Stephen White along today with his apprentice Mitch, primarily to fix the issues with the
RCD and UPS. While they were
at it, they also installed a couple of outside power points and took a look at the antenna.
To my surprise, there's quite a complicated masthead amplifier up there:
Stephen is not an antenna expert, and he doesn't have any signal strength equipment, but with the help of my Zuiko Digital
ED 70-300mm f/4.0-5.6 I was able to get enough of the amplifier to identify it. It's
a DIGIMATCH 10MM-UA30P:
That's quite an impressive device. I've complained about Jim Lannen in the past, and
there's another complaint coming below, but he certainly didn't cut corners with that
amplifier. A pity the reception is still so bad.
Unfortunately the web site doesn't want to know about it, presumably because it's last
year's model. But they have a very similar model, the 10MM-UA30PDF:
Even from the original, it's clear that it has adjustable gain. It also has a dedicated FM
output on the left, not connected. No wonder we have no radio reception. I suppose I'll
have to get a dedicated antenna specialist in to fix things up.
Apart from that, they came up with a good solution for the UPS. Despite my concerns, Jim
had laid a conduit between the shed and the garage that could only hold two cables, and
there was no space for a third. How do we separate the UPS and the other consumers in the
garage? I had thought of relocating the UPS into the garage, but Mitch, the apprentice,
came up with the idea of an RCD mounted inside a power point, so that could work for the
rest of the shed, while the UPS remains in place without an RCD in front of it. Total time
10 minutes. He seems a brighter spark(y) than most.
So Jim finished the power point installation with the connecting cable not connected. How
could he have missed that? I find it completely amazing. You could almost think that he
did it deliberately.
And finally, another small mercy: two switches in the toilet, one for the lights, one for
the (very noisy; thanks, Jim) fan. What a relief!
My photos of the antenna were taken with a couple of telephoto lenses. The second was taken
with the Zuiko Digital
ED 70-300mm f/4.0-5.6, severely reshaped and cropped:
The “equivalent focal length” (the focal length that would be needed to fill the frame with
that image) of the last image (natural size on the web page) is 4608 mm. This is an Olympus OM-D E-M1, so the
corresponding focal length for a full-frame camera would be 9216 mm. It's amazing that I
can read anything at all.
Finally decided to buy the photo stacking
software that I discussed a
couple of days ago. The good news: since I'm not in Europe, I don't pay
German Value-added tax (currently
19%) so instead of 69 € I only had to pay 57,98 €. And while paying I saw this:
That should be translated as “After confirming your order, you will be transferred to a
protected web site, where you can enter your credit card details”. But there's a typo in
the text: “protected” is „geschützt“, and not „geschätzt“, which means “valued”. Mildly
amusing in itself, but there's a story behind that: decades ago Fritz „Fliege“
Jörn, Tandem's publicity man,
wrote his texts so that they could be printed
on Qume Sprint\5 daisy wheel printers, which
used non-standard wheels where ä and ü were transposed compared to the then-standard. And
the funniest misunderstanding was exactly geschätzt/geschützt.
Mail from Adam Kranzel today about DDR3 RAM
compatibility. He pointed at this page, which describes “high density” and “low density” RAM configurations. I
had found many more. This one explains the difference:
All low density 1GB modules are made with 16 chips (8 chips on each side) using 64Mx8
device. All high density 1GB modules are made with 16 chips (8 chips on each side) using
128Mx4 device.
1 GB? Who uses 1 GB DIMMs any more? And there's the clue. Adam's reference was written 7
years ago, and here I read:
It may have something to do with memory density, but that's an old, old, old issue from a
few years ago and hasn't actually been an issue in a few generations.
That was written two years ago. It seems that the oldest Intel chipsets that supported
DDR3, such as the P35 Express, had
difficulty with high-density RAM configurations. The support page for the
Thinkcentre M71E 3132A8M confirms that it has an H61 Express chipset, which this page tells me was introduced in 2011.
So: so far it seems that there have been issues with Intel compatibility in the past, but
they're no longer relevant. Am I right?
Three days on and I still haven't made up my mind about new memory for dischord. Got
an email from Tim Bishop, who pointed me at this site selling DRAM specifically for my machine. I suppose that's some kind of
guarantee, but the site has a big problem: no 8 GB modules.
More reading brought me to this page, really relating to MSI products. But this got me thinking:
High density RAM is usually very slow anyway and are typically lower Binned chips too
hence they are slow and tend to have a very bad compatibility rate! (stick to X8 types 8
chips per side 16 total low density units)
That makes sense. The “density” (chip configuration) in itself doesn't make the chips any
more or less expensive. There must be some other reason why the “high density” chips are
cheaper, and speed or reliability are obvious candidates. And none of these offerings
specify the timing parameters. So independently of whether the cheap chips are compatible,
it seems reasonable to spend a few dollars more and buy reputable chips.
What choice do I have? Checking various sources, I have:
On the face of it, Megabuy offers the best price, and that would be true if I could go into
the shop and buy it. The same applies to MSY, whom I passed in Geelong only 2 days ago, but
I've stopped buying from them anyway. But Megabuy price themselves out of the market with
their ridiculous freight charges. For a couple of DIMMs it should be under $2. And the
last one, from eBay seller fr2242, is a well-known name and comes with $3 off the second
chip, so it's only $10 more expensive than the cheapest offering. It also does no harm that
the seller has 100% positive feedback, so I bought from them, which should also have the
advantage that it'll be here early next week.
There are other Australian suppliers like Megabuy. GameDude had marginally lower prices. But to find out
the shipping, I had to enter my complete billing information and navigate
a CAPTCHA. Sorry, people, if you want my
business you shouldn't go out of your way to make it difficult for me.
Off with Yvonne and the dogs
to Ballarat today, I with Nikolai to the Eureka Village
Hostel. Round again with Linda, and gradually both Niko and I are getting used to the
routine. So are some of the residents. One in “House 1”
(for dementia patients) followed us as
far as she was allowed before being escorted back; another (Bill) accompanied us the rest of
the visit.
Two weeks ago there were
fewer people there than four weeks
ago; today there were even fewer. This time Yvonne had fewer too, and we were both
finished after 45 minutes. Took the dogs back to Bill and a couple of smokers who were
sitting outside; it's nice to see just how much it makes their day.
Lately I've been subscribing to the RSS feed
of 43rumors, mainly to keep up to date with
new photographic equipment. But today I got another article claiming that DxO had released a new version of
DxO Optics “Pro”. If that were the case, I should have heard about it from DxO. Still, there's an easy
enough way to check—just ask the running program:
That's clear and straightforward enough. And wrong. Going to my customer account, sure
enough, there's a new version, and I can get it for a reasonable price. And it seems that
far from being a rumour, 43rumors are behind the times. I had already received mail from
dpreview, but deleted it because it's HTML
only. It contains another report with a link to DxO's own announcement. So why didn't DxO tell me? And why does the old version claim to be
up to date?
Should I buy it now? They haven't told me, but it seems there's a $20 discount on it,
probably to the end of the month. But I'll wait a week or two to see if they come up with
further discounts. In the meantime I can use it without restrictions for a month.
Downloading failed again! Clearly this is at least the fault of the web browser—I
should try ftp next time—but it happens continually with DxO. Again I downloaded it
to www.lemis.com and copied it from there. I wonder why it keeps failing.
Upgrades can be fun. Where did all my settings go? In fact it kept most of them, but I
still had to copy some configuration information, and there's a good chance I have missed
something. They claim that their noise reduction system (“PRIME”) is now “a new version of
PRIME that is better and faster than ever.” I suppose everything is relative, but
on dischord the current version takes about a minute per image. Possibly to
emphasize the speedup, they've changed their startup slogan from “Pushing the limits of
your patience camera” to “Reveal the RAW emotion”. Ugh.
Edwin Groothuis wants to cook a rendang,
something that has never been completely successful for me. Off to check alternative
recipes and found this one in “500 resep lazaaat makanan Indonesia”:
Preparing for dinner tonight, Yvonne sent me a link to a
recipe on Buzzfeed. I hardly bother with recipes on the web; so
few of them seem worthwhile. This one wasn't too bad, but not what I was looking for. But
in the noise around the recipe I saw a link “Can We Guess Your Exact Age With
These Food Questions?. That's so stupid that I had to try it. Basically it was a
test of my (almost non-existent) fast food preferences. Which pizza? Which burrito? Which
doughnut (no option to say “I don't like doughnuts”, and no plain doughnut)? Only one sane
choice: “how do you like your steak cooked?”.
And the result? My exact age is 31. But the funny thing is that the user comments show
that, though it was no more accurate for anybody else, most responders took the blame on
themselves rather than on the test. The astrology syndrome again, I suppose.
The lounge room is the second, and though it looks spacious in that image, it's quite
cramped. On the other hand, the dining room (which, confusingly, I've called lounge room;
maybe an omen) has plenty of space. So: time to change. It shouldn't be too difficult, but
we're going to need some wiring changes for power, antenna and network.
Nadirs revisited
Topic: photography, Stones Road house, general, opinion
Before rearranging the furniture, of course, I need to take some “before” photos.
Preferably a 360° panorama, something that has always been an issue because of issues with
the nadir. Lately I've had seen another
idea on the web, though I forget the link: hang
a plumb bob from the tripod and
rearrange the camera so that it's pointing down from the same location, and use a separate
tripod to mount the camera pointing down:
The important thing here is that the height of the mounting plate on the second tripod (on
the right) should be the same as the height of the entrance pupil (the optical axis when the
lens is positioned horizontally). After taking the “normal” images, mark the position of
the bob. Then place the lens assembly on the other tripod, hang the bob from it, and
position:
There are two obvious issues here: the bob doesn't hang from the optical centre in either
case. It's offset from the vertical axis of the tripod by 30 mm, and from the lens axis on
the second tripod by 60 mm in one direction and 10 mm in the other. So I need to position
the bob accordingly the second time round. The end of the remote control or tape is the
point where the bob hung from the first tripod, and I've oriented things so that the tape
goes under the real optical axis (at 3 cm length):
I need two images pointing down to be able to remove the tripod legs. Both cover the area
directly below the camera, so I can leave the tape measure there for next time:
With that, I had enough images to stitch my panoramas. Unfortunately, it didn't work as
well as I had hoped. Somehow I misjudged the relative positions of the tripod legs, and
they overlapped in places. But even apart from that, the control point errors were much
larger than I had expected. Have I made a misassumption somewhere? Until I can get
something to replace SaladoPlayer (now apparently defunct), I can't display the panoramas
anyway.
We don't eat much pork, but recently Margaret Swan gave Chris Bahlo a roasting joint of
pork, all 2.7 kg of it. I can't recall ever having cooked a normal joint of roast pork
before, and off to look for instructions, which varied widely. The big issue is cooking
temperature, of course. For once
Stephanie Alexander'sThe
cook's companion looked the most plausible: cook for 30 minutes per 500 g (why not 60
minutes per kilogram?) until the inside temperature reaches 65-71°. That seemed a wide
range, and on the low side. My own cooking times
page stated 78°. By contrast, “Joy of Cooking” suggests much
higher temperatures, 170° or 185°. That's these
horrible Fahrenheit units, of course,
and the conversion requirements make it really difficult to compare. But here goes, after
modernization:
Source
Cut
Start
Switch
Bake
End
Time/kg
temperature
after
temperature
temperature
Joy of Cooking
Loin
230°
0 min
160°
77°
55 - 77
Joy of Cooking
Shoulder
230°
0 min
160°
85°
55 - 77
Cook's Companion
Loin
220°
20 min
160°
65°
60
Cook's Companion
Shoulder
220°
20 min
180°
71°
60
my page
78°
That's a lot easier to read than the originals. In particular, Stephanie Alexander writes
for shoulder and adds the difference for loin as an afterthought. And I didn't know what my
cut was, though in retrospect it was almost certainly loin. So I went with Stephanie
Alexander for the times, my page for the temperature. Things didn't quite work the way I
expected. After about 10 minutes the skin had already browned considerably, and there it
was smoking. I carried on as advised, but after about an hour it was clear that the skin
was browning too much, so I turned the temperature down to 170°.
According to the cooking times specified, the joint should have required 2 hours, 42
minutes. In fact it was done after 2 hours. I left it in the oven to rest, where the
temperature continued to rise—as you'd expect—and finally reached 83°.
And the taste? Excellent! I even got Chris to carve, which she did better than I could
probably have done:
But I think it had probably had more cooking that was good for it. Was that because of the
continued temperature rise? That was primarily due to the cut of meat and its size, not to
the fact that it was in the (cooling) oven. Should I aim to cook big joints to lower
temperatures? I wonder when I'll have the occasion to try again.
Today was one of those horrible dark, dreary, moist days that I so hated in Europe, and that
happen so seldom in Australia. But it seems I was the lucky one: no power failures
or RCD issues for me.
But the weather in New South
Wales was different, and Andy Snow had both (an extended) power failure and fallen
trees, and Callum Gibson had an RCD that kept tripping.
Hugin 2016.2.0 beta has been
released. Today tried upgrading the FreeBSD port. Not as simple as it seems:
.../hugin-2016.2.0/src/hugin1/ptbatcher/FindPanoDialog.cpp:444:25: error: cannot initialize a variable of type 'int' with an rvalue of type 'void'
int newItem = m_list_pano->Insert(m_panos[selectedPano + 1]->GetItemString(m_start_dir), selectedPano + 1);
How I hate messages like that! It's clearly telling me that m_list_pano->Insert() returns void. But what type is
it? m_list_pano isn't even defined in that file! Where is it? Where's
my TAGS file? Ah, need to build it. Time for a function:
Yes, of course, it's obvious that you define program variables in header files.
wxCheckListBox looks like some weird library thing, and Google confirmed that it is
part of
wxWidgets. Lots of nice documentation including a diagram showing inheritance back
from the great-great-great-grandparents, with the arrows pointing backwards for my taste.
But where's the Insert() member? I went through all of them and couldn't find
them.
Never mind, it'll be in the preprocessor output. But this is made
with cmake. How do I find the
compiler invocation? I still don't know. Went back and checked the documentation, which
says “The wxWidgets GUI toolkit version >=2.7.0. 2.9/3.0 is supported.” OK, I have
2.8.2. But clearly there's a 3.0 out there as well, and that's probably what they used for
the build, so off to see how to do that. The ports Makefile contained:
USE_WX= 2.8+
What are the valid values? The Porter's Handbook tells me: 2.4, 2.6 or 2.8. No odd numbers, no 3.0. It also says
that the + at the end means “ascending range”, which I interpret to mean “2.8 or
higher”. OK, let's break the rules:
USE_WX= 3.0+
And how about that, it worked! So I have at least two bugs:
Despite the claims, Hugin 2016.2.0 doesn't (always) build with wxWidgets 2.8.
Off to Ballarat this morning to have a
tooth pulled—the first ever. Not a pleasant thought, especially when Mario told me—only
now!—of the possible side effects. It was tooth 7 on the right side of the upper jaw, and
it seems that the roots were particularly long, leaving little space between the end and
the maxillary sinus. With a bit
of bad luck, the bone in between could get damaged, and a hole develop between sinus and
mouth. As a result, I had a number of things to avoid in the next three weeks: don't
sneeze, don't sniff, don't spit, don't suck, don't...
The pulling was much as I expected; local anaesthesia, of course, and then the nurse held my
head down firmly. Pull. Crack. Root forceps. And it was out. I wanted to
take it with me to get some photos, but it's “biological waste”, and I wasn't allowed to.
One root attached, the other (crack) broken off. And a 3 mm abscess attached at the
end, which is a good sign.
Fortunately I didn't have a hole in the head, but it's still thin, and I need to be
careful. Back home, wasn't in any pain, but I didn't feel too active either, and spent the
rest of the day watching TV.
Bringing up kittens is always a bit unnerving. Rani is continually finding inappropriate things to play with. But today she found
something harmless that kept her busy for a long time:
DxO finally got round to informing me of the new
release of DxO Optics
“Pro” on Saturday, but it got classified as spam. Why? Who knows? Gmail does, maybe, but they're not telling. And yes, the
current price is a special valid until the end of the month.
Should I upgrade? Probably, but not because of the claimed improvements. The claim is that
their noise reduction system (“PRIME”) is now “a new version of PRIME that is better and
faster than ever.” In version 10, on dischord (an Intel
Core i5-2400, CPUMark 5827) it takes about one minute to process a single image. So
far my only attempt took about 60 seconds, not counting the 10 seconds it took for it to
even start to count.
What else do they have? Red-eye reduction seems good, though since I've all but given up on
using on-camera flash, I don't suffer from the problem. Full screen mode? Maybe, but I
haven't tried it yet.
Yvonne tried it later. Not so good. A repeatable cycle
trying to start it: first, the twirling circle started for a second, then stopped again.
Nothing started. Second, it got as far as the splash screen, then stopped. Only on the
third time did it start normally. If this sequence had happened only once, it might be
typical Microsoft gremlins, but twice?
What does that mean? It seems that all images were processed correctly, but DxO was dead in
the water, thus giving me the opportunity to start it again. All in all, not a good
advertisement for DxO.
Yvonne back from shopping today with 80 GB of memory: 16 GB
of DIMMs for dischord and a 64
GB microSDHC card, which must bring the
weight of data down to a new low. It weighs 160 mg, so 1 byte would weigh
2.5 pg. And clearly Andy Tanenbaum's old adage is out of date. Assuming a load of 480 kg (3 million cards or
192 PB), Yvonne could reach a bandwidth
of 4 Pb/s driving her station wagon to Chris Bahlo's round the corner, conveniently exactly
384 seconds away. I'm impressed.
I even managed to find a data sheet for the parts, containing the information:
This document describes ValueRAM's 1G x 64-bit (8GB) DDR3- 1600 CL11 SDRAM (Synchronous
DRAM) 2Rx8, memory module, based on sixteen 512M x 8-bit FBGA components. The SPD is
programmed to JEDEC standard latency DDR3-1600 timing of 11-11-11 at 1.5V.
That's a lot more information than I'd get on the no-name DIMMs. But when I put them in
dischord, it fired up, but there was no display, and after a few seconds I heard
a beep-beep. Then nothing. What was wrong?
Put the old DIMMs back in the box, fired up. Same thing! Surely even defective RAMs can't
kill the machine!
Tried firing up without any memory at all. That produced a very different result,
continuous beeping right from the start. Clearly my problem was something different.
Finally it occurred to me that the machine was on
a KVM that had caused problems in the
past. Connected it directly to the monitor, and sure enough, it went through the startup
sequence and stopped with the message:
Error 0164: Memory size decreased
Press F1 for setup or F2 to continue
That was with the 4 GB memory, so the message made sense. Replaced the 16 GB memory,
powered up and... the same message! Still, all I needed was to continue.
dischord has both VGA and
DVI connections. Maybe it's time to bypass
the KVM (which also confuses the hell out of Microsoft when trying to guess the dimensions
of the monitor (the KVM doesn't pass
the EDID information through).
Now that I have more memory in dischord, I can get back and look at
the focus stacking work I started
last month. On that
occasion I had to stop because of memory concerns. Now I have the memory, and looking at
the Task Manager shows that I did the right thing by buying 16 GB instead of 8 GB: it
immediately used up 10 GB of memory, making it one of the most memory-hungry programs I have
ever seen.
And the results? Here are two images. On the left the original, on the right the results
from Focus
Projects Professional. Run the cursor over either image
to compare with the partner, and click to see a larger version:
