Techniques for high speed (GHz) current sensing

Hi there - I am looking at methods of sensing current in an Avalanche PhotoDiode (APD). Typical current in them is along the lines of 100 nA to 1 mA. I'm trying to figure out how to sense current in the buggers. Here is what I think I need:

-Very low added resistance/inductance/capacitance on APD

-If using some sort of inline current sensing, I'd like to see a voltage drop across the shunt/transformer/etc. of less than a volt, ideally mV range (the APD will have a very large reverse bias (~100-200V), so its gain will be very sensitive to changes in supply voltage)

-bandwidth in the GHz region. I'm looking for a rise time from the circuit of ~10ps, which I believe puts the necesarry bandwidth at about 35GHz

When I saw the 35GHz number, I just about gave up - but figured I'd ask before doing that. What little analog stuff I've done has had bandwidth in the single digit MHz... So this is very new for me. I'm prepared to spend a lot of time learning, though.

Now, as for the output of this circuit, let me explain a bit about what I need: eventually I'll want it to go to an input of a comparator. I may end up only using a comparator, so in that case linearity would not be a big concern, as the other input to the comparator will be a constant, adjustable voltage. Ideally though, I'd have a nice linear output with a sufficiently high enough voltage so as to read from an ADC and do other fun stuff with. But the comparator function is much more important, so if having a highly non-linear output makes this more achieveable, than I will survive.

So - is this at all possible? Can anybody point me down the right path? I'm not even sure what sort of technique I should be using with this - resistive shunt, transformer, hall effect, etc.

Thanks!

-Michael

Reply to
Michael
Loading thread data ...

Wow, that sure is RF stuff. About the only way to do that on the cheap is to amplify with hotrods such as the BFP620. From a sense resistor, multiple stages, piece by piece. But be careful since those can oscillate at frequencies that even most fast scopes wouldn't show.

formatting link

Or use MMICs. Trouble there would be that it's hard to find any that could do wideband, depending on how far down in frequency you need to go. John Larkin would be the guys there.

--
Regards, Joerg

http://www.analogconsultants.com/
Reply to
Joerg

You are very confused about what APDs do.

First, although the avalanche process uses very fast physics, for the avalanche itself, the effect is seen rather slowly on the two output terminals of the detector. That's because they typically have a rather high capacitance, check your part, and the avalanche current is charging (discharging actually) that, and then you later see it with a sense resistor on the output side. What you'll see, depending on the part, is a modest pulse, lasting a relatively long time, like 5 to 15ns.

Second, for most APDs, after an event, which is initiated by just one photon, by the way, there's a refractory period where the APD detector doesn't work properly. Most detector circuits sense the inital pulse and fire a oneshot, for say 40ns, to insure that they don't pay attention to the APD until it's working reliably again.

Third, photons arrive randomly, so if your photon flux exceeds say f = 1 / 5*40ns = 5 x 10^6 per sec, then you'll start suffering from pulse-pair crowding losses and a nonlinear pulse-rate output with light flux. That sets the high end of the flux rate.

Fourth, low-cost APDs often have have a fairly-high dark count, because they're biased close to avalanche, and small noise events in the sensor cause avalanches. This sets the low end of your measurable flux rate. If you pay the big bucks, up to $5k per sensor, you can get quiet APD systems. They'll have to be cooled to achieve their performance specifications. If you want that supplied with the APD, it'll cost even more.

Reply to
Winfield

Hi Winfield - let me go full disclosure here:

I want to build a high accuracy low range time of flight laser range finder. I then want to put that device on an articulated 2DOF gimbal mount and make a 3D scanning laser rangefinder. My initial goal is 5 meter range with 1cm resolution and +-1cm accuracy. With that range, I believe an APD is overkill for all but the least reflective materials. (or would it be most reflective? I'm not sure which is harder to get a pulse back from - a mirror or a perfectly non-reflecting object).

However - my idea behind using an APD is that once I achieve 5M range, I can scale up the circuit fairly easily to higher and higher resolution without making massive changes to it. Mostly, I think I would just need to improve the optics in front of the APD and increase the APD's reverse bias voltage, and possibly actively cool the APD. I could also increase the output power of the laser diode, but I'd like to keep the device as safe as possible - I mean I'm quite happy having two functional eyes.

I was planning on driving the laser diode as quickly as possible. By keeping the current through the laser diode always just below the lasing threshold (except when it is on) I think it should be possible to make very very quick pulses - single digit nanosecond lengths should be very possible with enough work, I believe. Thus I was hoping to be able to take a range every 50ns or so. If I have to go a lot slower than that I'll survive - but I'm hoping for at least one hundred thousand samples per second. That is not as important of a parameter to me, though.

So - by my calculations it takes 67ps for light to travel 2cm (1 cm there and back). Thus I figured 10ps would probably be a good place to shoot for if I really want to acheive 1cm resolution. Next, to figure out the bandwidth necesarry - I tried to calculate the relationship between bandwidth and rise time of a RC low pass filter. My calculation gave me: bandwidth =3D ln(3)/(pi * rise time). I'm not sure if that is a valid calculation. Anyways, plugging 10ps into that gives me 35GHz.

Am I going about this in entirely the wrong direction? My understanding was that an APD was just a much more sensitive photo diode, and thus I thought it would make alot of sense to use one so that I could keep laser output power low. I thought I read that when you lower the reverse voltage across them they act just like a normal PIN photodiode. I know they are used for photon counting, but I thought that that was only done at very high reverse bias voltages with peltier cooling to keep the dark current down.

Thanks for putting up with me - I know I'm a bit clueless when it comes to some of this stuff. But what I lack in knowledge I make up for tenfold by an eagerness to learn new material!

-Michael

Reply to
Michael

You need to do a photon budget. Getting that 1 cm resolution will require a fair number of photons, and cranking up the APD gain won't improve the shot-noise limited SNR. Also APD speed depends on the multiplication gain--the ones I've seen are quite a bit faster than Win's--bandwidths of ~1 GHz or so--but it takes awhile for the avalanche to build up to its full glory, which slows things down. (And in some materials, you wind up with unstable avalanches since both holes and electrons can do the avalanching. Si and InGaAs are okay.)

It isn't that it can't be done, but you need to focus on how many photons you need--to measure a time delay of epsilon times the pulse width, you'll need several times 1/(epsilon**2) photons.

Cheers,

Phil Hobbs

Reply to
Phil Hobbs

Hi Phil - is it really possible to plan around how many photons the APD will be receiving? I mean, won't that depend heavily on the material that the laser is hitting and how far away it is? Thanks,

-Michael

Reply to
Michael

Precisely so. But physics being what it is, it's nonsense to go looking for good timing specs when you don't have enough photons for the job. Some back-of-the-envelope estimates for how many you're going to get would be very illuminating. ;)

The SNR of an optical measurement tends to vary all over the map with things like distance, surface finish, and incidence angle. One method that can help a lot, if your targets are cooperative, is to put some retroreflective tape on them. If you pick the right stuff, e.g. Scotchlite 2000X or 7610 (3M), you can get as much as a 70 dB signal level improvement compared with a Lambertian (i.e. diffuse) reflector, and a much bigger improvement compared with an oblique specular reflector.

Cheers,

Phil Hobbs

Reply to
Phil Hobbs

You might want to look into time of flight rangefinding methods that measure phase instead. Depending on the needed update rate, you might find that 100 MHz BW and ordinary PIN photodiodes can do the job. Paul Mathews

Reply to
Paul Mathews

What sort of laser do you plan to use? What wavelength? Affordable VCSEL or Fabry-Perot diode lasers can have 100 ps optical risetimes

*if* you drive them really hard and fast. A regular pin or avalanche photodiode will generally be in the hundreds of MHz, 1 ns or so, and get more expensive and harder to handle as speed goes up. 10 ps risetime is unreasonable, since you won't be able to process signals of this speed, and you can't buy a comparator that fast. 100 ps signal processing is difficult but posssible using conventional construction techniques.

Using pulsed time-of-flight is going to be extremely difficult here. Something like a sinewave phase-shift thing would be more practical.

What's the application?

John

Reply to
John Larkin

1cm = 30 ps ~ 50dB SNR if the bandwidth is 100 MHz. With the reasonable optics @ 5m distance, the incoming SNR ~ 0dB -> accumulation for ~100k pulses -> one measurement per millisecond. Not too bad. The short term stability of the clock should be at the order of 1e-8, which is not a problem either.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

formatting link

Reply to
Vladimir Vassilevsky

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.