For roughly the same money are you better off getting a used 640 (FLIR RS) scope with the larger pixel size or a new 320 scope (PT233)?

The new stuff coming out is leaving me in a bit of a dilemma.

Would like to know this as well and why....

640 for sure.
Once you go 640 you will never go back.

kinda like comparing gen 2 and gen 3

I've owned both.  I was pleasantly surprised by the mid-grade pulsar stuff even though it wasn't 640.  I previously had a 640 with a 75mm lense and it was great (and expensive).

I can't afford a 640/12 pixel when it comes out so that is a moot point; which leaves me trying to decide between a used 640 or a newer 320/12 pixel unit.

I own both 320 and 640 thermal devices. The 640 is by far the better scope. I do use the 320 quite often though, and it certainly is worth having.

One reads all sorts of technical info here regarding the very small pitch thermal sensors. Info regarding 320 devices rivaling older 640 ones (I remain unconvinced), much smaller lens sizes, and so forth. But what I have NOT seen is info comparing the newest, smallest pitch devices to older ones as regards thermal stability of the sensor core, figures comparing post production pixel failure rates in the old vs new sensors, response and recovery times of pixels in the sensor that are exposed to brief (or not so brief) high or low temp events, and, especially interesting to me, the unanswered question of whether the new, small pitch sensors require especially pure and unblemished objective lens material to realize its potential, and if so, any evidence that the lenses are being manufactured to the requisite specs (it doesn't take much of a defect to distort a significant portion of the pixels that remain available to see through after so much viewing real estate is already obliterated by symbology like the battery meter and other junk that is not what you wish to observe with a hyper spectral device).

Given a wide range of new and used gear to choose from within your budget, I suppose that I would DEMAND these things: Focus-adjustable eyepiece (those few pixels MUST be in perfect focus), One-to-one sensor-to-viewing screen pixel count (NO video processing required to adapt incompatible video format between sensor and eyescreen and generate artifacts that degrade what you thought you were buying), if you buy a 1X device, an objective lens that will accept a magnifier if you decide later to get one.

Beyond that, the battery type (internal rechargeable or field-changeable), rifle mount capability, objective focus range (I buy only close focus capable units so I can use them for equipment maintenance and other purposes), video/still photo capability, and the other characteristics become secondary, personal options.

It’s odd that we want a small sensor size here but the rest of the photography world wants a large pixel size.  Not many pros would opt for a crop frame 24MP camera over a full frame 24MP camera.  Yet the thermal marketing all makes it sound like the smaller pixels are better.  They would be better if you fit more into the same space, but keeping the number the same and shrinking the physical size of each one is the worst of all worlds, accept maybe cost/price.  That’s not to say that the new stuff is worse, just that people shouting about pixel size aren’t completely honest or informed.  Sensitivity, processing, etc can all make the image you see better, but all other things equal 320x240 sensor that’s made of larger pixels will gather more light and see smaller differences than a smaller pixel sensor.  The problem is that all things aren’t equal in reality and as other aspects advance, it’s impossible as a consumer to tell which change was the important one.

As for older 640 or newer 320, it would really depend on condition.  Take the RS series from Flir.  It’s probably not an issue that it has an internal battery, but what’s battery life lasting now?  What is the replacement cost out of warranty?  Are there features like one shot zero that matter?  If none of those are bad or none scare you away, then 640 for sure over anything newer and lower resolution.  If they do scare you, then I’d look newer.  I’d also consider Pulsar scopes.  Their sensors are 384x288.  That doesn’t sound like a ton more than 320x240 but it’s 44% more pixels.  It’s something to think about if you decide newer features matter too.  Ask just about anyone who’s looked through say the Flir Boson core stuff and the Pulsar 384 stuff.  The extra resolution shows up and the Pulsars typically get the nod for better image.  Even their Core series that is under $2000 is given the nod of a better image than the current 320 stuff, though there you sacrifice many of the nice features to make the price point.

I guess the short of it is, if the condition is good and the scopes type doesn’t have a history of issues (like say ATN stuff) I’d take the 640. If I wanted features, I’d probably look at the higher spec non-640 cores like Pulsar.  If I found an absolutely screaming deal that I was willing to gamble on and potentially completely write off, I’d grab a 640 ATN, especially the pre HD series.  That’s a big gamble as if anything is wrong or goes wrong, it’s to the trash.  I’ve had decent luck with ATN but I know that isn’t the common concensus here and my experience of three is far from being a significant sample.

I'm still learning myself.

My LGS owner tells me that he's gone through a lot of thermal (he hunts hogs regularly); and for the way he hunts he is using a Pulsar Apex 384x288 scope and is satisfied.
He doesn't record the hunts so he isn't interested in the recording feature; he also uses one of the Helion monoculars for spotting.
This guy can afford pretty much whatever he wants.
ymmv

After talking to some folks I'm going to skip the RS series for now.  I had a Zues 640 75mm that was great but I didn't like the narrow FOV.  I may end up with one of the newer 320 monoculars because I want to scan and have the ability to focus up close.

I have a friend that has that scope and it's pretty awesome.
He said he had detection at a mile away one night.  Had no idea what it was though.

One thing about the Pulsar is the wifi feature that links to your phone or tablet was pretty neat.  LGS owner showed me that once linked up, what was on the scope was shown on the phone or tablet and that you could also get into the scope menu via the phone or tablet.

Any 640 core scope is always going to have 4X the pixels of a 320 core unit with the same lens size for higher resolution and longer detection and positive ID.

I know you have a lot more experience than I do but that’s wrong.  A 640 core will have the same detection/ID range as a 320 using the same lens/focal length, if not slightly less.  You’ll have 4x the total field of view in the picture so you can see more of a scene at any given time, but the pixels per degree of view is the same in both setups.  The slightly less is due to the fact that the smaller sensor of the 320 has all the focused light on a smaller area, which should give a theoretically better image over that smaller FOV, assuming the lens is the same size/quality.  In practice, it seems companies will use the same lens on both size cores and as a result, the lens is probably not just covering the edges of the smaller core and you’d see no real world difference.

It’s very clear this is the case, just look at ATN’s Thor HD line.  The 640 magnification ratings are about 60% of the same focal length 384 option, which is the difference in sensor size.  If you had a 1.615x digital zoom on a 640 core, you’d have the same detail and same FOV as a 384of the same focal length with no digital zoom added.  Likewise, if you go to 2x digital zoom on a 640 you get the same image as a native 320 core using the same lens.  Increasing core size alone adds detail, but it does so by proportionally increasing FOV, not keeping FOV and increasing the detail in the same image.  That means there is no added detection or ID range, just more picture to detect or ID in.

I have no idea how manufacturers determine “detection range”..... but I’ll say this.  The other night we were out and I spotted a pig at a little over 2400yds. It’s a know distance to far hillside. I believed it to be a pig because of the way it was moving, not because I could actually ID.  I could see it with a trijicon, and a 384 pulsar.  Nobody else standing there ever saw it . It was just a faint blob of a heat signature. After a $10 bet that I was right, we went on to kill it . I tell this dumb story to point out that specs/mfg claims can be misleading.

1. Detection range? How is that tested? The faintest blob that can be seen by experienced user? Or heat signature that’s strong enough the cause 12 pixels in the screen to acknowledge its there?

2. Id range ? No way to qualify a statement like this. Period. No way to do any sort of controlled test.

Side note, the core of a 640/320 are the same size. The information that passes through the lens to the microbolometers is the same . The board itself has a smaller number of mb, covering the same area. So it presents that info, via voodoo magic, to the screen, in a less detailed, zoomed in image.

This is tough to explain without visual references. Somebody here may have some that could clear up the confusion on how that actually works

My numerous 640 cores have twice the ID range of my numerous 320 cores and function twice as good under high humidity conditions with almost twice the FOV.

To skypup......... your obviously well educated in the functionality of all this .  Do you have any visual reference material that explain/shows how the energy is transferred through the lens, onto the board, then converted to screen as image ? Something really simple .  I’d make a crude drawing, and put it up, but I’m not sure it would make sense.  But I think it would help people understand, and make choices accordingly .

Your right about 640 being better for all those reasons. But me and you just saying it’s better, doesn’t help anybody else. I’m nobody important. Why would anybody listen to me?

I did a crude test between some different systems last fall. I filled a milk jug with 105 degree water and placed it on a long straight Turnrow that I could see from as far away as over 1100 yards.

I had my flir vue 640/19mm rooftop filming and I also my IRD mark 2 640/35mm.

A month or so later I did the same test using my Flir PTS233 but I cannot for the life of me find any documentation from this test so I will redo it. Anyway the test should be redone since it would not be comparing thermal units on or under the same conditions.

Details from the previous link so you don't have to bother opening it are as follows.

detection range of milk jug

Flir Vue - approx 900 yards.
IRD - beyond length of road bed (1150 yards +)

PTS233 - I believe that I reported that I could no longer see the milk jug beyond approximately 450 yards. And at 450 yards if it were not moving around I probably would not have noticed it.

So the IRD 35 mm has the same field of view as the flir PTS233. They both have 12 field of view.

I know when I redo my test it will generate a similar result. The IRD 12 degree 640, 12 micron system will detect a hot water milk jug approximately two to three times as far away as my flir PTS233 12 degree, 12 micron system will detect the same object.

I intend to redo this test sometime soon and I will include footage from my pulsar XP50, as well as the flir breach. It would be nice to compare everything under the same conditions.

https://www.ar15.com/forums/armory/Milk-jug-distance-test/18-481714/

There are a couple of guys here presenting just plain wrong information.
i will leave it up to you to figure it out.

That's the driving factor, cost. The cost of R&D for a smaller sensor aside, manufacturing cost doesn't go up once the technology is refined. If anything, it may go down a bit as the yield of sensors per batch increases (can fit more sensors on a die, for example).

The main factor is the lower cost of optics. Maintaining the same FOV with a smaller yet comparably fast lens requires shrinking down the image scale. LWIR optics are fundamentally expensive and any sensor size reduction translates directly into very substantial savings on optics.

Ideally, larger aperture (and thus larger pixel size) is ideal for extracting as much signal as possible. Holds true for visible cameras, holds true for thermal.

For clarification above. My flir Vue 640 camera has a 19mm lens with a 17 Micron core.

The PTS233 is a 320 system which also has a 19 mm lens, but the core is 12 micron.

I hope to get some video footage documenting this over the weekend. But I am confident that the 640 Vue will detect a milk jug filled with hot water twice as far away as my PTS 233 will.

But for two grand, the PTS233 is no slouch and I have killed a bunch of shit with mine so far.  I don't bother posting videos from that unit anymore because there's plenty of them online and I've kind of gotten spoiled being able to download via Wi-Fi from my pulsar straight to my phone, but I use it all the time.

It's on my 9 mm A.R., and I killed a coyote with it just over 100 yards last night.

You get 4 times the pixels in your 640 cores, but the same focal length only adds FOV, nothing to pixel per degree.  That means you have no extra pixels to detect anything extra.  It’s simple and a concept that is proven in daylight cameras with the change from Full Frame to Crop sensors.  At the same focal length, humidity changes nothing for either as well, since once again pixel per degree is the same and you have no extra data in a given space to minimize the effects of humidity.  If focal length changes to give the same FOV, then you are 100% correct.

The IRD should detect roughly twice as far.  It’s roughly twice the focal length. The base FOV is the same and the IRD has double the resolution each way.  When you bump it up to 2x digital zoom, you end up with the same resolution and half the FOV, which again should equate to twice the detection/ID distance.  But it’s due to the focal length gain as well as any specific lens quality and image processing/sensitivity.

I know some of the industrial use stuff has been known to do that but it appears that even if they are using the same sensor and turning off 3/4 of the pixels, they are leaving on consecutive sections making the active sensor size still half the size in each direction making the math work.  Just look at ATN or the Flir RS line and you see at a given focal length, the FOV doubles moving from the RS32 to the RS64.

Assuming the milk jug is 1 gallon & the first google hit of dimensions of 15.24cm in width, 25.4cm in height are true or close enough, and assuming the sensor dimensions are pixel size * num of pixels, then the following would happen:

Milk jug occupies the following space in pixels on each sensor (assuming perfect focus obviously):

PTS233 19mm (11.5 x 9.2 deg)
w = 0.59px
h = 0.98px

FLIR Vue 19mm (32.0 x 24.2 deg)
w = 0.21px
h = 0.35px

IRD 35mm (12.5 x 9.4 deg)
w = 0.42px
h = 0.71px

Sounds like in each unit the size on sensor is less than one pixel (with Vue just 7.4% of the size of one square pixel). If that is the case then lens quality & focus issues will play big into this along with everything else, but when filling say hundreds of pixels then focus is not so much of a factor.

Why PTS233 would not detect it despite having the largest area of sensor filled with the the milk jug and being much closer so the environment doesn't play into it as much, I have no idea. If the differences were smaller then sure but that's way out from what I'd have assumed.

What's different for PTS233 is that the display doesn't match the sensor in aspect ratio, so either it does some resampling or cropping, or both. Not a big difference and shouldn't really affect the performance.

Could you please explain how the pixels per degree of view is the same for both sensor sizes?

A while ago we were having a similar discussion of detection range and one very nice user was gracious enough to illustrate the mathematical procedures for coming up with the number of pixels of a specific sized object at a certain range for each sensor size.

I saved a screenshot of what he wrote out into my phone for reference.  I think the actual post is in one of the two multi page Flir  PTS l233 discussions?

Attached File


According to the way this was explained to me a while ago, a coyote sized object shows up on a 640 system with about four times as many pixels than are available in the same field of view 320 system at a specific distance.

Am I totally not understanding this?

thanks.

Bingo, you are correct!

First, it needs to be the same sensor design or the math changes.  If the pitch is different, then the math changes.  Using sensors of the same design or at least same pitch makes the math simpler.  If the pitch is different, it's just another step to account for the sensor size difference.

Sensor size, the actual physical size not the pixel count, is what matters here.  Focal Length is like an hour glass, the closer the focal point is to the sensor, the wider the FOV.  This is because 19mm is where the image focuses to a point and then everything further out it is widening.  So a 38mm focal length focuses to a point twice as far from the sensor, cutting FOV in half each direction.  With this though, the sensor is looking at a smaller FOV and shows more magnification.

At the same time, the larger the sensor, the wider the FOV.  The lens projects the same size image, but you are only getting the middle chunk on a small sensor and you are getting a wider, fuller chunk with the physically larger sensor.  But it's the same image from the same lens.  The same generation core is proportionally larger for a 640 core compared to a 320.  There aren't more pixels packed into a smaller area for that image to shine on.  Its the same lens, just with a larger footprint sensor.  So the FOV widens but the image doesn't get a more dense pixel array to get more info from a given spot.

Some companies will cheap out (smartly) and use the same sensor for all devices and simply turn off some of the pixels in the firmware.  This was hacked in some of the Flir commercial stuff to turn a very low resolution device into a 320x240 device.  If this is the case, and the digital pixel killing is interlaced, then Skypup is correct, as the footprint of the sensor is the same.  Problem is, that isn't what is happening.  It's easy to see.  Look at the 35mm Flir RS32 and the 35mm Flir RS64.  Same lens, same generation and tech, one is simply 4 times the total resolution.  But what changes in the specs?  FOV.  It doubles in both length and width.  This is because they are either using the smaller core or they use the full core but delete all but a consecutive working 320x240 pixel section.  Either way, it goes back to the point that the pixels aren't more dense.  They are seeing the same image that a 640 core is seeing from the same lens, they are just the center set of pixels while the 640 extends out.  If you look at full frame vs crop frame dslr's you will see the same thing.  Problem there is they use different sensor sizes and it's not near the straight forward issue that it is here.

I'll ask this, what happens to the picture on a 1.1x Flir 35mm RS64 when digitally zoomed to 2x compared to a 2.2x Flir 35mm RS32 with no zoom?  I'll give you a hint...It's the same image.

EDIT:

Do the math.  The pixel count and FOV is taken directly from FLIR's site.

RS64 35mm scope has a FOV of 18* which is assumed to be diagonal since it's not listed.
RS32 35mm scope has a FOV of 9* which is again assumed to be diagonal.

The RS64 has 820 pixels diagonal over 18*.  That's 46 pixels per degree FOV.
The RS32 has 422 pixels diagonal over 9*.  That's 47 pixels per degree FOV.

The RS32 has half the diagonal pixels of the RS64.  And look at that, the FOV is half.  Which means?  The same pixels per degree and no more pixel density/detection range/ID range between the two.

EDIT #2:

Lets look at ATN as well.  Again from their site

The Thor HD 640 19mm scope has 640 pixels over 32* and 480 pixels over 25*  That's 21 pixels per degree and 19 pixels per degree.
The Thor HD 384 19mm scope has 384 pixels over 16* and 288 pixels over 12.5*.  That's 24 pixels per degree and 23 pixels per degree.

Again, no more pixels per degree.  I'd guess the difference here is marketing and rounding of FOV numbers.  I doubt the 384 is actually a slightly higher pixel per degree but it's within the rounding figures.  It still shows there is not 4 times the identification or detection range using the same focal length on a 640 core over a 320 core.  It's the same.  Pixel pitch along with focal length determine detection range.

Bumping to 640 is worthwhile.  It gives a much bigger picture and if focal length is adjusted to give the same FOV, then you get more detail in the same picture.  But not increasing the focal length as well gives you simply a wider FOV.

Just assume you were speaking to the village idiot here. I Appreciate you taking the time to explain this but I am still having trouble understanding.

How can you say the IRS 32 has half the diagonal pixels of the IRS 64, and then say they have no more pixel density?

Let me ask my question another way. Can you please illustrate the following question mathematically?

Let's say we are viewing a 12" x 12" x 12" cube shaped hot object 100 yards away with both of the flir systems mentioned in your example above. The RS64, 35mm and the RS32, 35mm.

Question 1. How many pixels are being utilized to display that 12 inch cube image from each of the 35mm system from your example above.

Thanks.

At 100 yards, 1 degree is 62 inches.  A 12x12 square has a 17 inch diagonal, which is .274 degrees (17/62).  With 46-47 pixels diagonal of both the RS32 and RS64 you have roughly 12.8 pixels diagonal which is about 9 pixels by 9 pixels.  This is for both the RS32 and RS64.   The difference is that the RS64 has a FOV of 92 feet diagonally where the RS32 has a FOV of 46 feet.  So you see a much wider and taller picture but the details in each degree, or 12x12 square are the same.

Pixel density is pixels per degree of view.  It’s the same from 320 to 640 with the same focal length as shown in my last post.  You have more pixels and a wider FOV but the detail in each foot is covered by the same number of pixels.  To have more detail/pixels in the same space you need to increase focal length.  That is how you gain clarity/resolution of a given subject.  Focal length is optical zoom.  If you want more detail, you need more zoom/focal length.  A wider picture by a 640 core doesn’t increase details of say that 12x12 box, it just shows more of what’s beside it.

For simplicity's sake let's think of that cube as a 12x12 square hot object instead so that viewing angle won't affect the 2D projection you (or a sensor) see of the 3D object. Also let's assume the imager & target are level and the lens is pointing directly to the center of the target.

Horizontal & vertical angular size of the target object at a distance
angularSizeHorizontal = 2 * atan((objectWidth / 2) / distance)
angularSizeVertical = 2 * atan((objectHeight / 2) / distance)

Sensor FOV. Not especially accurate as optics are complicated, but assuming the lens is good and designed for this certain sensor size in mind, then it's close enough
sensorFOVHorizontal = 2 * atan(((pixelSize * sensorResX) / 2) / lensFocal)
sensorFOVVertical = 2 * atan(((pixelSize * sensorResY) / 2) / lensFocal)

Utilized pixels in height and width. In other words a rectangle in which the target will fit as an image onto the sensor
utilizedPixelsWidth = angularSizeHorizontal / sensorFOVHorizontal * sensorResX
utilizedPixelsHeight = angularSizeVertical / sensorFOVVertical * sensorResY

Total area of the rectangle. If the situation is as described and the object is rectangular this holds true
utilizedSensorAreaInPixelsSquared = utilizedPixelsWidth * utilizedPixelsHeight

That value is below 1px^2 for the examples in @Wildfowler 's post where he described the max detecting ranges he noticed for a warm milk jug. Less than one px^2 just means the radiation is landing on one pixel, but the one pixel might still register the radiation.

You could flip the calculations the other way around to see what kind of focal length or sensor size is in theory needed to fill at least 1 pixel. Of course in real world the variables to determine the actual radiation strength (or difference to background / signal) are vast and no simple calculation will give a definitive answer to the max detection range. And sensors are different in sensitivity, optics can vary hugely, the software can be good/bad, output display good/bad, and so on.

If we wanted to know exactly how many pixels are being utilized, we'd need to know the shape of the target, translation and rotation between sensor & target and do much more complicated math. In estimating these things it wouldn't make a big difference.

Thank you for this further discussion. Sometimes you have to beat me over the head to get it to sink in.

Barring a difference in pitch. PTS233 and Vue with same size 19mm lens should give the edge to the PTS for detection.

27 versus 20 horizontal pixels per degree.

I may be getting ready to issue a retraction and issue a recognition of what a difference comparing systems under different conditions can do.

I hope to test this again tonight.

Thanks.

Yes, the same focal length with a smaller pitch will give a more magnified picture and with that comes longer detection/recognition/ID range.  The difference will be proportional to the percentage difference in pitch size.  The larger pitch sensor, if nothing else were different, could see slightly slightly smaller differences since ea h pixel is bigger and can gather more light under the same conditions.  But processing and sensor sensitivity are constantly improving and could easily hide the difference.

Mostly depends on what you plan on doing with it at what specific ranges of engagement, but for all intents and purposes a old 640 core scope will outperform any new 320 scope by a factor of 2X to 4X depending on the environmental conditions you are using it in.

You're comparing apples and oranges, but you forgot about the mangoes.  The Pulsar XQ line uses 384x288 cores.  These have 44% MORE resolution than the Flir 320s.  And some, like the RXQ30V, are even cheaper than the Flir.  Others to consider are Apex XQ38 and Trail XQ38.