Getting Oracle Linux on Amazon EC2

Edit 08-Nov-2017:

As per Oracle's instructions - https://linux.oracle.com/switch/centos/ adding the GPG key is no longer neededthe bad news is - the procedure documented by Oracle throws a "Broken pipe" error, and the OS left in an unusable state (all Yum configuration disappears):



How to fix (basically just nohup the script).

Start with the CentOS 6 (x86_64) - with Updates HVM on the AWS Marketplace (provided by CentOS.org). After logging in (as the centos user) run the following commands as root:

1
2
curl -O https://linux.oracle.com/switch/centos2ol.sh
nohup sh centos2ol.sh &

The script does some package updating that cuts the SSH connection to the EC2 instance, and when the connection gets cut, the script dies in the middle with a "Broken pipe" and leaves the OS in an unusable state. By using nohup we avoid the broken pipe issue.

After the conversion script completes, run the following:

1
yum distro-sync

Step 4 will no longer be required, and you can proceed to install the Oracle software prerequisites.

Old Information Below:

If you feel the need to roll your own Oracle Linux install on Amazon EC2 (since Oracle no longer provides an officially-supported AMI, and you may not be too keen on using one of the community AMI's):

(1) launch an EC2 image using the CentOS 6.5 AMI from the Marketplace (which is free..) then log in as root (not ec2-user)

(2) import the Oracle GPG key
1
2
cd /etc/pki/rpm-gpg/
curl -O https://oss.oracle.com/ol6/RPM-GPG-KEY-oracle

(3) use the Oracle specific rules to convert the CentOS 6 to OL from this documentSpecifically, run the following commands as root:
1
2
curl -O https://linux.oracle.com/switch/centos2ol.sh
sh centos2ol.sh

(4) Synchronize the yum repository
1
yum -y upgrade

Once the command completes, the end-user should have a fully-patched Oracle Linux 6.7.

If you intend to install Oracle Database 11gR2, you can apply all the necessary packages and kernel parameters with this command:
1
yum install oracle-rdbms-server-11gR2-preinstall -y

and if you will install Oracle Database 12c R1, use the following:
1
yum install oracle-rdbms-server-12cR1-preinstall -y

Canon EOS 6D + 35mm f/1.4L vs Fuji XE-2 + 23mm f/2

Introduction

I've been a Canon EOS user for almost twenty years, except for a brief foray with the earlier Pentax DSLR's (K10D and K20D) about ten years ago. My favorite camera for the past few years has been the Canon 6D full-frame, with the spectacular 16-35mm f/4L IS ultrawide zoom.

The 6D and 16-35mm f/4L IS is a large chunk of optics and electronics, and while it's the perfect travel combo, I have always been on the lookout for smaller alternatives. I did have a Panasonic GF2 five years ago (with the 14mm f/2.8 prime) and that experience nearly destroyed my opinion of mirrorless cameras.

Recently I was able to obtain a Fujifilm XE-2 (which is a circa 2013 body, but had a significant firmware upgrade in 2016) and the 23mm f/2 prime lens (equivalent to 35mm on full-frame). After overcoming the sticker shock (the Fuji lenses are almost without exception priced similar to Canon L glass) I figured it would be interesting to compare the XE-2 and the 23mm prime with the Canon equivalent - 6D with the famous 35mm f/1.4L.

I used a tripod, base ISO (100 on the 6D, 200 on the XE-2) on a sunny day. This resulted in 1/4000 second shutter on the 6D. For some reason the XE-2 also wanted 1/4000 ISO at the same aperture levels, even with higher ISO. Here's the entire image (the central and corner areas are highlighted):

Center Performance - Canon

I compared the Canon 35mm f/1.4L at f/2 (one step down) and f/4 (three steps down) with the Fuji 23mm at f/2 (wide-open) and f/4 (two stops down). I also threw in the Canon 16-35mm f/4L IS at 35mm and f/4 (wide-open).

Canon 35mm f/1.4L at f/2 - not very sharp, low contrast, and this is one stop down!

Canon 35mm f/1.4L at f/4 - sharper

Canon 16-35mm f/4L IS at 35mm f/4 (wide-open)

 A few takeaways:
  • the 35L is not sharp! no wonder the Sigma ART has soundly thrashed it
  • even the 16-35mm zoom is equal to or sharper then the 35L (at the same aperture)
  • it is possible that I have a bad copy of the 35L, since I bought it used
  • it's also possible that the 6D is mis-focusing with the 35L
  • seems the 35L's only value is for its bokeh wide-open, since the 16-35mm f/4L IS is equally sharp at f/4 and more versatile

Center Performance - Fuji

Fuji 23mm f/2 at f/2 (wide-open)

Fuji 23mm f/2 at f/4 (two stops down)
Maybe my eyes are fooling me, or it's that magic Fuji X-Trans sensor, but the 23mm f/2 even wide-open beats the Canon 35mm f/1.4Lstopped down to f/2, and in the center, where all lenses perform their best. Also, while the Fuji sharpens up at f/4 compared to f/2, there's not much difference (unlike the Canon, where the improvement going from f/2 and f/4 is quite obvious). I would even say that the Fuji 23mm f/2 beats the Canon ultra-wide zoom, which is a very modern design.

What about the corners?

Corner Performance - Canon

Canon 35mm f/1.4 at f/2 (one stop down) - corner

Canon 35mm f/1.4L at f/4 (three stops down) - corner

Canon 16-35mm at 35mm f/4 (wide open)

The Canon 35mm f/1.4L continues to underwhelm. The 16-35mm zoom continues to amaze.

Corner Performance - Fuji

Fuji 23mm f/2 at f/2 (wide open)

Fuji 23mm f/2 at f/4 (two stops down)
Again, Fuji performance is at par or better than the much larger Canon 6D and L lenses. Here closer to the corners however, the 23mm f/2 lens is visibly sharper at f/4 than wide-open. Overall, the Fuji has lower contrast in JPEG's than a Canon.

Conclusion

The Fuji XE-2 and 23mm f/2 prime is optically equal or better than (my copy of) the Canon 35mm f/1.4L (version 1) and Canon 6D at all equivalent apertures. And check out the size comparison:


The XE-2, even if it is quite a old body, is worlds away from the Panasonic GF2 in terms of usability. It turns on in half a second, has an excellent EVF (that is very usable even in the dark), and competently auto-focuses. The main LCD can be turned off with a function button to mimic DSLR behavior. The XE-2 does not AF as fast as the 6D but for most common photographic situations, it will do. The 6D focuses much faster and more effectively when it gets really dark, but in bright light the XE 2 and Canon 6D with the 35L (a ring USM lens) are effectively neck to neck.

An important caveat: all of these tests were done with JPEG, not RAW. And I used autofocus for both bodies; it's possible that the 6D was mis-focusing with the 35L, and manual focus with Live View would fix that. But nobody would use manual focus and Live View with the 35L in real life.

Furthermore, the Canon has automatic lens correction, and this was quite obvious with the 16-35mm f/4L IS, where rectilinear distortion was minimized. I believe the XE-2 also has some form of lens correction built in.

However, I think this use case (autofocus, JPEG) is a very common one for travelers and other casual users.

I must rent the 10-24mm Fuji ultrawide! if its performance matches the Canon 16-35mm f/4L IS, then that lens with the XE-2 would make the perfect travel kit.

If there's any disappointment with the XE-2, it would be
  • it doesn't turn on instantly like any DSLR does
  • the 23mm f/2 lens is huge (larger than a Canon 50mm f/1.8) - this is not a Summicron



















Polar Alignment With A Bubble Level

In low-latitude locations such as Singapore (1.3521° N) or anywhere that Polaris is not visible, polar alignment can be difficult and time-consuming to achieve. One method for doing initial alignment in altitude is to use the mount's latitude scale, or a digital level of some sort.

This method is not sufficiently accurate because the mount's latitude scale normally only has 2° increments, and digital levels (in spite of their supposed high accuracy) actually have a tiny pendulum inside which is insensitive to small angle changes and really only has a resolution of 0.5° which is 30 arc-minutes and insufficient for a good polar alignment.

Here we can see an EQ mount set to zero latitude (as indicated by the bubble level) but the digital inclinometer is claiming 0.50° angle (which can be zero'ed out, but illustrates the inaccuracy of digital inclinometers).


However, we can take advantage of a bubble level that has a 45° vial (or our untrustworthy inclinometer) to achieve more accurate angle measures.

The Starrett bubble level in the photo above has an accuracy of 1mm in 1m, or about 0.045° or 2.7 arc-minutes, hence we can be reasonably sure (within 5.4 arc-minutes or  1/11 of a degree) that the mount is indeed at zero latitude.

We then take note of the position of the altitude adjustment knob:


The next step is to keep rotating the altitude knob (counting rotations as we go) until the mount is at 45° latitude (as confirmed by our bubble level or digital inclinometer):


Again we have an uncertainty of 5.4 arc-minutes in this measurement; adding to the uncertainty in the zero measurement, yields a total potential error of around 11 arc-minutes or 0.2°. On the other hand, if we used the digital inclinometer to zero the mount and also measure the 45° angle, our total uncertainty is 1°.

It is obvious that the bubble level provides lower error than a digital inclinometer.

In the case of the mount pictured, it took 39.60 - 39.75 turns of the knob to reach 45° +/- 0.1° latitude; this means that every turn of the knob yields 1.1295 - 1.1338° degrees (67.77 - 68.028 arc-minutes, or an average of 67.89 arc-minutes) of altitude.

We can see that the mount's latitude scale indicates roughly 45° as well, however since we don't know if the mount and tripod are level, and the latitude scale has no vernier and only 2° increments, the latitude scale alone is insufficient for setting the latitude.

If we had used the digital inclinometer, with its systematic error of 0.5° (total 1° including uncertainty about the zero point) then over 45° +/- 0.5° every turn of the knob yields 1.119 - 1.149° (67.14 - 68.94 arc-minutes or an average of 68.04 arc-minutes) of altitude.

Most EQ mounts use a sort of tangent arm assembly to adjust the altitude; as the angle gets higher, there is an increasing error. At small angles, θ ≈ sin(θ) but as θ gets larger, the error also increases. For example, at 45° (0.785 radians), sin(45°) = 0.707, a difference of 11%.

Hence, we need to reduce our calculated arc-minutes per turn by around 11% - so 68.04 arc-minutes per turn of the knob is reduced to 61 arc-minutes per turn.

The Astro-Physics Mach1 has a very accurate altitude adjustment, and is spec'ed for 62 arc-minutes per rotation with 16 ridges on the altitude knob (3.875 arc-minutes per ridge); this is in very close agreement with our calculated 61 arc-minutes per turn.

It is obvious that by using the 0° and 45° points as reference, we can significantly reduce the effects of systematic error in the bubble level or digital inclinometer.

After resetting the mount to zero latitude (again confirming with the bubble level), we can set it to 1.3521° by rotating the knob (1.3521° * 60 / 3.875 arc-minutes) = 21 knob ridges.  If we had used our derived value of 61 arc-minutes per rotation (3.8125 arc-minutes per turn), (1.3521° * 60 / 3.8125) = 21 knob ridges which is the same as the (known for the Mach1) setting.

On the other hand, if we need to set the latitude to a higher value, say 30°, we would use the 45° angle as the reference.  In the case of 30°, we would zero the mount at 45°, then lower it by 15° (which is 900 arc-minutes) which would require 14.5 turns of the knob (14 turns and an additional 8 knob ridges).

This method would allow accurate altitude alignment to within (5.4 arc-minutes error of the level) + (8 arc-minutes from the knob uncertainty) of about 13 arc-minutes (0.22°) worst-case; actual error may be half that. Such a result is good but not great, a drift alignment can achieve better accuracy. With this amount of altitude error, there is approximately 3.5 arc-seconds of drift per minute, hence an unguided exposure of 1 minute with a typical DSLR will still show round stars.

In contrast, had we set the mount's latitude directly from the digital inclinometer, we could have been in error by 1° which would lead to a 15 arc-second drift in 1 minute, thus limiting maximum exposure times to around 20 seconds before star elongation would be visible.

Building and Using Astrometry.Net on MacOS

Why Astrometry.net on MacOS? because Macs are built to take bullets and have long battery life. Unfortunately building almost anything on MacOS is ten times as hard as doing it on Linux. There are some (old) instructions for building Astrometry.net on MacOS here. However they are dated 2012. The instructions for Homebrew work fine on Yosemite with python 2.7:


However, there are some missing steps:
With the above conditions met, Astrometry works fine.

Takahashi EM-1S RA Drive Analysis and Southern Hemisphere Modification

I had a look at the innards of the Takahashi EM-1S RA drive, because I wanted to see how hard it would be to convert it to center-positive. Quick answer: it's hard.

A second problem was how to use the EM-1S in the Southern hemisphere. Here is the outside of the EM-1S RA drive panel. It is obvious that there is no way to switch from northern to southern hemisphere tracking:


Inside, we see a Sanryusha P43G stepper motor with 24 pulses per revolution and a 1:500 gearbox. This is the same motor in my EM-11 Temma2 Jr. and presumably many other Takahashi mounts.


The circuit is fairly simple but full of obsolete components.


We see the following IC's:
  • IC1: OKI M5562, Google is not very helpful, but most likely this is a clock generator IC
  • IC2: Toshiba TC4013BP, dual D-type flip flop, probably the logic generator for the stepping waveform
  • IC3: TDG2004, my immediate guess was this is a variant of the ULN2004 stepper motor driver
Supposedly, there is a switch on the board to enable southern hemisphere tracking, but there is no such switch here. It's fairly apparent that in order to make this mount useful in the southern hemisphere, the wires from the motor to the ULN2004 will have to be switched around - a highly annoying chore.

With some tracing, we can determine that the topmost two wires from the motor are the commons (the motor is a 6-wire unipolar with split center tap) and the other four wires are the four phases.

Therefore, it should be possible to reverse the direction of rotation by swapping the four wires that go to pins 13, 14, 15, and 16 of the TDG2004.


To be more specific, assuming a stepping sequence of 1-2-3-4 (where the white wire from the motor is 1, blue is 2, black is 3, and yellow is 4) the motor should run in reverse with a stepping sequence of 4-3-2-1. In other words, swap 1 and 4, and 2 and 3.

To make this process simpler and avoid multiple soldering and de-soldering chores, I soldered some Berg pins to the board, and attached connectors to the motor wires. After some challenges (the #2 connection broke which prevented the motor from turning) I was able to validate that indeed, the motor now runs in reverse.


Repairing Astro-Physics GTO Hand Pad Cable

The cable on my Astro-Physics GTO hand pad cable (AP part number E0190CABLE-E) had deteriorated over time. The rubber insulation had peeled and cracked, exposing the shielding and conductors underneath.

I repaired it temporarily using black duct tape but the tape left a sticky residue and was pretty ugly. Astro-Physics wants $75 (plus shipping) for a replacement cable. Good to know I can buy the part, but I wanted to save some money.

While at Popular Bookstore last night, I saw a kiosk with a strange Play-Doh like adhesive, Sugru. A video was playing, touting various wonderful features. The adhesive was quite expensive (S$ 19.90 for eight tiny single-use packets) but I figured it was worth a try.

This adhesive can also be purchased on Amazon, for $19.58 - so the price at Popular was actually lower.

And.. 12 hours later.  The adhesive has hardened into a somewhat-flexible silicone rubber which feels pretty tough (the red material around the handpad cable at the bottom-left of the photo below). I would say this is a qualified success!


A Tale of Two Taks

I recently was able to buy a used Takahashi EM-1S from Yahoo Japan Auctions. Generally, Japanese sellers don't ship out of Japan, but From Japan made the process much easier. There's not much in the way of buyer protection, however. More on this later.


I already own a nice Takahashi EM-11 Temma 2 Jr. which I've reviewed in the past (the mount on the right in the above photo). It is supposed to be my travel mount (for when I travel to the Canary Islands.. someday..) but ironically I am not too keen on traveling with it because it is rather expensive and would be subject to the slings and arrows of outrageous fortune in the checked baggage.

Incidentally the EM-1 (the mount on the left) has a Canon 300mm f/4 L telephoto lens on it, and the Kenko Lens2Scope adapter, which is a 10mm eyepiece (of narrow field.. I estimate 40 degrees) and erecting 45-degree prism. The Lens2Scope is useful because when traveling, you can use the camera lens as both a small telescope and for astrophotography.

But I digress.

I got the EM-1S for the unheard-of price (by Takahashi standards..) of around $550. Based on the serial number it is a 1991 model, about fifteen years older than my 2006 EM-11 and making it 26 years old. It did not come with a counterweight shaft or weights, but I thought that was fine since I have an extra counterweight shaft and weights.

And that is where the good news ends. It arrived with a bent RA worm shaft. I tried to straighten the shaft on the RA clutch knob with a pair of pliers.. and the shaft snapped clean off. I was able to successfully straighten (more or less..) the RA worm shaft with a block of wood and some judicious whacks with a hammer. The shaft isn't perfectly straight but the worst of the RA spur gear wobbling has been taken out. I hope I didn't damage the worm with that Lizzie Borden activity..

The EM-1S has a tangent arm in declination, not a full worm. Oops! and it's quite hard to turn. I cleaned and re-lubed it but the DEC slow motion still binds a bit and does not feel smooth. Also, it seems to not have ball bearings, but rather sleeve bearings (like the Vixen Super Polaris and Great Polaris of similar vintage).

The worst bit is - my particular EM-1S is missing its famous Takahashi polar scope. Which is quite annoying. I do not know if I will get any satisfaction from the Yahoo Japan seller, since they stated that "only the parts shown in the photos are included" - and the photos showed the mount with its manhole polar scope cover screwed on, and no evidence of a polar scope.

In any case: the manhole covers on both the EM-1S and EM-11 fit each other. I suspect the EM-11 polar scope would fit as well but I don't want to remove it and potentially ruin the alignment. They both have the 35mm dual bolts on the top (I was able to use the Berlebach dovetail clamp from the EM-11 on the EM-1S). The tripod bases are also compatible. Even more impressive, both mounts use the same Sanryusha P43G stepper motor with 24 pulses per revolution and a 1:500 gear reduction.

I hear that the current small Takahashi mounts still use this model of motor. They must have a massive warehouse full of these motors somewhere.

The EM-1S does seem to have a larger-diameter RA worm wheel and is a bit lighter than the EM-11. It is powered by a rather large center-negative 6V battery pack, I will check if I can power it from a USB power bank.

Overall I am a bit underwhelmed. The old Tak isn't at the quality level of the new one, although admittedly it is better-built than Vixen mounts of the same age. And the lack of a polar scope peeves me - although luckily most of the time I won't actually be using the polar scope since I'm at a low latitude. I will try to see if getting a Takahashi polar scope is not out of the question cost-wise; if cost is prohibitive I will probably find someone to machine an adapter so I can use the cheap CG-4 or CG-5 polar scope in the EM-1S.

Was buying the EM-1S worth it? I will give a qualified yes.  P2Z, P2, and Space Boy mounts on Yahoo Japan (the mount I really wanted) sell for high prices (95000 JPY or $900 if you're lucky, but more typically up to $1500). I wanted a cheap Tak mount with setting circles for travel and got one.  Hopefully the periodic error is low (-er than my Vixen Super Polaris) because that was my goal.

Takahashi Mewlon 210 Vignetting on Full Frame

The Takahashi Mewlon 210 Dall-Kirkham Cassegrain only has an 18mm image circle at prime focus, and 39mm with the dedicated reducer.

Here are some images of daytime scenes taken with a Canon 6D full-frame DSLR.

Native focal length (2415mm, f/11.5)


With a Televue TRF2008 0.8X reducer/flattener for the TV85 (designed for 600mm radius of curvature) (f/9.2)

and with an Altair Astro 0.6X reducer/flattener (no longer available, but also sold by some other Europe-based sellers) (f/6.9)
 I will try to get some images of a large globular like Omega Centauri when I get the chance, but the illuminated field is not bad at all (an f/4 newtonian with a Paracorr has a comparable field to the Mewlon with the 0.6X reducer!)


Takahashi Mewlon 210 Maintenance

I've been looking for a Takahashi Mewlon 210 Dall-Kirkham Cassegrain since about 2010, but the new prices are pretty high and I didn't want to pay list price. I was able to find a used one (I am probably the third or fourth owner) on SingAstro for a pretty fair price. The downside was that it had some dings on the paint.

I had some enamel repair paint and some Canon touch-up paint for old grey L lenses (not a perfect color match for Takahashi hardware) but they suffice for now.

After a thorough cleaning (lots of black marks on the Takahashi hardware came off with the use of a rubber pencil eraser) I found several areas where there was paint loss. On the gloss-white tube, I used the enamel repair paint.

Before and after:

A large area of paint loss on the front of the tube:


A smaller area where the paint flaked off, also on the front of the tube:


 A ding on the finder bracket, and after patching with the Canon paint:



The Mewlon also came with a Vixen dovetail, but I wanted to mount it on my Mach1 mount and did not want to buy tube rings (which cost more, and add unwanted weight). I had a short Losmandy dovetail from FarPoint Astro, but it didn't have the holes in the right spots.

So I bolted the stock Vixen dovetail to the FarPoint dovetail and centered it so that I could use the Vixen dovetail as a drilling guide.


And after drilling. The second hole was not very well done; but it's not visible when clamped on the Losmandy saddle; and is definitely not visible at night!


Also, I found that the cheap reticle illuminator from a Seben reticle eyepiece actually fits in the reticle socket of the 50mm finder (not very well, however):


I did feel some grittiness when screwing in the illuminator plug afterwards; I think some metal came off the cheap Seben illuminator due to the not-quite-correct threading. So I probably have to buy the real Takahashi illuminator.

After maintenance, attaching the Losmandy dovetail, and mounting on the Mach1:


Canon 28-105mm f/3.5-4.5 USM II

TL; DR - use it at f/8 and it has (almost) L-class sharpness.

Back in 2001 or 2002 this was one of the lenses I wanted. It was too steep for me and I ended up with a Sigma 28-105mm f/2.8-4 (the infamously soft and bulky lens) which I used for some time on an EOS3000N and EOS50 until it got damaged (diaphragm stuck wide-open).

Fast-forward fifteen years and I have one from KEH for about $120 in "bargain" condition.

I have two lenses that cover (parts of) the range of this 28-105mm: the 16-35mm f/4 L IS, and the 70-200mm f/2.8 L IS. I was planning to compare these lenses but it turns out that The Digital Picture already has a lens comparison service. So here's the summary (so far as I can tell) on a 5D Mk III (which has the same sensor as my 6D):

Compared to the 16-35mm f/4 L IS (an $800 lens):
  • at 28mm and f/8, the 28-105 almost matches the 16-35mm wide-open (at f/4)
  • this is also true at 35mm
Compared to the 70-200mm f/2.8 L (a $1200 lens):
  • at 70mm and f/8, the 28-105 matches the 70-200mm at 70mm wide-open (at f/2.8) and at f/4, but notably the 28-105 has better corners than the 70-200 (note we are comparing f/8 to f/2.8 and f/4)
  • at 105mm and f/8, the 28-105 matches the 70-200mm at 100mm wide-open (at f/2.8) and at f/4, but the 70-200 beats the zoom in the corners even at f/2.8
The long and the short of it: the 28-105mm can produce almost-L class sharpness so long as you stick to f/8. However, the 28-105 has one massive feature that trumps these L lenses:
It is tiny (about the same size as a 35mm f/1.4) and not much larger than the Nifty Fifty.

My wife and I have traveled a lot with the 6D and 16-35mm f/4 L IS, and it is quite a large and bulky setup, which is why we ended up also buying a Canon G5X (which has a 24-105mm equivalent lens, with IS, and a 1" sensor). The G5X is tiny, but slow (slow to auto-focus and take photos) and has an EVF instead of an optical viewfinder.

I was stuck with the 16-35mm in Monterey in October 2016 when we went whale-watching, and 35mm is much too short for whales.  The 28-105mm would have been useful to have at that time: 105mm long end, f/8 is useful as there was full sun, and the 28-105mm doesn't add much weight or bulk to the camera bag.

Ultimately that's what I see the 28-105mm as: a useful adjunct to an ultra-wide L lens for travel. The 70-200mm is simply too large and bulky to be convenient when traveling. Furthermore, on a 5D Classic, I believe the gap between the 28-105mm and the L zooms would be even less.

I believe there's one reason the 28-105mm f/3.5-4.5 is unpopular and cheap, in spite of its FTM focusing, ring USM, and focusing scale: the zoom range is not very useful on reduced-frame (APS-C) DSLR's.

Amazon Polly Just For Laughs

If you've ever had a burning desire to create custom dialog for Consuela the housekeeper in Family Guy..

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
#! /usr/bin/python
import boto3
import io
from subprocess import call

client = boto3.client('polly', region_name='us-east-1')

# aws polly describe-voices --region us-east-1|grep Name
response = client.synthesize_speech(
    OutputFormat='mp3',
    SampleRate='16000',
    Text='<speak><p>Hola. Doggie a puera!</p></speak>',
    TextType='ssml',
    VoiceId='Conchita'
)

# Access the audio stream from the response
if 'AudioStream' in response:
    sb = response.get('AudioStream')
    data = sb.read()
    fd = io.open('hello.mp3', mode='wb', closefd=True)
    fd.write(data)
    fd.close()

    call(["afplay", "hello.mp3"])

Canon 70-200mm f/2.8L versus Pentax 200mm f/4 Super-Multi-Coated Takumar

I have compared two Canon lenses, the 10-20mm f/3.5-4.5 consumer zoom (on APS-C) and the 16-35mm f/4 L IS here. The results were predictable - the L zoom was consistently better than the consumer zoom, at about three times the cost (and increased weight).

Also, I used to collect old, Pentax screw mount manual focus lenses. These old lenses have a cult following, with various miraculous attributes being ascribed to them (glorious bokeh, an otherworldly quality of the image, etc. etc.) While some of these old lenses (particularly the old Leica lenses) do deserve the praise, I had always wondered how these old lenses compared to modern, non-consumer grade ones.

So in this post, I compare the older, non-IS Canon 70-200mm f/2.8L zoom with a decades-old, manual focus, Pentax screw mount 200mm f/4 Super-Multi-Coated Takumar. The comparison was done at f/4 (since the Pentax manual focus prime cannot open up to f/2.8) on a Canon 6D, which should have plenty of resolution. The target image was a neighboring building at (close to) infinity focus.  The Canon zoom was auto-focused, while the Pentax was manually focused using Live View (it ended up at the infinity stop anyway). Base ISO was used and a fairly high shutter speed (1/2000 second) on a tripod.

First the full image:


The highlighted areas are zoomed in (1:1) for comparison. They roughly cover the center of the frame, and the edge. The extreme edge was not used, because it turns out the Canon 70-200mm zoom is not really 200mm at the long end (it is a bit short).

Let's look at the center of the frame, where both lenses should be performing their best. The zoom is one stop down, so should show improved performance.. but it's a zoom, while the competition is a (wide-open) prime:

Canon 70-200mm f/2.8L at 200mm f/4, center:
Pentax 200mm SMC Takumar at f/4, center:
We can see that the Pentax is ever so slightly longer than the Canon, it has less contrast, and is less sharp. We're talking a $40 lens (at KEH) versus a $1300 lens. I wondered if the Pentax was not critically focused, but it was at the infinity stop, and Live View could not yield a better focus. In any case, "in the field" one would not have the luxury of tripod and Live View and would probably just use the prime at the infinity stop.

The difference is certainly obvious at pixel-peeping distances, but on a regular full-size image the two lenses would be indistinguishable.

Canon 70-200mm f/2.8L at 200mm f/4, edge:
Pentax 200mm SMC Takumar at f/4, edge:
again we can see better contrast with the Canon (better coatings perhaps) and also more detail in the fine vertical lines of the blinds. What is noteworthy is that both lenses don't really suffer much (if any) drop in resolution, going from the center of the frame to the (near) edge.

Is the $1300 Canon a better lens? of course it is - sharper, more contrast, autofocus - but is it thirty times better? that depends on the user.  If autofocus and zoom are important, then these are worth paying money for.

The Canon is quite bulky in comparison to the relatively compact Pentax prime, however.
I was whale-watching in Monterey in October 2016, and did not bring any long lenses (the Canon 70-200mm is very hard to travel with). Ended up with a lot of distant images of whales (using the 35mm end of the 16-35mm f/4L IS). At that time, I would have gladly used the small Pentax prime, in spite of all its shortcomings and non-L image quality.

In conclusion: yes, the Canon 70-200mm f/2.8L is a better lens at the 200mm end than the 1960's - 1970's Pentax 200mm f/4 Super-Multi-Coated Takumar.  But the SMC Takumar is more than good enough and I would say can more than hold its own image quality wise (so long as you don't pixel-peep at 100% magnification).