Nikon D1: FireWire Speed Testing
Bertho Boman

To be able to develop an efficient workflow, it is important to understand the limitations and features of the equipment and then optimize the procedure for best performance. For a photographer that has to go through a thousand photos, a 3.6 second speedup per processed picture will save one hour's worth of time. Being an amateur, I do not have that problem, but I am inpatient and do not want to wait any longer than necessary, and I like things to be efficient.

The first two measurement procedures evaluate the system using the D1 tethered to a PC or laptop computer while taking pictures. This is the typical mode of a product, or possibly a portrait photographer, and other persons that can have the camera tethered. Take one picture and evaluate it on the PC before taking the next one. The last measurement procedure evaluates Compact Flash card downloading via FireWire.

Test setup and equipment used:
Nikon D1. Manual focus, manual exposure time set to 1/125, NEF, 10 shot burst mode

Compact Flash Cards: Lexar 10X 160 MB and Delkins 32MB

OrangeLink "FireWire Cardbus PC Card" (PCMCIA)

Dell Inspiron 7000 laptop with a 300 MHz Pentium 2 processor and 128 MB RAM

Software: Windows 98SE, Nikon Capture 1.12, No curves or autosave.

All timing was done with a stopwatch with multiple lap splits and typically it was measured 10 to 15 times and the data was verified to be repeatable. Nikon's NEF file format was used for all measurements.

Note: There should absolutely not be any delay caused by the FireWire interface limiting the data rate. It is rated up to 50 MBytes/second so our 4 MB NEF would take less than 0.1 second if the camera and PC were fast enough.

1. Single Picture Mode:
The time to take a single picture, transfer it to the PC and display it was measured. The average was 14.1seconds. There are many factors affecting this. Here is a much more detailed breakdown on what happens:

Once the picture is taken, there is an initial delay until the transfer starts up. Most of that delay is presumably caused by the camera, but some delay is also caused by initial communication between the camera and the PC, that a picture is coming. The time was measured from shutter trigger until the PC started to display "Transferring" in the bottom task bar. For lack of a better name I have called this time "Processing" and it measured 2.7 seconds.

The next delay is the actual file transfer. The time was measured from the above mentioned "Transferring" until the thumbnail was shown. I called this time "Transfer time," and it was very consistent 6.1 seconds.

Afterwards, the PC "thinks" about the received picture. I really do not know what is happening in the software, but there is a well-defined delay of 1.9 seconds before the PC starts to "paint" the picture. Of course, I call this "PC thinking time".

The final delay is the actual "painting" of the picture. The time is called "painting"and typically was 3.4 seconds per picture.

If all delays are added up, we get a total of the 14.1 seconds to take one picture and display it. It is more interesting and educational to split the time in two sections: a camera delay of 8.8 seconds for the first two parts and a PC delay of 5.3 seconds. We cannot do much about the camera delay, but the PC delay depends on what computer is used. The measurement was taken on a reasonably good laptop but it is about a year old so there are better ones available today. A FireWire connection to a high end workstation should cut that delay to a second or so. On the other hand, a slower laptop might be much worse than the 5.3 seconds as tested on my laptop. Try it on your system and measure how long time it takes from thumbnail to a finished picture. That should give a good idea about your system's performance. My laptop is networked to a workstation but that part is not evaluated at this time.

2. Ten shot burst mode:
The time per picture to transfer a set of 10 pictures to the PC was 9.5 seconds per picture.

For this test, I shot a burst of 10 NEFs and started timing at the end of the burst until each of the 10 thumbnails appeared. There was big difference in the thumbnail to thumbnail timing, presumably caused by the Windows operating system that had to stop now and then to save the files on the hard drive. Typically values varied from 7.5 seconds to 12.4 seconds with the average of 9.5 second per picture. On the longer ones, I could hear extra disk activity. Unfortunately, I do not know why there was such difference in picture to picture timing.

The major difference in this operating mode is that we do not have to wait for the "thinking and "painting" of each image. It just transfers all of the NEFs continuously and then displays just the last one. Note, that both the Nikon Capture and the camera need to be set to continuos mode, not just the camera. For a long time, I never realized that, and I have been complaining, while taking studio pictures, about the long wait because the camera buffer did not work.

Ten pictures taken in the first mode require 140 seconds to process versus 95 seconds in continuos mode.

This continuous mode is advantageous to use even if you don't shoot in burst mode. In portrait work, you don't have to wait until the previous picture is done, just click away as long as the camera's buffer is not overflowing.

3. Downloading from Compact Flash:
To understand the system I also needed to find downloading time of Compact Flash Cards in the camera through the FireWire. In this case it was very repeatable, 8.5 seconds per picture with a maximum spread of +/- 0.2 seconds.

I did this test with both my fastest and slowest CF cards and there was no significant difference between them even though there was a two to one speed difference while writing to the cards. Apparently the data can be read from the CF cards faster than the camera can feed it to the FireWire. For CF speed evaluation, Click Here.

Use continuous mode in both camera and Nikon Capture, set the camera to one shot at a time to avoid accidental multi-shots.

The time to take one picture is 14.1 seconds on my system, and the burst time 9.5 seconds per picture. The computer speed can significantly affect the system performance.

Compact Flash transfer time was 8.5 seconds per picture.