Stable Diffusion v1.4 with webui, text to image prompt ‘SNOT ON A WINDOW’. Seed 2508499346, Classifier Free Guidance Scale 7.5, sampling steps 50. Sampler: k_lms
Snot on a Window
Tiger submarine
Stable Diffusion v1.4 with webui, text to image prompt ‘tiger submarine’. seed 1, classifier free guidance scale 9, sampling steps 82, sampling method LMS. Curated best and worst results from 50 image samples. Of course, ‘best’ and ‘worst’ is subjective, but this is a great example of the wide range of images that Stable diffusion will generate!
It is only fair (thanks Rob) to consider ‘submarine tiger’ too as the first image is clearly this rather than the original prompt, so let’s see what difference it makes. Spoiler alert – none of the set of fifty images contain any tigers!
The magic of Lego at none of the cost
The moment you break the seal on a new Lego set and start laying out the bags, you are in for a treat, but an expensive one. The price of Lego sets varies according to the size of the set, the complexity, and the licensing fees Lego had to pay to get the rights. This makes kit prices very variable when trying to apply a standard measure. Kits can end up anywhere between £50 and £180 per kilo of bricks.
The largest sets tipping 8KG, that can be a lot of dish. 75252 Star Destroyer, 75192 Millennium Falcon work out at over £500 each while being some of the best per kilo value at £50, and Lego architecture series working out at over £150 per kilo to be the worst value for money of all kits ever.
Lego Studio can be a great way to virtually build those kits that you can’t afford in real life. Here is a gallery of my builds. You can even find some rendered timelapse on my you tube channel.
I started this project because I wanted to simulate a Mega-Titanic. More on that another time, but in the meantime, here is a gallery of some renders!
Kerbal Space Simulator: Starlink
It seems that we will all have low latency, high bandwidth internet soon, regardless of where we are on the earth, all courtesy of SpaceX and Starlink. It seemed like a good idea at the time to make a Kerbal representation of the constellation just for shiggles over the xmas lockdown.
Step 1 – Design the bird
Rather than start from scratch, I had a search around (similar to a customer looking for a satellite bus format) and found this video from Gameplay review UK. After some messing about we end up with a lovely flatsat layout with a gigantor array and xenon power.
Our bird is designed flat so we can launch a stack of them. we base it on the OKTO2 and add mechjeb2 for additional control plus a reaction wheel for in service orientation (our normal orientation is radial out).
Launch weight is not too much of an issue so we have a full xenon tank and around 1200 delta v – enough for many years of faithful service and orbit raising.
We simulate the Starlink service with a pair of antennas – the Communotron 16-S for C&C and the folding SG-5 for the user internet traffic.
We put a Z100 battery pack on board so that small DV adjustments (<10 m/s) can be made even on the dark side of Kerbin. A couple of flat solar panels each side of the sat give us emergency power to deploy the gigantor.
Considerable tweaking is needed to get the COG aligned with the xenon engine.
Step 2 – Testing the bird
I built a SSTO rocket and fixed one satellite to it – we’ll call this the TinTinA launch. Launched to a nice low 100km orbit, the satellite works great – it’s controllable and maintains power throughout it’s orbit and has enough electricity for small DV changes. we have just enough DV in the booster to deorbit and our orbital debris is zero.
Step 3 – Orbital Parameters
Starlink orbits at 350km and at an inclination of 53°. Our constellation will use identical parameters even though that’s a much higher relative orbit. Looking at the offsets around Kerbin, we need to launch at around 7 minutes 30 offsets – this gives us about 49 separate orbital planes to fill, or around 2500 separate satellites (EEK)!
Step 4 – Extended orbital test programme
Extending the TinTinA launch, we send up 48 further single sat launches to test the concept and test the coverage. Again, we set these launches low, but to the correct inclination. This is successful and after the testing is complete, all test artefacts are deorbited.
Step 5 – Testing the stack
The first stack design is built and tested as a two stage to orbit dispenser. We release all the sats at once and deorbit the dispenser. This first attempt of 24 birds is messy – using the SpaceX technique of allowing the set to jostle a bit results in a few damaged birds – typically the gigantor array. it takes quite a while to deorbit the failed satellites using only the tiny fixed panel. Another problem with the all-at-once dispenser is that the satellites disperse too unevenly. Because they have to be individually positioned, they drift a long way in the orbit it takes to position each one.
Deorbiting the failed birds gives us a chance to test that they burn up completely in the atmosphere. They don’t, but it’s a bit late to do anything about this. As we have lot’s of delta V, we’ll try to controlled deorbit them into the ocean. Luckily Kerbin seems to be very sparsely populated.
The stack is redesigned slightly to release them individually and we try again. Now we have a stack of 38 on a two stage configuration.
It turns out that this is still a very messy configuration – the dispenser is aligned to prograde prior to sat deployment and then the idea is that the satellite can boost away from the 330 deployment orbit to the 350km operational orbit, but the stack is upside down and so there is considerable delay while the bird clears the stack far enough to invert it and deploy the antenna and solar array.
There is also a problem with the supports: many support struts were needed to prevent the stack flopping around and these were not all applied from the dispenser to the satellite. On separation, any bird with the strut base is unstable under thrust.
Step 6 – Testing one orbital plane
Testing the deployment of one entire orbital plane is hard work – we are working with a version one stack and so they are all upside down. Some of them have the detritus from the support struts and this means that they don’t produce thrust through the CoG. It takes about 8 hours (real time!) to deploy and we hit another snag: The last two birds are not on separation rigs and so are permanently attached to the dispenser. Another redesign!
Deployment process
Checklists for the win! It’s really easy to miss a step in the deployment for each satellite and as there will eventually be 2500 of them, probably important for a repeatable list.
Launch
Launch dispenser to 330km
Circularize
Check and adjust plane and offset from
Tune Apo and Peri to 330 exactly
Planning a train
The deployment process for each train needs careful timing and coordination – satellite spacing in the plane is determined by timed release of each sat. Once you have the apoapsis and periapsis well sorted and the inclination is solid, you need to warp to ‘dawn’ from the dispenser point of view: You are going to release the bird, boost it’s apoapsis and then circularize half an orbit later when it reaches apoapsis. you want to make sure bot of these happen in sunlight. After circularizing, you will flip back to the dispenser and repeat the exercise and so it should all be happening in daylight. If you screw this up, the whole train will be launched at night and the power budget is very tight!
The UT between each deployment is 30 minutes. That gives you time to release, align, raise Apo, wait for apo, circularize, rinse and repeat. Just. Every few launches, you will want to adjust the dispenser apoapsis back to 330 – those constant prograde sat deployments cause the dispenser to slow down!
Hints, tricks and tips
Creative use of the different ship type icons can help finds ways to declutter the screen *ahem* when in Tracking station and Map view. Perhaps use one for the dispenser, one for birds that have been deployed but not fine tuned, another for fine tuned, Another for on-orbit spares and another for the head of each train.
Name each satellite according to the launch, the plane and a unique number. I use starlink 430-33 where 430 is the launch time UT and 33 is the position in the train. I add ‘HEAD’ to the reference bird in each train and give it a different icon.
Each Starlink Sat Deployment
Point dispenser prograde
Select Sat
Deploy solar panel
Deploy Sat
Switch to Sat
Check naming
Unfold antenna
Activate engine
Switch on attitude hold
Align prograde
Change the Engine Max thrust to 50%
Use Mechjeb manoeuvre planner
Select ‘Change Apoapsis’
Select 350km (or whatever this train operation altitude is)
Select ‘After a fixed time’
At the appropriate time, hit ‘create and execute’
After the burn, use Mechjeb manoeuvre planner
Select ‘Circularize’
Select ‘at the next apoapsis’
Select ‘Create and execute’
Select 5% max engine power after the circularization burn
Minor apo/peri adjustment (no warp!) – Leave this until later!
After circularization complete orient ‘radial out’
Switch back to the stack and repeat.
Post Deployment Dispenser decommissioning
The empty dispenser is now useless. If you have not needed all the birds, keep the dispenser on orbit to keep them together. If keeping the dispenser, probably raise it’s orbit to that of the main plane for collision mitigation.
To deorbit it, you have choices. If you included chutes, you may want to go for a land return at KSC. It does have enough DV for alignment but not a powered landing. The dispenser is unlikely to survive re-entry unless designed quite carefully – it’s as aerodynamic as a house brick. Conduct your own risk assessment! Otherwise use Mechjeb and pick a nice uninhabited spot in the ocean. Bear in mind that a return to KSC from a 53° inclination may require a long warp.
You could of course boost the dispenser into deep space or go for a graveyard orbit. You may have the DV for either of these, but they are not very SpaceX!
Deployment Fine Tuning
Change the engine power to 5%
Orient prograde or retrograde as required
Warp to apoapsis -10 seconds or periapsis as required
Tweak orbit to 350km x 350km exactly. A couple of meters is acceptable. you may decide, based on your own collision avoidance measures to put each train at a slight different height, say 100 meters separation on each orbital plane. This would be the right time to chose that number and start tweaking.
Deactivate unused equipment (engine) to prevent any power drains and accidents. You may be able to deactivate mechjeb and other systems.
Respositioning
<to follow>
Gallery
Storing data for 1000 years
So we all know that books last a long while right? How about its modern equivalent the CD, DVD or Hard disc? Most of us commit a larg part of our life to digital storage these days and don't even spare a thought for how long this digital memory will last. In the case of hard discs, 5 years might be your limit.
This then is the classic problem with digital storage – longevity. All other problems vanish into insignificance when considering the persistence of storage.
Our current best hope is optical media, the best of which are believed to be good for around 50 years. (interesting to think that our CD and DVD collections are more temporary than we are!)
Recently, a team from University of California Santa Cruz have come up with a system called Pergamum which uses regular hard drives to store data for a projected 1000 years. The system, called Pergamum, uses a large number of inexpensive discs, most of which are spun down. Persistence is created by using multiple slow sync mirrors & stripes.
Clever caching and an even cleverer data distribution protocol ensures that data reads, writes, rebuilds and other operations make the least demand possible on the hardware whilst continuing to be power efficient.
As an example, it's estimated that a 10Petabyte system would consume only around £20 of power per year including HVAC requirements and would be capcble of maintaining that 10PB for arount 1000 years before a data loss occurred. Pretty good methinks!
More on the announcement in this tgdaily article.
Painless Phone Call Recording
Have you ever had to make one of those calls that you wish you could record but locating the voice recording feature on your mobile or answering machine is too hard or you can’t then get the finished recording into a useable format? This seems to happen to me about once a month.
Recently I discovered a service which makes the process of call recording extremely painless. After registering you are given a phone number, access code and a pin. After dialling these in easy succession, you get…. A dial tone. Any subsequent calls you make are recorded and the subsequent call recording is either emailed to you or is available for collection at the portal site. Making multiple calls is easy as the ‘#’ acts as a break-in code to tell the system you want to make another call.
In pay-as-you go mode the process uses a premium rate access number and you simply pay by the minute for the call to the access number. The onward part of the call is not charged back to you.
The service has other offerings suitable for more regular users and also features ideal for small businesses who need to regularly record outgoing calls to customers.
Additionally, it’s possible to have your own phone number for the service which then does not require the use of the access code and pin. This also allows you to record incoming calls as you can set the service up to call you on receipt of a call.
All in all, this is a really nifty service. Check it out at http://www.recordmyphonecall.co.uk/
Secret Wall Tattooing
Secret Wall Tattooing is a weird art form in which the artists deliberatly hide their light
under a bushel. In this case, that bushel is removing the pictures from
the wall of your hotel room, painting, drawing or scrawling some art,
then replacing the picture. Pretty cool!
Harbin Snow and Ice Fair 2005
The air is so cold it freezes your stinging tears to your face; the sun
is so low it escapes to leave you in darkness by mid-afternoon; the
trees are so gray, barren, and hard they could be concrete; the river
ice is so thick it actually supports entire buildings. This is
Harbin, China’s northernmost (and easternmost) metropolis. Over 17
years, as this eight-meter-high horse sculpture indicated, the festival
has grown in size, complexity, and elaborateness; where the snow
festival had a single massive sculpture before, a handful of these now
appeared. This year’s snow festival was officially called “The
17th Annual China Harbin Sun Island International Snow Sculpture Art
Fair.” more
Full Brain Simulation Courtesey of IBM
An effort to create the first computer simulation of the entire human
brain, right down to the molecular level has just been launched. The
“Blue Brain” project, a collaboration between IBM and a Swiss
university team, will involve building a custom-made supercomputer
based on IBM’s Blue Gene design. The hope is that the virtual brain
will help shed light on some aspects of human cognition, such as
perception, memory and perhaps even consciousness. more
Diamonds in double quick time
Researchers at the Carnegie Institution’s Geophysical Laboratory have
learned to produce 10-carat, half-inch thick single-crystal diamonds at
rapid growth rates (100 micrometers per hour) using a chemical vapor
deposition (CVD) process. This size is approximately five times that of
commercially available diamonds produced by the standard
high-pressure/high-temperature (HPHT) method and other CVD techniques.
In addition, the team has made colorless single-crystal diamonds,
transparent from the ultraviolet to infrared wavelengths with their
process. more
Honda ASIMO can run now!
I love this video of Honda’s new-and-improved Asimo robot running. From
the press release:
"The combination of newly developed high-response hardware and the new
Posture Control technology enables ASIMO to proactively bend or twist
its torso to maintain its balance and prevent the problems of foot
slippage and spinning in the air, which accompany movement at higher
speeds. ASIMO is now capable of running at a speed of 3km/hour. more
Fibre reinforced bendable concrete
A new type of fiber-reinforced bendable concrete will be used for the
first time in Michigan this summer- and University of Michigan
scientists hope that their new material will find widespread use across
the country. The new concrete looks like regular concrete, but is 500
times more resistant to cracking and 40 percent lighter in weight. Tiny
fibers that comprise about 2 percent of the mixture’s volume partly
account for its performance. more
Wire in the Blood!
A Japanese research team has developed a fuel cell that runs on blood
without using toxic substances, opening the way for use in artificial
hearts and other organs. The biological fuel cell uses glucose, a sugar
in blood, with a non-toxic substance used to draw electrons from
glucose.“Since the electron mediator is based on Vitamin K3, which exists in
human bodies, it excels in safety and could in the future generate
power from blood as an implant-type fuel cell,” the group said in a
statement.
Most other bio-fuel cells under study use a metal complex, spawning concern about harm if used for implants.
The newly developed cell in the size of a tiny coin is able to
generate 0,2 milliwatts of electricity, enough to power a device that
measures blood sugar level and transmits data elsewhere, the group said.
Two Wheeled Transport with an Axle?
With an 12mph top speed, the Segway could not be said to be fast. It is surely set to turn heads though.
The design is a cross between a child’s scooter and a pogo stick. The rider stands aboard a small platform between the two wheels. From the platform extends a long shaft with handlebars at the top.
To operate the device the rider simply tilts and the Segway tilt sensors detects what the rider intends and reacts accordingly. Steering is effected with a handlebar control.With a 15 mile range under ideal conditions, the Segway is obviously not for long journeys, but the device could easily find a niche.
The legal position is still to be sorted out for the UK, but most US states allow the use of the Segway.
Perhaps the biggest stumbling block though is the price: with £65M in development costs to recoup, the £3000 price tag is understandable but will limit the market considerably.