Going full circle with filter wheels, or how to accept your destiny without getting into too much of a spin….

By 11th May 2016Presidents Log

Once again I find myself writing a blog from Plymouth, this time for two reasons.  First, it’s much easier to find the time for this sort of thing here than it is back home in Faversham, Second and perhaps more importantly, this one is mainly going to be about what seems to have become a lifetime involvement with filterwheels, and this Plymouth outing effectively marks what is to all intents and purposes the completion of the Optospin IV project after a length of time that I can only regard as embarrassing.

Optospin VI

Optospin VI

However, before I embark on the entire filterwheel story in all its sordid detail, I thought I should plug a couple of other articles that I’d put together since my last contribution here [A tale of two buildings].   The first arose indirectly because we recently made a substantial update to our website, in respect of both content and usability, so that as well as including a lot more “stuff”, it now also formats itself nicely on other “viewing media” such as tablets, phones and digital watches (probably???), and with the direct brain implant version ready to go as soon as the hardware becomes more easily removable.  This activity appeared to be driven by what I can only describe as a Damascene conversion by one or two Cairn personnel who had previously been rather sceptical about the power of “digital marketing”, and although I’d have loved to have been more involved than I actually was, it seemed to me that the most sensible thing to do under these very welcome circumstances was just to let them get on with it!

However, that left me with the opposite of writer’s block, which for polite reasons I shall refer to by its German equivalent of Durchfall,  in that I had a very urgent need to contribute something myself somewhere.  As part of Cairn’s attempts to make our existence more widely known, James our Marketing Director had been negotiating with Laboratory News for placing product announcements and perhaps even an advert there.  This reminded me that all such publications are also potentially hungry for editorial input, a requirement that the person whom I’d helped get an audio amplifier company off the ground [link} in my pre-Cairn days had exploited to great effect, so I put the question.   It turned out that I asked at a good time, as they were in the process of putting together a digital supplement on microscopy, and so might consider including a 1,000-2,000 word article from me.  Well, it was one of those articles that just seemed to write itself, and although it came out much longer at something over 4,000 words, I figured that it would just make filling their editorial pages correspondingly easier, so I submitted it anyway.  Basically it’s about the most efficient use of photons in fluorescence microscopy, where of course the goal is to minimise the number of excitation photons you put in (because they are potentially harmful), while maximising the number of emission photons you can get out.  They did indeed take the whole thing, and it looks like being the headline article, so I’m pretty pleased about that.  It’s due for publication by around the time this blog is posted, so here’s the link

Every photon counts

Every photon counts

Second, a few months ago I was asked by the editor of Physiology News to write an article on alternative career opportunities for physiologists Physiology News – martin article which I had a fair amount of fun with in including some of my own personal experiences of trying to do some half-decent work in an industrial research environment that had an eventually successful deathwish.  My failure there made the establishment of Cairn essential as well as being something that I’d rather wanted to do anyway, so it was very much a “push-pull” situation, greatly facilitated by my prior experience with – filterwheels!  So here we go with the saga….

I described in a previous blog [link] how, rather to my surprise, I found myself with a postdoc position at Boston University School of Medicine way back in the late 1970s.  The research topic was in the general area of the effect of calcium ions on nerve cells, and the organism of choice (because some of its cells are relatively large as such things go) was the marine snail Aplysia californica, so named because it did indeed have to be shipped in from the other coast for us to work on them, and occasionally they even survived the journey.  My supervisor Tony Gorman was keen to try to measure the internal calcium concentration during electrical activity, using a then newly “discovered” optical absorbance indicator by the catchy name of Arsenazo III, so we decided to give it a go.  Little did I realise the long-term consequences of this decision!!!

Well, we knew that the absorbance changes we’d measure were going to be very small, and the standard trick under such conditions is to make the measurements in such a way as to maximise the rejection of potential background interference effects.  For optical measurements of this type, the solution is to make them differentially.  What you do is to pick a pair of wavelengths which are relatively close together but where there is also an appreciable difference in the absorbance change, which in this case arises when the indicator binds calcium ions.  Although absorbance is a logarithmic type of measurement, for small changes it approximates arbitrarily well to a linear relation, so in practice you just subtract one wavelength signal from the other, while changing their relative gains to get a zero output, and then crank up the gain.  Given enough light and sufficiently good detectors, even very small differential changes (1 part in 10,000 or even less) can be measured in this way.  However, both to cancel out any drifts or other errors in the instrumentation, and also in our case for practical reasons, the best way to do this is to apply the two wavelengths via a common optical pathway, but this means that they must be applied sequentially rather than simultaneously.  So how to do this?

Well, I am reminded of Lady Macbeth erroneously thinking after a quick murder or two that that “A little water clears us of this deed”, but which then of course went on to haunt her to the extent that she ended up killing herself.  In my case I thought that “a little filter wheel” would clear ME of the necessary deed here, but I too have been haunted ever since, although I haven’t quite been driven to suicide, in spite of it being touch and go from time to time.

Yes, I decided that the easiest and best way to change wavelengths rapidly was to put a series of optical filters in a wheel in the light path, and to spin it fast.  From childhood memories of torture by evil dentists, I remembered that they got their instruments of confession to spin fast by using compressed air, and since our lab had compressed air supplies already laid on, that requirement was quickly satisfied.  The only problem was a slight longterm speed instability as the compressors cut in from time to time, so we later went to using a smaller wheel (with 9mm rather than 12.5mm filters) which used air sufficiently frugally that we could use compressed air from properly regulated air cylinders instead.

filter wheels

One of our earliest filter wheels, dating from my Boston postdoc days, spun by compressed air

This was all very satisfying, as at low speeds the wheel would emit just a gentle whirr, but as you cranked the thing up it came to more and more resemble the sound of a 747 taking off from a similar distance away.  We could easily run at speeds in the 200-300 revolutions per second range (that’s 12,000-18,000 rpm) without too much risk of permanent hearing loss, although we did once get up to 700 while hiding in a room down the corridor (probably) without the thing quite tearing itself apart, but that lower range was as fast as we needed for effectively simultaneous absorbance measurements, actually at six different wavelengths rather then just two – those extra wavelengths were very useful for control and calibration measurements.

The work I did in that lab received a gratifyingly high level of interest in the USA, but I wanted to return to the UK, where by contrast the water was much colder.  In describing the reaction to my work back on these shores, the concepts of balloon and lead sprang to mind.  I could well imagine the discussions following my presentations.  “Not at all sure about this Thomas person, he just seems to spend most of his time building unnecessarily complicated equipment, and then the science he does with it could have been done by anyone who uses it.” “Exactly!  Very unsound!  He’s definitely not a real physiologist at all.  We’re all agreed that it’s a ‘no’ then?  Good!”  As in riddance of course, but with the Cairn story I think I’ve ended up having the last laugh in professional terms, so any broken bones have long since mended.

I did manage to secure a temporary academic position in order to get back home, but for the usual political sort of reasons on top of this general attitude, I felt I should perhaps look elsewhere for the longer term.  This led me to take up the offer I had from the (inevitably long-since defunct!) industrial lab to which I have also alluded in a number of previous blogs.  At the time the prospects there looked sufficiently exciting that I was expecting to leave those postdoc interests well behind me, so I gave no particular attention to the fact that I’d brought back most of a duplicate absorbance spectrophotometer system with me, and which I had made some use of in that brief interim period, so I had never expected to see another filterwheel ever again.  Wrong!!!

In order to complete a more “use anywhere” system, I had made a simple filterwheel with electric drive.  It consisted of a small mains-powered fan which had a motor in the hub, and I cut the blades off to leave a rotating barrel.  This spun the wheel via a step-up belt drive, and in order to use standard parts once again I went back up to using 12.5mm filters.  This combination meant that the wheel ran only at one relatively low speed, but it was good enough to do things with.

However, the “leading role” that I’d been offered in that industrial lab turned out in practice to be the back legs of the pantomime horse, where I couldn’t even see what was going on, let alone influence it any meaningful way.  I’m sure that all too many others have found themselves in similar situations, but it certainly wasn’t what I had expected when I joined!  It simply wasn’t a place for people who wanted to DO things, as a couple of examples I gave in that Physiology News article [Physiology News – martin article] should nicely illustrate.

No, I had to go, but British academia still seemed to view me with a certain degree of suspicion, and in any case I was increasingly wondering whether I would actually enjoy that sort of life anyway (sadly, all too many of my academic friends don’t nowadays).  It was while I was thinking about all this that I helped the person with that audio amplifier company [link] in what was effectively my increasing amounts of spare time, and which therefore introduced me to the world of the smaller business as opposed to the corporate behemoth with its toxic mix of politics and bureaucracy – which has turned out to be a very useful combination.

The general hopelessness of the place also gave me some sanity-saving time to write a book (“Techniques in Calcium Research”) that included further details of my postdoctoral exploits.  It’s still available secondhand via Amazon, at (as I write) prices varying all the way from 24p (Jim, who is editing this, has bought one at 24p therefore raising the minimum price to 25p) to £109.13.  Maybe I should buy the cheaper ones myself to push the average price up…. We show a photo of it here because the front cover shows the absorbance changes recorded during a train of action potentials, in which the incremental changes caused by calcium ion entry during each action potential could be clearly seen.  However, presumably for “artistic” reasons the publishers displaced the electrical and optical traces with respect to each other, thereby completely destroying the whole point of the illustration.  This is but one specific example of our species’ general tendency to place form before function, and it’s something I’m increasingly “hot” on, having had my fingers burnt on one or two Cairn projects as well in recent years!  It’s all part of the essential need to maintain good communications as the company grows….

Techniques in Calcium Research Book

The “Techniques in Calcium Research” book that I wrote during my tenure of that “wonder” job as part of my increasingly desperate attempts to retain some vestiges of sanity

It’s also nice to record that another little activity in my spare time during those “dark years” was to take that makeshift spectrophotometer system with me to Plymouth for a few days.  I’d got involved in a mini-collaboration in which we were using arsenazo III to measure calcium changes in the squid giant synapse, which turned out to be practically the only thing going on here at the time.  Thatcherism was at its malevolent height, which had left the place without enough money to do any meaningful research for a while.  Fortunately this lab as well as myself is now seeing rather better days, but having experienced a low point together, it’s yet another reason why I do feel a certain sense of affinity towards this place.

I’m often asked how Cairn managed to get established without any bank loans or outside investors, but the secret is to take it slowly, and not to give up the dayjob until your own business is making a decent profit.  I was helped in this by my attempts to put my dayjob activities on solid scientific foundations, which increasingly gave me the reputation of being an esoteric academic type, hopelessly ill-equipped to survive in the real world, so even though my Cairn activities were no secret, they were never taken seriously until it was time to say “so long and thanks for all the fish” [Hitchhiker link].  By then (we’re talking 1989 here) I’d already been employing someone else full time for a year, so we really hit the ground running.  But most importantly, those experiences that I’d just put behind me underlined the importance of being properly in charge of your activities, which if you have any outside investors you are almost certainly NOT.  It really is that simple.  Really!

So what was Cairn doing?  Actually I’d been trying to do something rather different, but a number of my academic friends wanted to use my filterwheel expertise to put together fluorescence spectrophotometry systems based on the calcium indicator fura2, which, being an excitation ratio indicator, required (preferably rapid) alternating illumination at two different wavelengths, and for which a spinning filterwheel was going to be ideal.

OptoSpin IV

Our first commercial Cairn version, effectively the Optospin I

Although we didn’t call it that at the time, this was historically the Optospin I, and in mechanical terms it was basically a better-engineered version of that first electric drive system.  The main difference was that we used a DC motor with a feedback loop to allow it to spin at a controlled but variable speed.  However, it included a lot of new electronics to control the data acquisition (via photomultipliers), to decode the output into individual channels for each filter wavelength, and then to convert them into ratios for subsequent analysis.  We found a very nice little market here for people who already had electrophysiology setups and who wanted to “add fluorescence” to their existing acquisition systems.  The electronics were in a plugin module format, and they remain available in an only marginally updated form to this day.

Troubleshooting was a breeze, as our electronics modules were pretty reliable, since we used socketed electronic devices for easy maintenance in the field (we still do wherever we can), and there were only two easily identifiable problems with the filterwheel.  Either the drive belt broke, in which case the wheel wouldn’t spin at all, or the “white stripe”, which was admittedly in a rather vulnerable position on the outside edge of the wheel, got smudged into oblivion, in which case the wheel would spin up uncontrollably to full speed, so both were pretty easy to identify!

The only problem was that this design wasn’t very suited to moving in a stepwise as opposed to a continuous fashion between the individual filter positions.  We just had a simple clock escapement type of mechanism which allowed the wheel to be stepped sequentially from one position to the next in a rather clunky fashion, but at least it did the job.  Still, after a few years of selling this version it was time to do something better.

What we came up with, which historically would have been called the “Optospin II” (although like it predecessor it was marketed as the “Cairn Spectrophotometer”) bore more than a passing resemblance to the current product, in that the motor was mounted in the hub.  However, like its predecessor it still used 12.5mm filters, so its diameter was correspondingly smaller than what we have now, but it had a series of both magnetic and optical sensors instead of just that single white stripe.  This meant that the control electronics “knew” rather more precisely where the wheel actually was, and hence could step directly from any one filter position to any other, as well as spinning continuously.

OptoSpin V

Our second commercial version, effectively the Optospin II, which could step as well as spinning continuously.  Quite a nice product in retrospect!

When this version comes back for repair, as it occasionally does, I tend to be much more impressed with it now than I was at the time.  The main problem was that the motor just wasn’t as powerful as we would have liked.  It was fine for spinning, but to get good stepping performance we had to drive it much harder.  This it could just about take, but the built-in internal position sensors were magnetic ones, and they got overloaded by the correspondingly stronger fields, so for stepping we had to add a duplicate set of optical sensors that read a series of stripes around the motor hub.  This was in addition to some other magnetic sensors that we needed to tell the electronics in which of four possible quadrants the wheel was currently positioned, and a further optical sensor, reading a series of slots around the edge of the wheel, that gave more precise timing information for the stepping but couldn’t give absolute position information.  I also felt that the control electronics ended up being rather “messy”, but what I felt to be the resulting abomination actually did the job rather well!

Clearly I was hoping to follow this up with a more personally satisfying version (which we do now finally have), but around that time TILL Photonics introduced a more versatile fluorescence excitation system based on a monochromator, which worked well enough for our customer base to request a Cairn version.  That’s another story, and when we released our Optoscan –  very different in design from TILL’s, and still available in essentially its original form whereas they went all the way up to a version 5, and now with a new lease of life following the discontinuation of the TILL product – it killed our filterwheel sales stone dead.  At last, filterwheels were behind me!

Er, no.  A whole new market awaited.  Until now, our market was in fluorescence photometry, so the light modulation (usually on the excitation side but also potentially on the emission one) could be carried out on a beam that could be focussed down to go through those 12.5mm filters.  However, by now (I guess we’re talking about the turn of the century here) camera technology was becoming good enough for imaging to become a viable alternative to photometric measurements.  Although fancier things can in principle be done, what people now wanted was a filter-changing wheel that could be put more or less directly in front of a camera that was taking snapshots at successive filter positions, with the wheel stepping around between them.

The evolution of the filter wheel

The evolution of the filter wheel

This configuration needed the filters to be at least as big as the camera sensor, requiring us to move up to the 25mm size.  However, as explained elsewhere [How do you make a filter wheel step more quickly?] the larger size incurs a very substantial speed penalty for discontinuous movement.  Unfortunately I got rather too carried away here, so what was effectively the Optospin III ended up as a mere curiosity in the Cairn Museum, as we sold only a couple of development products before beating a very hasty retreat!

The idea was pretty cool though.  Since my experience with the version II product suggested (incorrectly as we shall see) that a hub-driven motor couldn’t deliver enough power for half-decent stepping with 25mm filters, I hit on the idea of making the wheel itself the motor, so the rotating magnetic part was in the form of a ring around the outside of it, and the fixed coils were arranged around the edge.  This made for a nice thin “pancake” type of construction, although the coils did significantly increase the overall diameter of the assembly.

Each coil had a letter “C” type of configuration, with the magnets of the wheel spinning through the gap, so we basically had what is known as a disc motor.  However, a potential problem with this type of motor is that the configuration is on a “knife edge”, in that if the gap between the disc and one of the coil poles was less than the other, the magnets were disproportionately strongly attracted towards that side, and with the standard six-filter configuration we found it very difficult to make the system sufficiently rigid to prevent the wheel from eventually crunching into that nearer pole face.

Of course we could have made the design both more rigid and precise to get that problem down to more manageable proportions, but the thing was also rather too powerful in practice, generating an enormous countertorque that tended to spin the rest of the universe (especially anything to which it was attached) in the opposite direction, and all those coils around the outside also made it too big for comfort.  The more successful implementation was a smaller wheel with just three filters that were practically touching, and which just left room for the central axle.  However, although three filters may be enough for many applications in practice, most people want more!

So that experience paved the way for the Optospin IV, although the gestation period turned out to be embarrassingly longer than we had hoped.  During the intervening period a very interesting range of electric motors, designed for powering model aeroplanes, became available.  Thanks to the use of rare-earth (neodymium) magnets, that were very much stronger than the ones used in the Optospin II’s motor, their power-to-weight ratio was sufficiently higher to make a hub-mounted motor feasible for both spinning and fast stepping a fully viable proposition.

At last, the Optospin VI in all its glory. What took us so long? 2 x Optospin pictured

At last, the Optospin VI in all its glory. What took us so long? 2 x Optospin pictured

I could see a nice market opportunity here, as the competing products have tended to use “stepper motors” to drive their wheels.  Such motors can generate high torques, but aren’t suitable for fast spinning (think of a car being driven in first gear).  In retrospect though, what I should have done was to have developed a stepper motor design of our own while we worked on the Optospin IV, and we would have had something that was just as saleable as everybody else’s meanwhile.  Instead we had nothing  at all of our own to sell for some years – duh!!!

It was all too ambitious really, but we did get there eventually, although it was a battle all the way.  The design brief for spinning was to able to spin multiple wheels at a wide range of speeds (we can do anywhere in the 1-170Hz range, corresponding to a transition time going down to just 1 msec for each of the six filters in the wheel), but also in perfect speed and phase synchronisation.  That is to allow simultaneous excitation and emission wavelength changing with separate wheels, for example,  To meet that requirement we use a synchronous drive system, in which the motor coils are sequentially energised at precisely the right times for the wheel(s) to spin at the required speed.  However, except at the very lowest speeds the wheel(s) can’t suddenly spin at that speed, so would just sit there and sulk as the electromagnetic field rapidly flies around them.  This means that we have to start slowly and ramp things up until the required speed is reached, which requires correspondingly more complex programming.

However, it was the stepping that proved the real challenge!  You can read more about the tricks we had to resort to here, but in control terms we ended up with a threeway combination of feedforward, feedback and learning, so we really had to throw the kitchen sink at the thing in order to get fast and “wobble-free” stepping.  And of course we wanted to be able to step multiple wheels simultaneously (we can do up to four), but with each one being able to go to different positions independently.  We have stayed with using “only” six filters in the wheel in order to keep the size down, as the physics says that stepping time is going to be at least a square-law function of the diameter.  However, we can interdigitate two wheels within the same physical and optical space, which gives you a choice of ten positions if you leave one position free in each one.  This combination will always substantially outperform a single ten-position wheel.  There are other nice two-wheel possibilities too, our favourite one (memories of the ill-fated Optospin III here), being to step the two wheels simultaneously in opposite directions, which very nicely cancels out the reaction torque if that is an issue for you.

All this seriously tested my sanity, and there was some electronics to develop too, but that gave me some very welcome light relief one all-too-long evening.    I was travelling from Washington DC to Boston by train, which should have taken “only” four and a half hours, but the drive belt on their motor had broken, so we instead made the journey on Amtrak’s equivalent of impulse power, which took at least twice as long.  After about seven hours, by which time my fellow business passengers had run out of windows to break and seats to tear up in their frustration, their attention turned to this strange Englishman who was serenely doing something or other with some odd pattern of multicoloured lines on his laptop.  “Heeey, izzis sam sordda GAAYME?” asked one of them in their best east-coast variant of our not-so-common tongue, in reply to which I had to explain that I was using a CAD package for electronic circuit board design, for I was indeed laying out the circuit for the first version of the Optospin IV’s control electronics.  They all retreated in suitable awe and carried on with smashing the toilets (well, that’s what they all should have done under the circumstances, I think).

Dan and myself teaching students in Plymouth during the imaging course

Dan and myself teaching students in Plymouth during the imaging course

So, back to Plymouth!  As I write this, Jez and our Royal Society Research Fellow Dan Mulvihill are working with real live yeast cells, simultaneously labelled with mCherry and the GFP-like fluorophore neongreen.  Normally there would be a crosstalk problem here, because the excitation spectra of the two fluorophores overlap, but thanks to the latest feature of the Optospin, we can get around this problem!   The Optospin is spinning continuously (which gives way faster wavelength changing than stepping it would), and it has emission filters optimised for each fluorophore.  As the wheel spins, it generates control signals that switch on an OptoLED optimised for neongreen excitation when the neongreen filter is in the emission lightpath to the camera, and another OptoLED optimised for mCherry excitation when the mCherry emission filter is in the emission lightpath.  Even though we say so ourselves, this is seriously cool!  Here we show a sample of the results that we are obtaining literally right now.  The wheel is currently spinning at 20Hz, but we could go much faster if we wanted to.  And to further gild the lily, we have an Optosplit in the emission light path too.  This simultaneously gives us separate images for the neongreen and mCherry emission wavebands, which is even cooler of course.

(Optosplit and Optospin in series mounted on c-mount before the camera) shows different images acquired when only the GFP LED is triggered (Left); the mCherry LED is triggered (middle); or both are sequentially triggered (right)

(Optosplit and Optospin in series mounted on c-mount before the camera) shows different images acquired when only the GFP LED is triggered (Left); the mCherry LED is triggered (middle); or both are sequentially triggered (right)

Combined image above shows the data. Majority of protein signal co-localises (yellow), with red only seen on cytokinetic ring. However due to lack of bleed through this system gives the researcher 100% confidence that the colocalisation is real.

Combined image above shows the data. Majority of protein signal co-localises (yellow), with red only seen on cytokinetic ring. However due to lack of bleed through this system gives the researcher 100% confidence that the colocalisation is real.

Leave a Reply