Hi Guillaume, good to hear from you! (Slashdotters - do yourself a favour and visit Guillaume's website and have a look at some of his amazing photos). We aren't currently running anything at Dome C. Dome A is likely to have similar seeing to Dome C above the boundary layer, but the layer is expected to be lower, possibly touching the ice. That is one of PLATO's prime goals - to measure the height of the boundary layer with a sonic radar.
The Chinese took an Australian Antarctic Division AWS to Dome A in 2005.
Yes, the reliability of our equipment continues to improve. It is now even better than the stuff we took to Dome C!
We mounted everything on shock absorbers to survive the 1200 km sled trip. There was no damage.
The traverse team should arrive back in Zhongshan station today.
Has your team considered puting the hard drives in a pressure vessel of some sort?
Yes, we have thought about this a lot, and have never had the time to complete the design! It is the best solution, and we should be doing it. It is much easier nowadays that IDE interfaces are going away and serial connections mean that fewer cables need to leave the pressure vessel.
If we use a USB interface, I'm a bit nervous about the reliability of Linux USB storage, or perhaps it is the controllers that interface the drive with USB. I've had many examples of filesystem corruption with external USB drives. And USB flash disks seem to have problems too. Quite often during boot a drive will give all sorts of error messages and will require power cycling to fix it. Googling for these problems show that they are common, but with no solutions that I have found.
Our particular PC/104 computer has both USB 1 and 2 interfaces, but we can only boot reliably off USB 1, and we see intermittent failures if we use both USB 1 and 2. The flakiness of USB for storage is a major frustration.
SATA wasn't an option on our computer. These are low-power embedded systems, so they tend to lag a bit with some of the newer interfaces.
For your interest, here is some information on how PLATO got to Dome A.
The PLATO modules were built at the University of New South Wales in Sydney, Australia. Instruments were provided by our collaborators at a number of universities in China, the US, and the UK.
In late November 2007 PLATO was trucked 3912 km to Perth, where it joined a Chinese icebreaker for a two week trip to Zhongshan station on the edge of Antarctica. A helicopter then lifted the modules off the ship and about 100km inland where they joined a traverse for the ~1200 km journey to Dome A.
The traverse was an amazing feat. 17 people, 5 tractors. PLATO itself weighted about 10 tonnes. The traverse moves at speeds of 5-10 km per hour each day for 10 hours, and then rested for 14 hours. After three weeks of this, they arrive at Dome A. I am told that the undulating motion of the tractors over the ice can give you "sled sickness", an unpleasant variety of seasickness.
The team spent 10 days at Dome A, and did a fantastic job of installing the experiments and getting everything working. The temperatures were around -30C, which isn't much of an issue at low wind speeds. The altitude (4090m) is more of a problem, as it makes physical work exhausting, and there are difficulties with sleeping, mental acuity, etc.
The engines are Hatz 1B30, we use two different generators: four are made by eCycle, and two by Mavilor. Each puts out about 1kW at 120VDC.
To start the engines we have two banks of Ultracapacitors. These are amazing devices, 3000 Farads each, charged to 2V, with 12 in each bank arranged to give 12VDC. They can turn over the engines very quickly. We haven't had to crank an engine for more than 2 seconds yet, although we haven't dropped the engine temperatures below 0C.
We tested the system in a pressure tank at UNSW to simulate the roughly 0.5atm pressure. The engines still work well at this altitude.
With two modems going flat out we could theoretically transfer 40MB per day. In practice the link isn't all that reliable and we would be lucky to achieve half of that. Still, it is enough to control the experiments and return reduced data to verify that everything is working. All of the health and status information fits into 12KB per day.
The bulk of the data will be physically returned by the next Chinese traverse team, this time next year.
The decision to use "jet fuel", specifically Antarctic grade kerosene, was made by the Australian team after much consideration of all reasonable alternatives. Environmental issues were foremost in our minds. PLATO produces a microscopic addition to the kerosene usage in Antarctica. We are using efficient diesel generators, and have over 200mm of additional insulation lining both modules of PLATO. Every 15 watts of heat we put in raises the internal temperature by 1 degree C.
We have 1kW of solar panels, which provide most of the power during summer. However, when the sun is down, and with the very low windspeeds at Dome A, the choices become limited.
We will eagerly embrace hydrogen fuel cells when they become practical. However, they are not there yet.
Actually, Dome A is one of the least windy places on Earth, typically just 2-3 metres per second. Dome A is the highest point in the centre of the Antarctic plateau, and this is where the katabatic winds start from. The winds accelerate as they head towards the coast, and that is where they can reach 100's of kph.
As one of the University of New South Wales people involved, I thought slashdot might like some information on the computer systems that PLATO uses.
PLATO uses two redundant PC/104 form factor computers running Debian Etch. The computers boot from a 4GB flash disk (we tested 5 different models in the lab, and found one that worked reliably to -60C, despite only being spec'ed to -25C; all the other models worked to -40C, but had problems below that).
We use a readonly filesystem, with/home,/etc, and/var being created on boot in a ramdisk. This works really well, and it is nice to be able to turn off the power at any time without being concerned about filesystem corruption. Needless to say, with no possibility of any human being on-site for the rest of the year, we have thought very carefully about reliability.
Bulk data storage is provided by terabytes of conventional disks, with the most precious data being backed up on ~64GB of USB flash disks. Conventional disks don't handle the altitude very well, so we don't like to rely on them.
Communication is via two Iridium satellite modems, running at 2400 baud. We can push software updates by sending a set of "Short Burst Data" messages of up to 2000 bytes at a time. We can also login to PLATO using ssh, and I'm logged in as I'm typing this and running experiments.
There is a CAN (Controller Area Network) bus running throughout PLATO and linking the two modules: the Instrument Module, and the Engine Module, 45m apart. Each of 11 nodes on the bus has a small Atmel board that can turn power on/off to experiments, digital and analog I/O, etc.
Well, you have to start somewhere! Remember that Dome A is completely remote. There is no station there, and PLATO is running without human intervention for as long as a year. The amount of fuel we could take in dictated the available power, and that in turn limited the size of telescope we could take in. Still, we have four 14.5cm telescopes, a 1.5m sonic radar, two sky cameras, 4 webcameras, a 15-m tower, and a 450 micron wavelength telescope, several terabytes of disks, a dozen computers, about 64GB of flash storage, two Iridium satellite modems.
I'm one of the 4 UNSW scientists who designed PLATO. We certainly are using the fuel efficiently. When the sun is up we get over 1kW from solar panels, and we run one diesel generator at a time with just enough heat output to stop the fuel from getting too cold and turning to gel. Interestingly, the solar panels are considerably (about 30%) more efficient than you would expect from temperate site measurements - the colder temperatures (-50C at the moment) help, as does sunlight reflected from the snow.
there is still some glaring room for innovation that I don't expect anytime soon under present industry conditions
I agree strongly with that. I have been amazed at the glacial progress in speech recognition in the decade I have been using ViaVoice and Dragon Naturally Speaking. Bugs in Dragon remain uncorrected for years, despite user complaints. As is common with proprietory software, there is no dialog between the users and the anonymous programmers.
The lack of any Linux alternative is also a major problem - Dragon is the only reason I still run Windows. The last time I checked (6 months ago), Dragon didn't work with Wine. With Microsoft owning, AFAIK, a substantial fraction of Nuance, and with Dragon being the only game in town, I see a very bleak future for Linux-based speech recognition.
If only the current generation of elite open-source programmers would realize that unless they get speech recognition working in the next decade, they will be in trouble when they themselves develop repetitive strain injury.
It is easy to think you are invincible when you are young and your fingers can hammer away at lightning speed on the keyboard for 18 hour programming sessions, but, reality hits after 5, 10, perhaps 20 years, when you suddenly find that typing is painful and even clicking a mouse button is difficult.
Antarctica:
Con:
Can only ever view half the sky.
Unusable during summer.
Very expensive.
Poor accesibility: Only during summer, only at great expense.
Hostile environment: extreme cold. Possible build up of ice by sublimation deposition.
Viewing half the sky is not a problem for most astronomical projects, particularly since most of the interesting objects are in the Southern Hemisphere anyway!
In summer when the sun is up you just do IR/submillimeter astronomy.
Antarctica is not as expensive as you might think. The cost to get a telescope to Dome C is only US$3/kg. Overland traverses from the Antarctic coast can easily take hundreds of tonnes of equipment. And rather than driving a truck up to your door, at the South Pole you can land a Hercules aircraft right next to the observatory.
True, the stations are only accessible over summer, but it is not excessively expensive to get there
As for the extreme cold, this is something you have to put up with to have the best site for IR astronomy. Both the sky and the telescope emit in the infrared, so you want to have them as cold as possible to achieve the best performance. Existing observatories might have shirt-sleeve environments, but they have IR background levels that are between 10 and 100 times greater than in Antarctica.
Issues of ice build-up are understood through over a decade of work at the US South Pole station. Sub-millimeter telescopes are working there now, and IR telescopes (SPIREX) have done so in the past.
When the sun is up (summertime) you can observe in the infrared and submillimeter. Hubble's observing efficiency is about 50% due to the requirement to avoid the Earth, the South Atlantic Anomaly, slew time, etc.
The limitation is sky coverage is not important for many astronomical programs. Important regions such as the Galactic Center, the Magellanic Clouds, and the South Galactic Pole, are all visible.
Some electronics operates below its specified minimum operating temperature, and some doesn't. For example, we had some solid-state disks that were rated to -40C, but that failed at -20C. Mostly we have found that PC/104 computers, memory, etc work fine at -60C. M-Systems solid-state disks have been very reliable.
You want to avoid spinning up a hard disk at -85C though! The altitude (4000m equivalent) also tends to be rough on hard disks (both due to the cooling problems and the smaller head-gap), which is why we avoid hard disks in critical applications. Actually, one of our experiments is running on a Dell laptop with a normal 2.5-inch IDE drive and RT Linux. It has worked fine for two years now.
Slashdot Mirror
Hi Guillaume, good to hear from you! (Slashdotters - do yourself a favour and visit Guillaume's website and have a look at some of his amazing photos). We aren't currently running anything at Dome C. Dome A is likely to have similar seeing to Dome C above the boundary layer, but the layer is expected to be lower, possibly touching the ice. That is one of PLATO's prime goals - to measure the height of the boundary layer with a sonic radar.
The Chinese took an Australian Antarctic Division AWS to Dome A in 2005.
Yes, the reliability of our equipment continues to improve. It is now even better than the stuff we took to Dome C!
We mounted everything on shock absorbers to survive the 1200 km sled trip. There was no damage.
The traverse team should arrive back in Zhongshan station today.
Has your team considered puting the hard drives in a pressure vessel of some sort?
Yes, we have thought about this a lot, and have never had the time to complete the design! It is the best solution, and we should be doing it. It is much easier nowadays that IDE interfaces are going away and serial connections mean that fewer cables need to leave the pressure vessel.
If we use a USB interface, I'm a bit nervous about the reliability of Linux USB storage, or perhaps it is the controllers that interface the drive with USB. I've had many examples of filesystem corruption with external USB drives. And USB flash disks seem to have problems too. Quite often during boot a drive will give all sorts of error messages and will require power cycling to fix it. Googling for these problems show that they are common, but with no solutions that I have found.
Our particular PC/104 computer has both USB 1 and 2 interfaces, but we can only boot reliably off USB 1, and we see intermittent failures if we use both USB 1 and 2. The flakiness of USB for storage is a major frustration.
SATA wasn't an option on our computer. These are low-power embedded systems, so they tend to lag a bit with some of the newer interfaces.
For your interest, here is some information on how PLATO got to Dome A.
The PLATO modules were built at the University of New South Wales in Sydney, Australia. Instruments were provided by our collaborators at a number of universities in China, the US, and the UK.
In late November 2007 PLATO was trucked 3912 km to Perth, where it joined a Chinese icebreaker for a two week trip to Zhongshan station on the edge of Antarctica. A helicopter then lifted the modules off the ship and about 100km inland where they joined a traverse for the ~1200 km journey to Dome A.
The traverse was an amazing feat. 17 people, 5 tractors. PLATO itself weighted about 10 tonnes. The traverse moves at speeds of 5-10 km per hour each day for 10 hours, and then rested for 14 hours. After three weeks of this, they arrive at Dome A. I am told that the undulating motion of the tractors over the ice can give you "sled sickness", an unpleasant variety of seasickness.
The team spent 10 days at Dome A, and did a fantastic job of installing the experiments and getting everything working. The temperatures were around -30C, which isn't much of an issue at low wind speeds. The altitude (4090m) is more of a problem, as it makes physical work exhausting, and there are difficulties with sleeping, mental acuity, etc.
Much more information, and a diary of the trip by the Chinese team members, is at http://mcba11.phys.unsw.edu.au/~mcba/plato.
The engines are Hatz 1B30, we use two different generators: four are made by eCycle, and two by Mavilor. Each puts out about 1kW at 120VDC.
To start the engines we have two banks of Ultracapacitors. These are amazing devices, 3000 Farads each, charged to 2V, with 12 in each bank arranged to give 12VDC. They can turn over the engines very quickly. We haven't had to crank an engine for more than 2 seconds yet, although we haven't dropped the engine temperatures below 0C.
We tested the system in a pressure tank at UNSW to simulate the roughly 0.5atm pressure. The engines still work well at this altitude.
With two modems going flat out we could theoretically transfer 40MB per day. In practice the link isn't all that reliable and we would be lucky to achieve half of that. Still, it is enough to control the experiments and return reduced data to verify that everything is working. All of the health and status information fits into 12KB per day.
The bulk of the data will be physically returned by the next Chinese traverse team, this time next year.
The decision to use "jet fuel", specifically Antarctic grade kerosene, was made by the Australian team after much consideration of all reasonable alternatives. Environmental issues were foremost in our minds. PLATO produces a microscopic addition to the kerosene usage in Antarctica. We are using efficient diesel generators, and have over 200mm of additional insulation lining both modules of PLATO. Every 15 watts of heat we put in raises the internal temperature by 1 degree C.
We have 1kW of solar panels, which provide most of the power during summer. However, when the sun is down, and with the very low windspeeds at Dome A, the choices become limited.
We will eagerly embrace hydrogen fuel cells when they become practical. However, they are not there yet.
Actually, Dome A is one of the least windy places on Earth, typically just 2-3 metres per second. Dome A is the highest point in the centre of the Antarctic plateau, and this is where the katabatic winds start from. The winds accelerate as they head towards the coast, and that is where they can reach 100's of kph.
So, unfortunately, wind power was not feasible.
As one of the University of New South Wales people involved, I thought slashdot might like some information on the computer systems that PLATO uses.
PLATO uses two redundant PC/104 form factor computers running Debian Etch. The computers boot from a 4GB flash disk (we tested 5 different models in the lab, and found one that worked reliably to -60C, despite only being spec'ed to -25C; all the other models worked to -40C, but had problems below that).
We use a readonly filesystem, with /home, /etc, and /var being created on boot in a ramdisk. This works really well, and it is nice to be able to turn off the power at any time without being concerned about filesystem corruption. Needless to say, with no possibility of any human being on-site for the rest of the year, we have thought very carefully about reliability.
Bulk data storage is provided by terabytes of conventional disks, with the most precious data being backed up on ~64GB of USB flash disks. Conventional disks don't handle the altitude very well, so we don't like to rely on them.
Communication is via two Iridium satellite modems, running at 2400 baud. We can push software updates by sending a set of "Short Burst Data" messages of up to 2000 bytes at a time. We can also login to PLATO using ssh, and I'm logged in as I'm typing this and running experiments.
There is a CAN (Controller Area Network) bus running throughout PLATO and linking the two modules: the Instrument Module, and the Engine Module, 45m apart. Each of 11 nodes on the bus has a small Atmel board that can turn power on/off to experiments, digital and analog I/O, etc.
More info, photos, and links to the health and status data are at http://mcba11.phys.unsw.edu.au/~mcba/plato
Well, you have to start somewhere! Remember that Dome A is completely remote. There is no station there, and PLATO is running without human intervention for as long as a year. The amount of fuel we could take in dictated the available power, and that in turn limited the size of telescope we could take in. Still, we have four 14.5cm telescopes, a 1.5m sonic radar, two sky cameras, 4 webcameras, a 15-m tower, and a 450 micron wavelength telescope, several terabytes of disks, a dozen computers, about 64GB of flash storage, two Iridium satellite modems.
I'm one of the 4 UNSW scientists who designed PLATO. We certainly are using the fuel efficiently. When the sun is up we get over 1kW from solar panels, and we run one diesel generator at a time with just enough heat output to stop the fuel from getting too cold and turning to gel. Interestingly, the solar panels are considerably (about 30%) more efficient than you would expect from temperate site measurements - the colder temperatures (-50C at the moment) help, as does sunlight reflected from the snow.
I agree strongly with that. I have been amazed at the glacial progress in speech recognition in the decade I have been using ViaVoice and Dragon Naturally Speaking. Bugs in Dragon remain uncorrected for years, despite user complaints. As is common with proprietory software, there is no dialog between the users and the anonymous programmers.
The lack of any Linux alternative is also a major problem - Dragon is the only reason I still run Windows. The last time I checked (6 months ago), Dragon didn't work with Wine. With Microsoft owning, AFAIK, a substantial fraction of Nuance, and with Dragon being the only game in town, I see a very bleak future for Linux-based speech recognition.
If only the current generation of elite open-source programmers would realize that unless they get speech recognition working in the next decade, they will be in trouble when they themselves develop repetitive strain injury.
It is easy to think you are invincible when you are young and your fingers can hammer away at lightning speed on the keyboard for 18 hour programming sessions, but, reality hits after 5, 10, perhaps 20 years, when you suddenly find that typing is painful and even clicking a mouse button is difficult.
Viewing half the sky is not a problem for most astronomical projects, particularly since most of the interesting objects are in the Southern Hemisphere anyway!
In summer when the sun is up you just do IR/submillimeter astronomy.
Antarctica is not as expensive as you might think. The cost to get a telescope to Dome C is only US$3/kg. Overland traverses from the Antarctic coast can easily take hundreds of tonnes of equipment. And rather than driving a truck up to your door, at the South Pole you can land a Hercules aircraft right next to the observatory.
True, the stations are only accessible over summer, but it is not excessively expensive to get there
As for the extreme cold, this is something you have to put up with to have the best site for IR astronomy. Both the sky and the telescope emit in the infrared, so you want to have them as cold as possible to achieve the best performance. Existing observatories might have shirt-sleeve environments, but they have IR background levels that are between 10 and 100 times greater than in Antarctica.
Issues of ice build-up are understood through over a decade of work at the US South Pole station. Sub-millimeter telescopes are working there now, and IR telescopes (SPIREX) have done so in the past.
When the sun is up (summertime) you can observe in the infrared and submillimeter. Hubble's observing efficiency is about 50% due to the requirement to avoid the Earth, the South Atlantic Anomaly, slew time, etc.
The limitation is sky coverage is not important for many astronomical programs. Important regions such as the Galactic Center, the Magellanic Clouds, and the South Galactic Pole, are all visible.
Some electronics operates below its specified minimum operating temperature, and some doesn't. For example, we had some solid-state disks that were rated to -40C, but that failed at -20C. Mostly we have found that PC/104 computers, memory, etc work fine at -60C. M-Systems solid-state disks have been very reliable.
You want to avoid spinning up a hard disk at -85C though! The altitude (4000m equivalent) also tends to be rough on hard disks (both due to the cooling problems and the smaller head-gap), which is why we avoid hard disks in critical applications. Actually, one of our experiments is running on a Dell laptop with a normal 2.5-inch IDE drive and RT Linux. It has worked fine for two years now.