From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from vsmx002.dclux.xion.oxcs.net (vsmx002.dclux.xion.oxcs.net [185.74.65.108]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by lists.bufferbloat.net (Postfix) with ESMTPS id 184EF3B2A4 for ; Wed, 27 Jul 2022 17:03:07 -0400 (EDT) Received: from proxy-2.proxy.oxio.ns.xion.oxcs.net (proxy-2.proxy.oxio.ns.xion.oxcs.net [105.163.34.62]) by mx-out.dclux.xion.oxcs.net (Postfix) with ESMTPA id 26EBF8C0332 for ; Wed, 27 Jul 2022 21:03:03 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=simple/simple; d=dclux.xion.oxcs.net; s=mail1; t=1658955786; bh=g+Y/mxQqBvnrxzfLDtNIDyqqmU2jHQcv7/BHwaNKLYo=; h=Date:From:To:In-Reply-To:References:Subject:From; b=BK+itYGOJs7mqFsaLgXi1Ou5BwcmN44Vk4UvHwn83eR5FkOUFFMePLKykOWRpQ4nk F3Jn7rmSSz2jcSrFclJoj6PcJHwlrrFNsNTyDgcw/IgoBQzvH3Th/TtZwQR3JZevWz hwPj2sSUVLG4rxSqih8Umu0YnSQMLBRppGUiIru/c4DyV+CBfXSyUybUVWU86HiRdF AY/yOQFgxnKTeFLN1BjEu19gSJnl9RbXP6E4qHyLTnbnGLileXIBjU8k5L2bJPe2S3 s31X/gzYSrwS2Zk5P9pBZwmpek523FEm8ANpGRswb7x4z9U/52ZJ0mHCa7zp2qEHQo lyV0zmZ6hl9LQ== Date: Thu, 28 Jul 2022 00:02:55 +0300 From: Mike Puchol To: starlink@lists.bufferbloat.net Message-ID: <7c498f9c-f1f3-4689-b33c-1caf8cde02e0@Spark> In-Reply-To: <2f1b21a5-1aac-caa5-adb7-770a436d61de@auckland.ac.nz> References: <13620203-c86f-d6d3-ed46-841109de105f@auckland.ac.nz> <3d8ae0b7-3f37-417c-9e54-76e6f9e08991@Spark> <2f1b21a5-1aac-caa5-adb7-770a436d61de@auckland.ac.nz> X-Readdle-Message-ID: 7c498f9c-f1f3-4689-b33c-1caf8cde02e0@Spark MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="62e1a804_288f1a34_d91b" X-VadeSecure-Status: LEGIT X-VADE-STATUS: LEGIT Subject: Re: [Starlink] starlink extensions over uk X-BeenThere: starlink@lists.bufferbloat.net X-Mailman-Version: 2.1.20 Precedence: list List-Id: "Starlink has bufferbloat. Bad." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 27 Jul 2022 21:03:07 -0000 --62e1a804_288f1a34_d91b Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline I understand your points - Starlink, at most (where license allows), can = use 10.7 to 12.7 GHz in downlink, split in 8 250 MHz channels, with an em= issions designator that uses 240 MHz of each channel. =46rom my initial o= bservations, the signal is O=46DMA, using the entire 240 MHz. In uplink, = they can use at most 14.0 to 14.5 GHz, split in 8 62.5 MHz channels. In u= plink, from my measurements, they also use O=46DMA and take the entire ba= ndwidth by a single terminal when available: NGSO constellations are regulated at a disadvantage compared to GSO, as t= he latter cannot move or use alternative paths between satellite and term= inal, whereas NGSO can adapt. Thus, GSO protection arc, EP=46D limits, et= c. - all designed to protect the GSO operators. It is unlikely that out-of-band interference is an issue, as I=E2=80=99m = sure both satellites and terminals have gone through a fairly rigurous de= sign and test process. Things that I have seen in terrestrial networks, w= ith DC carriers emanating from 3 GHz point-to-point links, which totally = killed 2.4 GHz Wi=46i, are unlikely to take place. Starlink will only operate in the channels it is authorized over every sp= ecific service region, so the only realistic impact is primarily from in-= band interference, if, for example, GSO protection is not implemented or = enforced well, where both GSO operators and Starlink are licensed to use = the same frequencies (TV LNBs cover almost exactly the entire Starlink do= wnlink band, so you can see concerns from GSO operators in downlink). Ka band is usually more controlled, and gateway to satellite discipline i= s better enforced, plus the use of Cassegrain antennas reduces beamwidth = and sidelobes considerably, compared to the ESAs in Ku band. =46inally, the GSO protection is calculated by the operator, based on the= GIMS and Transfinite calculations, taking their system=E2=80=99s input p= arameters. Telesat, for example, filed with O=46COM with only 4.5=C2=BA o= f GSO protection, due to their particular link characteristics. I do think we are yet to see real world effects from multiple constellati= ons and their additive effects, ESA side lobes, etc. - you can only simul= ate so much before the wonderful and mysterious world of R=46 takes over.= Best, Mike On Jul 27, 2022, 16:37 +0300, Ulrich Speidel via Starlink , wrote: > On 27/07/2022 1:35 am, Mike Puchol via Starlink wrote: > > > > > The interference is directly in-band, as the Ku and Ka bands are used= both by satellite TV, GEO data systems, and Starlink. Thus, if a Starlin= k satellite in-line between a satellite TV customer and the GEO satellite= , the satellite TV customer would experience considerable interference. T= his is why Starlink satellites, gateways, or user terminals cannot transm= it anywhere between 10=C2=BA above and 10=C2=BA below the GSO arc (18=C2=BA= previously). > Hang on a second - there's a terminology issue here relating to the ter= m =22band=22. =22Ku band=22 refers to frequencies between (roughly) 12-18= GHz and =22Ka band=22 to frequencies between about 26-40 GHz. =22Out-of-= band interference=22 refers to signals emanating from a transmitter that = are outside its intended signal bandwidth - and that signal bandwidth is = normally a lot smaller than the bandwidth of the Ku or Ka band - unless w= e're talking UWB applications here, and we're not. > A Starlink satellite or ground station will only ever transmit (wanted = signal) within a small sub-band within these bands. =46or example, in NZ,= Starlink only holds licenses for five 500 MHz sub-bands of Ka between 27= .5 GHz and 30 GHz, and half a dozen 250 MHz subbands of the Ku band. So w= e can safely assume that any transmissions from Starlink ground stations = will take place within these limits. We can also safely assume that an in= dividual transmission between a ground station and a single satellite wil= l occupy at most one of these sub-bands, and for Dishys it's likely to be= a lot less than that. > Any GEO sat listening to that sub-band from behind the transmission's t= arget Starlink bird would indeed experience in-band interference. Any GEO= sat listening to other sub-bands of Ka or Ku that the one used for uplin= k may also experience interference, but in this case it's out of band bec= ause it isn't within the sub-band that the Starlink ground station transm= its in. > Normally: > > =E2=80=A2 A transmitter will be designed so as to minimise emissions ou= tside its core wanted signal spectrum in the sub-band / channel in which = it's operating. Why=3F Because power outside that core bandwidth is waste= d, and can be a source of interference to others. This is achieved with a= band pass filter around the transmit frequency that is designed to let t= he wanted signal pass and hold anything else back. > =E2=80=A2 Similarly, a receiver listening to a signal on a different fr= equency will use a band pass filter to keep signals outside that wanted f= requency range out. > > Think of ideal signals as being vehicles on a road that stick to their = own lane so they don't bump into each other. So where's the problem=3F Th= e problem is twofold: =46irstly, receivers are designed to handle very we= ak signals and amplify them to the point where they =22mean something=22 = if there's any meaning to be had in them. Secondly, those band pass filte= rs aren't perfect. They suppress signals in the unwanted part of by a dec= ent amount but not completely. How =22decent=22 an amount depends on the = construction of the band pass filter, and that in turn depends on size, p= recision, material, number of filter elements, resistive losses etc. It a= lso depends on how far that frequency of interest that you want to have s= uppressed is from the band that that band pass filter will let pass. The = further the better. > =46or the GHz microwave spectrum in which the Ku and Ka bands lie, band= pass filters are essentially just appropriately shaped pieces of metal. = Size isn't so much the issue as wavelengths are very short, but precision= and material / resistive losses (skin effect) are. There's only so much = gold you want to put into a Dishy and gold only gets you that far. > So in order to interfere with a GEO sat receiver, a Starlink ground ter= minal will have to either: > 1) Produce a spurious signal in an unwanted part of the spectrum (not t= o be confused with antenna side lobes, that's a different type of unwante= d signal - any part of the spectrum but wrong direction of propagation an= d normally at a much lower power level). In any decent transmitter with b= and pass filter, this is likely to be several orders of magnitude below t= he wanted signal. That unwanted signal has to propagate along the much lo= nger path to the GEO sat, where it is subject to path loss. It then has t= o arrive at the GEO sat with sufficient signal strength in order to raise= the noise + interference level at the GEO sat's receiver to a point wher= e the signal-to-noise-plus-interference ratio at the receiver falls below= the minimum required for that sat's mission. > 2) Have its wanted signal communicate well beyond its intended target (= extra path loss again) to have it suppressed to a good extent by the band= pass filter at the GEO sat receiver that is there to keep out of band si= gnals out. If the residual signal from the Starlink terminal is still str= ong enough to raise the noise + interference floor at the receiver enough= to lower the signal-to-noise-plus-interference ratio at the GEO sat, the= n again we have a problem. > Sadly, both are possible - however as I've argued, it's probably less o= f an issue with Starlink than with other types of NGSO services. > Side lobes are usually orders of magnitude below the main lobe, so tend= to be less of an issue than the main lobe as they project a much lower s= ignal to start with. Plus these signals aren't correlated when coming fro= m different transmitters, so it's a matter of =22powers add=22 not =22amp= litudes add=22, and powers drop by one over distance squared, so... I'm n= ot sure how seriously one should take these complaints. > In the R=46 world, not everything is as it seems. When they ask you to = turn your mobile (cell) off on the plane, the usual grounds proffered are= some nebulous claims of interference with navigational systems - which h= asn't actually stopped people from putting base stations into airliners, = although their means of navigation haven't changed all that much. Plus, a= lot of planes fly with active cellphones on board. The reason why airlin= es don't want this is because the mobile networks aren't designed for han= dovers at several hundred knots from 30,000 feet, where a phone can keep = a large number of base stations very busy with handovers. So the mobile o= perators require airlines to minimise this nuisance - and that's why they= play on your fear of flying, and why it's become an accepted part of fly= ing culture that your plane might crash if you leave your phone on... ;-)= > > > > This presentation on the subject was shared by a friend earlier, it i= s a really good read on the topic:=C2=A0https://www.itu.int/en/ITU-D/Regi= onal-Presence/AsiaPacific/Documents/Events/2017/Aug-ISS2017/PAPER=5FWorks= hop=5FS3=5FTimur.pdf > > > > Phased array antennas are notoriously prone to generating considerabl= e sidelobes, unlike e.g. a Cassegrain dish. A good article on these (and = mitigations) can be found here:=C2=A0https://www.mwrf.com/technologies/sy= stems/article/21143497/analog-devices-phasedarray-antenna-patterns-part-6= sidelobes-and-tapering > > > > These sidelobes contribute to the interference, and have been a major= source of complaints by the likes of Viasat and others - whereby they cl= aim SpaceX doesn=E2=80=99t consider the additive effects of sidelobes fro= m dozens or hundreds of ESAs on satellites and terminals, against a singl= e victim earth station or satellite. > > > > Best, > > > > Mike > > On Jul 26, 2022, 16:10 +0300, Ulrich Speidel via Starlink , wrote: > > > So then the difference really is just in the GSO protection setting= s I guess. > > > The GSO protection is - to an extent - also patch protection. After= all - who'd need satellite TV if everyone could watch the same TV via LE= Os=3F But consider that: > > > > > > =E2=80=A2 Interference to GEO sats from Starlink & Co. is out of ba= nd. I'm not sure what the out-of-band emissions profiles of dishys are, b= ut I'd imagine we'd be looking at the usual few dozen dB below peak. > > > =E2=80=A2 Distance to GSO is around 64 times (2=5E6) larger than to= Starlink's orbits - GEO sats see around 1/4000th ((2=5E6)=5E-2) of the p= ower from a ground station that the Starlink satellite in front of it see= s. That's 36 dB in extra separation. > > > =E2=80=A2 Dishy is comparatively small in cross-section, and that s= everely limits its gain. Most serious GEO uplink applications that I'm aw= are of use dishes more like 8 times dishy's cross section. That's another= 9 dB or so in separation between a Starlink dishy signal and signal dire= cted specifically at a GEO sat just from the overall size. Uplinks from g= ateways are likely to be much more of an issue (always on and probably hi= gher power as well as =7E6 dB higher antenna gain judging from the photos= I've seen), but then again these point away from the GSO when serving bi= rds further north. > > > =E2=80=A2 My understanding (correct me if you think I'm wrong here)= : It's also possible that Dishy's nature as a phased array helps here. Wh= y=3F Out-of-band interference results from intermodulation and other unwa= nted emissions from the power amplifiers (PA) of transmitters. Where thes= e are fed into a dish from a single PA, their amplitudes get amplified by= the gain of the dish. Now say we're trying to replace that dish by an ar= ray with N antennas and N associated PAs that feed at the appropriate pha= se. Then each element (individual PA with associated antenna) needs to co= ntribute P/N of the total transmit power P of the big PA & dish. Now reme= mber that power is proportional to the square of the amplitude. As long a= s the wanted signal components from the PA are correlated - and they have= to be for the phased array to work - their N amplitudes add up, meaning = the total output power of the wanted signal across all elements is propor= tional to N=5E2. So each element only needs to contribute an amplitude pr= oportional to 1/N in order to produce the wanted output at the correct po= wer. However, now each PA produces its own dirt signal. But unlike the wa= nted signal, these unwanted signals aren't necessarily all correlated bet= ween the elements. As a result, the amplitudes of the unwanted signals fr= om the PAs will partially cancel out as the signal combines into the beam= that is being formed. So only the powers of these uncorrelated unwanted = signals add up, but their amplitudes don't, meaning we now have an unwant= ed signal power that is proportional only to N rather than N=5E2, resulti= ng in a higher ratio of wanted signal to unwanted signal. Again that'd he= lp a lot with separation in an array with lots of elements and associated= PAs. That said, published research into the unwanted emissions of phased= arrays is still a bit in its infancy, and I can't profess to understand = enough about Dishy's innards or the level of correlation between unwanted= emissions in the array, but I'd consider it possible that this allows fo= r some relaxation on the GSO protection parameters when it comes to Starl= ink. > > > > > > On 26/07/2022 9:06 am, Mike Puchol via Starlink wrote: > > > > Your calculations are =E2=80=9CBack of the Envelope Approved=E2=80= =9D :-) > > > > > > > > My simulator can be found at https://starlink.sx and runs on your= browser (desktop only). I have just released v1.10.6, which adds a slide= r for setting the GSO protection. This is the difference in Sweden from 1= 8=C2=BA (as used until recently) versus 10=C2=BA (as found in recent ITU = filings for STEAM-1B): > > > > > > > > <18=5Fvs=5F10=5FGSO.png> > > > > > > > > Minimum elevation in both cases is 25=C2=BA, so that variable is = unchanged. > > > > > > > > Best, > > > > > > > > Mike > > > > On Jul 25, 2022, 11:35 +0300, Ulrich Speidel via Starlink , wrote: > > > > > I haven't got Mike's nice software but I got snail mail today, = so let's try the back of one of the envelopes. > > > > > 53 degrees north (where the existing constellation tops out) is= roughly the latitude of Nottingham or Stoke-on-Trent. Note that this is = also the latitude with the largest number of birds per km (or mile, if yo= u so prefer) of parallel. They sit almost cheek to jowl there. > > > > > Now the northern tip of the UK's main islands is at around 59 d= egrees north - that's six degrees more. One degree of latitude is roughly= equivalent to 10,000 km / 90 degrees =3D 111 km, so we're talking around= a devilish 666 km to the north of Nottingham here as the crow flies. > > > > > Lets assume Dishy points itself due south at 59 degrees. The bi= rds are at about 550 km. So doing a flat earth approximation for the mome= nt, we're having to point at a satellite 550 km up from 660 km away - tha= t gives an elevation of about 40 degrees (=3Darctan(550/660)). Ballpark. = Now reduce that by the 6 degrees of difference to account for the curvatu= re of the earth and we still have around 34 degrees, allowing for a bit o= f leeway either side if the closest bird isn't actually due south but a b= it to the east or west. Again, ballpark, but we're nowhere near 25 degree= s yet. > > > > > So I'd say that should cover it nicely=3F Where does my envelop= e err=3F GEO arc protection aside for the moment, of course. And of cours= e just because Starlink offers you service doesn't mean that it's actuall= y continuous. > > > > > On 25/07/2022 8:59 am, Dave Taht via Starlink wrote: > > > > > > =46or those of you that don't follow mike's twitter feed... > > > > > > > > > > > > https://twitter.com/mikepuchol/status/1551288485713149952 > > > > > > > > > > > > -- > > > > > > =46Q World Domination pending: https://blog.cerowrt.org/post/= state=5Fof=5Ffq=5Fcodel/ > > > > > > Dave T=C3=A4ht CEO, TekLibre, LLC > > > > > > =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F > > > > > > Starlink mailing list > > > > > > Starlink=40lists.bufferbloat.net > > > > > > https://lists.bufferbloat.net/listinfo/starlink > > > > > -- > > > > > ***************************************************************= * > > > > > Dr. Ulrich Speidel > > > > > > > > > > School of Computer Science > > > > > > > > > > Room 303S.594 (City Campus) > > > > > > > > > > The University of Auckland > > > > > u.speidel=40auckland.ac.nz > > > > > http://www.cs.auckland.ac.nz/=7Eulrich/ > > > > > ***************************************************************= * > > > > > > > > > > > > > > > > > > > > =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F > > > > > Starlink mailing list > > > > > Starlink=40lists.bufferbloat.net > > > > > https://lists.bufferbloat.net/listinfo/starlink > > > > > > > > =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= > > > > Starlink mailing list > > > > Starlink=40lists.bufferbloat.net > > > > https://lists.bufferbloat.net/listinfo/starlink > > > > > > > -- > > > **************************************************************** > > > Dr. Ulrich Speidel > > > > > > School of Computer Science > > > > > > Room 303S.594 (City Campus) > > > > > > The University of Auckland > > > u.speidel=40auckland.ac.nz > > > http://www.cs.auckland.ac.nz/=7Eulrich/ > > > **************************************************************** > > > > > > > > > > > > =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F > > > Starlink mailing list > > > Starlink=40lists.bufferbloat.net > > > https://lists.bufferbloat.net/listinfo/starlink > > > > =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F > > Starlink mailing list > > Starlink=40lists.bufferbloat.net > > https://lists.bufferbloat.net/listinfo/starlink > > > -- > **************************************************************** > Dr. Ulrich Speidel > > School of Computer Science > > Room 303S.594 (City Campus) > > The University of Auckland > u.speidel=40auckland.ac.nz > http://www.cs.auckland.ac.nz/=7Eulrich/ > **************************************************************** > > > > =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F > Starlink mailing list > Starlink=40lists.bufferbloat.net > https://lists.bufferbloat.net/listinfo/starlink --62e1a804_288f1a34_d91b Content-Type: text/html; charset="utf-8" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline
I understand your points - Starlink, at most (where= license allows), can use 10.7 to 12.7 GHz in downlink, split in 8 250 MH= z channels, with an emissions designator that uses 240 MHz of each channe= l. =46rom my initial observations, the signal is O=46DMA, using the entir= e 240 MHz. In uplink, they can use at most 14.0 to 14.5 GHz, split in 8 6= 2.5 MHz channels. In uplink, from my measurements, they also use O=46DMA = and take the entire bandwidth by a single terminal when available:

NGSO constellations are regulated at a disadvantage compared to GSO, as t= he latter cannot move or use alternative paths between satellite and term= inal, whereas NGSO can adapt. Thus, GSO protection arc, EP=46D limits, et= c. - all designed to protect the GSO operators.

It is unlikely that out-of-band interference is an issue, as I=E2=80=99m = sure both satellites and terminals have gone through a fairly rigurous de= sign and test process. Things that I have seen in terrestrial networks, w= ith DC carriers emanating from 3 GHz point-to-point links, which totally = killed 2.4 GHz Wi=46i, are unlikely to take place.

Starlink will only operate in the channels it is authorized over every sp= ecific service region, so the only realistic impact is primarily from in-= band interference, if, for example, GSO protection is not implemented or = enforced well, where both GSO operators and Starlink are licensed to use = the same frequencies (TV LNBs cover almost exactly the entire Starlink do= wnlink band, so you can see concerns from GSO operators in downlink).

Ka band is usually more controlled, and gateway to satellite discipline i= s better enforced, plus the use of Cassegrain antennas reduces beamwidth = and sidelobes considerably, compared to the ESAs in Ku band.

=46inally, the GSO protection is calculated by the operator, based on the= GIMS and Transfinite calculations, taking their system=E2=80=99s input p= arameters. Telesat, for example, filed with O=46COM with only 4.5=C2=BA o= f GSO protection, due to their particular link characteristics.

I do think we are yet to see real world effects from multiple constellati= ons and their additive effects, ESA side lobes, etc. - you can only simul= ate so much before the wonderful and mysterious world of R=46 takes over.=

Best,

Mike
On Jul 27, 2022, 16:37 +0300, Ulric= h Speidel via Starlink <starlink=40lists.bufferbloat.net>, wrote:
On 27/07/2022 1:35 am, Mike Puchol via= Starlink wrote:


The interference is directly in-band, as the Ku and Ka bands are used bot= h by satellite TV, GEO data systems, and Starlink. Thus, if a Starlink sa= tellite in-line between a satellite TV customer and the GEO satellite, th= e satellite TV customer would experience considerable interference. This = is why Starlink satellites, gateways, or user terminals cannot transmit a= nywhere between 10=C2=BA above and 10=C2=BA below the GSO arc (18=C2=BA p= reviously).

Hang on a second - there's a terminology issue here relating to the te= rm =22band=22. =22Ku band=22 refers to frequencies between (roughly) 12-1= 8 GHz and =22Ka band=22 to frequencies between about 26-40 GHz. =22Out-of= -band interference=22 refers to signals emanating from a transmitter that= are outside its intended signal bandwidth - and that signal bandwidth is= normally a lot smaller than the bandwidth of the Ku or Ka band - unless = we're talking UWB applications here, and we're not.

A Starlink satellite or ground station will only ever transmit (wanted= signal) within a small sub-band within these bands. =46or example, in NZ= , Starlink only holds licenses for five 500 MHz sub-bands of Ka between 2= 7.5 GHz and 30 GHz, and half a dozen 250 MHz subbands of the Ku band. So = we can safely assume that any transmissions from Starlink ground stations= will take place within these limits. We can also safely assume that an i= ndividual transmission between a ground station and a single satellite wi= ll occupy at most one of these sub-bands, and for Dishys it's likely to b= e a lot less than that.

Any GEO sat listening to that sub-band from behind the transmission's = target Starlink bird would indeed experience in-band interference. Any GE= O sat listening to other sub-bands of Ka or Ku that the one used for upli= nk may also experience interference, but in this case it's out of band be= cause it isn't within the sub-band that the Starlink ground station trans= mits in.

Normally:

  • A transmitter will be designed so as to minimise emissions outside it= s core wanted signal spectrum in the sub-band / channel in which it's ope= rating. Why=3F Because power outside that core bandwidth is wasted, and c= an be a source of interference to others. This is achieved with a band pa= ss filter around the transmit frequency that is designed to let the wante= d signal pass and hold anything else back.
  • Similarly, a receiver listening to a signal on a different frequency = will use a band pass filter to keep signals outside that wanted frequency= range out.

Think of ideal signals as being vehicles on a road that stick to their= own lane so they don't bump into each other. So where's the problem=3F T= he problem is twofold: =46irstly, receivers are designed to handle very w= eak signals and amplify them to the point where they =22mean something=22= if there's any meaning to be had in them. Secondly, those band pass filt= ers aren't perfect. They suppress signals in the unwanted part of by a de= cent amount but not completely. How =22decent=22 an amount depends on the= construction of the band pass filter, and that in turn depends on size, = precision, material, number of filter elements, resistive losses etc. It = also depends on how far that frequency of interest that you want to have = suppressed is from the band that that band pass filter will let pass. The= further the better.

=46or the GHz microwave spectrum in which the Ku and Ka bands lie, ban= d pass filters are essentially just appropriately shaped pieces of metal.= Size isn't so much the issue as wavelengths are very short, but precisio= n and material / resistive losses (skin effect) are. There's only so much= gold you want to put into a Dishy and gold only gets you that far.

So in order to interfere with a GEO sat receiver, a Starlink ground te= rminal will have to either:

1) Produce a spurious signal in an unwanted part of the spectrum (not = to be confused with antenna side lobes, that's a different type of unwant= ed signal - any part of the spectrum but wrong direction of propagation a= nd normally at a much lower power level). In any decent transmitter with = band pass filter, this is likely to be several orders of magnitude below = the wanted signal. That unwanted signal has to propagate along the much l= onger path to the GEO sat, where it is subject to path loss. It then has = to arrive at the GEO sat with sufficient signal strength in order to rais= e the noise + interference level at the GEO sat's receiver to a point whe= re the signal-to-noise-plus-interference ratio at the receiver falls belo= w the minimum required for that sat's mission.

2) Have its wanted signal communicate well beyond its intended target = (extra path loss again) to have it suppressed to a good extent by the ban= d pass filter at the GEO sat receiver that is there to keep out of band s= ignals out. If the residual signal from the Starlink terminal is still st= rong enough to raise the noise + interference floor at the receiver enoug= h to lower the signal-to-noise-plus-interference ratio at the GEO sat, th= en again we have a problem.

Sadly, both are possible - however as I've argued, it's probably less = of an issue with Starlink than with other types of NGSO services.

Side lobes are usually orders of magnitude below the main lobe, so ten= d to be less of an issue than the main lobe as they project a much lower = signal to start with. Plus these signals aren't correlated when coming fr= om different transmitters, so it's a matter of =22powers add=22 not =22am= plitudes add=22, and powers drop by one over distance squared, so... I'm = not sure how seriously one should take these complaints.

In the R=46 world, not everything is as it seems. When they ask you to= turn your mobile (cell) off on the plane, the usual grounds proffered ar= e some nebulous claims of interference with navigational systems - which = hasn't actually stopped people from putting base stations into airliners,= although their means of navigation haven't changed all that much. Plus, = a lot of planes fly with active cellphones on board. The reason why airli= nes don't want this is because the mobile networks aren't designed for ha= ndovers at several hundred knots from 30,000 feet, where a phone can keep= a large number of base stations very busy with handovers. So the mobile = operators require airlines to minimise this nuisance - and that's why the= y play on your fear of flying, and why it's become an accepted part of fl= ying culture that your plane might crash if you leave your phone on... ;-= )

&=23160;
This presentation on the subject was shared by a friend earlier, it is a = really good read on the topic:&=23160;https://www.itu.int/en/ITU-D/Regional-Presence/AsiaPacific= /Documents/Events/2017/Aug-ISS2017/PAPER=5FWorkshop=5FS3=5FTimur.pdf<= br />
Phased array antennas are notoriously prone to generating considerable si= delobes, unlike e.g. a Cassegrain dish. A good article on these (and miti= gations) can be found here:&=23160;https://www.mwrf.com/technologies/systems/article/211434= 97/analog-devices-phasedarray-antenna-patterns-part-6sidelobes-and-taperi= ng

These sidelobes contribute to the interference, and have been a major sou= rce of complaints by the likes of Viasat and others - whereby they claim = SpaceX doesn=E2=80=99t consider the additive effects of sidelobes from do= zens or hundreds of ESAs on satellites and terminals, against a single vi= ctim earth station or satellite.

Best,

Mike
On Jul 26, 2022, 16:10 +0300, Ulric= h Speidel via Starlink <starlink=40lists.bufferbloa= t.net>, wrote:

So then the difference really is just in the GSO protection settings I= guess.

The GSO protection is - to an extent - also patch protection. After al= l - who'd need satellite TV if everyone could watch the same TV via LEOs=3F= But consider that:

  • Interference to GEO sats from Starlink & Co. is out of band. I'm = not sure what the out-of-band emissions profiles of dishys are, but I'd i= magine we'd be looking at the usual few dozen dB below peak.
  • Distance to GSO is around 64 times (2=5E6) larger than to Starlink's = orbits - GEO sats see around 1/4000th ((2=5E6)=5E-2) of the power from a = ground station that the Starlink satellite in front of it sees. That's 36= dB in extra separation.
  • Dishy is comparatively small in cross-section, and that severely limi= ts its gain. Most serious GEO uplink applications that I'm aware of use d= ishes more like 8 times dishy's cross section. That's another 9 dB or so = in separation between a Starlink dishy signal and signal directed specifi= cally at a GEO sat just from the overall size. Uplinks from gateways are = likely to be much more of an issue (always on and probably higher power a= s well as =7E6 dB higher antenna gain judging from the photos I've seen),= but then again these point away from the GSO when serving birds further = north.
  • My understanding (correct me if you think I'm wrong here): It's also = possible that Dishy's nature as a phased array helps here. Why=3F Out-of-= band interference results from intermodulation and other unwanted emissio= ns from the power amplifiers (PA) of transmitters. Where these are fed in= to a dish from a single PA, their amplitudes get amplified by the gain of= the dish. Now say we're trying to replace that dish by an array with N a= ntennas and N associated PAs that feed at the appropriate phase. Then eac= h element (individual PA with associated antenna) needs to contribute P/N= of the total transmit power P of the big PA & dish. Now remember tha= t power is proportional to the square of the amplitude. As long as the wa= nted signal components from the PA are correlated - and they have to be f= or the phased array to work - their N amplitudes add up, meaning the tota= l output power of the wanted signal across all elements is proportional t= o N=5E2. So each element only needs to contribute an amplitude proportion= al to 1/N in order to produce the wanted output at the correct power. How= ever, now each PA produces its own dirt signal. But unlike the wanted sig= nal, these unwanted signals aren't necessarily all correlated between the= elements. As a result, the amplitudes of the unwanted signals from the P= As will partially cancel out as the signal combines into the beam that is= being formed. So only the powers of these uncorrelated unwanted signals = add up, but their amplitudes don't, meaning we now have an unwanted signa= l power that is proportional only to N rather than N=5E2, resulting in a = higher ratio of wanted signal to unwanted signal. Again that'd help a lot= with separation in an array with lots of elements and associated PAs. Th= at said, published research into the unwanted emissions of phased arrays = is still a bit in its infancy, and I can't profess to understand enough a= bout Dishy's innards or the level of correlation between unwanted emissio= ns in the array, but I'd consider it possible that this allows for some r= elaxation on the GSO protection parameters when it comes to Starlink.
On 26/07/2022 9:06 am, Mike Puchol via= Starlink wrote:
Your calculations are =E2=80=9CBack of the Envelope= Approved=E2=80=9D :-)&=23160;

My simulator can be found at https://starlink.sx and runs on your browser (de= sktop only). I have just released v1.10.6, which adds a slider for settin= g the GSO protection. This is the difference in Sweden from 18=C2=BA (as = used until recently) versus 10=C2=BA (as found in recent ITU filings for = STEAM-1B):

<18=5Fvs=5F10=5FGSO.png>

Minimum elevation in both cases is 25=C2=BA, so that variable is unchange= d.

Best,

Mike
On Jul 25, 2022, 11:35 +0300, Ulric= h Speidel via Starlink &l= t;starlink=40lists.bufferbloat.net>, wrote:

I haven't got Mike's nice software but I got snail mail today, so let'= s try the back of one of the envelopes.

53 degrees north (where the existing constellation tops out) is roughl= y the latitude of Nottingham or Stoke-on-Trent. Note that this is also th= e latitude with the largest number of birds per km (or mile, if you so pr= efer) of parallel. They sit almost cheek to jowl there.

Now the northern tip of the UK's main islands is at around 59 degrees = north - that's six degrees more. One degree of latitude is roughly equiva= lent to 10,000 km / 90 degrees =3D 111 km, so we're talking around a devi= lish 666 km to the north of Nottingham here as the crow flies.

Lets assume Dishy points itself due south at 59 degrees. The birds are= at about 550 km. So doing a flat earth approximation for the moment, we'= re having to point at a satellite 550 km up from 660 km away - that gives= an elevation of about 40 degrees (=3Darctan(550/660)). Ballpark. Now red= uce that by the 6 degrees of difference to account for the curvature of t= he earth and we still have around 34 degrees, allowing for a bit of leewa= y either side if the closest bird isn't actually due south but a bit to t= he east or west. Again, ballpark, but we're nowhere near 25 degrees yet.<= br />

So I'd say that should cover it nicely=3F Where does my envelope err=3F= GEO arc protection aside for the moment, of course. And of course just b= ecause Starlink offers you service doesn't mean that it's actually contin= uous.

On 25/07/2022 8:59 am, Dave Taht via S= tarlink wrote:
=46or those of you that don'= t follow mike's twitter feed...

https://twitter.com/mikepuchol/status/15512= 88485713149952

--
=46Q World Domination pending: https://blog.cer= owrt.org/post/state=5Fof=5Ffq=5Fcodel/
Dave T=C3=A4ht CEO, TekLibre, LLC
=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F
Starlink mailing list
S= tarlink=40lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/starlink
--   =20
****************************************************************
Dr. Ulrich Speidel

School of Computer Science

Room 303S.594 (City Campus)

The University of Auckland
u.speid=
el=40auckland.ac.nz   =20
http://www.cs.auckland.ac=
.nz/=7Eulrich/
****************************************************************
=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F
Starlink mailing list
S= tarlink=40lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/starlink

=5F=5F=5F=5F=5F=5F=5F=5F=5F=
=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=
=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F
Starlink mailing list
S=
tarlink=40lists.bufferbloat.net
https://lists.buf=
ferbloat.net/listinfo/starlink


--  =20
****************************************************************
Dr. Ulrich Speidel

School of Computer Science

Room 303S.594 (City Campus)

The University of Auckland
u.speid=
el=40auckland.ac.nz  =20
http://www.cs.auckland.ac=
.nz/=7Eulrich/
****************************************************************
=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F
Starlink mailing list
Starlink=40lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/starlink

=5F=5F=5F=5F=5F=5F=5F=5F=5F=
=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=
=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F
Starlink mailing list
Starlink=40lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/starlin=
k


-- =20
****************************************************************
Dr. Ulrich Speidel

School of Computer Science

Room 303S.594 (City Campus)

The University of Auckland
u.speidel=40auckland.ac.nz =20
http://www.cs.auckland.ac.nz/=7Eulrich/
****************************************************************
=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F= =5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F=5F
Starlink mailing list
Starlink=40lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/starlink
--62e1a804_288f1a34_d91b--