1

I need to turn on an LED (NTE 30030) for 100 milliseconds, turn it off and then immediately (< 1ns or as close to zero as possible) take measurements on a nearby (< 10mm away from LED) photodiode at 1 ns intervals for 5 ns. Those measurements can be put in a specialty buffer chip that can be retrieved later with an Arduino device.

Is this possible with specialty chips? Can Arduino do this? What kind of literature do I need to read to understand this problem?

EDIT: For clarity, this is for atomic fluorescence. The biomarker we are using continues to fluoresce for only 2.6 ns after excitation cut off. I can also use a laser if that makes it easier. For cost reasons, we cannot go down the route of photomultiplier tubes or monochromator style of setups. We're trying to make this accessible but it may not be feasible to do this reading some of the comments below. I am not an electronics expert.

ejkitchen
  • 111
  • 2
  • 1 ns ? no not possible or even close. – Tony Stewart EE75 Sep 28 '16 at 20:13
  • What phenomena are you trying to measure? I would like to be convinced that < 1ns response time and 1GHz sampling is actually required for your project – Jon L Sep 28 '16 at 20:20
  • @JonL It sounds very much like a fluorescence or (less likely) phosphorescence measurement process to me. But the OP will need to say, I suppose. (Hmm. 575nm.) – jonk Sep 28 '16 at 20:22
  • 1
    @JonL The reason I think it must be atomic fluorescence is that the timescales are about right. Atoms (the timing and the energy is atomic and not lattice/molecular) absorb and re-emit photons on about the scale we are talking about here. To compute the \$\tau\$ one may need this kind of device capability. It also could just be direct measurement of the re-emission of any atom. – jonk Sep 28 '16 at 20:29
  • 1ns is about the half life of aspirin excited at 240nm converted to 380 nm – Tony Stewart EE75 Sep 28 '16 at 20:41
  • @TonyStewart: GHz scopes exist... so I have to believe this physically possible. Not that it is going to be easy... – Andrew Spott Sep 28 '16 at 20:43
  • cost alot more than an Arduino – Tony Stewart EE75 Sep 28 '16 at 20:47
  • OP heeds to do some more research, e.g. http://electronics.stackexchange.com/questions/86717/what-is-the-latency-of-an-led – Ale..chenski Sep 28 '16 at 21:06
  • @AndrewSpott the problem is not sampling at 1 Gsps. The problem is getting the LED to turn off fast enough to not be brighter than the phosphoresence you're trying to measure. – The Photon Sep 28 '16 at 21:24
  • It is for atomic fluorescence. The response curve for the biomarker is peak fluorescence up until 2.6ns after cut off. This may not be possible to do as I can see from the comments here. The reason we are trying this method is to reduce costs to make something portable and reasonably priced (ie no photomultiplier tubes or monochromators) so that people will be able to build their own cheaply. I can also use a green laser as input. – ejkitchen Sep 28 '16 at 22:04
  • @AliChen I do not have the vocabulary to have found the article "What is the latency of an LED?". But that was not my question. However, now I realize that this is an important aspect to my question. I will do more research. But this is why I asked. – ejkitchen Sep 28 '16 at 22:21
  • Do you need to measure amplitudes, or is this signal going to be a pass/fail binary signal for each of the 5 samples? Digital hardware operating at GHz speeds is not trivial, but analog readings at those speeds on the cheep will really call for a specialist. Is it possible to bring an electronics expert onto your team? If you're worried about per-unit cost, someone who knows both high speed electronics and the specifics of your task may be invaluable. – Cort Ammon Sep 28 '16 at 22:30
  • I would need some degree of amplitudes. Perhaps not high resolution but a scale of at least 0-256. I think you're right that I do need to speak to an expert. I had not realized this and I'm way over my head as I can see. Where could I find & hire online a high speed electronics expert that could answer some and more of these questions. That must be a specialty difficult to find. – ejkitchen Sep 28 '16 at 23:03
  • I know people that do this for a living in the UK, but they are really not cheap. If you've got several tens of thousands $ to spend I could probably put you in touch. – pjc50 Sep 28 '16 at 23:17
  • @ejkitchen Thanks for the confirmation. This is an area where I specialize. So it just looked like it, to me! Appreciated. I'm actually interested in the details. But I suppose that's not appropriate here. – jonk Sep 28 '16 at 23:26

2 Answers2

2

1 ns ? no not possible or even close.

You can turn it on in 30 ns but off takes longer with high impedance conventional switch, perhaps turn off < 500ns with a shunt driver.

But to detect the light with this speed an extremely low capacitance photo detector with very low impedance low gain amplifier because the current source slew rates slows down with lower light. A GaP PD has a slew ON/OFF time of 1 ns / 140 ns e.g. Thorlabs FGAP71

You might need to get into lasers and advanced photonics for these decay times with fused silica photo detectors 1ns/1ns e.g. FDS010 ~$60 and then expensive amplifiers with microwave rugged assemblies for stability.

https://www.thorlabs.com/_sd.cfm?fileName=0636-S01.pdf&partNumber=FDS010

Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182
  • 1
    If I used a laser, do you think I could cut down the shut off time? – ejkitchen Sep 28 '16 at 22:16
  • certainly fa$ter than an LED – Tony Stewart EE75 Sep 28 '16 at 22:21
  • 1
    When you say more expensive are you talking about $1 versus $14 or do I need a special kind of laser? I can use a laser to do the excitation and some cheap ones are near what I need in terms of wavelength. That could work. – ejkitchen Sep 28 '16 at 22:28
  • laser costs in $100 range pulsed from 100% to <10% might give the desired response depending on wavelength and expensive optical filters. .. you need to search. – Tony Stewart EE75 Sep 28 '16 at 22:34
  • 1
    We've got everything working with laser/LED and expensive filters plus other fancy stuff to reduce noise. Now we're simply trying to make a cheap portable device using off the shelf components. That's our struggle right now and it may not be possible to do this. Thanks for the feedback and responses. Much appreciated. – ejkitchen Sep 28 '16 at 22:49
0

Do you need to take each sample with any kind of amplitude resolution or just temporal resolution?

Accurate amplitude measurement at 1GHz will not come cheap but if all you need is a really accurate stopwatch then something like the THS788 might work. It is a high precision counter that has an LSB of 13pS and an MSB of +/- 7 sec (although the minimum detectable pulse width is 5nS so you may need to add some latching comparators).

Couple it with an appropriate photodiode amplifier and you'll be able to measure the time where your signal crosses a set of 4 thresholds (the THS788 has 4 input channels and if each has a comparator well... you get the idea). As a sort of added bonus, the THS788's got a sync input which sets t=0 which you could hook up to your LED source to trigger the counter.

But, like Tony said, anything involving sub-nanosecond edges will cost many hundreds of dollars (e.g. $60 for the PD, $160 for the 788, an ADCMP566 is ~$9 and will respond to pulse widths down to 200pS, that's already ~$250 and that's without a high speed PD amp, a high speed driver for the LED or any other supporting circuitry). But if you need to take amplitude measurements as well, expect a 4 figure price tag at least.

Sam
  • 3,639
  • 13
  • 16
  • Thanks for the feedback! I think I may not be able to do this. If I could get it to a few hundred that would work, but over a thousand this no longer is feasible for our goals. It's an open source project and it has to be reasonably priced for poorer countries to use. – ejkitchen Sep 28 '16 at 22:15
  • @ejkitchen For fluorescence, you'd also need a beam splitter, thin film filter or two, and an assembly for all this -- if I'm projecting my thoughts correctly. There are a lot of things, other than the electronic parts alone, that make this difficult to do. Any electronic circuit will do this anyway, when talking about these speeds, but I think you may need to consider the integral of the process instead of a direct measure. You can, perhaps, take the ratio of two integrals? You need to examine other ways to bring the price down. And explore the math here for ideas. – jonk Sep 29 '16 at 03:10
  • @ejkitchen don't be too disheartened. The price scales with the sample rate and the sample resolution, so if you can get by with slower pulses, the price will drop right down, if you can get by with 10nS samples, then I can easily see this thing being done for under $1k. The transition from hundreds of MHz to GHz has always been expensive as it requires a whole different design methodology, as you're getting into the microwave realm, all kinds of RF voodoo starts to mess with your circuit. – Sam Sep 29 '16 at 22:57