Quote from: OTV Booster on 09/20/2019 04:36 pmThe shorter lifetime due to lower orbits would be less of a hit.If satellite lifespan at 450km is an actual problem you can add years to it by adding 1 kg of krypton to each bird, so 4000kg total upmass. Quite a bit cheaper than adding an extra layer of satellites.If routing is a problem you can add hierarchical routing or share the routing load without an extra physical tier. In any case don't try to simplify software by complexifying hardware unless the software group asks for it. And then give some thought to what they asked for. The number of times I asked for cache and got flops and a thermal management problem ... .
The shorter lifetime due to lower orbits would be less of a hit.
Quote from: RedLineTrain on 09/20/2019 09:03 pmQuote from: ninjaneer on 09/20/2019 04:48 pmQuote from: woods170 on 09/20/2019 07:22 amQuote from: ninjaneer on 09/19/2019 04:14 amHas there been any mention since the May media call of whether Starlink will be sold directly or resold through ISPs?SpaceX is not in the business of reselling its product thru third parties.That applies to Musk's other businesses as well.I very much hope that applies, but has there been any actual evidence to support that claim? If they do intend to start selling modems next year, then there should be some trace signs of customer support and call center development. A best effort lead time would be 180-270 days.The evidence is that there have been no announced marketing deals for Starlink.No need to make this overly complicated. Musk will announce it on Twitter with a sign-up/order page on Starlink.com. They will get a million takers depending on the required deposit, from which they can pick and choose geographies for a beta-ish service.At first, Starlink customer service can be co-located with Tesla's in Las Vegas. Built-out in Las Vegas or elsewhere as necessary.As there is the southern border of Canada to consider in their first iteration of Satellite Service, after 6 launches, do you think that they may utilize Musk's experience with Paypal to enable cross border sign up and payments... Or something similar... or is there some international barrier that I am not aware of?
Quote from: ninjaneer on 09/20/2019 04:48 pmQuote from: woods170 on 09/20/2019 07:22 amQuote from: ninjaneer on 09/19/2019 04:14 amHas there been any mention since the May media call of whether Starlink will be sold directly or resold through ISPs?SpaceX is not in the business of reselling its product thru third parties.That applies to Musk's other businesses as well.I very much hope that applies, but has there been any actual evidence to support that claim? If they do intend to start selling modems next year, then there should be some trace signs of customer support and call center development. A best effort lead time would be 180-270 days.The evidence is that there have been no announced marketing deals for Starlink.No need to make this overly complicated. Musk will announce it on Twitter with a sign-up/order page on Starlink.com. They will get a million takers depending on the required deposit, from which they can pick and choose geographies for a beta-ish service.At first, Starlink customer service can be co-located with Tesla's in Las Vegas. Built-out in Las Vegas or elsewhere as necessary.
Quote from: woods170 on 09/20/2019 07:22 amQuote from: ninjaneer on 09/19/2019 04:14 amHas there been any mention since the May media call of whether Starlink will be sold directly or resold through ISPs?SpaceX is not in the business of reselling its product thru third parties.That applies to Musk's other businesses as well.I very much hope that applies, but has there been any actual evidence to support that claim? If they do intend to start selling modems next year, then there should be some trace signs of customer support and call center development. A best effort lead time would be 180-270 days.
Quote from: ninjaneer on 09/19/2019 04:14 amHas there been any mention since the May media call of whether Starlink will be sold directly or resold through ISPs?SpaceX is not in the business of reselling its product thru third parties.That applies to Musk's other businesses as well.
Has there been any mention since the May media call of whether Starlink will be sold directly or resold through ISPs?
Quote from: DigitalMan on 09/21/2019 04:10 pmQuote from: mlindner on 09/21/2019 02:02 pmIn case it wasn't mentioned before, with the potential close approach to ESA's satellite. That weekend was labor day weekend in the US with the close approach right on labor day, which with the on-call system failure (very commonly used in the software world) there would be literally no person in place who would actually have seen any email. So it's a perfect storm of unfortunate situations (and ESA was still wrong to directly attack SpaceX for it and push that idea in the media).No SpaceX and ESA are not lacking communication. No SpaceX is not acting incorrectly. They're acting just as well as anyone else in the space industry with regards to space junk, if not better than most. They're just new to the process and their systems aren't operational mode yet which will explain any oversight that takes places.I found it interesting Starlink performed 21 autonomous avoidance maneuvers to avoid objects that had no propulsion. Odd that the media isn't talking about it especially since it is a new thing, no?To be honest, I didn't know that either and I can't find it via google. Got a link?
Quote from: mlindner on 09/21/2019 02:02 pmIn case it wasn't mentioned before, with the potential close approach to ESA's satellite. That weekend was labor day weekend in the US with the close approach right on labor day, which with the on-call system failure (very commonly used in the software world) there would be literally no person in place who would actually have seen any email. So it's a perfect storm of unfortunate situations (and ESA was still wrong to directly attack SpaceX for it and push that idea in the media).No SpaceX and ESA are not lacking communication. No SpaceX is not acting incorrectly. They're acting just as well as anyone else in the space industry with regards to space junk, if not better than most. They're just new to the process and their systems aren't operational mode yet which will explain any oversight that takes places.I found it interesting Starlink performed 21 autonomous avoidance maneuvers to avoid objects that had no propulsion. Odd that the media isn't talking about it especially since it is a new thing, no?
In case it wasn't mentioned before, with the potential close approach to ESA's satellite. That weekend was labor day weekend in the US with the close approach right on labor day, which with the on-call system failure (very commonly used in the software world) there would be literally no person in place who would actually have seen any email. So it's a perfect storm of unfortunate situations (and ESA was still wrong to directly attack SpaceX for it and push that idea in the media).No SpaceX and ESA are not lacking communication. No SpaceX is not acting incorrectly. They're acting just as well as anyone else in the space industry with regards to space junk, if not better than most. They're just new to the process and their systems aren't operational mode yet which will explain any oversight that takes places.
I imagine SpaceX will need government approval to sell the ground portion (umm.. the antenna?) in every country they want to sell it in. Anyone have a different opinion?
other cases might require all traffic to go through a gateway in that country.
Quote from: Barley on 09/20/2019 07:33 pmQuote from: OTV Booster on 09/20/2019 04:36 pmThe shorter lifetime due to lower orbits would be less of a hit.If satellite lifespan at 450km is an actual problem you can add years to it by adding 1 kg of krypton to each bird, so 4000kg total upmass. Quite a bit cheaper than adding an extra layer of satellites.If routing is a problem you can add hierarchical routing or share the routing load without an extra physical tier. In any case don't try to simplify software by complexifying hardware unless the software group asks for it. And then give some thought to what they asked for. The number of times I asked for cache and got flops and a thermal management problem ... .They expect the lower sats to have a short lifespan. I think the numbers were ~3-4 years for the low ones and ~5-7 for the high ones. If they’re cheap, de nada.Ouch. Sounds like the teams weren’t integrated as tight as they might have been. My experience with networking is about 16 years out of date and mostly simple local area stuff. I’m really qualified enough to only have a general idea of routing issues, so take my thoughts and questions as coming from a firm foundation of ignorance. On a ground based network ISTM the routing issues would be qualitatively different than StarLink. In either starlink architecture routing would be not just logically but physically dynamic. A ground based router might have several ports but they always physically connect to the same end point and routing is mostly a logical process. starLink has a strong physical routing element so the physical aspect has to be addressed. The two architectures are different ways of addressing this. I’m not so much trying to substitute hardware for software at one point in the system as looking looking at the overall system and thinking about overall system optimization.In what I propose some things get more difficult but others seem to get easier and probably less expensive. Some of the difficulties have been commented on in ways I understand to one degree or another but honestly, I don’t feel like my understanding of the issues has been expanded very much.Ce la viePhil
Semmel, sorry for jumping in late here, but I noticed some things about the link you gave to SpaceMicro's OISL product page.1) They are using two parallel beam expanders, one for transmit and one for receive. This makes a lot of sense to me: using a beamsplitter to combine the two paths into a single optic still leaves you with the problem of calibrating the alignment between the two devices to either fork of the beamsplitter. For calibration you can emit a beam at a diffuse wall at some reasonable distance, and measure what beam offset you receive. Subtract the stereo angle between the two optics at that distance.2) The first picture looks like a CAD render of two little Cassegrain telescopes. The second pictures looks more like a refractor. Maybe the beam expander is two stages, as getting from 9 microns to 5+ cm is almost 4 orders of magnitude.3) The tracking appears to be a 4-cell receiver. I presume this is fine tracking and coarse tracking uses some other method.4) They are using telecom 1550nm lasers. I've seen experimental systems with 40 channels of 40 Gb/s each, it seems like these would just swap in if they can be packaged into the power/weight numbers needed. The power numbers seem suprisingly high (FPGAs doing some kind of FEC?), I suspect that's the limit here.4b) I'll also note that commercial systems would let you pass some/most DWDM channels from one receiver to another transmitter without ever dropping into the digital domain. Using add/drop muxes like this, you could have each satellite use just four laser links, but have very near speed-of-light (lowest possible latency) channels to as many as 160 other satellites. Why bother routing in digital on every hop? You could probably arrange for most packets to go up, make one routing decision, get routed optically to a destination satellite, back to RF and down. That would minimize control latency.5) The gimbal looks like a commercial gimbal which is wholly unsuited for flat-packing into a StarLink Pizza. But note that they tilt/tip the entire optical train down to a fiber pigtail instead of just a flat mirror. This makes more sense to me, since high speed tracking is not really necessary. When targeting a satellite in an adjacent orbital plane, the angle needed will sweep through +/- 20 degrees or so every 90 minutes.6) A commercial gimbal is probably going to have ball bearings. These have been problematic on the ISS solar panels (Google Solar Alpha Rotary Joint), and suck for high accuracy pointing because the balls make detents in their races. The better way in high-accuracy optics is flexure joints, which can last forever, don't need lubricants, and have little hysteresis. You'd need compound flexure joints to get +/- 20 degrees for zillions of cycles, but that's possible.Commercial 1550nm SFP+ laser modules are 1.5 watts, and transmit 2 dBm and receive -20 dBm. This means they need to receive 1% of the power transmitted. That won't work for OISL. Here's an off-the-shelf amplifier that will boost output to +37 dBm (5 watts), but that gets spread across all the DWDM channels. Note: the amp burns 170 watts. So if you've got 10 channels, you'd need to keep your loss to 47 dB. There will be a bunch of other losses in the system, so figure you need 43 dB.The new StarLink proposal has 22 satellites in a plane, at 550 km altitude. That's approximately 2000 km between satellites. Two 16 cm apertures passing 1550 nm light would have 43 dBm loss.I noticed that the 25 Gb/s per lane transceivers have roughly the same specs: transmit 2.5 dBm and receive -20 dBm. 15 watt (42 dBm) amplifiers can be had. So probably the upper limit that can be achieved with off the shelf equipment right now is 42 dBm total transmit power spread across 40 channels with 18 cm diameter apertures, giving 1 Tb/s transmit and receive in each of 4 links. It'll take a lot of RF to keep up with that.
Eventually, we’ll expand this to connect people around and between other heavenly bodies.
I mean this news may be interesting - 10 minutes ago I received mail:
5) The gimbal looks like a commercial gimbal which is wholly unsuited for flat-packing into a StarLink Pizza. But note that they tilt/tip the entire optical train down to a fiber pigtail instead of just a flat mirror. This makes more sense to me, since high speed tracking is not really necessary. When targeting a satellite in an adjacent orbital plane, the angle needed will sweep through +/- 20 degrees or so every 90 minutes.6) A commercial gimbal is probably going to have ball bearings. These have been problematic on the ISS solar panels (Google Solar Alpha Rotary Joint), and suck for high accuracy pointing because the balls make detents in their races. The better way in high-accuracy optics is flexure joints, which can last forever, don't need lubricants, and have little hysteresis. You'd need compound flexure joints to get +/- 20 degrees for zillions of cycles, but that's possible.
On October 7th, the FCC made 20 new, independent filings on behalf of SpaceX. They comprise 30,000 new satellites as broken down below:1500 sats at 97.7°, 580 km1500 sats at 85°, 539.7 km1500 sats at 80°, 532 km1500 sats at 75°, 524.7 km1500 sats at 70°, 517.8 km4500 sats at 53°, 498.8 km4500 sats at 40°, 488.4 km4500 sats at 30°, 482.8 km3000 sats at 53°, 345.6 km3000 sats at 40°, 334.4 km3000 sats at 30°, 328.3 kmSource: https://www.itu.int/ITU-R/space/asreceived/Publication/AsReceivedThe filings starting with "USASAT-NGSO-3" all come with a letter from the FCC stating "The operating agency for the network is Space Exploration Technologies Corp."