The Starfish camera was originally designed as an auto-guider camera for astrophography uses. Since there are several other fine auto-guider cameras on the market today, we frequently get questions about how the Starfish camera compares to our competitor's offerings.
We are not trying to disparage any of our competitors with this comparison. On the contrary, all of the cameras we compared are fine instruments and should serve their user's well. We only wish to point out some of the qualities that make the Starfish camera stand apart from the other cameras. We have tried to be as objective as possible in our tests.
All of these cameras are capable of being used as an auto-guider. However, some of them have been simply pressed into service and were not originally intended to be used in this capacity. For instance, some cameras do not have an integrated guide port and may require additional hardware to be purchased by the user. Others only have USB 1.1 interfaces that have very slow image upload speeds that may hinder performance.
For a given optical setup, the field of view (FOV) will be determined by the physical size properties of the image sensor in the camera. Larger pixel sizes and higher number of pixels in the sensor's array will increase the FOV of the resulting image. As can be seen from the table above, the FOV of the test cameras varies quite a bit. The following image will graphically show the differences between the cameras tested.
We tried to keep the support equipment and environment variables constant in our tests. Only varying the camera itself for the comparison pictures. For instance, we used a common optical train and telescope mount for the tests. We also used MaximDL as our image capture application since all of the cameras we tested were supported by this application. We used the latest camera drivers available at the time the tests were run. Lastly, we used the Delta2 Lyrae star field in all of the test pictures.
The following equipment was used in our test setup:
Image sensor sensitivity is specified in many different ways by different manufacturers. The most widely used method is to specify the Quantum Efficiency (QE) parameter. This is a measure of the efficiency in which the image sensor can convert light photons into electrons that can be measured by the A/D converters in the camera. The cameras that use Sony image sensors cannot be easily compared to the others because of the way Sony measures sensitivity of their image sensors.
For our tests, I imaged the target star field using several exposure times for the picture. Exposure times were chosen to be fairly short intervals since those are typical of what is used while guiding. Longer exposures will allow you to guide on fainter guide stars but the longer exposure times will result in less frequent guide corrections being sent to the mount. Because of seeing conditions, you will usually not want to have too short of an exposure time, otherwise you will be trying to have your mount chase the apparent position of the guide star due to seeing turbulence and not the actual position of the guide star.
There is a fine balance that needs to achieved when determining the best exposure time to use for a given set of equipment and seeing conditions. Exposure durations of from 0.5 sec to several seconds are common.
In the set of pictures taken, only a single dark frame was subtracted from each picture. The resulting images were contrast stretched at approximately the same amount so as to present a consistent relative star brightness for all of the images produced by the cameras at a given exposure setting.
The pictures can be seen here: http://www.fishcamp.com/starfish_shoot_sens.html
As can be seen form the pictures, the CamE camera is the clear winner in terms of sensitivity. This is not surprising from the QE specification for its sensor in the above table. This sensor did not have anti-blooming circuitry in it so all of the images from the CamE camera show significant blooming on all of the bright stars. You would not want to select any stars that exhibited this blooming characteristic as your intended guide star. The CamD camera was the least sensitive of the cameras tested.
Many times people will make a statement like "the image sensor determines the performance of a camera". Well, to a large degree, the image sensor used in a camera will determine many of the performace characteristics of a camera. However, the design of the camera electronics surrounding the image sensor can have a profound effect on performance. Seeing the two image animations above, side by side, helps illustrate the point. Both cameras are using the same image sensor but the final images produced by the cameras are quite different.
In this case, the CamA camera was designed to be a low-cost, minimalist guide camera. The Starfish camera, however, was designed as a higher-end guide solution. It has some dedicated image processing, in hardware, that detects and corrects certain dynamic noise components like the line offset variations visible in the CamA image.
A dynamic noise measurement commonly made on cameras is a 'read noise' measurement. This is done by taking a pair of bias frames and flat frames that are then used to compute the cameras read noise. Programs like IRAF have routines (FINDGAIN) that simplify the computation of the read noise of a camera. We have measured the read noise of the cameras being tested with the following results:
Fixed pattern noise is routinely eliminated by astrophotographers from their images through the process of image calibration during post capture image processing steps. However, this is sometimes not practical when using the camera as an auto-guider since the images are being exposed and operated upon in somewhat real-time. For guide applications, it is best not to have to deal with this kind of anomaly in the image.
For this test, we used MaxIm DL's ability to set specific parameters of the guide camera. Specifically, we set the minimum and maximum guide move to be 100mS. This guaranteed that every guide correction made by the camera should be exactly 100mS in duration. What we discovered is that there was a lot of variability in the length of the guide pulses generated by the cameras.
We took some oscilloscope pictures of the test results and present them below:
CamA, in general, had good pulse characteristics. Levels were good and the pulse duration was consistent from command to command. However, there was a peculiar duration offset of about 10% visible. Each pulse command had the exact same offset added to it giving the pulse a duration of ~110mS vs the desired 100mS nominal pulse width.
CamB was not tested since it did not have a built in guide port.
CamC had the most variation of the cameras tested. The pulse duration varied wildly based upon processor load. As a matter of fact, just moving the mouse around the screen while guiding caused a noticeable variation in pulse width from nominal.
CamD had good pulse width repeatability but had a peculiar amplitude offset. Ideally, a pulse signal should be driven to the ground reference level to register a guide command. In this case, the guide signal would only go to 0.8V above ground level. It is not certain if any telescope mounts would have a problem with this. If the mount controllers utilized standard TTL logic levels on the guider interface, this would be precariously close to the TTL logic level low signal level which could , perhaps, cause uncertainty in the command state of the signal.
The Starfish camera and CamE both had excellent pulse characteristics. The Starfish camera has on board pulse generation hardware that guarantees consistency in pulse duration from command to command. It actually has four such sets of hardware so that each guide direction signal can operate independently and concurrently with the other directions. This is important if you have a mount that can handle simultaneous guide commands in multiple directions.