Sunday, November 27, 2011

Comparing JPEG Compression Quality - Nikon vs PhotoShop

The Gory Details

Was curious as to how Nikon's D50 in-camera JPEG compression compared to that of PhotoShop Elements 6 (PSE6). According to the D50 manual, Nikon's "Fine" jpeg compression is 1:4 and their "Basic" compression (i.e. smaller file size) is 1:16. I was thinking of using Nikon's "Basic" compression quality to save a few steps in PSE6 when making images for auctions or quick forum posts. So, I took two identical shots of the trusty Citizen Wingman, with the only difference being the selected image size: everything else set manually, and no change in lighting. Here they are both, "Fine" image on top, click either one to see it's full-size image.

The above are the images as delivered by the camera without any editing other than cropping and re-sizing to act as clickable thumbnails. As you can see, they look much of a muchness. However, at the pixel level, a different story unfolds . .

Above is the logo from the Fine image, blown up 16X time fullsize in PSE6 (but not resized, I used Windows screen capture to get the image). Not bad, really. The Basic image detail below, however, came as a bit of a shock! (scroll up and down to compare the two).

The color rendition of the logo is pretty poor and there may even be an artifact or two to be seen. Gone is any notion of cutting out jpeg compression from my auction image workflow in PhotoShop!

Next, out of interest, I used PSE6 jpeg compression (level 7, whatever that means) on the Fine image and, for a somewhat smaller resultant file size than that given by Nikon's "Basic", got this:

Pretty good, I thought. So, at least in the compression department, PhotoShop beats Nikon handsomely, IMHO.

Best regards, xpatUSA

Friday, November 4, 2011

Nice Lady's Antique Wristwatch

1900's Stadler Mabel Enameled Solid Silver Lady's Watch
Front View
White porcelain dial with no hairline cracks. Breguet-style numerals with red 12. Classic poire shaped hands. Sunken seconds sub-dial. Purple enameled bezel. Clear mineral glass crystal with no dings or major scratches.
Inside View
Fine Swiss hallmarked 0.935 silver case by the Stadler Watch Co. in Solothurn, Switzerland sometime between 1916 and 1930. The silver used is a little finer than Sterling which is itself only 0.925.
Back View
Solid silver snap-on back with purple basse-taille enamel. Basse taille is the rarer technique of first engraving the metal and then firing a smooth coat of glass enamel on top. This makes the engraving stand out with much nicer reflections and eliminates cleaning wear on the engravure.
Typical Swiss hand-decorated 7-jewel movement, also by Stadler. Sets correctly, runs and keeps good time for an antique watch. The silver wire lugs were straightened after this picture was taken.
Side View
Enameled silver back and bezel snap firmly on to the case. Comes with a 10mm lightly-used Swiss oiled leather band with matching silver-tone buckle. Onion-style crown makes the watch easy to wind and set. Watch is 27½mm diameter, excluding the crown. 9mm thick, from front to back, and weighs 22½ grams (¾ oz),including the band.
Best regards, xpatUSA

Thursday, September 22, 2011

Watch Valuation the Hard Way

You want to sell a watch but you're not sure what to ask for it. Too high - it may not sell; too low - you lose out. If you have the time, here's a scientific way to figure a market value.

First go to your country's eBay and do a search for the same model as your watch. Then, staying in that search, click on the checkbox that says "completed listing only" or words to that effect. Now start writing down all the prices where the watch actually sold (at auction, not buy-it-now). You might like to weed out prices for watches that are not your actual model or are not in a similar condition as yours. But a little diversity can be beneficial, for example when there are only a few watches of your exact model, dial color, etc. Something like 20 prices should do it, more if you have the patience. Less than 6 prices would only give you a rough idea.

Now you select some price ranges, somewhere between 10 and 20% of the total range of the prices that you found. For example, if the "spread" (max minus min) is 600-100 = $500 and you found a good few sold items then your ranges could be 500/10 = $50, i.e. 100 to 149, 150 to 199, and so on up to 600-649 (just in case a watch was actually $600, if not - the last range would be 550-599).

Next you count up the prices that fall within each range and list them with an "X" for each watch found:
You can see that the 'X's form a horizontal bar chart which shows quite clearly that the range with the most sales is the $400-449 range. The other ranges indicate either worse or better condition, if the market***  is to be believed. So, if you consider your watch to be in average condition, a selling price in the $400-449 range would give you a reasonable chance of a sale. The higher ranges would be appropriate for a better condition and vice-versa. The shape of this particular  bar chart tells you more about the model of watch. The majority of the 'X's are at the higher end of the total price, indicating a sought-after model for whatever reason - quality, name, rarity, etc.

***Remember, "the market" in this example is eBay in the country or countries where you intend to sell. If instead you had researched prices obtained by Retailers in the open market, prices would be higher and less realizable in a different market.

Here's some results from a search for Luminox "Navy Seals". About 20 results were found, of which 6 were rejected for various reasons - like, one item was a Luminox box but no watch.

Column A shows the prices found on eBay, sold recently, and searched for as Luminox "Navy Seals". The average of the 14 prices is shown in row 16. At $119, this value falls into price range 3 (100-129) but the chart shows that the average is quite a bit higher than price range 2 (70-99) which has the highest count. This is because the chart values are "skewed" toward the lower end of the total price range, indicating a much less well-regarded watch than our other example above, The chart is telling you that, for a watch in average condition, a selling price selected from price range 2 would have a better chance of selling. It is also implying that simply taking the average of the prices as the sole criterion for your sales price would have put you too high. Worse, had you been selling an expensive, high quality watch, your price could have been 10-20% too low.

Row 17 shows a price of $110 as the "median" a word much-loved by statisticians and much avoided by the rest of us. However, this median value is a lot closer to price range 2 than the average is, so there must be something to it. Having said that, the median is not easy to figure out unless you use a spreadsheet function like in the above. Simplistically, it's a value which is in the "middle" of an ascending range of prices. With the 14 prices as shown above, it lies somewhere between the 7th and 8th price, i.e. the median value is between $95 and $125, QED.

In conclusion, it's worth at least picking some ranges and filling in the 'X's. Picking the longest bar gets you closer to establish a good criterion by which you can judge and price your own watch for the market. Simply figuring the average from the prices can throw you off quite a bit, either way. Doing a spread-sheet is fun for geeks but is quite un-necessary.

Best regards, xpatUSA

Saturday, September 17, 2011 Web Site Updated

In the beginning, there was simple HTML which stands for HyperText Markup Language. Couldn't do a lot with it but it was nice to be able to include "hyper" links to related documents or to be able to link to other places within a long document. The ability to link to pictures stored elsewhere was especially welcome as means of creating an illustrated document without having to learn how to "embed" pictures or drawings in it.

Following on from those good old days, HTML has developed beyond all recognition to the point of being almost unintelligible. Not to mention the emergence of XHMTL with it's Draconian rules and it's incestuous relationship to XML, XSL. There also appeared scripting languages such as JavaScript, which allow a page to be messed with once your Browser of choice has downloaded it and displayed it on your screen.

One good development was something called CSS which allowed you to write up your subject and then to apply rules separately on how it should look. It was also a pleasure to discover that HTML5 has appeared and sounded the death knell for XHTML in all it's horrible varieties.

Meanwhile, like the good techie that I am, I had developed my website into a paragon of scripting such that you couldn't even view my pages properly without having JavaScript enabled in your browser. That meant that every bit of content showed by JavaScript in my pages had to be repeated as plain old HTML with "NOSCRIPT" tags around it - double the pain for no gain.

"Enough already" as they say. I've just changed to HTML5 for my collection pages. Completely separate CSS files now say how they look. Any scripting is used only for minor functions, like "back to the top" links which have little effect if they don't work.

I also took the opportunity to show the collection in thumb-nail format with each thumb-nail being a clickable link to a page showing bigger pics and more information. Previous visitors will notice a simpler horizontal navigation bar up top instead of at the side. Makes room for more watches!!

So, click here and see how it looks now.

Best regards, xpatUSA

Friday, August 12, 2011

An Experiment in Oil Migration on Metal Surfaces

I've often read that oil can spread around inside a watch and that proper watch oils are designed to "stay in place". I was a little skeptical that oil could move around all by itself until I tried an experiment involving a comparison between Moebius 9010 (a Swiss oil designed for general use in wristwatches) and "Mobil 1" 5W-30 (an automotive engine oil). Here are the results:

The areas of interest are indicated by purple arrows. At left, a test to see if an excessively-oiled metal-to-metal bearing would leak oil out onto the surface surrounding the bearing hole. At right, a test to see if drops of oil deposited onto a metal surface would migrate (move or spread). The barrel at right was mounted vertically to show any effect due to gravity.

This is a shot of the comparative test set-up after 36 hrs. The Moebius oil drop at right has stayed in place with a very slight spread delineated by the purple lines. The Mobil oil at left shows some unexpected behavior. The main part of the oil drop has sagged downward due to gravity, a lower surface tension, and a somewhat larger initial drop size than that of the Moebius oil. However, a thin film has formed and continued to grow, apparently following the circular machining marks on the metal surface. It appears that gravity has aided the migration of the lower part of the film. There remains a line of demarcation between the film and the edge of the original drop.

This is an external shot, after 17 hours, of the barrel arbor bearing hole, the inside of which was oiled excessively with the Mobil 1 automotive oil. The purpose of this test was to see if any oil is able to exit by overcoming the capillary force induced by the bearing clearance gap. It is seen that no spreading has occurred so far, although the oil "ring" itself is clearly visible. The small blob of Mobil 1 at left was added later to test spreading of a drop on a horizontal, smoother surface.
I have drawn no firm conclusions yet and am still in research mode. Therefore, conclusions will be published later as additions to this post.

Best regards,


Saturday, July 30, 2011

Domed versus Flat Crystals

Flat crystals have their place in the scheme of things, I suppose, but the one annoying thing is reflections - especially from nearby lamps. By comparison, a domed crystal makes a reflection smaller and, if the underside is equally curved, it's reflection is also smaller. For example, this Stocker & Yale model 490's flat glass crystal shows no mercy when it's under a desk lamp.
The standard flat crystal on this watch is 30mm diameter and 1.5mm thick. A 30mm, 2mm thick, "double domed" mineral glass crystal was purchased for $7, installed, and the result was excellent:
The two pictures were taken at the same distance and in the same position. The lamp distance was about a foot above the watch. The domed crystal has an unpolished chamfer on the top and the crystal projects about a 1/2 mm from the anodised aluminum flat-topped bezel:
The side view is much improved. The angularity of the bezel is relieved by the domes' curve, which itself complements the curved chamfers formed on the side of the plastic body.
 Was thinking of selling this watch, but now it's keeper!!
Best regards, xpatUSA

Tuesday, June 14, 2011

Exposing the Lume

Some watch enthusiasts are proud of how bright the luminous paint is on their watches. They publish highly edited photos or use heavy exposure settings and then post a "lume shot" with the full intention of impressing the reader. Others, of a more pedantic persuasion, might wonder exactly how bright their lume is, as measured in (usually) milli-candelas per square meter i.e. mcd/m2. And not all watches are that bright - for example, older military watches that have the tritium paint can be quite dim after 20+ years. Like this 25-year old Altus:
Not quite up to the blinding images usually posted on watch fora by lume freaks, even though it's exposure was increased by 4 stops in PhotoShop! So, how could we measure the actual brightness of the hands and markers of this watch? There are several ways, some more technical than others - but in this post we will examine how a good camera might do the job, the example chosen being, of course,  my trusty Nikon D50 armed with a Nikkor 60mm f/2.8D Macro lens.
At first, it seemed easy . . just take a pic, note the camera's exposure settings and do a quick calculation of some sort. Note the vagueness implied by "some sort". (Like: - to get to the moon, we'll build "some sort of rocket"). The first problem was that the exposure meter found it quite difficult to measure what was, to it, almost total darkness. It was also difficult to mess with the resulting image so as to find the relative area of the fuzzy luminous paint to the surrounding area of darkness. It could all be done but it wasn't much fun ;-), so much not fun that I gave up on that approach.
Then it was thought that using the "spot" metering mode would allow a direct measurement without even having to take a pic. Fine in theory until it was found that watch hands aren't generally wide enough to fill the spot metering area, even with the lens set to maximum close-up. (The D50's spot metering field of view is two degrees which translates to 3.5mm at maximum close-up setting). Attempts to interpose magnifying lenses failed miserably due to a) focusing problems and b) a much dimmer image, it seemed. Therefore I gave up on that approach, too, although it did work on a WWII aircraft clock that I have by my bed.
So an approach was sought that involved no particular camera exposure setting and brought into use a picture's image pixel brightness but without those tedious lume versus total area calculations or equivalent editor trickery. The thought was that, if an image pixel brightness was known, then it should be possible to calculate the exposure at the sensor that resulted in such a pixel brightness.
It's quite hard to find detailed information about cameras. The great majority of websites only skim the surface of the subject, carefully avoiding any mention of scientific units, performance data, etc., that would actually be helpful in this task! However, some individuals such as Doug Kerr have written useful articles aimed at  less knowledgeable folks such as myself. With Doug's help, the following method was developed  . . .
Step 1)
As Mrs Beeton might have said "first, take your picture". Armed with the D50, a tripod and a completely dark room, I took this rather unimpressive shot:
As can been seen, the lume is not real bright on this 25-year old military watch. The original shot was taken in the RAW (Nikon's .nef) format, with the camera set manually to an arbitrary f/5.0, 13 seconds at ISO 400. The file was opened in PhotoShop Elements and everything that the camera had done was reset to zero. The white balance was reset to 6500K with no tinting. At this point, the on-screen image was color-picked within the hour hand area, and RGB readings of 5, 31, 6 were noted.
Step 2)
Now we estimate what photometric exposure  Ho  (i.e. at the face of the camera's CCD sensor) was necessary to cause the RGB readings noted above. Doug provided a spreadsheet which produced a relative luminance factor (y) which could be applied to the maximum or saturated exposure Hsat in order to determine the exposure Ho caused by the lume. Hsat  is derived from ISO 12232:2006 as being: Hsat = 55.56/Si, where Si is the camera's ISO setting. At a setting of ISO 400, Hsat comes out as 0.139 lux-seconds (lx.s). Cameras can vary in their implementation of the ISO standard, so 55.56 may not be correct for your camera . Running Doug's spreadsheet with RGB = 5, 31, 6 plugged in gave a value for the relative luminance as y = 0.0103. Therefore,  Ho =  Hsat times y = 0.139 x 0.0103 = 0.001431 lx.s.
Step 3)
Having found the photometric exposure at the face of the sensor, and knowing the camera settings, it is now possible to to work a photometric equation backwards to thereby find the object's luminance  Lo . The particular equation that relates Ho to Lo is:

Ho = 4/pi * t/N^2 * L* q  where t = exposure time, N = aperture setting (f/N),
and q = T * Vθ * cos^4(θ) where T = lens transmittance, V = vignetting factor and θ = the angle of the object from the lens' axis. An arbitrary value of 0.88 for q may well suffice if the object is less than 10 degrees off-axis.
The equation can be re-arranged to provided our desired result, namely  the object's luminance Lo   in cd/m^2:
Lo =  H/ (4/pi * t/N^2 * q)
Substituting Ho = 0.002575, t = 13, N = 5 and q = 0.88 gives  Lo = 0.0024561 cd/m^2
Therefore, the luminance of the watch lume is 2.46 mcd/m^2
For those who are not into heavy calculations like the above, I've added them to Doug's spreadsheet here.
Here's a shot of a Marathon military watch I have. It has 12-yr old tritium vial markers and does not need to be "charged up" under a lamp. It still glows well in the the dark.
Best regards,  xpatUSA

Sunday, May 29, 2011

SandY Seconds Hand - is it balanced?

The later Stocker & Yale military watches had this ugly seconds hand with a huge paddle on it:
I've read that the paddle acts as a counter-balance to the much longer arm with the arrow on it - but you can't really judge by just looking at it. However, it can be calculated if you know the areas of the various geometrical figures that go to make up the seconds hand's outline.
So I took a macro photo of the hand, brought it into PhotoShop and measured the needed dimensions (in pixels). Then, like a pilot checks the Weight and Balance of an aeroplane before taking off, I calculated the moments of area for each arm of the seconds hand. It's OK to use area instead of weight because the hand is made of thin sheet steel which means that, if the areas balance, so will the weights. The units end up as pixels cubed - because area was measured from the image as pixels squared and the distance of each geometrical figure from the center was also in pixels. Ordinarily, moments are given in units of force x distance - for example Newton-meters.
The paddle side came out at 1.73 Mpx3 and the arrow side came out at 3.29 Mpx3, almost twice as much. So, it's not balanced, even though it looks like it. The specification for the movement, a Ronda 715, states that the out-of-balance for the seconds hand is to be less than 0.04 μNm. So, some unbalance is allowed but, to check compliance, it would be necessary to use real units of mass instead of the above pixel-area short cut. I'll do it one day, just out of curiosity.
Best regards, xpatUSA

Thursday, April 28, 2011

Smooth Sweeping Seconds - an alternative

Many of us like to see a smoothly-moving seconds hand, as opposed to that of a quartz watch which clunks it's way around the dial, only occasionally aligning with the marks thereon. And a good few of us think that the higher the beat rate the smoother it will be - which is true to an extent, but why? One could be forgiven for thinking that our so-called "persistence of vision" has something to do with it but, sorry to say, that is not the case. There might be a smooth running appearance if the beat was over 16 per second or even 24 per second like old movies but that would mean a beat rate of 86,400 bph!!
Yep, the hand does seem to run smoother on a 28,800 bph watch but only because it moves in smaller steps and takes more steps to go once round the dial. You can still see the steps, though. That got me to wondering how small a step would give apparently smooth motion. Well, apart from persistence of vision, another property of vision is that "visual acuity". Put simply, this is the ability of the eye and brain to distinguish between two objects that are very close together. Like the bars on the letter "E" when you visit your optician or optometrist.
Apparently most folks can distinguish objects quite easily if they are more that a certain angle apart. By "angle", I mean relative to the eyeball. That magic angle is 1/60 of one degree, also known as 1 minute of arc. At 20 feet distance, it is no coincidence that a person with 20/20 vision (6/6 in metric countries) should be able to read an "E" that is 0.35" tall at that distance because it is apparently 5 minutes of arc in size and each bar or gap is 1 minute of arc as seen by the eye.
However, when we glance at our watch, it's a little closer than 20 feet! Switching to metric, let's say 300mm, just less than a foot. This is the only information we need in order to calculate how far apart are two successive step positions of the tip of the hand that can be easily distinguished. Trigonometry tells us that the distance is 0.087mm. Less than a tenth of a millimeter!
Next, let's have a sweep seconds hand that is 13mm long from center to the tip. More trigonometry tells us that the tip of this hand sweeps a circumference of 81.681mm every minute. From here, we can go several ways but it might be instructive to calculate the beat rate in bph for the step distance of 0.087mm established in the preceding paragraph. It comes to 938 beats/min which is 56,267 bph. This tells us that, even at this very high beat rate, a person with normal vision will still see the seconds hand ticking!
Well now, what about those old-fashioned sub-seconds hands? They seem to run smoother too, huh? I have a watch with a sub-seconds hand 3mm long. Ergo, it sweeps a circumference of 18.85mm. That works out, for the same step distance, to only 13,000 bph. The watch actually ticks at 15,000 bph, so it is no surprise that the hand motion (to my less that perfect eyes) seems quite smooth with no apparent ticking discernible.

Best regards, xpatUSA