PMT General Discussion.

[Panrock]

I'll be following this with interest Steve, all the more so because clearly I have a stake in your success with noise-abatement!

You may wish to move my recent other new thread on PMTs: "new life...etc" - here, since I posted it before noticing this thread.

Steve O

[Steve Anderson]

...This is a continuation of the thread 'Preliminary Testing for A-K' which in recent times has diverged into PMTs, so here we'll start afresh.

PMTs have a reputation as being noisy and 'contrasty', this really shouldn't be so. Their output is as linear as any photodiode with a noise floor that beats the pants off semiconductors. Although as time goes on the gap is narrowing.

Steve O uses PMTs in his colour camera and within the confines of 30-lines all appears reasonably well except on two counts, first is colour shading which I presume is down to optical factors and what is now a small amount of residual noise which would be nice to eliminate.

I have said in another thread that to gain some kudos here I really should build a mechanical device, I've gone for a camera. Not having used PMTs before they've become the choice of photosensor. Some are on their way to me as I type this, standard 931As which are generally a good price and moderately plentiful, as are the sockets for them.

I'll only be using monochrome so I can 'get away' with having no red sensitivity. Steve's red sensitive R446 PMTs are around at just shy of US$100. The 931A is on average is US$40-50.

Initially to investigate this noise phenomenon I'll be knocking up the test-bed attached, the supplies are well and truly decoupled to eliminate noise. I reverted back to a resistor chain such that I can vary the supply voltage to the PMT and for the time being done away with the dynode 4 & 5 gain adjustment. I will re-introduce a gain adjustment at some future point, it would appear that there are better ways to do this.

I will also try a LT1022 in place of the first TL071 as this is ideal for current sources which a PMT is. I was planning to use a OPA128 or a AD549 as well, but at anything from GBP12.00 to well over 100, forget it!...for a single op-amp!

The input protection diodes are now back-biased by 6V or so to reduce their capacitance from 4pF to under 1pF. This might introduce some noise, we'll see. I've made one or two small changes to the circuit that Steve is currently using, but none of any significance.

The blue LED is to irradiate the PMT and will be run at very low currents with a lot of physical attenuation in the way!

More anon...

Steve A.

[Steve Anderson]

Something that I've been wondering about... Since the greatest overload will only have occurred at the final dynode or dynodes, could a 'ruined tube' in fact still be used, but with a lower output, by simply taking the signal off at an earlier dynode rather than the anode ? Steve O

This is covered in the Hamamatsu manual I posted a few days ago, page 172, "Output Contol Circuit". This arrangement is based on a good tube, whether it's applicable to a 'ruined' one I'm not sure. I'm not going to encourage anyone to deliberately ruin a tube to find out...

Steve A.

In combining two threads into one Steve O's original posting (quoted above) had to be deleted. The software for moderating is not that easy to use, it's easy to split threads, but seemingly impossible to merge them.

[AncientBrit]

Steve A,

In the Hamamatsu manual you attached a few days ago there was an interesting circuit arrangement where the dynodes were fed directly off the taps in a Cockcroft Walton multiplier, thereby removing the need for a tapped divider. I haven't seen that before.

Also I must dig out a reference from my "library" to a push-pull version of the Cockcroft Walton multiplier which claimed superior performance in terms of output impedance. The"down-side" is that you have to use a centre tapped transformer.

Must try and locate the circuit,

Regards,

Graham

[Steve Anderson]

In the Hamamatsu manual you attached a few days ago there was an interesting circuit arrangement where the dynodes were fed directly off the taps in a Cockcroft Walton multiplier, thereby removing the need for a tapped divider. Regards,
Graham

I too mused over the Cockcroft-Walton multiplier, it's impressive in it's simplicity and it only has to provide power that the PMT itself uses, not a wasteful 'bleeder' string.

However, I simulated it on the PC and using a 50Hz feed the capacitors became quite large, it would of course be better fed with a higher frequency, several kHz probably. This means winding or finding a suitable transformer and drive circuit. In addition each stage is of equal voltage, fine for the PMTs like the 931A and the such used in the 'standard' configuration, but if you wish to alter inter-dynode voltages for whatever reason it all starts to lose it's advantages.

Having said that it is popular and Hamamatsu make sockets for their PMTs with all of the circuitry built in...apply low volts...off you go...

I found a more recent version of the Hamamatsu manual, the changes to the first few chapters are minor, but the later chapters have been expanded mainly to flog their products. I attach it here for completeness but the first version meets all the requirements we really need.

For those that have masochistic tendencies also attached is the Burle PMT manual...good bed-time reading material...

Steve A.

[AncientBrit]

>Steve A,

Attached is the parallel fed voltage multiplier circuit.
Purves Prescott, not Cockcroft Walton BTW.

It has the disadvantage that the caps voltage ratings increase as you go down the chain, but "all the values are the same compared to the C-W circuit"

And that it needs a P-P drive.

The implication is for the C-W circuit one should use graded cap. values although I've never seen that in practice.

Regards,

Graham

[Steve Anderson]

In my forlorn hope to actually build a mechanical camera sometime before I expire I recently purchased some 931A PMTs. This was partly prompted by Steve O's observation of low-level noise in his colour camera and monitor system. With the exhibition of this apparatus at the weekends convention I wonder how many noticed it?

Now, I intend to use a higher resolution than Steve's current 30-lines, but only in monochrome, he's also considering a move to something like 50-lines and still in colour...but that's up to Steve.

If there's low-level noise at 30 lines, it'll become more noticeable at 50 lines when there's only some 30-40% of currently available light reaching the PMTs.

To that end I'm in the process of knocking up a noise test-bed for standard PMTs like the 931A of which the enhanced red-sensitive R446 is a relative. However, red sensitive PMTs are inherently more noisy due to the multialkali photocathode as opposed to the bialkali photocathode 931A.

So I intend to split out the PMT noise from the pre-amplifier noise and try and determine which is the culprit...it might of course be a bit of each.

The blue LED light source I intend to use has a rise/fall time of around 100ns and is flat to the limit of my audio signal generator at 150kHz, the distortion at around 90% modulation is under 5%, all down to the non-linearity of the LED, but good enough for this test. These are measured figures.

More to come.

Steve A.

Footnote...I'll not be posting as frequently as I used to, but I'll still have a look around once every few days...time is tight now...

[Panrock]

This was partly prompted by Steve O's observation of low-level noise in his colour camera and monitor system. With the exhibition of this apparatus at the weekends convention I wonder how many noticed it?

I think many did, Steve. The noise is a worthwhile price to be paid (IMO) for the greatly increased sharpness. Not helped in this instance by the rather feeble lighting I was using. Less dynode-divider gain, and more light would have removed it.

Steve O

[Steve Anderson]

PMTs have a reputation as being noisy and 'contrasty', this really shouldn't be so. Their output is as linear as any photodiode with a noise floor that beats the pants off semiconductors. .

Now where in the heck did I get that impression? At the very low light-levels involved it's true, but I have to admit I was shocked by the level of noise. True, most semiconductor devices wouldn't even produce a signal so the PMT still wins, but I'm still shaking my head in disbelief.

The test-bed constructed was largely as per the circuit diagram above without the second-stage TL071. A 20kHz Butterworth filter was added to the output with a slope of 24db/octave. (See below)

Follows are chunks lifted from e-mails over the last few weeks between Chris Long, Jeremy Jago, Steve Ostler and myself...but I'm only quoting myself here...

15/05/09
Gents,

This is very preliminary, as a result of the first 931A tube being fired up. At first there was a quite large dark current of some 2uA which reduced to 280nA after 15 minutes, then 75nA after one hour, 40nA after 4 hours and finally to 33nA after eight hours. In spec but two things to consider, first my ambient temperature is some 15 degrees above the usual 20-25 Celsius test conditions, and these tubes haven't seen any Volts since March 1970. So this first tube is in spec, but not 'typical' as Burle would put it.

Noise was fairly proportional to dark current, no irradience as yet. Baseline noise was below what I can measure, but what might be either shot-noise or scintillation due to the close proximity of metal earthed chassis (but not in contact with tube) might have something to do with that.

To which I have fitted an aluminium foil 'condom' to the tube, I'll let it dark adapt overnight and repeat the tests tomorrow.

Attached are two traces from this virgin and naked tube in the dark, one showing an individual pulse whose shape and overshoot is simply down to the Butterworth 20kHz filter I'm using, the other showing that most in dark current are much lower. The larger pulses were generally seconds apart. (See tek0013.gif and tek0020.gif below).

Also 15/05/09

Very quick resultant of same tube with electrostatic condom...but after a short period of time...note trigger threshold reduced and less noise pulses...but very interim...

Much quieter...more tomorrow... (See tek0026.gif below)

16/05/09

Gents,

A third fire-up of this individual PMT after a rest time of 18 hours resulted in an initial dark current of 70nA which fell to 40nA after an hour or so. So the current at first switch-on after some 40 years seems to be a once-off event.

17/05/09

so far the results have exceeded my expectations but I have yet to irradiate the tube with the blue LED. I'm doing one last eight hour dark current test today which will be the fourth. Bearing in mind the ambient temperature here this 40 year old tube is in spec but still some way above the typical value for dark current. It will be interesting to see how the others compare.

I'm starting to wonder if these random and widely spaced large pulses are actually mains-borne interference getting into the test-bed. I have a separate earth rod for the workshop cum office and two line filters in series. I'll have to see if there is any correlation between these and mains spikes.

Sure is an old design, over 60 years, one would hope that in 1970 the materials and consistency of the tubes would have improved somewhat. No hot cathode or heater agreed, but some process must be in action to drop the dark current from 2uA to 33nA over the course of eight hours, a 60:1 drop. Subsequent tests after a rest period of around 18 hours with no exposure to normal light, the dark current initially is around 70nA which drops to 30-40nA after an hour or so, the majority of the drop occurring in the first five minutes.

I've heard something about 'light bias' and likewise I cannot quite see how it would help in our intended application. No 'fancy' bleeder strings, simply 10 new equal value 0.5W 1% metal film resistors from cathode to ground well bypassed so they cannot be blamed as a source of noise. The supply can be varied to any value up to -1300V, it too is well smoothed, as you can see on the traces there's no hint of hum or ripple. To start with I'm sticking to the usual 1000V for this tube, if I start varying it I'll only end up confusing myself. All transformers are at least an arms length away from the tube.

I have the RCA datasheet for the 931A dated 1/9/1950 plus quite a few from other manufacturers, also handbooks/manuals from Burle and a number from Hamamatsu, all of which are quite thorough on the care and feeding of PMTs of all flavours and uses. More to come.

Steve A.

21/05/09

Gents,

After a lull in the PMT tests I've been back on the case today....

Well, you've had the data for the first tubes' dark current and noise, the second one was better being a slightly lower baseline current and the noise pulses some 50% less, no correlation between them an mains spikes as it turned out. It's from within the tube.

But when I turned on the blue LED I was horrified by the amount of noise! Three examples attached are from the first tube at 10uA, 20uA and 50uA. The second one did no better. Initially there was a moderate amount of visual attenuation between the LED and the PMT but to get an anode current of 10uA through the tube only required 3.4uA, yes uA through the LED! Thinking that perhaps at these very low currents LEDs became noisy in their output I dramatically increased the attenuation such that it required some 20mA of LED current to get 10uA of tube current. The supply to the LED was well regulated and thoroughly decoupled. No improvement, if anything it was slightly worse.

Unless I'm doing something very wrong or I've been sold a duff batch of tubes (they are 40 years old) I've decided to abandon these tests as they are taking up a lot of time. I've e-mailed Burle and Hamamatsu to see if they've go a local distributor and if the price is 'reasonable', I'll buy a brand new one, preferably the 931B which is better specified and in current manufacture by both companies.

I know all three of you have produced functioning cameras using these or very similar devices, but with what I've seen thus far I'm disinclined to pursue it further. I'm tempted to see if a BPW34 with a suitable pre-amp would produce better results...but of course in the red part of the spectrum.

Steve, put it this way...the noise you see is not coming from the electronics!

Steve A.

And today, 22/05/09

Gents,

Thanks for your thoughts...even with Chris' assurance that the noise is exactly what he would expect I'm still agog at the level of it. I expected a moderate amount but nothing like this level, if it were say one tenth of its current level (-20db) I would have thought, "That looks about right."

All the data and specifications I have for not only the 931A but many others too neatly omit DC noise levels. It has to be said that the major uses for these tubes are for pulse counting where s/n ratio is not so much of an issue, or for simple on/off switching.

Years ago Crow Broadcast in the UK used to make a 35mm slide-scanner for 625/525 line TV, in colour. It used PMTs. The company went bust decades ago and I cannot find any data on the slide scanner. The last one I saw was in Qatar in the mid-80s. This would have had to have met some stringent requirements to see service with the likes of the BBC and ITV companies. So there must be other PMTs out there far better suited to this application than the 931 series. Yet in the listing of recommended tubes for flying spot scanners in the Burle handbook is the 931A along with eight others.

Although I cannot be absolutely certain these tubes are unused it would appear so. Each one has a serial number which matches that on the box, with the yellowing of the adhesive tape sealing them they haven't been opened in a long time. They're RCA tubes.

Depending on the response from Burle and Hamamatsu I'm putting this on the shelf for a while and getting on with something else. These tubes would be fine for broadband data over an optical link, their sensitivity is amazing and those that I have tested are very close in sensitivity supporting my belief that they're unused. But being 40 years old Chris' point regarding a dubious vacuum and Helium penetration could well be spot-on.

My suggestion r.e. using a BPW34 was somewhat of a knee-jerk reaction, I'm not serious about that.

Thanks again gents for your input,

Steve A.

So that's the status at the moment. If I can get a brand new tube at a 'reasonable' price I'll resume these tests. The PMTs are doing their job detecting light levels that semiconductors would have no chance of doing so...but I'm still stunned by the noise...

Steve A.

The electrostatic 'condom' was connected electrically to the photocathode.

[Steve Anderson]

For those interested in PMTs the future looks dire....follows is a press release from Photonis (Burle)...

PRESS RELEASE

In the course of the recent years PHOTONIS has faced huge and recurrent difficulties with its “photomultplier” (PMT) branch of business activity. Unfortunately it is now very clear that the market has entered into a deeply depressing phase and that the financial environment is worsening month after month.

Since the beginning of last year PHOTONIS has been struggling to find a long term solution in order to keep this activity at a viable economical level. We have implemented drastic organizational measures such as a significant reduction of the headcount in France, the closing of our production facilities in Lancaster USA, the overall reduction of costs, the rationalization of our product portfolio and optimization of our industrial deployment worldwide.

Unfortunately the market situation continues to degrade quickly and despite our constant and best efforts, this activity is no longer economically viable.

As a consequence, and in order to avoid placing the whole Group at risk, PHOTONIS has no alternative but to stop all its activities in relation to the designing, manufacturing and commercialization of Photomultiplier tubes.

This has to be done as quickly as possible with a targeted closing date at the end of July 2009. We deeply regret being forced into this drastic position especially as PHOTONIS is one of the only two major players in this sector, having accumulated over many years a strong and recognized know how and experience in the designing and manufacturing of photomultiplier tubes.

The management confirmed that there will be no “social plan” implemented in France as a result of this decision. The non temporary dedicated personnel will be allocated to the other activities of the group.

The PHOTONIS Group will refocus its efforts and knowledge on other photodetection and imaging needs.

The end is nigh...

Steve A.

[Klaas Robers]

Steve, if I look at the way PMTs are configured, I think they are perfectly matched to detect cosmic radiation. So if you find spikes now and then in the output signal, while the tube is in total darkness, this may be cosmic radiation hitting the cathode or some dynode.

Secondly: if you illuminate a PMT with very low level light, e.g. an LED at a few micro amps, you might see noise due to the individual photons. The light then is not a continuous stream, but individual photons that hit the cathode is an unordered way. This generates a certain noise.

If you increase the light level by a factor of two (6 dB) the noise also increases by a factor of 1.4 (3 dB). So the higher the light level, the better the signal to noise.

I hope this helps you....

[Steve Anderson]

....if you illuminate a PMT with very low level light, e.g. an LED at a few micro amps, you might see noise due to the individual photons. ..

The more I delve into this the more likely that the noise under irradiation is due to the random nature of arriving photons. I'm simply surprised that it's so easily detectable. I know a PMT has a huge amount of gain and the light levels I have been using are almost undetectable by the dark-adapted eye, but it still comes as somewhat of a shock.

As you increase the light level the s/n ratio does improve to what approximates a square root function, but there is a limit on the average current a PMT can handle. For the 931A it's either 100uA or 1mA depending on whose datasheet you're reading. I took it only as far as 50uA as that would appear the maximum Steve O is likely to ever get out of his camera.

As for the large pulses in the dark current...well, it could be cosmic radiation but the pulses are infrequent and effectively will be obscured by the noise in the light current.

Recalling the Crow slide-scanner mentioned above, I do remember that the CRT used to provide the scanning raster was very bright, so bright it was painful to look at, even in standard office lighting, this would have helped with the noise.

So returning to where this all started, Steve O has simply got to ramp up the illumination of the subject to keep the noise to an acceptable level. Remember Bairds' subjects were almost roasted alive by the lights. Here that can be reproduced simply by sitting outside in the sun at midday...

Steve A.

[Panrock]

So returning to where this all started, Steve O has simply got to ramp up the illumination of the subject to keep the noise to an acceptable level. Remember Bairds' subjects were almost roasted alive by the lights. Here that can be reproduced simply by sitting outside in the sun at midday...

Not even that is necessary Steve. At the convention I was just using just a bit of dim window light augmented by a fairly weak lamp. Ask Klaas - he was televised!

Steve O

[Steve Anderson]

Well, as an interim conclusion (I know, an oxymoron) there isn't a lot one can do regarding the PMT noise except keep the light levels high enough and aim to keep the average PMT current at the limit of 100uA.

Currently you seem to be getting about 30uA peak PMT current, this will drop to around 10uA should you decide to go to 50 lines. So the light levels need ideally to be ten times higher for the same optics and f-stop.

Some form of portable home-brew studio lighting might be in order, run off DC to eliminate hum bars. A few 12V car headlights spring to mind. During the day I'll have no problem with light, maybe too much!

Steve A.

[Steve Anderson]

OK Folks...It's hands-up time! I dropped off!

As I reviewed my results from these PMT tests I was staring at the pre-amp circuit diagram, all of a sudden it came to me that I hadn't fitted the 22pf cap across the 100k feedback resistor in the pre-amp! This would at best led to gain peaking outside the bandwidth of interest, although the 20kHz filter should quell that I was concerned that perhaps the first stage was going into saturation, oscillating or 'squegging' in some manner.

So I have now fitted said cap and re-doing some of these tests to see if there is any difference. Thus far in terms of dark noise and current, no real change, but the DC level at switch-on of dark current has reduced even further, perhaps simply due to the tube having been energized several times recently.

More to come.

Yours embarrassed,

Steve A.