In a previous job I was working on designing a cooled CCD camera somewhat like this. It was a really tough job, one that we never completed, so we stuck with the off the shelf cooled cameras we were already using. The two biggest issues were frost on the CCD, and getting reliable vacuum feedthroughs. We were originally doing a gas filled camera, and any tiny amount of moisture will condense into frost when the CCD is at -20C. The plan was to flow dry nitrogen through the chamber while baking, then hermetically seal it, along with desiccant inside. Vaccuum/hermetic feedthroughs were a major problem to source and attach to the case. We ended up having a company that specializes in this kind of thing build the CCDs, peltiers, and temp sensor into modules that were vacuum packaged. The case was all brazed, made out of Kovar I believe, which is a metal that has the same coefficient of thermal expansion as glass so you can braze a metal plated glass window straight onto the metal. It ended up looking like a giant version of this with a window on top: www.maxwell.com/images/documents/hsn1000_rev3.pdf. The company folded before we were able to test the new packaged chip :( I really wonder if they would have worked, I think we were very close. The sampling caps on the analog front end are probably for the CDS (correlated double sampling) of the CCD output. Normal procedure is to sample the analog level, clock the CCD to dump the charge out, then sample again. CCD outputs often have variable DC offset, so the difference is proportional to the amount of charge on that pxiel. Proper analog front end chips have that function built in.
If you don't mind my asking: why were you trying to do it in-house when you could buy the cameras from a company that'd already solved the problems? I ask because I remember you saying you've worked for a medical imaging company and I would have thought that the _application_ of the cooled camera was the problem you'd be more interested in solving (as with the 454 DNA sequencer), rather than making the camera itself. Or was it that the off the shelf cameras didn't meet the performance specs you wanted?
We wanted to get the cost down, and have a more elegant system. The off the shelf (Apogee Alta) cameras weren't too bad at about $8k, but we would have been able to build them for about $2k. The Alta camera was USB and only had Windows drivers, and at the time we wanted to use a SOC CPU running Linux like a raspberry pi instead of an embedded PC. The camera dictated Windows 7 running on an industrial mini PC motherboard on the production systems.
The 4 "conditioning" areas before the ADCs are a correlated double sampling circuit. From the look of it given the nice film caps I'd say that it's using dual slope integration. The loads of linear regs and the DSP are generating the CCD clocking sequence and bias lines (VOD, VRD, VOG, image clock, register clock, possible summing well etc).
The primary reason for the heater behind the reaction chamber in the DNA sequencer is to hold the DNA polymerase reaction at 37 deg C. It seems to have worked out well that it pulls double duty keeping condensation away from the CCD light-coupling fibres.
The CCD is a CCD6161(edited as per part 2). Before you pump it down, you probably want to dry the desiccant out by warming it up to 40-45C and slowly sucking the air out(really slowly, on the order of 1l/hr). The vacuum seal is quite rare, but it is what you've described - a folding metal seal, possibly made out of indium or out of a hard alloy with a soft outer coating. I couldn't find anything remotely useful online in terms of prices, but it's not going to be cheap. Just stick a rubber o-ring in for now. Make sure it's close to the thickness of the groove, but slightly undersize. That'll ensure it squeezes out. The silicone vacuum grease is a must in either case, and that £15 tube of grease you mentioned will last you three lifetimes. You only need a very thin layer of slime. Oh, and also - loss of vacuum may occur through the valve. Make sure to flip the power the other way around when you're done, don't just disconnect it.
Interesting. You've both answered a question I've wondered for a while: why are genomic DNA sequencers so bloody slow?! In an ordinary DNA copy operation in the lab, you can copy ~1k bases in ~20 sec or so (and that's lots of margin for error, it can go much faster). These machines can take hours or days to run and they're only copying a couple hundred bases at the most. The photographing-the-slide part is the bottle neck that holds the process up and now I understand why :)
It's the combination of the requirements - large sizes, high resolutions, and low noise. Take one out, and you can do it much faster, but as I'm not an expert on DNA sequencing(I have no idea how it works), I assume all three are a must for near-100% correct sequence outputs.
Yes and no. The large size, high res and low noise will certainly give you the best results, but you _can_ do it with less. Basically, when the sequence is assembled from millions of individual pieces of genes being sequenced at the same time, there will be many many copies of each gene being sequenced together, so we can actually average out noise, copy errors and other issues. The algorithm that does the assembly looks at the overlapping pieces of sequence and "calls" a base based on what all the little snippets of sequence that overlap that part of the gene say. If a couple of them are wrong and say "A" when a few hundred others say "T" then they'll be ignored and that position will be called as a T. Obviously, as errors mount, we lose accuracy, so high res low noise systems give us much more _depth_ of coverage, which means we can be more confident on particular bases, or, more commonly trade off that depth to sequence more DNA (you can actually load multiple genomes onto these chips and trade off depth for shear sequence volume). I used to work on bacteria, which have much smaller genomes than humans, so we could routinely put six or eight samples on one run and still get usable results out of it. Without the low noise, high res camera, that probably would've been impossible.
That is one sexy CCD. Same amount of pixels as my Fuji XT-1. When I took apart an image intensified camera (I think military origins) the coupling between the CCD sensor and tube was a fiber optic. The front of the tube is also a fiber optic, still need to power it up, reverse engineered the driver, just need to pump a signal into the oscillator pin from a function gen.
That CCD sensor makes me think of the Trilogy Systems WSI (Wafer Scale Integration) attempts by Gene Amdahl back in the early 1980s. They attempted to basically build a "supercomputer on a wafer", as an attempt to solve the problem of latency of signals between chips. I have one of those prototype wafers. Absolutely intriguing thing... An entire wafer consisting of a single massive circuit, not configured in a segmented die format. Sadly, yield and scale bankrupted them. They never managed to succeed in creating a functioning unit, and went down in history as nothing but a vacuum for investor's money. The wafer photos on Wikipedia are all from my wafer. I think it might be fitting to see this wafer scale CCD sensor get added to the Wikipedia page, considering this one is a _SUCCESSFUL_ WSI device!
Those C seals are crazy expensive. They are often made from a high-performance alloy such as Inconel 625 or 718. They're also typically coated with a soft metal like silver or gold. The coating is responsible for 'flowing' into the surface finish of the surrounding components, thereby making the seal airtight. These types of seals are quite finicky and almost guaranteed to only work once when used for vacuum. During assembly they are crushed beyond yield which is how they conform to the sealing surface. A common o-ring should work just fine as a replacement. C-seals were possibly used to avoid contamination or outgassing into the vacuum over time.
PRO SKUB is correct they have a soft surface on them so they can take up any imperfection in the mating faces of the components we a using then at work for a new type of vale there is to types one for internal pressure sealing and one for external pressure sealing as they are activated by pressure so i would think when you pull the vacuum external atmospheric pressure will help the seal to activate. don't know if o rings will do the job as they require pressure to activate they might work by just filling the groove mechanically
The company would sell a lot more if they reduced the price to a more reasonable level, e.g., $50 for the sensor, and $75 for the rest of the camera housing/ components.
Razor2048, you clearly have no concept of the economics of silicon yields. There is no possible way to bring prices for this unit that low, no matter how many they made. Those cheap chips that you enjoy buying are cheap cause they can make hundreds of chips on each silicon wafer (a nearly perfectly polished disc of silicon). The more chips you can pack into a wafer, the better. It means that for the cost of one wafer, you get many more chips when you shrink them down. Often times, there are defects though, and a defect means a chip is scrapped. That's cost to make, that's not recovered. The good news, is if you pack a lot of chips into one wafer, there's less waste, and more good chips as a result. Here's where the problem comes in... That CCD sensor _IS the entire wafer_! If there's a defect... They throw the WHOLE WAFER away. There's no saving it. They have to make wafer after wafer till they get one that's _PERFECT_ Consider that the biggest chips you probably will encounter will be the CPU in your computer, and the GPU in it's graphics card, and consider even those chips are multiple chips on a wafer, and can cost between $100-1000 each... You can imagine how much it costs when you have to discard multiple failed wafers, till you get one good wafer out of them. Those CCD sensors probably cost a couple dozen grand a piece, not even counting the cost of all the other mechanical and electronic supporting components! I would _not_ expect them to even be able to make these for _$7500_, even if they were able to make them in large quantities! These units will remain five to six figures in total cost, and that's not likely to change any time soon. There's just simply no economical way to make that sensor. It's a prohibitively expensive part.
IBM had a similar design cooler for the CPU of the earlier 370 mainframes. Each MCM had twenty or so dice and each one had a coupling piston behind it filled with silicone oil to thermally couple them to the heatsink itself.
I guess they are kept so far apart from each other intentionally (If I am not mistaken they are from Australia, The USA and The UK). Imagine how the world would look now if these geniuses combined their intellectual resources O.o
Im surprised they had a hard time getting vacuum tight electrical feedthroughs, there exist a lot on the vacuum market. I even have some that would have worked.
The vacuum seal is most likely Indium. I've used that for sealing windows on a cryostat. It's well suited for seals with thermal cycling (the cryostat was mostly run from ambient to liquid N2, but could have gone down to liquid He as well). And yes, it is a one-time seal that has to be replaced. What is weird that you had some grease there as well. There is special vacuum grease (don't won't outgas), but as far as I remember the Indium seals should be applied to a completely cleaned surface and not greased.
That is the largest direct image sensor i have ever seen. With the right optics it would make an absolutely sick astro or high mag microscope camera. Preformance would be close to an intensified ccd but with less speckle noise. Build something cool with it 😁❤🤓
Hope you didn't blow the output transistors of the CCD, there are no intrinsic diodes for performance reasons. I found, the hard way, a balanced ion blower is needed to handle the devices; they're sensitive down to 50 V/M.
Peltier can be quite efficient up to a COP of 2 or 3 if the element is used with much less than rated power (3 to 5%). They are so inexpensive now that I do not understand why they are not used more under rating to make efficient solid state cooling devices. Two or 3 stage will be needed for good efficiency and decent temperature delta but that can be done easily with same type single stage peltier module supplied with different amount of power. My small peltier fridge designed to use in cars at 12V about 50W I use at 5V about 2A so 10W and is close to COP of 1. It will freeze the content is used at 12V and using a thermostat that turns off the power at set temperature will make it even less efficient since the peltier needs to run all the time else heat will be conducted trough the peltier module.
I guess, as a temporary sollution, instead of using rubber gaskets, you could try putting some leaded solder wire under or over that existing metal gasket. When the casing gets tightened, it will compress the solder in between the gasket and the metal cover itself. That might just do it. I't thinking some 0.5, 0.7 mm wire should be appropriate
Awww....How are you going to mix a pug with a manatee now that the 454 is gone? But that gorgeous monster CCD would be spectacular fitted to a telescope. I watched this before the 454 teardown... yeah, guys with big telescopes would love this.
Absolutely gorgeous tear down and analysis. Do you see any more 'obsolete' devices like this from other fields that might be coming to e-bay that would have other possible repurposing?
The CCD device looks like the Fairchild CCD486 Front illuminated version, based on your package and pin out: 2:12 www.tug.tubitak.gov.tr/dokumanlar/T100/CCD486DataSheetRevB.pdf
How pure does the vaccum have to be? Instead of the seal leaking, it might have been some of your solder flux evaporating? Also sometimes by careful bending action, these metal seals can be reused a second time.
He could've just stuck a paper shim under. I've done that a lot in testing on vacuum rigs with metal o-rings. You don't want to break an expensive o-ring trying to get the proper result, so you reuse the damned thing for all its worth, tho, i admit, it's a bit fiddle to cut the exact paper shape.
Can you find coiled copper tubing of the right thickness that would fit in the channel? if you preserve the radius of the tubing, carefully cut a circle, and prepare/solder/braze the ends, would that make an equivalent gasket?
You know it occurs to me that some vacuum grade epoxy might work to seal that again. I admittedly don't have a lot of experience in the area of electronics that operate in a vacuum. I do know a bit about pressure seals though and epoxy seems the logical choice. There's also the possibility of O rings being a good option to affirm your idea. Especially those which are considered high compression. One with a large diameter should provide a sufficient seal. I think a nice thick one is a good solution.
That is reusable seal, i have used that one in high vacuum conditions (2x2000l/Diff pumps) many times between diff pumped chambers... it is flexible enough to seal few times. although in my case the seal was light, system did had a hole (on purpose) for ion beam...
as a heads up, the camera will probably give you fits images. you can open them in a program called fits liberator, or fitswork (have not tested that, I'll do so tomarow).
I'm pretty sure this camera dumps raw readout data, unless explicitly programmed to do otherwise through the PC interface. This is miles beyond amateur astronomy stuff. All the actual data handling and formatting is done on the PC (as Mike pointed out).
Mythricia I stated the above, after reading a bit about spectral instruments cameras. Also Fits is not exclusively for armatures. Its used by most large space and ground based telescopes, such as Hubble. Check the 800 series rev B Brochure: "Software included with every camera; SI Image software suite for camera control, data manipulation and archiving. Native file format is FITS..."
I think what they mean by that is that the entire suite is based on a FITS workflow. Which makes sense. But I was just saying that the camera doesn't just pop out a FITS file on the other side - that FITS file is created _by_ the software suite, after it receives the data. Which is unlike a lot of astro cameras, they create the full file in-camera and spit it out over USB or Serial. The above is true for Hubble and all the large telescopes of the world too - they transport and handle the images in FITS format, but the actual cameras they use are very very much unaware of what FITS is, or any format at all for that matter. Worth mentioning that the full resolution images from the HiRISE camera onboard the Mars Reconnaissance Orbiter are stored as JPEG2000. FITS is not that universal in the big-boys game! They have far too specific needs.
This would be great for dark sky imaging with the appropriate lenses. I wonder what other products which can be found on eBay/scrap sites would have similar cooled CCDs... :)
for tight spaces you should get one of the cheapo boroscopes to connect to you phone from china.Should do the trick.I am unsure if you can actually record video with them
Also don't bother with lead or solder seals. Won't work, it neither self-amalgamates or conforms well enough and the flux core in solder will cause loads of problems. (also i tried it, it doesn't work ahah). it has to be indium or gold
Now what's the light sensitivity on that! The ccd is like 4+ times as big as a 35mm / full frame. GL finding a lens for that :D I'd Imagine you'd need something from Boeing (spy satellites)
Eh, there's a few large lenses around, besides, for something not really worth going full combat in to, you could probably rig up something from various parts. I have seen, for example, a small surveillance lens (something russian, the name escapes my mind now, but it's pretty common there) mounted on a Canon DLSR with the help of an eBay adapter, and the results were pretty good (the idea was to get this tiny lens, in OD, through a very small hole, it was an architectural survey). I mean sure, it's not the closest comparison, but i think it's doable.
Lots of medium format film lenses are designed for >= 60x60mm e,g, Hasselblad V system or Mamiya. You could make a wicked digital back for a medium format camera...
Andrew Basterfield If the budget is lower than Hasselblad or Mamiya then there are always Pentacon 6 lenses or Kiev 88 lenses....they are the eastern block versions.
Cool.. I would probably never have the currency to buy anything with medium format (digital), so I'm pretty oblivious about the lenses.. Good that they are available :P
Large format analogue lenses may be a cheaper option. Old process cameras also had decent lenses. There is also the possibility that this or some similar systems used direct contact to read fluorescent markers in a gel plate which is another possible reason for having the optical fibre face place, lets you use non-imaging optics.
It is unusual to have grease on a metal vacuum seal. normally hard knife edges compress something like a copper gasget. rubber o-ring seals will have grease (a tube of which i have... but i won't be able to get to it until late feb next year :( ) Be careful with cheap ebay o-rings though, i had a set once that were full of some awefull smelly brown oily stuff that outgassed all over the place. thought "ha! i'll soak them in acetone to remove it all" yeah well, that did remove the oil, and they turned hard as a rock and brittle. www.lesker.com/ were good last time i used them ~13 years ago and will sell you any size of o-ring you want. Another option if you really can't find a good replacement for the original seals is indium wire (lesker sell that too) be sure to thoroughly degrease the surfaces first though with some suitably evil solvent ;)
I am somehow surprised there is no heat spreader and metal mechanical support at the back of the sensor. The fact that the cylindrical copper cool block only touches the center with a spring tension, is weird. I imagine there will be some temperature gradient on the sensor because of that.
Yes - also a bit surprised to see Meanwell PSUs in the PSU box - not too bad but for that price I'd have expected better. At least they're easy to replace.
I've designed Meanwell supplies into consumer storage gear and while they're very conservatively rated and leagues better than alternatives from Shenzhen, I was floored to see them used in a bit of kit costing half a million.
It looks like they are using row after row of SMD capacitors standing on end to attach each cooler to the next? Looks like lager Caps and smaller caps on the bottom board. Is that what they are?
How long of an exposure do they need in order to get a decent image, and will that sensor be able to be adapter for any other kind of lens system since it seems to designed for taking super macro shots.
It's extremely sensitive, so it'll need shorter exposures than any consumer camera to get the same brightness. Probably about half the time, compared to a consumer DSLR. This is where CCD's shine - they make CMOS look pretty horrendous in comparison.
"how many wafers before a good die" made me think about some giant image sensors canon made for a telescope that occupy a full 300mm wafer... I just wonder how much do they cost... You can see it here:www.canon.com/technology/future/cmos.html
What is the sensitivity of this camera compared to a more modern CCD sensor? Does it's performance still exceed consumer gear? The reason I ask is that consumer 35mm style cameras appear to have incredibly sensitive sensors these days. For astronomy use there are of course special requirements such as very long exposure times...perhaps the cooled camera would still be superior to a consumer grade camera??
This camera is vastly, vastly, VASTLY more sensitive than any consumer camera sensor. They are toys in comparison, it's not even close. CCD sensors have always ruled in the sensitivity and noise game - for almost as long as they've been around. And this is not just a CCD, it's a back-illuminated CCD, there's no such thing in the consumer sensor world. The only exception would be the Sony a7R II, which in 2015 was the world’s _first_ Back-Illuminated 35mm Full Frame Sensor used in a consumer camera. But it's still less sensitive, because it's not a CCD, and it's a colour-array sensor, this is not, this is purely monochrome I'm pretty sure. That distinction alone will make this sensor at least 30-40% more sensitive than the Sony. And, you know, there's also the whole thing with the pixels on this sensor being literally over three times as large, making them much less prone to noise. The inherent signal-to-noise ratio of an individual pixel increases by the square root of the area, so with 9 times the area, you'll have 3 times better Signal/Noise. Ish. Roughly. Napkin math. But it'll be a lot better... Oh and I guess the fact this is cooled, as well, reducing the noise yet again by a huge amount. Which no, absolutely no consumer DSLR/Full Frame/Any other format camera does. Basically, the answer is: This camera makes any and all consumer grade sensors look like toys. Not even close.
The grease may be something like Apiezon. www.apiezon.com/products/vacuum-greases they offer a low creep, low outgassing grease for high vacuums. They also have a cryogenic formulation. They claim they get used by NASA.
Wonder how this would perform as a IR/night vision camera, noise is a really big issue with that sort of thing. Would be a good test when you get the camera up and running.
Should be great for that - very large pixel pitch, extremely low noise, extremely high quantum efficiency, etc... So, see no reason why it wouldn't be incredibly good for low light or IR, with an appropriate filter. Even consumer DSLR's with removed IR filters are perfectly good for that sort of thing.
cutoff wavelength probably too low to observe in thermal IR. for this you would go with CMOS actively cooled to cryogenic temperatures. peltier stack wouldn't make it
Now when some camera nut (who doesn't properly know how to use the thing except in auto mode) braggs about the size of the image sensor in their new camera and he whips this beauty out.
You want someone to send you some grease because its quite expensive... whilst you are sitting there with a piece of kit you just said was worth 20grand! What are you, Scottish?
agreed, but someone like me, with access to a 0.8 meter telescope (and a 14"), that camera would probably work for those. However, on my own telescope, there is no chance of using that camera. Also, you need to keep in mind, you need narrow band filters that are larger then the camera sensor, to even get color out of it.
Average? No, certainly not, tho, i can envision a frame to mount it on a small telescope (well, i say small, i mean decently sized). Nothing is impossible.
aserta keep in mind, the focusers have a max carry weight, if you exceed it, the will break, and the camera will fall. The focuser on my tiny 6" wont be able to hold it. Im not certain, but i think the C14's focuser might. The focuer of that 0.8 meter telescope i was talking about, will defiantly be able to hold it. Also keep in mind, the mount will also need to be able to carry the camera as well.
Custom frame could solve that probably, although I fail to see much reason to put this on anything smaller than a 14" tube. I mean yeah, it'd be cool to take some wide field ultra-low-noise, ultra-high-sensitivity shots, but y'know... With a camera like this you better go hunting for those 13 billion light years away galaxies to set some new amateur record or something. Although I think the amateur record is already over 13B LY....
In a previous job I was working on designing a cooled CCD camera somewhat like this. It was a really tough job, one that we never completed, so we stuck with the off the shelf cooled cameras we were already using. The two biggest issues were frost on the CCD, and getting reliable vacuum feedthroughs. We were originally doing a gas filled camera, and any tiny amount of moisture will condense into frost when the CCD is at -20C. The plan was to flow dry nitrogen through the chamber while baking, then hermetically seal it, along with desiccant inside. Vaccuum/hermetic feedthroughs were a major problem to source and attach to the case. We ended up having a company that specializes in this kind of thing build the CCDs, peltiers, and temp sensor into modules that were vacuum packaged. The case was all brazed, made out of Kovar I believe, which is a metal that has the same coefficient of thermal expansion as glass so you can braze a metal plated glass window straight onto the metal. It ended up looking like a giant version of this with a window on top: www.maxwell.com/images/documents/hsn1000_rev3.pdf. The company folded before we were able to test the new packaged chip :( I really wonder if they would have worked, I think we were very close.
The sampling caps on the analog front end are probably for the CDS (correlated double sampling) of the CCD output. Normal procedure is to sample the analog level, clock the CCD to dump the charge out, then sample again. CCD outputs often have variable DC offset, so the difference is proportional to the amount of charge on that pxiel. Proper analog front end chips have that function built in.
If you don't mind my asking: why were you trying to do it in-house when you could buy the cameras from a company that'd already solved the problems? I ask because I remember you saying you've worked for a medical imaging company and I would have thought that the _application_ of the cooled camera was the problem you'd be more interested in solving (as with the 454 DNA sequencer), rather than making the camera itself. Or was it that the off the shelf cameras didn't meet the performance specs you wanted?
We wanted to get the cost down, and have a more elegant system. The off the shelf (Apogee Alta) cameras weren't too bad at about $8k, but we would have been able to build them for about $2k. The Alta camera was USB and only had Windows drivers, and at the time we wanted to use a SOC CPU running Linux like a raspberry pi instead of an embedded PC. The camera dictated Windows 7 running on an industrial mini PC motherboard on the production systems.
Hm, I thought most Apogee systems offered ethernet control, at least as an option on most of their cameras.
Ah, okay. More flexibility and a quarter of the price seems like a good reason to give it a go :)
The 4 "conditioning" areas before the ADCs are a correlated double sampling circuit. From the look of it given the nice film caps I'd say that it's using dual slope integration. The loads of linear regs and the DSP are generating the CCD clocking sequence and bias lines (VOD, VRD, VOG, image clock, register clock, possible summing well etc).
The primary reason for the heater behind the reaction chamber in the DNA sequencer is to hold the DNA polymerase reaction at 37 deg C. It seems to have worked out well that it pulls double duty keeping condensation away from the CCD light-coupling fibres.
Haven't even watched the vid and it has already brightened my day!
I'll walk the dog first, so I can sit back and enjoy.
Cheers Mike.
Mike, to seal the unit you are looking for a hollow crush gasket. Hope this helps.
Exciting :)
This is some serious camera. I hope you get that running.
He did.
The CCD is a CCD6161(edited as per part 2).
Before you pump it down, you probably want to dry the desiccant out by warming it up to 40-45C and slowly sucking the air out(really slowly, on the order of 1l/hr).
The vacuum seal is quite rare, but it is what you've described - a folding metal seal, possibly made out of indium or out of a hard alloy with a soft outer coating. I couldn't find anything remotely useful online in terms of prices, but it's not going to be cheap. Just stick a rubber o-ring in for now. Make sure it's close to the thickness of the groove, but slightly undersize. That'll ensure it squeezes out.
The silicone vacuum grease is a must in either case, and that £15 tube of grease you mentioned will last you three lifetimes. You only need a very thin layer of slime.
Oh, and also - loss of vacuum may occur through the valve. Make sure to flip the power the other way around when you're done, don't just disconnect it.
Wow, they make a 10k x 10k chip, almost 10x10cm! At the recommended 50kHz low noise clock speed, it takes 2 minutes to read out.
tesla500 That's the physically biggest commercially available sensor as far as I know. There are higher resolution ones from GPIXEL.
Interesting. You've both answered a question I've wondered for a while: why are genomic DNA sequencers so bloody slow?! In an ordinary DNA copy operation in the lab, you can copy ~1k bases in ~20 sec or so (and that's lots of margin for error, it can go much faster). These machines can take hours or days to run and they're only copying a couple hundred bases at the most. The photographing-the-slide part is the bottle neck that holds the process up and now I understand why :)
It's the combination of the requirements - large sizes, high resolutions, and low noise.
Take one out, and you can do it much faster, but as I'm not an expert on DNA sequencing(I have no idea how it works), I assume all three are a must for near-100% correct sequence outputs.
Yes and no. The large size, high res and low noise will certainly give you the best results, but you _can_ do it with less. Basically, when the sequence is assembled from millions of individual pieces of genes being sequenced at the same time, there will be many many copies of each gene being sequenced together, so we can actually average out noise, copy errors and other issues. The algorithm that does the assembly looks at the overlapping pieces of sequence and "calls" a base based on what all the little snippets of sequence that overlap that part of the gene say. If a couple of them are wrong and say "A" when a few hundred others say "T" then they'll be ignored and that position will be called as a T. Obviously, as errors mount, we lose accuracy, so high res low noise systems give us much more _depth_ of coverage, which means we can be more confident on particular bases, or, more commonly trade off that depth to sequence more DNA (you can actually load multiple genomes onto these chips and trade off depth for shear sequence volume). I used to work on bacteria, which have much smaller genomes than humans, so we could routinely put six or eight samples on one run and still get usable results out of it. Without the low noise, high res camera, that probably would've been impossible.
I love this teardown. Thanks for uploading it
That is one sexy CCD. Same amount of pixels as my Fuji XT-1. When I took apart an image intensified camera (I think military origins) the coupling between the CCD sensor and tube was a fiber optic. The front of the tube is also a fiber optic, still need to power it up, reverse engineered the driver, just need to pump a signal into the oscillator pin from a function gen.
That CCD sensor makes me think of the Trilogy Systems WSI (Wafer Scale Integration) attempts by Gene Amdahl back in the early 1980s. They attempted to basically build a "supercomputer on a wafer", as an attempt to solve the problem of latency of signals between chips. I have one of those prototype wafers. Absolutely intriguing thing... An entire wafer consisting of a single massive circuit, not configured in a segmented die format. Sadly, yield and scale bankrupted them. They never managed to succeed in creating a functioning unit, and went down in history as nothing but a vacuum for investor's money.
The wafer photos on Wikipedia are all from my wafer. I think it might be fitting to see this wafer scale CCD sensor get added to the Wikipedia page, considering this one is a _SUCCESSFUL_ WSI device!
I love the pro tips like pulling the wire. Thank you very much!
Something I was taught back in '75 and have passed it on many times since. This is the first time I've seen it demonstrated online.
Those C seals are crazy expensive. They are often made from a high-performance alloy such as Inconel 625 or 718. They're also typically coated with a soft metal like silver or gold. The coating is responsible for 'flowing' into the surface finish of the surrounding components, thereby making the seal airtight. These types of seals are quite finicky and almost guaranteed to only work once when used for vacuum. During assembly they are crushed beyond yield which is how they conform to the sealing surface. A common o-ring should work just fine as a replacement. C-seals were possibly used to avoid contamination or outgassing into the vacuum over time.
These were greased so I don't think any funny surface stuff was going on
PRO SKUB is correct they have a soft surface on them so they can take up any imperfection in the mating faces of the components we a using then at work for a new type of vale there is to types one for internal pressure sealing and one for external pressure sealing as they are activated by pressure so i would think when you pull the vacuum external atmospheric pressure will help the seal to activate. don't know if o rings will do the job as they require pressure to activate they might work by just filling the groove mechanically
26:15 that is a brilliant life hack! Im going to start doing that to all my rolls of wiring thanks for the pro tip!
Damn, what is the manufacturing yield on that sensor?
If you have to ask, you can't afford it.
Bear in mind feature size is pretty big ( ~14um pixel) so possibly less sensitive to flaws than other stuff.
low
The company would sell a lot more if they reduced the price to a more reasonable level, e.g., $50 for the sensor, and $75 for the rest of the camera housing/ components.
As Mike mentioned, it's wafers per functional die rather than functional dies per wafer.
Razor2048, you clearly have no concept of the economics of silicon yields. There is no possible way to bring prices for this unit that low, no matter how many they made. Those cheap chips that you enjoy buying are cheap cause they can make hundreds of chips on each silicon wafer (a nearly perfectly polished disc of silicon). The more chips you can pack into a wafer, the better. It means that for the cost of one wafer, you get many more chips when you shrink them down. Often times, there are defects though, and a defect means a chip is scrapped. That's cost to make, that's not recovered. The good news, is if you pack a lot of chips into one wafer, there's less waste, and more good chips as a result.
Here's where the problem comes in... That CCD sensor _IS the entire wafer_! If there's a defect... They throw the WHOLE WAFER away. There's no saving it. They have to make wafer after wafer till they get one that's _PERFECT_ Consider that the biggest chips you probably will encounter will be the CPU in your computer, and the GPU in it's graphics card, and consider even those chips are multiple chips on a wafer, and can cost between $100-1000 each... You can imagine how much it costs when you have to discard multiple failed wafers, till you get one good wafer out of them.
Those CCD sensors probably cost a couple dozen grand a piece, not even counting the cost of all the other mechanical and electronic supporting components!
I would _not_ expect them to even be able to make these for _$7500_, even if they were able to make them in large quantities! These units will remain five to six figures in total cost, and that's not likely to change any time soon. There's just simply no economical way to make that sensor. It's a prohibitively expensive part.
IBM had a similar design cooler for the CPU of the earlier 370 mainframes. Each MCM had twenty or so dice and each one had a coupling piston behind it filled with silicone oil to thermally couple them to the heatsink itself.
Reminds me of the ccd's that are used in professional motion picture equiptment for wide screen imax productions.
I think this CCD is a bit bigger than that
Applied Science (Ben) did a video for metal "o-rigns" for vacuum.
I guess they are kept so far apart from each other intentionally (If I am not mistaken they are from Australia, The USA and The UK). Imagine how the world would look now if these geniuses combined their intellectual resources O.o
Im surprised they had a hard time getting vacuum tight electrical feedthroughs, there exist a lot on the vacuum market. I even have some that would have worked.
The vacuum seal is most likely Indium. I've used that for sealing windows on a cryostat. It's well suited for seals with thermal cycling (the cryostat was mostly run from ambient to liquid N2, but could have gone down to liquid He as well). And yes, it is a one-time seal that has to be replaced.
What is weird that you had some grease there as well. There is special vacuum grease (don't won't outgas), but as far as I remember the Indium seals should be applied to a completely cleaned surface and not greased.
That is the largest direct image sensor i have ever seen. With the right optics it would make an absolutely sick astro or high mag microscope camera. Preformance would be close to an intensified ccd but with less speckle noise. Build something cool with it 😁❤🤓
Hope you didn't blow the output transistors of the CCD, there are no intrinsic diodes for performance reasons. I found, the hard way, a balanced ion blower is needed to handle the devices; they're sensitive down to 50 V/M.
Very interesting hardware and good mechanical tips Mike! Thx
Mike I think those type of seals are called "flexitallic"
This must be an astronomer's wet dream.
Peltier can be quite efficient up to a COP of 2 or 3 if the element is used with much less than rated power (3 to 5%). They are so inexpensive now that I do not understand why they are not used more under rating to make efficient solid state cooling devices. Two or 3 stage will be needed for good efficiency and decent temperature delta but that can be done easily with same type single stage peltier module supplied with different amount of power.
My small peltier fridge designed to use in cars at 12V about 50W I use at 5V about 2A so 10W and is close to COP of 1.
It will freeze the content is used at 12V and using a thermostat that turns off the power at set temperature will make it even less efficient since the peltier needs to run all the time else heat will be conducted trough the peltier module.
If it ain't broke, take it apart and find out why not!
If it ain't broke, take it apart and make sure it stays that way.
Very interesting, love your vids.
Really hope you can get this working!
I think I'll buy one of these sequencers, just for one of those cameras.
Seems to be a few around - one at $99 in the US at the moment
mikeselectricstuff
i cant seem to find it, could you PM me a link?
I guess, as a temporary sollution, instead of using rubber gaskets, you could try putting some leaded solder wire under or over that existing metal gasket. When the casing gets tightened, it will compress the solder in between the gasket and the metal cover itself. That might just do it. I't thinking some 0.5, 0.7 mm wire should be appropriate
Awww....How are you going to mix a pug with a manatee now that the 454 is gone? But that gorgeous monster CCD would be spectacular fitted to a telescope. I watched this before the 454 teardown... yeah, guys with big telescopes would love this.
Absolutely gorgeous tear down and analysis. Do you see any more 'obsolete' devices like this from other fields that might be coming to e-bay that would have other possible repurposing?
The CCD device looks like the Fairchild CCD486 Front illuminated version, based on your package and pin out: 2:12
www.tug.tubitak.gov.tr/dokumanlar/T100/CCD486DataSheetRevB.pdf
Real fullframeness.
How pure does the vaccum have to be? Instead of the seal leaking, it might have been some of your solder flux evaporating? Also sometimes by careful bending action, these metal seals can be reused a second time.
He could've just stuck a paper shim under. I've done that a lot in testing on vacuum rigs with metal o-rings. You don't want to break an expensive o-ring trying to get the proper result, so you reuse the damned thing for all its worth, tho, i admit, it's a bit fiddle to cut the exact paper shape.
aserta paper? I would use some metal foil
Yeah, I was thinking a few layers of aluminium foil cut into continuous rings. I also didn't take care to clean the seals and faces before reassembly
Can you find coiled copper tubing of the right thickness that would fit in the channel? if you preserve the radius of the tubing, carefully cut a circle, and prepare/solder/braze the ends, would that make an equivalent gasket?
You know it occurs to me that some vacuum grade epoxy might work to seal that again. I admittedly don't have a lot of experience in the area of electronics that operate in a vacuum. I do know a bit about pressure seals though and epoxy seems the logical choice. There's also the possibility of O rings being a good option to affirm your idea. Especially those which are considered high compression. One with a large diameter should provide a sufficient seal. I think a nice thick one is a good solution.
I like star grounding a lot. You don't want a ground plane acting like an analogue computer.
bit out of the loop here, why is the camera in a vacuum? is that just about stability?
That is reusable seal, i have used that one in high vacuum conditions (2x2000l/Diff pumps) many times between diff pumped chambers... it is flexible enough to seal few times. although in my case the seal was light, system did had a hole (on purpose) for ion beam...
you should consider breaking the mini desoldering tutorial out into its own video, quite interesting
It's easy to zap those large CCDs - they are very finicky about static. Be careful!
Hook it up to a telescope. Low noise + telescope = great pictures
Lead will make a good seal.
TheBarret No
yes
I almost yelled at my screen when I saw you drag that magnet past the CCD. Weren't you worried about damaging it with eddy currents?
Can you not use a flat screwdriver to prise the seal back out again so it gets crushed again? Or does it feel like it's use once?
Looks like they did versions with coated optics to enhance more towards UV.
as a heads up, the camera will probably give you fits images. you can open them in a program called fits liberator, or fitswork (have not tested that, I'll do so tomarow).
I'm pretty sure this camera dumps raw readout data, unless explicitly programmed to do otherwise through the PC interface. This is miles beyond amateur astronomy stuff. All the actual data handling and formatting is done on the PC (as Mike pointed out).
Mythricia
I stated the above, after reading a bit about spectral instruments cameras. Also Fits is not exclusively for armatures. Its used by most large space and ground based telescopes, such as Hubble.
Check the 800 series rev B Brochure:
"Software included with every camera;
SI Image software suite
for camera control, data manipulation and archiving. Native file
format is FITS..."
I think what they mean by that is that the entire suite is based on a FITS workflow. Which makes sense. But I was just saying that the camera doesn't just pop out a FITS file on the other side - that FITS file is created _by_ the software suite, after it receives the data. Which is unlike a lot of astro cameras, they create the full file in-camera and spit it out over USB or Serial.
The above is true for Hubble and all the large telescopes of the world too - they transport and handle the images in FITS format, but the actual cameras they use are very very much unaware of what FITS is, or any format at all for that matter. Worth mentioning that the full resolution images from the HiRISE camera onboard the Mars Reconnaissance Orbiter are stored as JPEG2000. FITS is not that universal in the big-boys game! They have far too specific needs.
Mythricia
yes, with that clarification, i fully agree with you.
This would be great for dark sky imaging with the appropriate lenses. I wonder what other products which can be found on eBay/scrap sites would have similar cooled CCDs... :)
for tight spaces you should get one of the cheapo boroscopes to connect to you phone from china.Should do the trick.I am unsure if you can actually record video with them
Looks like the Dalsa or Kodak CCDs used in digital medium format photography
I want to see astrophotography images captured on that sensor!
That seal could be what's called a Wills Ring
Also don't bother with lead or solder seals. Won't work, it neither self-amalgamates or conforms well enough and the flux core in solder will cause loads of problems. (also i tried it, it doesn't work ahah). it has to be indium or gold
I wonder what the image quality on these would be? Any idea what the bit depth is on the CCD? Seems like it could be great for astrophotography.
Now what's the light sensitivity on that! The ccd is like 4+ times as big as a 35mm / full frame. GL finding a lens for that :D I'd Imagine you'd need something from Boeing (spy satellites)
Eh, there's a few large lenses around, besides, for something not really worth going full combat in to, you could probably rig up something from various parts.
I have seen, for example, a small surveillance lens (something russian, the name escapes my mind now, but it's pretty common there) mounted on a Canon DLSR with the help of an eBay adapter, and the results were pretty good (the idea was to get this tiny lens, in OD, through a very small hole, it was an architectural survey).
I mean sure, it's not the closest comparison, but i think it's doable.
Lots of medium format film lenses are designed for >= 60x60mm e,g, Hasselblad V system or Mamiya. You could make a wicked digital back for a medium format camera...
Andrew Basterfield If the budget is lower than Hasselblad or Mamiya then there are always Pentacon 6 lenses or Kiev 88 lenses....they are the eastern block versions.
Cool.. I would probably never have the currency to buy anything with medium format (digital), so I'm pretty oblivious about the lenses.. Good that they are available :P
Large format analogue lenses may be a cheaper option. Old process cameras also had decent lenses. There is also the possibility that this or some similar systems used direct contact to read fluorescent markers in a gel plate which is another possible reason for having the optical fibre face place, lets you use non-imaging optics.
It is unusual to have grease on a metal vacuum seal. normally hard knife edges compress something like a copper gasget. rubber o-ring seals will have grease (a tube of which i have... but i won't be able to get to it until late feb next year :( ) Be careful with cheap ebay o-rings though, i had a set once that were full of some awefull smelly brown oily stuff that outgassed all over the place. thought "ha! i'll soak them in acetone to remove it all" yeah well, that did remove the oil, and they turned hard as a rock and brittle. www.lesker.com/ were good last time i used them ~13 years ago and will sell you any size of o-ring you want. Another option if you really can't find a good replacement for the original seals is indium wire (lesker sell that too) be sure to thoroughly degrease the surfaces first though with some suitably evil solvent ;)
I've got some high vacuum silicone grease lying around somewhere....
really cool!
Request vs iphone camera test!
I am somehow surprised there is no heat spreader and metal mechanical support at the back of the sensor. The fact that the cylindrical copper cool block only touches the center with a spring tension, is weird. I imagine there will be some temperature gradient on the sensor because of that.
Doubt it. In vacuum with a powerful Peltier element, it'll all get to equilibrium fairly quickly.
How soft is the metal seal? If you're lucky it could be indium!
not very - feels like aluminium
Surprised to see a crappy noisy 40mm fan on something like that, would have expected something better.
Yes - also a bit surprised to see Meanwell PSUs in the PSU box - not too bad but for that price I'd have expected better. At least they're easy to replace.
I've designed Meanwell supplies into consumer storage gear and while they're very conservatively rated and leagues better than alternatives from Shenzhen, I was floored to see them used in a bit of kit costing half a million.
If they can squeeze $5 out of a BOM for a half mil piece of kit, they will.
How much did you pay for all that equipment in your workshop?
At the end it looks like you opened it up from the wrong end again. How did you do that without breaking the wires again?
I started at the right end this time, unplugged the PCB from the CCD then opened the front. Wires a little bent but not broken
I am surprised there is not a copper or aluminum plate on the back of the CCD to equalize the temperature over the whole CCD.
Ceramic is pretty conductive, and the die is probably on a metal paddle
Hear the sustain on this thing.. Was anyone self thinking spinal tap?? Does it go up to 11!?
-Listen to it!
-I don't hear it...
-Well you would, if i was playing
It looks like they are using row after row of SMD capacitors standing on end to attach each cooler to the next? Looks like lager Caps and smaller caps on the bottom board. Is that what they are?
They are the peltier elements
Cool I was wondering what they looked like inside. An array of alternating n- and p- type semiconductors. I should have just Googled it.
How long of an exposure do they need in order to get a decent image, and will that sensor be able to be adapter for any other kind of lens system since it seems to designed for taking super macro shots.
It's extremely sensitive, so it'll need shorter exposures than any consumer camera to get the same brightness. Probably about half the time, compared to a consumer DSLR. This is where CCD's shine - they make CMOS look pretty horrendous in comparison.
Its a Helicoflex hasket: www.prophysik.li/index.php?article_id=9&clang=1
"how many wafers before a good die" made me think about some giant image sensors canon made for a telescope that occupy a full 300mm wafer... I just wonder how much do they cost... You can see it here:www.canon.com/technology/future/cmos.html
can anyone tell me how much would this cost?
I hope it didn't break just because you took it apart.
What is the sensitivity of this camera compared to a more modern CCD sensor? Does it's performance still exceed consumer gear? The reason I ask is that consumer 35mm style cameras appear to have incredibly sensitive sensors these days. For astronomy use there are of course special requirements such as very long exposure times...perhaps the cooled camera would still be superior to a consumer grade camera??
I don't know - it may be that the main advantage is ability to do long exposures
This camera is vastly, vastly, VASTLY more sensitive than any consumer camera sensor. They are toys in comparison, it's not even close. CCD sensors have always ruled in the sensitivity and noise game - for almost as long as they've been around. And this is not just a CCD, it's a back-illuminated CCD, there's no such thing in the consumer sensor world. The only exception would be the Sony a7R II, which in 2015 was the world’s _first_ Back-Illuminated 35mm Full Frame Sensor used in a consumer camera. But it's still less sensitive, because it's not a CCD, and it's a colour-array sensor, this is not, this is purely monochrome I'm pretty sure. That distinction alone will make this sensor at least 30-40% more sensitive than the Sony. And, you know, there's also the whole thing with the pixels on this sensor being literally over three times as large, making them much less prone to noise. The inherent signal-to-noise ratio of an individual pixel increases by the square root of the area, so with 9 times the area, you'll have 3 times better Signal/Noise. Ish. Roughly. Napkin math. But it'll be a lot better... Oh and I guess the fact this is cooled, as well, reducing the noise yet again by a huge amount. Which no, absolutely no consumer DSLR/Full Frame/Any other format camera does.
Basically, the answer is: This camera makes any and all consumer grade sensors look like toys. Not even close.
Mythricia Thanks, that is exciting news about the sensitivity! What do you think the ASA/ISO rating would be?
The grease may be something like Apiezon. www.apiezon.com/products/vacuum-greases they offer a low creep, low outgassing grease for high vacuums. They also have a cryogenic formulation. They claim they get used by NASA.
Wonder how this would perform as a IR/night vision camera, noise is a really big issue with that sort of thing.
Would be a good test when you get the camera up and running.
Should be great for that - very large pixel pitch, extremely low noise, extremely high quantum efficiency, etc... So, see no reason why it wouldn't be incredibly good for low light or IR, with an appropriate filter. Even consumer DSLR's with removed IR filters are perfectly good for that sort of thing.
cutoff wavelength probably too low to observe in thermal IR. for this you would go with CMOS actively cooled to cryogenic temperatures. peltier stack wouldn't make it
why is vacuum needed?
To stop heat convection warming up the CCD. Not relevant for this unit but for non-fibre coupled units it also prevents condensation
Why do these need a vaccum?
Normally to stop heat convection warming up the CCD and prevent condensation.
That looks bloody expensive.
Now when some camera nut (who doesn't properly know how to use the thing except in auto mode) braggs about the size of the image sensor in their new camera and he whips this beauty out.
You want someone to send you some grease because its quite expensive... whilst you are sitting there with a piece of kit you just said was worth 20grand! What are you, Scottish?
that's a great CCD camera!but i dont think average astronomy amateurs can use it.too heavy to mount on a small dia telescope
btw that heat sink sounded beautiful!
agreed, but someone like me, with access to a 0.8 meter telescope (and a 14"), that camera would probably work for those. However, on my own telescope, there is no chance of using that camera. Also, you need to keep in mind, you need narrow band filters that are larger then the camera sensor, to even get color out of it.
Average? No, certainly not, tho, i can envision a frame to mount it on a small telescope (well, i say small, i mean decently sized). Nothing is impossible.
aserta
keep in mind, the focusers have a max carry weight, if you exceed it, the will break, and the camera will fall. The focuser on my tiny 6" wont be able to hold it. Im not certain, but i think the C14's focuser might. The focuer of that 0.8 meter telescope i was talking about, will defiantly be able to hold it. Also keep in mind, the mount will also need to be able to carry the camera as well.
Custom frame could solve that probably, although I fail to see much reason to put this on anything smaller than a 14" tube. I mean yeah, it'd be cool to take some wide field ultra-low-noise, ultra-high-sensitivity shots, but y'know... With a camera like this you better go hunting for those 13 billion light years away galaxies to set some new amateur record or something. Although I think the amateur record is already over 13B LY....
hdhd