Microsecond air-gap flash photography

October 2nd, 2011

To take photo’s of really fast events, you need a very short duration flash. A normal camera flash lasts about 1-3 milliseconds at full power. These pictures were taken using a home built microsecond flash, a thousand times faster.
Example: a bullet travels at 1000 feet per second (305 m/s), so in 1 millisecond it would be blurred out to a 1 feet (305 mm) long blur. Using a microsecond flash, the blur would be 1/1000 of a feet or 0.3 mm.

This card deck is incomplete now (AR15, 9mm bullet)

No flames like .44 magnum flames

Bullet just left the S&W 686 .38 Special at about a 1000 feet per second.

A classic: 1911 makes for a messy shot

Microflash has more "stopping power" than a hollowpoint. Measured at 329 m/s or 1184 km/h using a chronometer. It rotates at 80120 RPM, once ever 9.7 inch.

Small balloon collapses very fast, even compared to bullet speed.

Chronometer showing 9mm hollowpoint speed in m/s (1080 fps).

Another balloon hit.

.177 pellet exiting the muzzle of an air pistol

Soap bubble hit by pellet

The air-gap flash was built from readily available scavenged and new electronic components, very similar to the EG&G MicroFlash 549.
The flash was triggered by an Arduino microprocessor with user selectable delay of about 1 millisecond using a piezo microphone to detect the shot.
Main inspiration/information for (and warning against) building a similar flash unit can be found here.

Complete flash system with tubular flash housing, Arduino in a control box and microphone.

Flash stripped from drain tube showing from left to right: spark gap, capacitor, trigger bobbin, trigger circuit and flyback driver.

Business end of the flash

Close-up of the 20mm spark gap and trigger wire.

Arduino triggered lightning photography

June 28th, 2011

- Arduino
- a simple photodiode
- a cheap shutter release remote cord from Ebay to get a plug for the Nikon D90 remote port (2.60$)
- a major thunderstorm after sunset

- DSLR set to manual 1s exposure time, diaphragm wide open , ISO low
- Arduino waits for light flash detected by phototransistor
- On the detected light flash the port connected to the remote cord goes LOW
- Shutter opens after 69 ms (that is the Nikon D90 shutter lag)
- since lightning is not an instantaneous event you still catch the final discharge and/or afterglow

Bahtinov Grabber goes 64 bit (and 32 bit update)

June 23rd, 2011

This is the introduction of a 64 bit release of the grabber. Download Bahtinov Grabber 64 bit (6619) here.
It was tested in combination with ASCOM 5.5b. Make sure you have that installed.

This same release is available in 32 bit as well: download Bahtinov Grabber 32 bit (5370) here.

Main changes:
- generic, more accurate calculation of focus error. Also works for masks other than the standard 20 degrees
- choice which R,G,B channel(s) should be part of the line detection
- audible feedback

Polar alignment check using a single plate solve

October 31st, 2010

This is how:

- set up your mount
- point telescope or camera towards Polaris
- take long (30s?) exposure of which the first 10s are static and the last 20s rotating the RA axis for 180 degrees

This results in an image that both show a star pattern and a set of circles.
By “plate solving” (finding the celestial coordinates) of the star pattern and drawing a RA, DEC grid you can see how much your polar alignment is off. Perfect polar alignment will result in concentric star trails and RA DEC grid.

Plate solving can be done with astrometry.net.


September 22nd, 2009

Yet another non-rigid registration and stacking tool.
This time using the “demons” technique for non-rigid registration (matching) of AVI frames.

I hope the enforced workflow makes the use of the software clear by itself:

download: RegistaC# (1566)
.NET 2.0 required if not already installed

The application was entirely written in C# (VS2005 express), making use of the FFTW library for FFT (wavelet filtering and global shift detection) and the AVI library from Codeproject.

The idea is

1. open avi and select a nice frame
2. align all other frames to that frame
3. measure quality of all aligned frames in preferred region of interest
3. average best X% frames
4. correct all sharp frames for seeing using “demons”
5. stack again for final result
6. apply wavelet sharpening

after that you can re-iterate alignment, seeing correction, frame selection etc. quite ad-lib.

Here is an explanation of the demons technique: demons powerpoint

feedback more than welcome!


3D images from 2D

September 14th, 2009

You can turn a simple 2D image of a planet or the moon into a “3D” image using this 3Dfication.

Make sure you make square crop that fits the planet or moon with only a small margin, of the image you want to use for 3Dfication. Using animation, you can make movies like this and this. AVI’s need the DivX codec.

download: 3Dfication (1793)


Periodic error and Polar alignment error model

September 8th, 2009

Using linear algebra, a model for calculating the trails that stars produce when making astrophotos from an equatorial mount was created.

An ideal mount that is perfectly parallel to earth rotation axis results in photo’s with pinpoint stars.
In the real world, the stars will always produce trails. Main causes are

- imperfect polar alignment
- periodic error of the (worm) drive
- RA drive rate that is different for siderial rate

Download Star Trails (1772) here.

All of these parameters can be played with in this model.

What I learned from the simulations:

For DEC=0 and a small polar alignment errors, the stars will describe a vertical line over 24h:

24h DEC 0

For DEC>0 and a small polar alignment errors, the stars will describe an ellipse, over 24h. The ellipse will be wider for larger DEC:

24h DEC 30

The periodic error will always add a wiggle in RA to the curve:

24h DEC 30 plus PE

Automatic planetary animation beautyfication in Matlab

August 26th, 2009

Using a “simple” Matlab script it is possible to fully automatically enhance planetary animations:

- shift between frames is corrected using “center of mass”
- color balance/brightness of all frames balanced is matched against the best frame

An example based on work by Mike Salway.

Jupiter Animation Mike Salway

Jupiter Animation Mike Salway

Jupiter Animation by Mike Salway improved

Jupiter Animation by Mike Salway improved

Jupiter animation Mert

Jupiter animation Mert

Jupiter animation Mert improved

Jupiter animation Mert improved

Neptune and Jupiter

August 23rd, 2009

Two 30s exposures, ISO200, F/8, f=200mm are enough to reveal their apparent motion across the stars.

Neptune and Jupiter photographed 4 days apart on August 19 and 23 2009.

Neptune and Jupiter photographed 4 days apart on August 19 and 23 2009.

Periodic error of my Kenko Skymemo

August 15th, 2009

kenko skymemo

I have a Kenko Skymemo, which has proven to be a portable mount with a good polar scope, which makes for nice images during trips to dark places. I had to replace the tripod though, since it is not designed for lattitudes as high as mine (51 degrees). This was solved with manfrotto 410 with geared head.

Looking in the internet for a specification of the the priodic error, I could not find any real proof of its size. Samir Kharusi has the most detailed review I could find, but still no details.

So I set out the measure it myself. This is what I did

- mount my Nikon D90+200mm+2x teleconverter on
- manfrotto tripod with geared head
- accurate polar alignment using the scope
- then spoil the polar alignment by tilting the kenko to the west, to induce a drift of the stars so the periodic error shows as sine-wave line star trails in the photo
- exposure of 10 minutes on Altair (DEC=8 degrees, chosen close to DEC=0, the most critical area in space with respect to star trailing due to periodic error and polar alignment error).

Some calculations:
pixel size of D90 is 5.5 microns, so at f = 2x 200mm , this corresponds to 3 arc seconds per pixel. The found periodic error in the images must be divided by cos(DEC_Altair), but since cos(8 degrees)=0.99, this factor is not significant.

The wiggle found in the 10 minute expose was 14 pixels peak-to peak, thus the PE=14×3=42 arc seconds peak to peak. A bit disappointing to me as I read somewhere it was around 15 arc seconds. This would mean that using a D90, PE would be invisible at any exposure time and any declination, using f=400mm/14pix = 29 mm focal length. Hardly a tele-lense value.

trails of Altair (bright) and a close by star in a 100% crop of a D90 image at f=400mm, exposure=607s

trails of Altair (bright) and a close by star in a 100% crop of a D90 image at f=400mm, exposure time=607s

The good news is that the curve is very smooth. This means that during short exposures of 1-2 minutes, the chances of getting a trail-free image is high, depending though on where you are on the sine wave of the periodic error (peak, valley=good, in between is bad).

Altair in the south at during the PE measurement

Altair in the south at during the PE measurement

UPDATE: Hutech, vendor of the Skymemo (by person of Ted Ishikawa) confirms via email that this is a normal PE value, quote: “Yes, the performance seems to be quite normal.”