday, March 25, 2008 8:22 PM To: Discussion of precise time and frequency measurement Subject: Re: [time-nuts] Close-in phase noise measurements Shane wrote: Wenzel has a setup you can purchase at low cost. http://www.wenzel.com/pdffiles1/PNTS%201000/BP-1000-SC.pdf Phase noise test sets can be pricey... $200K Shane Their calibration method is somewhat problematic at the low frequency end where the effect of the PLL and the audio amplifier low frequency cutoff may be significant. The NIST calibration technique: http://tf.nist.gov/timefreq/general/pdf/1000.pdf is far superior. Bruce ___ 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.
Re: [time-nuts] Close-in phase noise measurements
I am continuing my phase noise measurement quest. I gathered equipment (HP 8662a/11729C/8568B/multipliers) to measure 100Hz+ from the carrier. I now need to get a grip on the 0.1-100Hz range, which is where most of my applications are. What is the suggested measurement methodology for this range? My first idea would be to squeeze out the most of the above equipment...maybe add a dynamic signal analyzer like the 35660/3561/3562? What is a good HP LF analyzer? This combined with multiplication (to 1GHz) and EFC locking, could take care of the 1-100HZ range? Anythings else needed? No ideas for 0.1-1... maybe by counting with the 5370B? I am sure this has been answered before but the archives are difficult to search... So, I've started to look into the sub-100 Hz regime lately, but haven't had time to get very serious about it. Here's my take on the question: 1) The TSC-51xx analyzers are sure nifty if you have the budget and can live with being limited to HF measurements. They offer high performance with less fiddling around than any other PN-measurement platform I've seen. 2) The next best thing would be a sound-card FFT option that works in conjunction with an HF analyzer. PN.EXE will do this at some point, using the 11729's front-panel LF analyzer output. It's really just a matter of writing the UI code to support it. 3) The next best thing, after that, is a dedicated FFT analyzer, perhaps in conjunction with an HP 3047A or 3048A system (i.e., an FFT analyzer plus a 35601A or 11848A interface). Boxes like the HP 3561A and 3562A are neat, but they are complete technological relics. They are limited to about 80 dB of dynamic range in a 100-kHz bandwidth, they take up quite a bit of space, and they require either a lot of custom coding or some obsolete HP software. The 3582A is in the same boat, only more so. An 11729B/C plus a simple sound-card interface will be the clear winner once the software support issue is resolved. Disclaimer: I do have a 3561A and a 3562A. They are really cool boxes, but if I ever write any code to support them in PN.EXE, it'll be because I think they're cool, not because I think they're the right answer for any FFT-analysis applications today. Technology has left them behind in a drastic way that hasn't happened with RF spectrum analyzers. An FFT analyzer can still be used at offsets below 1 Hz, but at that point people usually want to see Allan-deviation graphs rather than PN graphs. It becomes a different problem, since you most likely want to use a counter rather than a spectrum analyzer. -- john, KE5FX ___ 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.
Re: [time-nuts] Close-in phase noise measurements
John Miles wrote: I am continuing my phase noise measurement quest. I gathered equipment (HP 8662a/11729C/8568B/multipliers) to measure 100Hz+ from the carrier. I now need to get a grip on the 0.1-100Hz range, which is where most of my applications are. What is the suggested measurement methodology for this range? My first idea would be to squeeze out the most of the above equipment...maybe add a dynamic signal analyzer like the 35660/3561/3562? What is a good HP LF analyzer? This combined with multiplication (to 1GHz) and EFC locking, could take care of the 1-100HZ range? Anythings else needed? No ideas for 0.1-1... maybe by counting with the 5370B? I am sure this has been answered before but the archives are difficult to search... So, I've started to look into the sub-100 Hz regime lately, but haven't had time to get very serious about it. Here's my take on the question: 1) The TSC-51xx analyzers are sure nifty if you have the budget and can live with being limited to HF measurements. They offer high performance with less fiddling around than any other PN-measurement platform I've seen. The range can be extended with a couple of external mixers (plus low pass IF filters) and a low noise offset source. The offset source is mixed with each of the signals to be compared to produce a pair of IF frequencies within the 0-30MHz range. 2) The next best thing would be a sound-card FFT option that works in conjunction with an HF analyzer. PN.EXE will do this at some point, using the 11729's front-panel LF analyzer output. It's really just a matter of writing the UI code to support it. Most, but not all, sound cards have a low frequency cutoff of 20Hz or so. Some (but not all) sound card ADCs can dc coupled. A high resolution dc coupled ADC may be more effective for frequencies below 20Hz. 3) The next best thing, after that, is a dedicated FFT analyzer, perhaps in conjunction with an HP 3047A or 3048A system (i.e., an FFT analyzer plus a 35601A or 11848A interface). Boxes like the HP 3561A and 3562A are neat, but they are complete technological relics. They are limited to about 80 dB of dynamic range in a 100-kHz bandwidth, they take up quite a bit of space, and they require either a lot of custom coding or some obsolete HP software. The 3582A is in the same boat, only more so. An 11729B/C plus a simple sound-card interface will be the clear winner once the software support issue is resolved. Sound card support appears to be something of a minefield, baudline thinks my 16 bit 48kHz motherboard sound system is a 24 bit 192kHz system. This probably means that the frequency scale and consequently FFT filter noise bandwidths are unreliable. However with a low frequency noise calibration source and set of marker harmonics derived from a crystal these calibration issues can be resolved. Windows software fares little better and some crashes when set to sample at 192kHz (the windows machine has a sound system with a 192kHz 20 bit ADC system). Disclaimer: I do have a 3561A and a 3562A. They are really cool boxes, but if I ever write any code to support them in PN.EXE, it'll be because I think they're cool, not because I think they're the right answer for any FFT-analysis applications today. Technology has left them behind in a drastic way that hasn't happened with RF spectrum analyzers. An FFT analyzer can still be used at offsets below 1 Hz, but at that point people usually want to see Allan-deviation graphs rather than PN graphs. It becomes a different problem, since you most likely want to use a counter rather than a spectrum analyzer. -- john, KE5FX A finite beat frequency is required when using a counter. With low beat frequencies a lower noise zero crossing detector than the counter input trigger circuitry is required to minimise system noise. A dual mixer time difference system can have a lower noise floor than a single mixer system. Bruce ___ 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.
Re: [time-nuts] Close-in phase noise measurements
Wenzel has a setup you can purchase at low cost. http://www.wenzel.com/pdffiles1/PNTS%201000/BP-1000-SC.pdf Phase noise test sets can be pricey... $200K -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Bruce Griffiths Sent: Tuesday, March 25, 2008 5:43 PM To: Discussion of precise time and frequency measurement Subject: Re: [time-nuts] Close-in phase noise measurements John Miles wrote: I am continuing my phase noise measurement quest. I gathered equipment (HP 8662a/11729C/8568B/multipliers) to measure 100Hz+ from the carrier. I now need to get a grip on the 0.1-100Hz range, which is where most of my applications are. What is the suggested measurement methodology for this range? My first idea would be to squeeze out the most of the above equipment...maybe add a dynamic signal analyzer like the 35660/3561/3562? What is a good HP LF analyzer? This combined with multiplication (to 1GHz) and EFC locking, could take care of the 1-100HZ range? Anythings else needed? No ideas for 0.1-1... maybe by counting with the 5370B? I am sure this has been answered before but the archives are difficult to search... So, I've started to look into the sub-100 Hz regime lately, but haven't had time to get very serious about it. Here's my take on the question: 1) The TSC-51xx analyzers are sure nifty if you have the budget and can live with being limited to HF measurements. They offer high performance with less fiddling around than any other PN-measurement platform I've seen. The range can be extended with a couple of external mixers (plus low pass IF filters) and a low noise offset source. The offset source is mixed with each of the signals to be compared to produce a pair of IF frequencies within the 0-30MHz range. 2) The next best thing would be a sound-card FFT option that works in conjunction with an HF analyzer. PN.EXE will do this at some point, using the 11729's front-panel LF analyzer output. It's really just a matter of writing the UI code to support it. Most, but not all, sound cards have a low frequency cutoff of 20Hz or so. Some (but not all) sound card ADCs can dc coupled. A high resolution dc coupled ADC may be more effective for frequencies below 20Hz. 3) The next best thing, after that, is a dedicated FFT analyzer, perhaps in conjunction with an HP 3047A or 3048A system (i.e., an FFT analyzer plus a 35601A or 11848A interface). Boxes like the HP 3561A and 3562A are neat, but they are complete technological relics. They are limited to about 80 dB of dynamic range in a 100-kHz bandwidth, they take up quite a bit of space, and they require either a lot of custom coding or some obsolete HP software. The 3582A is in the same boat, only more so. An 11729B/C plus a simple sound-card interface will be the clear winner once the software support issue is resolved. Sound card support appears to be something of a minefield, baudline thinks my 16 bit 48kHz motherboard sound system is a 24 bit 192kHz system. This probably means that the frequency scale and consequently FFT filter noise bandwidths are unreliable. However with a low frequency noise calibration source and set of marker harmonics derived from a crystal these calibration issues can be resolved. Windows software fares little better and some crashes when set to sample at 192kHz (the windows machine has a sound system with a 192kHz 20 bit ADC system). Disclaimer: I do have a 3561A and a 3562A. They are really cool boxes, but if I ever write any code to support them in PN.EXE, it'll be because I think they're cool, not because I think they're the right answer for any FFT-analysis applications today. Technology has left them behind in a drastic way that hasn't happened with RF spectrum analyzers. An FFT analyzer can still be used at offsets below 1 Hz, but at that point people usually want to see Allan-deviation graphs rather than PN graphs. It becomes a different problem, since you most likely want to use a counter rather than a spectrum analyzer. -- john, KE5FX A finite beat frequency is required when using a counter. With low beat frequencies a lower noise zero crossing detector than the counter input trigger circuitry is required to minimise system noise. A dual mixer time difference system can have a lower noise floor than a single mixer system. Bruce ___ 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.
Re: [time-nuts] Close-in phase noise measurements
Shane wrote: Wenzel has a setup you can purchase at low cost. http://www.wenzel.com/pdffiles1/PNTS%201000/BP-1000-SC.pdf Phase noise test sets can be pricey... $200K Shane Their calibration method is somewhat problematic at the low frequency end where the effect of the PLL and the audio amplifier low frequency cutoff may be significant. The NIST calibration technique: http://tf.nist.gov/timefreq/general/pdf/1000.pdf is far superior. Bruce ___ 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.
Re: [time-nuts] Close-in phase noise measurements
Do you know much about the RS FSUP50? http://www2.rohde-schwarz.com/en/products/test_and_measurement/product_categ ories/spectrum_analysis/FSUP-%7C-Key_Facts-%7C-4-%7C-966.html -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Bruce Griffiths Sent: Tuesday, March 25, 2008 8:22 PM To: Discussion of precise time and frequency measurement Subject: Re: [time-nuts] Close-in phase noise measurements Shane wrote: Wenzel has a setup you can purchase at low cost. http://www.wenzel.com/pdffiles1/PNTS%201000/BP-1000-SC.pdf Phase noise test sets can be pricey... $200K Shane Their calibration method is somewhat problematic at the low frequency end where the effect of the PLL and the audio amplifier low frequency cutoff may be significant. The NIST calibration technique: http://tf.nist.gov/timefreq/general/pdf/1000.pdf is far superior. Bruce ___ 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.
Re: [time-nuts] Close-in phase noise measurements
Most, but not all, sound cards have a low frequency cutoff of 20Hz or so. Some (but not all) sound card ADCs can dc coupled. A high resolution dc coupled ADC may be more effective for frequencies below 20Hz. True; I'm assuming that anyone using a sound card for these purposes is either going to bypass the coupling capacitor in front of the ADC, or calibrate out the highpass response by adding an inverse function. (A QEX article gave an example of the latter technique not too long ago.) I'm still hoping to get a 24-bit, 2.5-MSPS ADC chip hooked up via USB 2.0 fairly soon. That will solve a multitude of problems, eliminating the need for both a sound card and an HF analyzer. I have C code on the PC that's talking to the FPGA, but haven't yet tried to bring the ADC up with it. Sound card support appears to be something of a minefield, baudline thinks my 16 bit 48kHz motherboard sound system is a 24 bit 192kHz system. This probably means that the frequency scale and consequently FFT filter noise bandwidths are unreliable. However with a low frequency noise calibration source and set of marker harmonics derived from a crystal these calibration issues can be resolved. Windows software fares little better and some crashes when set to sample at 192kHz (the windows machine has a sound system with a 192kHz 20 bit ADC system). Yeah, I think it'd be better not to even use the sound-card drivers if possible. At 10 million bytes per second (32 bits/sample at 2.5 MSPS) they won't be an option for the hardware I'm looking at. A dual mixer time difference system can have a lower noise floor than a single mixer system. I wonder if those are still covered by patents in various corners of the world, the way the TSC's dual-ADC architecture appears to be... -- john, KE5FX ___ 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.
Re: [time-nuts] Close-in phase noise measurements
John Miles wrote: Most, but not all, sound cards have a low frequency cutoff of 20Hz or so. Some (but not all) sound card ADCs can dc coupled. A high resolution dc coupled ADC may be more effective for frequencies below 20Hz. True; I'm assuming that anyone using a sound card for these purposes is either going to bypass the coupling capacitor in front of the ADC, or calibrate out the highpass response by adding an inverse function. (A QEX article gave an example of the latter technique not too long ago.) I'm still hoping to get a 24-bit, 2.5-MSPS ADC chip hooked up via USB 2.0 fairly soon. That will solve a multitude of problems, eliminating the need for both a sound card and an HF analyzer. I have C code on the PC that's talking to the FPGA, but haven't yet tried to bring the ADC up with it. Sound card support appears to be something of a minefield, baudline thinks my 16 bit 48kHz motherboard sound system is a 24 bit 192kHz system. This probably means that the frequency scale and consequently FFT filter noise bandwidths are unreliable. However with a low frequency noise calibration source and set of marker harmonics derived from a crystal these calibration issues can be resolved. Windows software fares little better and some crashes when set to sample at 192kHz (the windows machine has a sound system with a 192kHz 20 bit ADC system). Yeah, I think it'd be better not to even use the sound-card drivers if possible. At 10 million bytes per second (32 bits/sample at 2.5 MSPS) they won't be an option for the hardware I'm looking at. That would be the AD7760 then? A dual mixer time difference system can have a lower noise floor than a single mixer system. I wonder if those are still covered by patents in various corners of the world, the way the TSC's dual-ADC architecture appears to be... -- john, KE5FX John Unlikely, the dual mixer time difference technique has been around so long that most patents like: United States Patent 5128909 (issued 7 July 1992, expired on 7 July 2004). which is an extension of the technique to multiple channels, have expired. Bruce ___ 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.
Re: [time-nuts] Close-in phase noise measurements
Shane wrote: Do you know much about the RS FSUP50? http://www2.rohde-schwarz.com/en/products/test_and_measurement/product_categ ories/spectrum_analysis/FSUP-%7C-Key_Facts-%7C-4-%7C-966.html Shane Only whats specified on the datasheet. Its performance appears to be several 10's of dB worse than the state of the art (at low frequency offsets) when using the internal reference oscillator even when using cross correlation. Not clear how much this is improved with an external reference oscillator. It would be nice to know what the instrument noise floor (exclusive of its internal reference) is. Bruce ___ 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.
Re: [time-nuts] Close-in phase noise measurements
Shane wrote: Do you know much about the RS FSUP50? http://www2.rohde-schwarz.com/en/products/test_and_measurement/product_categ ories/spectrum_analysis/FSUP-%7C-Key_Facts-%7C-4-%7C-966.html This is a nice but expensive equipment. Nice: can do PN measurement by phase quadrature(where you need to oscillators) _and_ can doe cross-correlation measuremnts (where you need only one external oscillator). Expensive: depending on options abt US-$ 100,000 to $150,000 If you are ready to spend so much money, consider the Agilent Signal Source Analyser 5052B, who does use the cross-correlation method and has a lot of other features. But both of them are no ham gear yet... Regards Bernd Neubig __ AXTAL GmbH Co. KG Facility MOS Wasemweg 5 D-74821 Mosbach / Germany fon: +49 (6261) 939834 fax: +49 (6261) 939836 www.axtal.com ___ 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.