Digital Cameras

[Martin Chalfie]

Thanks to everyone who responded to my request for information about digital 
cameras.  Here is the summary of the responses sent to me or placed on the C. 
elegans or fluorescent protein newsgroups (with some editing)
- Marty Chalfie

From: Philippe-Alexandre Gilbert <Philippe-Alexandre.Gilbert at sit.ulaval.ca>

I just bought the Quicktake 200 from Apple. The quality is very surprising, 
but I do not know if you can plug it into a PC. Here's some features: 640x420 
dpi, capacity of 28 pictures (3.2 Meg), Video plug out, screen for preview, 
JPEG file format, etc. etc. I'm using this camera for taking pictures of 
plates and equipment in the lab. Perfect for presentation with Powerpoint.


From: Anne Hart <hart at helix.mgh.harvard.edu>

We (Sander and I) recently bought a SenSys CCD camera from
Photometrics- and are using a Mac to capture the images- which are directly
exportable to Photoshop. So far it works well- the images are high enough
resolution to obviate the need for film photography for publication...the
only difficulty is that leaving the Zeiss microscope cover atop the camera
when the camera is on causes the main circuit board to overheat- a 500$
repair. But, otherwise the camera and program are easy to use and are
working well for us. The whole thing was ?20K excluding computer - I think.


From: "Dr. George McNamara" <george_m at image1.com> Universal Imaging  
Corporation 

  The following are the digital cameras I have been using for GFP and other 
fluorescence imaging. All of these also work great for DIC and other 
transmitted light modes. Most of the CCD chips are available from both 
Photometrics and Princeton Instruments, and can be purchased as either 1 MHz 
(1 million pixel per second) readout (SenSys, MicroMax, respectively) or 5 MHz 
(Quantix, PentaMax, respectively). Photometrics and PI also make "less cooled" 
Kodak KAF-1400 CCD cameras that cost a little less but do not perform as well 
on long exposure times (long = 10 seconds, a long time for a GFP exposure!).  
I'll also mention that Hamamatsu makes digital cameras, but I don't have 
enough experience with their new C4742-95 camera.  QE is Quantum Efficiency, 
the % of photons detected/reach the camera, and varies with wavelength. All 
the cameras use a PCI interface card to transfer data to our Windows95 
software.

Kodak KAF-1400 full frame CCD: 1317x1035 pixels, 12-bit data (4096 intensity 
levels), each pixel is 6.8x6.8 um, 20% QE at 450 nm (BFP), 30% at 500 nm, 40% 
at 550-650 nm. This has been the "standard" digital CCD we have been selling 
to customers for several years.

SITe 512x512 back-illuminated full frame CCD: 512x512 pixels, slow readout 
16-bit data (65536 intensity levels), optional 12-bit fast readout is 
available in some cameras for fast focussing, each pixel is 25x25 um, high QE 
throughout the visible range (70-90% !!!). Outstanding senstivity with slow 
readout (i.e. 100 KHz = 100,000 pixels per second). A SITe 1024x1024 pixel CCD 
is also available but this is wasted by the standard field of view of a 
microscope. In principle, this camera can detect single fluorochromes, but it 
would have to sit still for a second or more.

PI 1000x800 back-illuminated CCD: 1000x800 pixels, dual speed 12- and 16-bit 
readouts, each pixel is 15x15 um, high QE throughout the visible range [give 
me a call if you want my opinion of my head to head test with a 512x512 SITe]. 
Only available from Princeton Instruments. Takes 8 seconds to transfer all 
800,000 pixels in slow 16-bit readout mode (but the data looks great). There 
is a choice of antireflection coatings, one of which is much better for BFP 
imaging because it keeps the QE at 450 nm high.

PI MicroMax/Interline CCD: 1300x1030 pixels, 12-bit readout (4096 gray level, 
about 11-bit data quality because of small well capacity), each pixel is about 
the same size as the KAF-1400. The critical two features of this CCD compared 
to the Kodak KAF-1400 are:
  1) no mechanical shutter (eliminating all risks of vibrating the 
microscope).
  2) QE specifications claim better performance at 450 nm (~30% QE), and 
00-550 nm (a few percent better than the KAF-1400). Also has somewhat ower QE 
than the AF-1400 in the far red (Cy5, far-red DIC) and does not image in the 
infra-red (so no infra-red DIC for imaging thick specimens; but also minimizes 
the amount of noise related to heat in the microscope).   The same CCD is 
available in the Hamamatsu C4742-95.

Two other cameras I'll mention, though neither is ideal for one or more of: 
large field of view, high spatial resolution and high dynamic range.

Kodak KAF-1600: 1536x1024 pixels, 12-bit data (14-bit when binned 2x2), each 
pixel is 9x9 um. Using the same optics (60x 1.3 NA plan apo objective on an 
Axiophot 2) and specimen (human metaphase chromosomes stained with FISH probes 
and DAPI) the KAF-1600 produced less crisp images than the KAF-1400. Not quite 
a fair test for the -1600 since I did not adjust the intermediate relay optics 
for the -1600. I took the images at a course at CSHL last October and one of 
the student's - an experienced cytogeneticist - greatly preferred the KAF-1400 
image quality.

EEV37 Frame transfer camera: 512x512 pixels, 12-bit data, each pixel is 15x15 
um. The QE is not awesome but the frame transfer mode allows the camera to 
acquire images in rapid succession. Typically used for high speed digital 
camera fura-2  ratio expriments for Ca++ measurements. I recently acquired 
1000 images in 27 seconds with this camera (25 ms exposure times, binned 
10x10), with only 2 ms "dead time" between each image.

Gen IV intensified CCD: The recent introduction of the Generation IV 
intensifier makes this an interesting challenger for digital cameras. I refer 
to this intensifier as a "no lag SIT" camera in terms of sensitivity and 
spatial resolution. However, since it is video, you get 640x480 pixels, 8-bit 
dynamic range, which is much less than the digital cameras described above. 
The Gen IV is great for live cell physiology such as fura-2 imaging of Ca++ 
responses. One caveat is that I don't know how sensitive it is for BFP 
emission. See http://www.image1.com/newsletter/v8n1/gen-III-IV.html for 
details.

The two back-illuminated CCD's are outstanding cameras if sensitivity is the 
most critical issue. With the advent of brighter GFP mutants that are 10x to 
100x brighter than wtGFP, a two-fold difference in QE of back (70-90%) vs 
front illuminated (20-40%) does not seem that compelling.

The MicroMax/Interline is likely to be the most useful camera for GFP::BFP 
work, though the KAF-1400 has performed well for many of our customers for 
several years.
Everyone's imaging needs are different.

 The manufacturer's I mentioned have the following web addresses:

Photometrics:           http://www.photomet.com/
Princeton Instruments:  http://www.PrInst.com/  includes downaloadable PDF
file of manual.
Hamamatsu Photonics     http://www.hpk.co.jp/hpke.htm   C4742-95 camera is a
few levels down


From: David Spector at Cold Spring Harbor Laboratory <spector at cshl.org>

Photometrics makes excellent B & W digital cameras with which you can input 
your images into a PowerMac or PC and then using image analysis software you 
can manipulate the images and make plates for publication etc.

SenSys Camera is about $15,000
NU200 Series is about $25,000

Both are excellent for fluorescence imaging, however the NU200 may be better 
for Nomarski, you should check with Photometrics.

From: Petrus Tang <petang at cguaplo.cgu.edu.tw>

The digital camera game is not as simple as you imagine. The whole system can 
be separated into two parts: image capture and printout. Digital camerars can 
be the simplest Kodak DC series equipped with different lens for stero- and 
light- microscopes or the expensive 3CCD cameras for image capture. The Kodak 
grade digital cameras usually come with a simple software for image managing. 
For the CCD cameras, you need an extra Capture card installed in your 
computer. No matter which type of cameras you use, the images looks excellent 
on the computer monitor if you set the resolution to 16 bits (true color).

The real pain is the printouts. You will be disappointed even you use the 
highest resolution settings on an ordinary color printer. An acceptable 
quality will be at least 1440 dpi on a glossary photopaper for ink jets. If 
you want a better one, you can use either colour laser printer or a thermal 
sublimation printer (both are really expensive). 

A dream system will be a 3CCD camera attached to a microscrope, a capture 
card; a computer with at least 2MB dispaly, 32 MB memmory, 2 GB hard disk; a 
thermal sublimation printer and boxes of photopaper for printer.


From: "Alex Bogdanov"abogdanov at helix.mgh.harvard.edu

We've got SenSys from Photometrics (tel. (520) 889-9933) and attached it to 
Axiovert. Approx $15K with a software package (IP Lab Spectrum). Excellent 
stuff. The company is currently upgrading these cameras with new sets of 
adapters. 

From: rtsien at UCSD.EDU ("Roger Y. Tsien")

On the subject of cooled CCD cameras, we envy Dr. Bogdanov's success with the 
SenSys. We've had our Sensys 0400 for less than a year, during which it has 
had to go back to Photometrics no less than 3 times for repair, each time for 
the same problem -- the window on the chip starts to leak, so that atmospheric 
moisture seeps in (even though San Diego is not considered a particularly 
humid environment) and condenses when the chip gets cold. Each drop shows up 
on the image, just like viewing through a rain-streaked window. Also, there is 
a recurrent but intermittent bug in the Photometrics
software so that roughly half the times one switches it on, the images are 
crazy with only about one bit gray scale instead of 12 bits. Fortunately 
Photometrics at least fixes the moisture problem for free each time, but after 
three tries I fear they can't solve either problem permanently. Fortunately 
other software, e.g. Axon Instruments' Imaging Workbench, reads the chip more 
reliably. Caveat emptor!


From: "Alex Bogdanov"abogdanov at helix.mgh.harvard.edu

Indeed, we had no  moisture condensation problem and almost no problems with 
IP Lab spectrum software. This must be different from Photometrics' own 
software package. IP Lab's  running  fine on Power PC.   I guess now we'll 
have to prepare for the worst and wait till we see sweating cells on our 
monitor screen.


From: "Albrecht von Arnim" vonarnim at utkux.utcc.utk.edu

My lab is using a Princeton Instruments MicroMax/IPLab system with the KAF1400 
black and white CCD (1350x1000 pixels) on an Axiovert scope. 

I have some experience with a Sony 3-chip color CCD camera (Model 72, 8-bit, 
about $10,000 with fluorescence imaging extension package).  Color is nice, 
but the sensitivity, dynamic range, and resolution were all insufficient for 
our work in subcellular imaging.

What I like about the Micromax:

1. Low noise due to cooling to -18oC, and therefore high sensitivity.  We 
haven't been limited by the sensitivity of this camera.

2. Very high resolution, due to the small size of the chip.

3. High dynamic range (12 bit), which is more useful than I imagined, because 
our cells contain very bright and very dim signals.

4. The IPLab image acquisition program, which comes with the Mac-version of 
the camera, is quite powerful.  We are only scratching the surface. 

What I don't like about it:

1.Like most CCDs, the chip has low sensitivity in the blue, and very high 
sensitivity in the infrared, which requires specially blocked emission filters 
for our work in plants.  Since we mostly image GFP (which you kindly sent to 
us, thank you again), the reverse would be more ideal. Can't have everything.  
I have not tried 'back-illuminated' or 'UV-coated' CCDs.

2. The system must run better on a PC than on a Mac.  We are stuck on the Mac 
right now.

3. There are quite a few bugs in the system that we try to sort out the hard 
way.

4. My computer wiz here clasps his hands over his head when contemplating the 
design of the system:  It converts a digital signal on the chip into an analog 
signal in the controller, and back into a digital signal in the computer.

-> If I had to buy a system again, I might possibly end up with the same one 
($15,000 including a 2nd grade CCD [good enough] but without the video option 
that you may hear about).

My advice: When you invite reps over to demonstrate the equipment, grill them 
carefully, and make sure the system works bug-free in your environment.
 
From: MARTIN PERRY <73061.3336 at CompuServe.COM>

Jennifer Kramer of Scanalytics asked me to contact you regarding your inquiry 
on digital imaging. We offer a number of solutions in digital formats. For 
Black and White we offer the Photometrics SenSys or Quantex camera systems for 
12 bit imaging.
In color we have the New SPOT digital color camera using a 1K x 1K sensor for 
high resolution digital image RGB color files. Out put of both cameras can be 
printed out utilizing a dye sublimation type printer such as the PrimeraPro 
Elite which yeilds a 8>5 x 11 print in a little over 4 minutes.


From: Roger McCleary <mccleary at netacc.net>

I have personal experience with two high resolution digital camera systems:
For fluorescence photography we use an Apogee cooled ccd camera (-15 C is 
adequate for this use) with the Kodak KAF1600 detector. Fast, sensitive and 
low noise 12-bit images. Apogee has a variety of cameras with different chips 
depending on requirements for cost, resolution, sensitivity, bit depth, etc. 
http://www.apogee-ccd.com/

For brightfield microscopy we use a Kodak Megaplus 1.4 camera. This is an 8 
bit camera from which our image analyzer (Quantimet 600) grabs 1K x 1K images. 
Newer cameras are available: http://www.masdkodak.com/mplusmtrx.html

You may need an output device. We use a Kodak Colorease thermal printer; newer 
models are available that are faster. Unlike some "popularly priced" inkjets 
we have tried, color reproduction and photographic look were very good on the 
first try. 
http://www.kodak.com/cgi-bin/webCatalog.
pl?section=Share&category=Continuous-Ton e+Color+Printers  


From: Thomas Burglin <burglin at ubaclu.unibas.ch>

I have been looking into the Kodak DCS 420  (9000$), which can be had with 
Nikon or Canon mounts, 1000x1500 pixels, with plug-in into SCSI port and 
software for loading it directly into Photoshop.  Haven't used it though.

Overall, the resolution looks nice in the demo pictures, though hardly any 
camera will do the high resolutions that current photographic films can 
deliver. Also, work with huge pictures on a computer is also quite a 
bottleneck.

(if I remember correctly, the Minolta was basically the same as the Kodak).

You can also go the videocamera-framegrabber route.  Info for that is on John 
White's pages on 4d-microscopy.


From: kraev at bc.biol.ethz.ch (Alexander Kraev)

I have been using a Minolta RD-175 (also known as Agfa Action Cam)  for almost 
a year now and I am happy with the results.  It has interchangeable lenses 
fitted to an (almost) regular Minolta AF SLR body (I think it is  450si in the 
US) and so you can have real professional quality images with its macro lens 
(tested to be one of the best SLR macrolenses) and it intergrates well into a 
Mac dominated environment ( I have no opinion for a PC-version). The pictures 
are transferred to a Power Mac using the software coming with the camera and 
then processed with Adobe Photoshop, and finally stored on 128 or 230 MB 
optical disks.  We got ours with two lenses, a flash and some small 
accessories (like orange and  yellow filters, for EtBr and SYBR Green, resp.) 
for just under $10000 as an intro offer. I think it is the only 
semi-professional compromise at the moment.  Cheaper cameras have a limited 
close-up range and mediocre lenses, and thus are not much use in the lab,  
better cameras  are way too expensive. It may help browsing the last five 
issues of any photographic magazine to get  an overview: these cameras are 
"hot" on the  consumermarket right now.


From: roman at bcm.tmc.edu (Greg Roman)

About two years ago we purchased a Kodak DCS 420 with IP labs software for 
about $10K (No Affliations with Cited Manufactors).  This Camera has a Nikon 
N-90 body and easily mounts our Axiophot. We have a semi-dedicated PowerPC to 
handle the images with Photoshop (128 Meg RAM; Radius monitor and 24 bit vid 
card etc...).  The images can be beautiful, and printed on dye-subs without 
loss of resolution. Definitely publication quality.


From: Ralf Baumeister <bmeister at lmb.uni-muenchen.de>

We are also looking into this issue at the moment. The best cameras we have 
seen so far is from Princeton Instruments, a cooled -20 deg 12 bit camera, in 
Germany named MicroMAX Digital TE/CCD-1400 Grade 1. The price in Germany is 
pretty high, about $25,000. The cooled cameras have much less background and 
are definitely worth their high price. The disadvantage with the 12 bit 
cameras is, that their very high resolution slows down the data transmission  
(about 5-10 seconds per picture).
Therefore, you cannot use them for videotaping worm movements. 

From: Susan Mango <smango at genetics.utah.edu>

I've been looking at the Micromax from PI recently. What I liked about it was 
a higher QE for GFP (around 40%) and a lower price tag than some of the other 
cooled digital cameras ($19,500). Although it's only a 1 MHz camera, it 
actually seemed to be reasonably fast. I'm trying to decide between that and 
an analog system (e.g. Hamamatsu C2400) with a good processor. Some people 
swear you can't get as good a Nomarksi image digitally as you can with analog. 



From: Michael Herman <mherman at ksu.edu>

I am also looking into digital cameras.  Right now I have one pretty good 
resource, it's an article my Zeiss rep gave me that was in the April 1995 
American Laboratory: "A guide to selecting electronic cameras for light 
microscope-based imaging."  I've also heard that a large S/N ratio, 55-58 dB, 
is important and to buy as many bits as you can afford.