SpaceX F9 : Bangabandhu-1 : KSC 39A : May 11, 2018 : Discussion

[RoboGoofers]

San Diego had a famous incident in 2012 where the pyro vendor for 4th of July ran the wrong program at the start of the show. Instead of the synced-to-music program, they ran a wiring continuity check. The diagnostic program was written with minimum timing delays, and should have been run only with shorting plugs at the far ends of the wiring runs instead of the actual mortar igniters. 7000 shells, 4 barges, 30 seconds.

Lesson: Make sure the pre-check routine follows the Hippocratic Oath and does no harm.

and here's what that looked like-


yes it's OT but how can you have that description without a video!

[Hitech]

Did anyone else notice the overall poor quality of on board video? Lots of "jello" and the IR spectrum was excessive. and lots of camera movement on reentry.

[gongora]

Did anyone else notice the overall poor quality of on board video? Lots of "jello" and the IR spectrum was excessive. and lots of camera movement on reentry.

If you look back in this thread I think you'll find other people did notice that.

[Hitech]

Thanks I had missed those. It reminds me of the early days of flying Quad Copters. before the "good" cameras came out. I sure hope that was an anomaly and not the "new standard"

[litton4]

Consequence of move from CCD to CMOS sensors. Get used to it. (probably)

[JamesH65]

Consequence of move from CCD to CMOS sensors. Get used to it. (probably)

Note that CMOS sensors use CCD's (which are really just arrays of diodes anyway IIRC), but generally include silicon on the die to get the data off rather than a separate chip.

The effect being seen is caused by using rolling shutters, and there are CMOS camera available that do not use the rolling shutter and will produce very good images - global shutter cameras. There are also cameras with fast read out that suffer less from rolling shutter - the faster it's done, the less the effect. Not sure why SpaceX have moved to a less capable system, seems odd. Perhaps they just don't need high quality video any more?

[kevin-rf]

Consequence of move from CCD to CMOS sensors. Get used to it. (probably)

CMOS with rolling shutter vs. CMOS with global shutter. Rolling shutter CMOS is cheaper... At least that's what my camera vendor claims to justify his premium.

[edzieba]

Consequence of move from CCD to CMOS sensors. Get used to it. (probably)

Note that CMOS sensors use CCD's (which are really just arrays of diodes anyway IIRC), but generally include silicon on the die to get the data off rather than a separate chip.

The effect being seen is caused by using rolling shutters, and there are CMOS camera available that do not use the rolling shutter and will produce very good images - global shutter cameras. There are also cameras with fast read out that suffer less from rolling shutter - the faster it's done, the less the effect. Not sure why SpaceX have moved to a less capable system, seems odd. Perhaps they just don't need high quality video any more?

Ehhh... It's a bit of an artefact of using "CMOS" as a colloquial term. Broadly:
CCD sensors (Charged Couple Devices) are 'photon buckets', capacitor arrays that get charged up as photons hit them. Voltage bars are then use to 'jump these charge buckets down the sensor row-by-row, then feed out that last row bucket-by-bucket into a single amplifier and digitiser.
CMOS sensors have an amplifier and digitiser for each pixel, next to the pixel on the sensor.
Because a CCD has only the voltage bars on the sensor itself, most of the surface area is dedicated to capturing photons. For CMOS, the actual photon capture area is squished in along with all the other circuitry, so the effective sensitivity is lower.
Either type of sensor can exhibit 'rolling shutter' (really rolling scan-out as often there will not be a shutter at all), CMOS just makes it more obvious then CCD because the rolling readout of a CCD occurs over a much shorter time (e.g. 95% of a frame spent capturing photons all at the same time, 5% spent shifting charges and reading out where the 'wrong' charge can build up in cells as they are moved about) while a CMOS sensor is limited by output bandwidth so captures AND reads the sensor a row at a time over the whole frame.

As for why the switch from CCD or global-shutter CMOS to rolling shutter CMOS? Possibly a combination of price for custom cameras (they probably want to re-use them too, so they may want to just stick in a camera they don't need to touch again rather than a cheap GoPro they need to swap out every mission) and savings on mass and power. The days of the video footage being diagnostic are probably over now Block 5 is the 'fixed' configuration and stages are regularly recovered, so not as much mass or money budget needs to be spent on a purely 'pretty pictures' component.

[Lars-J]

The days of the video footage being diagnostic are probably over now Block 5 is the 'fixed' configuration and stages are regularly recovered, so not as much mass or money budget needs to be spent on a purely 'pretty pictures' component.

That makes no sense. We now can reuse boosters, so we will use cheaper and worse components?  No.

They might have other reasons for the switch, but that would not be it. Because as confidence in recovery has grown, they have started adding more expensive parts that they are confident they can get back. The titanium grid fins is a perfect example.

[edzieba]

The days of the video footage being diagnostic are probably over now Block 5 is the 'fixed' configuration and stages are regularly recovered, so not as much mass or money budget needs to be spent on a purely 'pretty pictures' component.

That makes no sense. We now can reuse boosters, so we will use cheaper and worse components?  No.

That's not what I said. If you can put on your existing COTS cameras but need to replace them after every launch because they can't handle the thermal cycling, they are 'worse' than custom cameras that have a lower picture quality but can be left in place for multiple re-uses. Extra effort put into higher picture quality would not provide value beyond shinier PR pictures, but effort put into longevity means another barrier to rapid re-use is removed.

[EeeVee3]

/unlurk  . Ive been using both cmos and ccd cameras on my UAV's for years for 1st person view flying... to me.. ccd cams have much better WDR (wide dynamic range) handling. CMOS have usually none. Though some in recent years have implemented a software solution which is ... mmm ok ish.
WDR brightens darker parts of the view and dulls off brighter parts of the scene. Space is VERY BLACK or very Bright so WDR is a GIVEN. IR filters which (dont seem to be fitted on the banga flight) produce greater better colours.. (SX may have done this on purpose looking for IR hot spots, on the S2 nozzle I can understand, but on the S1 ?  )

/back to lurking 

[kevin-rf]

Either type of sensor can exhibit 'rolling shutter' (really rolling scan-out as often there will not be a shutter at all), CMOS just makes it more obvious then CCD because the rolling readout of a CCD occurs over a much shorter time (e.g. 95% of a frame spent capturing photons all at the same time, 5% spent shifting charges and reading out where the 'wrong' charge can build up in cells as they are moved about) while a CMOS sensor is limited by output bandwidth so captures AND reads the sensor a row at a time over the whole frame.

A small nit on that, Global shutters on CCD's are usually done with either interline or frame transfer masks, not rapid readout of the sensor. The charge for the entire chip is transferred under a light blocking mask and then read out.  With an interline mask every other row is a mask and the charge for the whole frame gets shifted sideways one pixel then readout. This gives you a true snapshot and global shutter. Classic example is the Sony ICX-285AL, heavily used in the scientific and microscopy imaging (Though I am really loving new discontinued Sony ICX-825). Frame transfer involves one side of the sensor being under the mask and the charge being quickly transferred sideways under it. CMOS will work in a similar manner with a mask pattern that the charge is quickly transferred under. Not sure of the pattern Sony is using with it's new CMOS, but I believe they are burying the mask in the gate structure under the pixel.

Now the reason global shutter sensors cost more is it requires twice the silicon area of a rolling shutter (and with interline you also have micro lenses above each non masked row to increase the fill factor from 50% to closer to 100%).

We know return you to the usual discussions of SpaceX, Rockets, and how many Falcons can dance on the head of a barge.

*btw The writing is on the wall for CCD sensors, Sony (largest manufacture) has announced it will switch to CMOS over the next 5 years. Actually, Sony polled it's vendors, made a final prediction and built all the CCD's it will ever make back in 2016, it then stuck all in a wharehouse promising to have enough inventory to last through 2025. In 2016 it then decommissioned the CCD line (The floor space has been replaced with a new CMOS line). But the good news is my suppliers have promises from Sony that I will have the CCD's I use until 2025. For CCD's this just really leaves E2V, Hammatsu, On Semi (bought Kodak's line). CMOS, it is the future!!!