[AstroDave]

I recently acquired a Pentax *istD on Ebay. It was listed as non-working, but the price was right.

Other than a cracked rear display screen cover, however, it is in reasonably good shape and works fine. I suspect the seller did not know how to set the camera to shoot without an A lens. It is also quite fussy about batteries - regular AAs would last for about 3 pictures and I couldn’t get it to function properly on external power. I finally sprang for some Eneloops and all is well.

Ultimately, I hope to remove the Bayer color filter array from the sensor, so I have a proper black & white camera.

Since I may well wind up destroying the camera in that process, I figured I should make some measurements of linearity and shutter speed first, using techniques I’ve developed for my much newer K-3 and K-1 and other cameras that I have. I was not sure how well a ~15 year-old camera would fare when subjected to some rigorous tests.

Quite well, it turns out, although there is a shutter speed anomaly, which persists in recent Pentax DSLRs.


LINEARITY OF THE SENSOR

First, I measured the linearity of the camera sensor response, in the same fashion as I tested my newer cameras ( How Linear Are Some Pentax Camera Sensors? - PentaxForums.com ). Basically, I take RAW frames over a range of shutter speeds for a gray card which has the same illumination throughout the tests. I then use RawDigger software to read out the R, G, and B average values over a small rectangle.

When corrected for true shutter speeds (see below), a plot of sensor value versus exposure length should be a nice straight line. And, indeed it is:



Included in the plot is the result of fitting power law curves to the R,G,B data. If the camera is linear (pixel data values correspond exactly to the amount of light which fell on the sensor during an exposure), the power law exponent should be 1.000. The *istD results are very close - to a per cent or so, and comparable to the fits for the K-3 and K-1. The goodness of fit, as represent by the R-square values on the plot, is very nearly 1.000 for all three colors, which would be the result for a perfect fit.

The data cover shutter speeds from 1/125 sec to 8 seconds, or a range of 1000 in light intensity, equal to 10 stops. So, the elderly *istD is just as linear as its newer brothers.


SHUTTER SPEEDS

I measured the camera shutter speeds using the latest iteration of my LED flasher. This version features 4 8x8 LED sub-arrays, giving a total of 256 LEDs that can be illuminated one by one sequentially. Conceptually, the flasher is quite simple: it uses only 4 integrated circuits: two binary 4 bit counters and a pair of binary-in, 16-out decoders. All those lines out, though, need a lot of wires to the 8x8 displays, resulting in a bit of a ratsnest appearance. Here’s a photo of the flasher system



I’ve also made a short video of the flasher in action, to give a feel for how it actually works. Take a look here: http://photodave.us/ShutterSpeedTesting/LED-Array.mp4

To measure a shutter speed, I set the flash rate for the LEDs so that there should be some expected number of flashes (lit-up LEDs) during a particular exposure. For this array, I typically use a 240 or 250 flash rate. The 250 rate makes, perhaps, the most sense, except that for the 1/3 EV shutter settings, there is a systematic error of +/- 3 per cent, or 7 or 8 flashes. In that case, for the slightly long speeds, 256/257/258 flashes could occur during an exposure, and I would not get a correct count. Thus, I used a setting of 240 counts for the 1/3 EV step measurements.

Here are a couple of shots of the array in action, for a quarter second exposure. There should be 250 LEDs lit up if the shutter speed is correct. In the first shot, there a exactly 6 “missing” LEDs, all in the same column. In the second, there are again 6 missing LEDs, but some are in one column and the rest in the next column (where “next” happens to be in a different 8x8 array, because of how things are wired). The missing LEDs can be anywhere in the entire array (but always in the same column or “adjacent” columns) - that just depends on how the flashes occurred during an exposure.






For all shutter speeds, I assume the mathematically expected shutter speed when I calculate what flash rate to set (see https://www.scantips.com/lights/fstop2.html for more explanation and True Shutter Speeds compared to Nominal Shutter Speeds - Actual Measurements - PentaxForums.com for some of my earlier measurements).

For half-EV steps, each shutter speed should be the square root of 2 bigger or smaller than the adjacent value. A shutter speed of 1 second appears to be generally as expected for all the cameras I have measured (Pentax *ist D, K-3, K-1, and Q7 as well as Canon G15 and Rebel XS). Up to whole-value “standard” speeds of 8 seconds (1, 2, 4, and 8 seconds and reciprocals 1/2, 1/4, 1/8 second), the mathematics is trivial since the factors of 2 are exactly as indicated on our shutter speed displays. However, at, say, 15 second exposure, our “standard” speeds are no longer correct. The true value is 16 seconds (= 2 x 8). For intermediate half-step values, say, 3 for the speed displayed between 2 and 4 seconds, the value really is 2 times the square root of two, or 2*1.4142 = 2.828 seconds (not too far from, but definitely not, 3).

OK, here are the results for the half-EV shutter speeds:



The results are exceedingly close to expected. The whole step speeds are all within 0.5% of expected, corresponding to at most one LED different from expected. (In some cases, I can estimate to half an LED - i.e. an LED is partially lit / less bright than a fully lit up LED.) The half step speeds are comparably close, with perhaps a slight tendency to be slightly longer than expected, but still less than half a percent in error, except for the very shortest speed I measured (“1/45" second - the speed between 1/30 (really 1/32) and 1/60 (really 1/64)).

At even shorter speeds, I see shutter motion effects - the shutter leaves are moving across the LED array at speeds approaching the dot change interval. This effect can be lessened by moving the camera farther from the array, but then the dots get small. A linear (one dimensional) array across the camera frame would give the best results, but I have not bothered to try making one - the wiring would get quite tedious for a row of 256 LEDs!


Now, for some weird results - One Third EV Step Measurements

Now, the shutter speeds should change by the cube root of 2 (very close to 1.256) as we go from one speed to the next, with the “standard” values (1, 2, 4, 8, 15 (really 16!), and reciprocals) still being the values as those for the half-step case.

This where I used an expected value of 240 flashes during an experiment, and the results show why I had to choose a value smaller than 250 - there are very systematic errors in the speeds:




For all the speeds measured, when I measure the next longest exposure above a standard value (i.e. “1.3 seconds"), the measured shutter speed is very nearly 3 percent GREATER/LONGER than it should be!!!!! And, when I go another one third step longer, the values are all about 3 per cent LESS/SHORTER than they should be!

I have no idea why this occurs. I see the same behavior for my latest Pentax DSLRs, the K-3 and K-1. I do NOT see this effect for my Pentax Q7. It has only the 1/3 step option, and the speeds are all quite close to expected. Similarly, for my Canon G15 and a friend’s Canon Rebel XS, the 1/3 step values are all very close to expected.

It seems that some mathematical error has affected the shutter speed calculations from the earliest Pentax DSLRs. When I commented on this situation previously, photoptimist suggested that the camera programmers might have just taken 1.3 as an adequate approximation to the proper cube root of 2. I sorta doubt this, since the error does not quite match: 1.3/cube root 2 = 1.032, whereas my biggest measured/expected values are a bit less than 1.03 . The reciprocal (i.e. 1/1.3), however, 0.969, is rather closer to my measured values. But, could be - Hmmm - were they just lazy (or didn't have a scientific calculator lying around)?

While this error is small (less than 0.05 stop), it just doesn’t seem right! Why is it wrong in some Pentax cameras, but not others (my Q7), and is not a problem in the two Canons I’ve measured? If the camera firmware can accurately calculate speeds for the half steps to better than a per cent, it should not be a challenge to get the third steps equally close.

I suspect this could be correctable in firmware update - just a number stored somewhere.

It would be very interesting for someone with Pentax contacts to inquire as to the origin of this effect - somebody there (perhaps a programmer way back in some dark corner?!) must know about it!

I wonder if this effect occurs in the Pentax film SLRs. Can some one check whether or not these cameras can be set for third-step shutter speed intervals? If so, I think measurements similar to those I did for my ME Super ( https://www.pentaxforums.com/forums/8-pentax-film-slr-discussion/409300-opti...ter-speed.html ) could discern the 3 per cent difference between true and actual speeds.

Turns out, I have an old Pentax SF1 which I acquired on Ebay when I bought it for the lens that it came with. I will try to kludge up some power for it and see what it displays for shutter speeds, and maybe make some laser beam timings. (If anybody wants that camera afterwards, the price will be right: $0.0 plus postage. PM me)


I hope to soon have a presentation of my shutter speed measurements for the newer Pentax cameras.

---

Last edited by AstroDave; 07-14-2020 at 03:48 PM. Reason: fix a reference