Noisy amplifiers put the Kepler exoplanet detection spacecraft on hold until 2011

til 2011:

..

In some ways it sounds fairly simple. A planet will cover part of the face of the star in a repeated pattern once per orbit. Depending on the exact nature of the noise, it may not be too hard to pull out the shape of the transit from the noise. This assumes that the noise isn't just like the dimming the planet causes.

Typically known as spurs.

Sure, if the baseline noise is up where the observed data is, one is going to experience false and/or errant data sets.

If it is an amp, it is analog, but if noise is an issue, it is probably very high frequency. Maybe the communications amplifier?

If it takes two years to get hands onto an acceptable unit, it must be a very precise, and/or complex amplifier.

We had several hundred that operated in the 10GHz range that did not even get warm, but would fail due to an overheated small transistor package *inside* the unit. It was a remake of a previous design where a certain particular of thermal management was overlooked due to the use of a different transistor vendor.

They repaired or replaced them all, but it took them a while to even characterize the nature of the problem since they were not getting very hot, bur were injecting spurs into the signal, and due to the fact that the same design had worked for years previously. Then they nailed it as the new vendor item. Weird.

Sounds like that 100 Watt bulb does heat the box less than the 100 Watt heating coil does... HAHAHAHAHA! Absolute affirmation, in fact!

Anyway, it cost us several hundred man hours to remove the 2 units per rack module on 53 modules that were already installed into rack, and they each weighed over 90Lbs. What a pain that was. They are only 1 inch by

Most likely out sourced it out from china! what do you expect!

Taking an Earth sized planet and a sun sized star it is worth doing a quick sum on how much the occultation changes the stars apparent brightness. And how often - typically once a year.

Sun ~ 1400000 km Earth ~ 12740km

About 110x bigger on linear dimensions so you need 4 decades of signal to noise to get a 1 sigma detection at an unknown orbital period. It is a distinctly nontrivial problem and they are already using state of the art methods to extract what little signal there is from the noise. And they would probably want to get a 3 sigma detection to be comfortable with making the claim of detecting and Earth like planet.

When you consider that real stars also have limb darkening and sunspots that are comparable or even larger in size and go around with rotational period of their latitude on the star it gets even hairier. This isn't an easy problem on a nice flat baseline. It it like trying to detect insect splats on the front of a car headlamp placed on the moon.

The next transit of Venus in 2012 gives you a chance to see first hand what they are up against. And viewed from infinity Venus/Earth would appear 1/3 the linear size or 1/10 the area during transit.

I suspect the problem is that the noise is somewhat more than will allow them to get a workable 10000:1 SNR on typical targets.

I am a bit surprised that they cannot just accept all the data on the downlink and then process it on the ground where we have essentially unlimited computing power and storage capacity compared to a typical satellite.

Regards, Martin Brown

I think their down link bandwidth is much narrower than the detector bandwidth. This is very often the case with NASA.

They DO include enormous amounts of FEC with their streams.

Too many missing bits, however (bit error rate), and even that fails as entire packets begin being dropped.

Not on the receiving end. Some of the Microdyne supplied telemetry equipment was capable of 40 MB/S

Hahahaha... He said _bit_... :-)

On most of the NASA stuff the limitation is at the probe and is often the transmitter band width. They try not to lift any more weight than needed. The transmitters tend to be designed to exactly match the expected data load.

You were way more polite than i would have been. I would have pointed out that available transmitter power and bandwidth prevent transmitting the total data available at the satellite (which is much farther away than geosynchronous satellites). Shannon's law tells us just how much we can reasonably expect to obtain. It even helps us make better decisions on how much (and which kinds of) data compression and forward error correction we may use most effectively for any well described best/worst case link.