I don’t think that’s quite true.

There are actually some interesting patents related to the star sensor in the devices, particularly related sky background gradients. Northrop’s patents are more interesting than Lockheed’s, IMHO. In all cases the patents mention using an IR-pass filter to improve contrast.

The sensor is essentially an analog lock-in/synchronous detector with a rotating shutter and a wedge prism that nutates the star field about the boresight. the patented component in most cases is the shutter (there are different patterns, and Northrop came up with some clever designs).

You end up with a frequency modulated signal from the photomultiplier tube (carrier is from the shutter, modulation frequency the difference between the prism and shutter).

By measuring the phase and magnitude of the modulated signal, you can steer the telescope onto the star, and the “coding gain” from the lock-in is substantial.

Edit, previous comments:

https://news.ycombinator.com/item?id=27084261

https://news.ycombinator.com/item?id=23238437

I have recently been bitten by the modular synthesis bug, and it is wild to me how these terms seem to have correspondence even though I am learning it from the audio domain. I guess a signal is a signal and that is the power of analog.

For context: learning about carriers, modulation frequency in relation to audio synthesis [acknowledging there is more than one type of 'FM' in relation to audio but nonetheless similar]).

> when you're at 85,000 feet (as noted, the SR-71's cruising altitude), the concept of a blue "daylight" that blocks the view of space only exists below you, not above you

The Astro-Nav System didn't work only at cruising altitude, it tracked stars even when the SR-71 was on the ground. [1]

[1] https://airandspace.si.edu/webimages/collections/full/NAS-14...

> The Astro-Nav System didn't work only at cruising altitude, it tracked stars even when the SR-71 was on the ground. [1]

>

> [1] https://airandspace.si.edu/webimages/collections/full/NAS-14...

I don't necessarily doubt you, but linking to a 76-page PDF without any indication of where the reader is supposed to find support for your claims is not nearly as convincing as you think it is.

> I don't necessarily doubt you, but linking to a 76-page PDF without any indication of where the reader is supposed to find support for your claims is not nearly as convincing as you think it is.

It's funny you say that, because based on almost all the assertions you read here on HN which are taken for granted without any links to support them, you'd think linking to an authoritative PDF would be more convincing than if I hadn't linked to it. But based on your comment it seems like it works the opposite way (unless I point to the exact place supporting my assertion, presumably).

But sure, fine, I'll take your bait.

I only skimmed the document, so I can't point you to the best place where to look for it. But I can point you to page 31, paragraph 10A-69 which says the following: "When alignment is performed in a hanger, there is a possibility that the system will track false stars (ceiling lights, etc). To prevent this from happening, the INERTIAL ONLY mode is selected and enabled after completion of an alignment. After the airplane taxies into the open, ASTRO INERTIAL mode is selected and enabled". I'm pretty sure this supports my assertion.

> It's funny you say that, because based on almost all the assertions you read here on HN which are taken for granted without any links to support them, you'd think linking to an authoritative PDF would be more convincing than if I hadn't linked to it. But based on your comment it seems like it works the opposite way (unless I point to the exact place supporting my assertion, presumably).

Not at all, your comment was strictly better for including that link, thanks for including it. It was neat to briefly skim the PDF, even if I don't have time to read the 76 pages of technical documentation.

> I only skimmed the document, so I can't point you to the best place where to look for it. But I can point you to page 31, paragraph 10A-69 which says the following: "When alignment is performed in a hanger, there is a possibility that the system will track false stars (ceiling lights, etc). To prevent this from happening, the INERTIAL ONLY mode is selected and enabled after completion of an alignment. After the airplane taxies into the open, ASTRO INERTIAL mode is selected and enabled". I'm pretty sure this supports my assertion.

Maybe. But that whole paragraph is hard to understand. How were they doing alignment in a hangar to start with, were these roofless hangars?

> How were they doing alignment in a hangar to start with, were these roofless hangars?

I don't think so.

You can read more about alignment in page 7, paragraph 10A-23 and the following pages. It'd usually involve using test equipment and manually inputting the current coordinates (e.g. heading, position, altitude, date/time, ...) into the system, from what I understand.

Regardless, the paragraph I mentioned in my previous comment clearly instructs to enable ASTRO mode after the alignment is completed and the airplane is taxiing in the open. It also mentions that in ASTRO mode, the system can confuse the ceiling lights of the hangar for stars (hence why it needs to be disabled inside the hangar), which means that ASTRO mode does track stars on the ground.

I wonder, however, whether it could track stars from the ground during the day. Might well only have worked at night.

The first paragraph of page 40 says: "Stars are normally tracked before takeoff, both at night and during the day (...)".

Now that's a money quote. The following sentence, which says that "If star tracking has not commenced before takeoff, it should start at an altitude where cloud cover and sky brightness conditions have improved.", leads one to believe that it was possible for the sky to be too bright to track stars from the ground sometimes (since it basically says exactly that), but it now seems clear ground tracking during the day was possible. Maybe it depended on the latitude, time of day, and position of the Sun relative to the stars (low latitude in the middle of the day without sufficiently bright stars far enough away from the Sun = sky too bright, higher latitude early or late in the day with sufficiently bright stars far enough away from the Sun = track from the ground, any other combination = results may vary). Actually, I just saw that the bottom of page 42 describes how to handle situations where the sky was too bright for specific stars.

The whole document supports the claim, it details ground astroinertial startup which would not occur if it could not see stars from the ground. The fact they made this document available is very useful to those who are actually invested in the subject.

They also need an initial fix before take off and that could be day or night depending on the mission.

I think that in order to calculate your latitude and longitude position using what amount to high-tech sextant readings you have to know your local down vector to a high degree of accuracy. That is, the system judges the position of the Blackbird by comparing its local down vector with the angles to the star sightings. If the down vector measurement is off by an arcsecond, the measured latitude and longitude position on the earth will also be off by an arcsecond (at sea level, about 600 feet).

It seems to me that there's an interesting problem that the Blackbird goes so fast that its down vector changes dramatically over the course of a typical flight, so I'm wondering how it accurately updates its down vector reference. Jets can go through arbitrary and wobbly 3-d trajectories, accelerating in any direction, so it can't be a simple measurement of gravity. It would have to be a accelerometer combined with inertial and gyro and elevation readings all summed up.

What does 'down vector' mean in this context?

A very precise estimation of which direction gravity is pulling down towards. Which is almost, but not exactly, the vector towards the center of the earth. And it also needs to be disentangled from the suborbital centrifugal/centripetal forces experienced by the movement of the Blackbird.

Can it just look at the front and back horizon to estimate the local down vector?

If you can measure the front and back horizon accurately you can measure the left and right horizons too, and that will give you an accurate down vector.

But I don't think it's possible to measure a horizon that accurately. There are temperature and pressure dependent atmospheric effects at low angles that effect how light refracts and curves through the atmosphere, and these effects will have to be compensated for to measure the angle accurately. Additionally, there are mountains and valleys to compensate for.

if you use a SWIR camera stars are pretty visible in the daytime even at ground level. Scattering is related to a fourth power of frequency so by the time you're down to IR the amount of scattered light is pretty low.

(which is also why the sky is blue-- or really UV, if we could see it)

The software for plate solving is pretty good. I've had some fun taking movie screenshots and figuring out if the stars are real, and if they're the right hemisphere. (can't be used for location without the time and orientation of the camera)