Processing power usage is measured in Watts/Hz (or for modern embedded processors MHz/microwatt). Adding a processor performing the same task doubles your power usage. On the other hand Moore's law roughly doubles the processing per Watt every 1.5 years. You either maintain the same power and double the processing (common on desktop systems) or for embedded systems you maintain the same processing and half the power. Curiousity is out roughly 10 years on Moore's law which is 6.7 doublings or a factor of 100 times.All other factors equal you can add 100 modernly equivalent processors and have the same power usage. In reality you only need something like SpaceX's setup with 3 sets of computers each having 2 processors voting against each other.(I made a few simplifications, Moore's law more refers to the density of transistors, but the parallels are quite similar.) For reference, the CPU Curiosity is using is in the processing level as the old blue iMac computers of 10 years ago.
Table 9 A few of the resources that drive science operations and the required minimum capabilities built into the design of MSLSome resources that drive science operations Required capabilityEnergy available for science activities250 Wh/solDownlink volume (two UHF passes)250 megabits/solRover awake time6 h/solTraverse distance50 m/solNote: The actual capability on Mars may exceed the requirements for certain environmental conditions or other favorable conditions. The requirements may not be met during anomalous conditions. Science activities include rover traverses. A small fraction of the downlink volume will be used to transmit rover health and housekeeping data
Quote from: mlindner on 03/27/2013 02:39 amProcessing power usage is measured in Watts/Hz (or for modern embedded processors MHz/microwatt). Adding a processor performing the same task doubles your power usage. On the other hand Moore's law roughly doubles the processing per Watt every 1.5 years. You either maintain the same power and double the processing (common on desktop systems) or for embedded systems you maintain the same processing and half the power. Curiousity is out roughly 10 years on Moore's law which is 6.7 doublings or a factor of 100 times.All other factors equal you can add 100 modernly equivalent processors and have the same power usage. In reality you only need something like SpaceX's setup with 3 sets of computers each having 2 processors voting against each other.(I made a few simplifications, Moore's law more refers to the density of transistors, but the parallels are quite similar.) For reference, the CPU Curiosity is using is in the processing level as the old blue iMac computers of 10 years ago.All other factors are not equal; Watts/Hz is not necessarily the figure of merit, especially when there is a hard upper bound on Watts (among other things). As you say, "Moores law" says nothing about Watts/Hz, only transistor density and they are *not* "quite similar" (otherwise Intel's horrid and abandoned NetBust architecture would rule today).And SpaceX's approach--or any other LEO architecture--is not comparable; "in reality" the conditions are far different and what works in one environment does not necessarily mean it will work in another. Such "voting logic" is not new, and there are reasons why such a simplistic approach alone is not used for high value deep space missions.Now can we get off this tangent please?
Quote from: Dalhousie on 03/27/2013 01:29 amThere are lots of reasons why the rover is slower that terrestrial autonomous vehicles.Power for the wheels is one. Limited number of CPUs because of power issues is another. Curosity has a few hundred watts available, a terrestrial vehicle can have a kilowatt or more of electrical power available.CPU performance inherent in having a Mars-capable unit. In particular there is radation resistance to consider. Then there is the navigation question. Primary route finding for terrestrial autonomous vehicles is GPS, which of course is not available for Mars. Then there are issues of the terrain roughless and potential hazards, which would slow the rover down.Two cents here. Power is only a problem for calculation because radiation is a problem. They're using decade old (or older) computer electronics because of the radiation protection which makes them lose out a lot on Moore's law. They also redesign the radiation proofed CPUs to use extra computation checking for radiation effects. All in all, if several modern non-radiation hardened CPUs were used in a voting scheme, you'd have much faster performance at lower power, while maintaining the radiation resistance.
There are lots of reasons why the rover is slower that terrestrial autonomous vehicles.Power for the wheels is one. Limited number of CPUs because of power issues is another. Curosity has a few hundred watts available, a terrestrial vehicle can have a kilowatt or more of electrical power available.CPU performance inherent in having a Mars-capable unit. In particular there is radation resistance to consider. Then there is the navigation question. Primary route finding for terrestrial autonomous vehicles is GPS, which of course is not available for Mars. Then there are issues of the terrain roughless and potential hazards, which would slow the rover down.
From "Mars Science Laboratory Mission and Science Investigation" http://link.springer.com/article/10.1007%2Fs11214-012-9892-2QuoteEnergy available for science activities250 Wh/solTraverse distance50 m/solScience activities include rover traverses
Energy available for science activities250 Wh/solTraverse distance50 m/solScience activities include rover traverses
If only 250 Wh/sol is available for roving then I find energy being limiting quite plausible. That leaves another question however. The total energy available per sol is about (110 W) * (24.6 hours) = ~2700 Whr. Where does the other 2000+ Whr/sol go?
After descent, the rover only receives the approximately 110W (BOL) ... MMRTG power is used to charge the rover’s two Li-ion batteries, which are the primary source of electrical power for operations. Rover power utilization ranges from 45—70W while sleeping, to approximately 150W minimum when awake, up to greater than 500W peak while “roving” or moving.
Watts/Hz is a fine figure of merit because computation can be stopped or even duty cycled to reach any power level needed even with very fast power hungry processors.
Following the above power conversation:Quote from: mlindner on 03/27/2013 03:52 amWatts/Hz is a fine figure of merit because computation can be stopped or even duty cycled to reach any power level needed even with very fast power hungry processors. How can Watts/Hz be a fine figure of merit? When the processors are in energy saving mode, their functionality is always reduced.
According tohttp://trs-new.jpl.nasa.gov/dspace/bitstream/2014/38066/1/05-0708.pdfthe heat removal system removes 150 W of heat from the avionics. This heat presumably comes from electricity usage, so this must be where the 150 W minimum waking power usage goes. Why the avionics use 150 W is beyond me however.
Today's takeaway is that for the moment, there's no "activity", other than testing and validation of systems, until Mars pops out from the other side of the Sun?
Will the rover be stationary all that time?
The rover will move!!!2.75m I'm bored, I want to see pictures of the landscape as they trundle along.Perhaps this rover has too much instrumentation? The scientists have gone mad with glee and want to analyse everything in the immediate vicinity in great detail.That is their right but it's not very exciting for those of us that want some cool images for our desktop.