Hunchback's major (mis)understanding of Apollo TV tech.

[ka9q]

This tells me it is a fairly recent complaint, born of a generation who have supercomputers in their homes and, now, their pockets.

It's amazing what you can do with a limited computer when you know what you're doing. And if you don't, having a supercomputer won't help.

[ka9q]

Hunchbacked's amusingly naive belief that computers always boot slowly actually reveals an interesting qualitative difference between the AGC (and other dedicated computers of that era) and modern computers: their speed/memory ratios. Because they were so small, old computers could read their entire memories much more quickly.

Last Monday I toured Johnson Space Flight Center. The highlight was a visit to the Apollo MOCR in building 30. We got to sit at the consoles, so naturally we all had our pictures taken in Gene Kranz's chair. (I added Sy Liebergot's for good measure.) It was surreal to be so intimately familiar with a place I'd never been before, but as a kid I spent many, many hours watching that room (and its twin) on TV.

I chuckled when one of our guides said that modern laptop computers were "five times more powerful" than the Apollo onboard computer. I worked up some newer numbers. Modern computers are roughly 1,000 times faster than the AGC but have a million times as much RAM and ten million times as much permanent data storage (though it's a little dodgy to compare a core rope with a hard drive).

The interesting thing is how the ratio of speed to memory has changed. A modern computer is a thousand times faster than the AGC but it takes far longer to process all of its memory. It took the AGC only 0.4 seconds to read all of its core rope; it takes my computer well over 7 hours to read one of its 4 TB hard drives at full disk speed.

[Noldi400]

I tink whoever created that image screwed up..

Try this one, where I've swapped the two images over, L-R.

Added - BTW, I suspect they still got the angles a bit wrong - there seems to be over-exaggerated depth - hurts my eyes!  (I do these sort of stereograms myself (on normal terrestrial topics), and if you get the angle/spacing wrong with your originals, they can look quite weird and just not work..

Personally, I learned the technique looking at stereopair X-rays - not that I ever got very good at it. Like someone upthread, it seems to work best for me if I'm kinda tired.  It's as if my eye muscles are fatigued and unfocus more easily.

Yeah. Witness the battle I'm currently having with him on Youtube. He keeps claiming the obvious asymmetry of the LM ascent stage was a "joke" by the engineers to show that it couldn't possibly fly; the RCS thrusters would quickly run out of propellant and crash.

It may be that that touches on the visual comprehension thing again.  He doesn't seem to grasp the difference between the centroid of a shape and the center of mass. In his animations, he also misses a technical detail - during ascent, all forward-firing RCS thrusters were disabled because they didn't want any impulse, however slight, reducing the net thrust.

It's amazing what you can do with a limited computer when you know what you're doing. And if you don't, having a supercomputer won't help.

Yes, it's amazing how well the simple processors that control traffic lights work, even though they can't do anything else.

On a side note, he also scoffs at the notion that "little old ladies" could accurately thread the rope memory modules.  I worked in a textile plant in my younger days and some of the highest paid workers in my department were the "hookers"; women who threaded ends through a thing called a heddle, which controlled the pattern woven into the cloth.

Something like this:



A thousand or more ends and they all had to be right or when the loom started up the resulting product would be unusable. You can believe that the ones who kept their jobs got it right first time, every time.

So this...

...was just another day at the office - probably actually easier than their usual job.

[cjameshuff]

The interesting thing is how the ratio of speed to memory has changed. A modern computer is a thousand times faster than the AGC but it takes far longer to process all of its memory. It took the AGC only 0.4 seconds to read all of its core rope; it takes my computer well over 7 hours to read one of its 4 TB hard drives at full disk speed.

Another factor is that embedded systems often execute directly from storage. Some don't even have the ability to execute from RAM. On startup, they don't need to read or write their memories, they just need to run some code to initialize their working memory. In some cases, the most time consuming task is waiting for the power supply to stabilize.

In the case of the AGC, there was read-only core rope memory and and the writable but still non-volatile core memory. There was no executable code that needed to be loaded into memory on startup, as everything was already there ready to be executed.

[ka9q]

Another factor is that embedded systems often execute directly from storage. Some don't even have the ability to execute from RAM.

Right, that's why I said it was a little dodgy to compare core rope with a hard drive, as the AGC could directly execute out of the former. They also differ in other ways besides size.

Even flash memory is more like a hard drive than core rope, as you can't execute directly out of it either (not the NAND type, anyway). You have to copy it into RAM and execute it from there, and to do that you need some true ROM that you can execute from directly. General purpose computers don't use it for much more than the initial bootstrap, but as you say embedded systems use it much more.

It's fun to think that the first microcomputer I built circa 1976 with the Intel 8080 was very roughly comparable to the AGC in speed and memory, but the cost was already many orders of magnitude lower. Of course it didn't meet the same reliability specs...

[ka9q]

In his animations, he also misses a technical detail - during ascent, all forward-firing RCS thrusters were disabled because they didn't want any impulse, however slight, reducing the net thrust.

Did you mean disabling the upward firing RCS thrusters from operating directly against the main engine? Still, I suppose that to most efficiently maintain attitude during both ascent and descent it would be even better to use only the four downward firing thrusters, one or two at a time, to provide roll and pitch corrections. The net effect would be to move the main engine laterally a little depending on the duty cycles of the four thrusters. You might still have to fire a lateral thruster or two if you needed to make a yaw correction, but you wouldn't have to fire any upward thrusters.

On a side note, he also scoffs at the notion that "little old ladies" could accurately thread the rope memory modules.

Yup. It's a good example of his many arguments from incredulity. This is a very popular denier tactic that I think derives from simple hubris, as well as a lack of imagination. He's the smartest person on the planet (why, he has a degree in aeronautics!) so if even he can't figure out how to do it, why, no one else possibly could either!

[cjameshuff]

Even flash memory is more like a hard drive than core rope, as you can't execute directly out of it either (not the NAND type, anyway). You have to copy it into RAM and execute it from there, and to do that you need some true ROM that you can execute from directly. General purpose computers don't use it for much more than the initial bootstrap, but as you say embedded systems use it much more.

The smaller embedded processors do execute directly from NAND flash. Some of the fancier ones have controllers that read and buffer multiple instructions at once, but this is still all in the flash controller hardware, not in RAM. On many of these, instructions are read via a physically separate bus that's dedicated to the flash, and it's actually impossible to execute code from RAM...Harvard architecture as opposed to von Neumann architecture.

It's only when you get to relatively large and fast systems that you need to load code into RAM before execution.

[ka9q]

Ah, I see. I know NAND can't be randomly accessed a word at a time, so you need something between it and a CPU.

Modern CPUs heavily decouple their RAM operations through their caches, with I/O being done a line at a time, so it makes sense that something similar would be used with flash.

[Noldi400]

In his animations, he also misses a technical detail - during ascent, all forward-firing RCS thrusters were disabled because they didn't want any impulse, however slight, reducing the net thrust.

Did you mean disabling the upward firing RCS thrusters from operating directly against the main engine?

I should know better than to speak English around engineers. Yes, it was the +X thrusters that were disabled. I don't remember at this moment where I read that, but I'm sure it was either a NASA document or an account by one of the astronauts.

Hunchy also gets hung up on the two-second guidance cycle which was designed to (as I understand it) allow any errors to get large enough to be correctable; he uses that number to claim that the AGC's memory cycle time was two seconds (it was actually 11.7 milliseconds), which would have been hopelessly slow. I still can't decide whether he's a <insert mentally-challenged epithet of your choice>  or deliberately deceptive.  It's just hard for me to accept that someone who seems to spend as much time as he does poring over NASA documents in search of "jokes" has no comprehension of what he's reading.

Side Comment: BTW, Hal Laning really should have been given a Nobel Prize in Engineering (if there were such a thing)  for the software he basically made up from scratch to largely error-proof the AGC.

[cjameshuff]

Ah, I see. I know NAND can't be randomly accessed a word at a time, so you need something between it and a CPU.

You can read as little as you like, there's nothing stopping you from taking just a subset of the bits available. This doesn't work for erase and write operations, though. (But those are relatively rare and unusual operations for these chips.)

The chip on this board is the smallest I've encountered with something resembling an actual cache:
http://www.st.com/web/en/catalog/tools/FM116/SC959/SS1532/PF252419

A full 1 MB of flash, and even has a FPU. 168 MHz, much too fast to read each instruction individually, so it's got a fancy "ART Accelerator", an instruction prefetch queue and branch cache.

Modern CPUs heavily decouple their RAM operations through their caches, with I/O being done a line at a time, so it makes sense that something similar would be used with flash.

Interestingly, that chip does the opposite. The RAM is divided into two contiguous but independent 112 KB and 16 KB blocks on the system bus and a separate 64 KB block coupled directly to the core. Peripherals can't access the last one for DMA, but the core doesn't have to compete with peripherals for access...you could have DMA transfers going on with both main blocks while the core's working on something else. (there's also a 4 KB battery-backed RAM block tucked away with the RTC registers)

For comparison: http://www.ti.com/ww/en/launchpad/tiva_c_head.html
At 40 MHz or below, it just reads each instruction from flash as needed, and uses a simple prefetch queue at higher speeds.

[ka9q]

I should know better than to speak English around engineers. Yes, it was the +X thrusters that were disabled.

Those would have been the -X thrusters, then. The nozzles pointed up along the +X axis, but the thrust they produced was in the -X direction.

Hunchy also gets hung up on the two-second guidance cycle which was designed to (as I understand it) allow any errors to get large enough to be correctable; he uses that number to claim that the AGC's memory cycle time was two seconds (it was actually 11.7 milliseconds), which would have been hopelessly slow.

I've noticed that too. I think the 2 second SERVICER cycle had to do only with numerical integrator that updated the state vector (estimated position and velocity). The accelerometers themselves had built-in integration so on each 2 second cycle you'd accurately account for the total impulse experienced over the previous two seconds even if it fluctuated during that time. But I'll have to check this.

The response time of the RCS to attitude errors was clearly faster than 2 sec because we can see it in the 16mm movies of the ascents.

I think the memory cycle time was 11.7 microseconds, not milliseconds.

I still can't decide whether he's a <insert mentally-challenged epithet of your choice>  or deliberately deceptive.  It's just hard for me to accept that someone who seems to spend as much time as he does poring over NASA documents in search of "jokes" has no comprehension of what he's reading.

Me too. Most of the time I think he's sincere, but occasionally he really stretches my willingness to follow Hanlon's Razor (never attribute to malice that which can be attributed to stupidity). Like right now, in his claims about the asymmetry of the ascent stage being a deliberate joke by the engineers. He seems to be going well out of his way to misunderstand my very simple explanation: that the ascent fuel and oxidizer had very different weights, forcing the tanks to be mounted asymmetrically to bring the stage back into balance.

[Noldi400]

I should know better than to speak English around engineers. Yes, it was the +X thrusters that were disabled.

Those would have been the -X thrusters, then. The nozzles pointed up along the +X axis, but the thrust they produced was in the -X direction.

Picky, picky, picky.  I meant the thrusters whose nozzles pointed in the +X direction witch would, yes, have produced thrust in the -X direction.

I think the memory cycle time was 11.7 microseconds, not milliseconds.

Right, microseconds. My fingers stuttered.

Me too. Most of the time I think he's sincere, but occasionally he really stretches my willingness to follow Hanlon's Razor (never attribute to malice that which can be attributed to stupidity). Like right now, in his claims about the asymmetry of the ascent stage being a deliberate joke by the engineers. He seems to be going well out of his way to misunderstand my very simple explanation: that the ascent fuel and oxidizer had very different weights, forcing the tanks to be mounted asymmetrically to bring the stage back into balance.

Hunchy reminds me of those religious fanatics who take a single line of scripture and build up a complex theological principle from it. (My father used to call it "Making a lot of stew from one oyster".)  He has fixated on the phrase from David Hoag's GNC Progress Report (E-2411):

"...the system estimates the torque arising from the offset of the main engine thrust from the center of gravity."

along with the visual asymmetry of the ascent stage and decided that the CoG and thrust axis were deliberately misaligned.  He totally disregards the fact that the CoG is a bit of a moving target, being affected by equipment stowage, crew movement, propellant usage, etc.

His whole argument seems to be that the design (that he imagines) would have been wasteful of RCS fuel and would have carried the risk of running out of propellant, leaving none for maneuvering for rendezvous. Even IF it were designed that way, don't you think they would have allowed for that in their fuel calculations?

As you say, he does seem sincere. But knowledgeable he ain't.

[JayUtah]

I've noticed that too. I think the 2 second SERVICER cycle had to do only with numerical integrator that updated the state vector (estimated position and velocity). The accelerometers themselves had built-in integration so on each 2 second cycle you'd accurately account for the total impulse experienced over the previous two seconds even if it fluctuated during that time. But I'll have to check this.

Correct; SERVICER performed dead reckoning at 0.5 Hz.  In accelerated flight the guidance loop operated at 10 Hz.

Me too.

And me.  It's still difficult for me, after more than a decade of this debate, to see how anyone who professes to be a qualified professional can be so devastatedly obtuse and not have that be a deliberate behavior.  Yes, I'm ignoring the possibility of mental illness, but that's deliberate.  I don't know of any specific illness with these symptoms, and I have no evidence for that cause.

[JayUtah]

along with the visual asymmetry of the ascent stage and decided that the CoG and thrust axis were deliberately misaligned.  He totally disregards the fact that the CoG is a bit of a moving target, being affected by equipment stowage, crew movement, propellant usage, etc.

Yes, he totally disregards the entire science of spacecraft dynamic control.   Where to begin...

First of all, the conceptual control axes for the LM were not orthogonal, as was the case in the other spacecraft.  This bias was intended to correct for the fact that the LM flew "forward" for a great deal of the landing.  And nowadays we are perfectly happy with non-orthogonal control axes; we deliberately do not align reaction wheels along cardinal axes or in orthogonal planes precisely so that we can use a subset of them to maintain control after the failure of a wheel.

Second, no RCS design or control law, or indeed any accelerated-flight control law, requires the spacecraft to have "aligned" mass properties.  Such a thing is impossible to achieve in practical design, and further impossible (if that concept means anything) to maintain in normal operation or powered flight.  No method of flying a spacecraft has ever relied on delicately balancing the mass properties so as to require no imposed pointing control.  Even the basic skyrockets from thousands of years ago required arranging the center of (aerodynamic) pressure to provide a corrective moment.

Third, the RCS system provides control moments.  The jets do not fire continuously during powered flight.  The spacecraft, as all spacecraft do, wanders around its deadband until a control moment is required.

Fourth, on the ascent the RCS was operated in pulse mode because it was too powerful for the largely empty spacecraft.  It was sized to provide adequate control moments for the docked and fully-fueled spacecraft.

[dwight]

How does Hunchy, with all his claimed brilliance, miss these small basic things?