Reminds me of da palmy days of me yoot, when I wasted an entire day trying to make an op amp phase shifter that had the opposite phase slope...in other words, as I eventually realized, a time machine.
Cancelling a pole with a zero is good medicine sometimes, if you don't need too much accuracy. In the present case, of course, doing that gives you back the raw PWM signal. All the application examples are analogue, e.g. a thermometer, and there's a good reason for that.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
The initial single pole is silly, I did a P+D circuit last year to speed up a response time... Does the author point out that the speed up comes at the cost of more noise? If not then, he's missing a crucial point. I like this line,
"e is an irrational constant approximately equal to 2.718281828"
They were trying to claim that that was the only irrational thing in the article. ;)
On the other hand, you can use a gated integrator or a T/H to get a PWM to settle in one period. It's usually better to use a DAC at that point, but a Joergesque solution could get by with a CD4007 and two caps, maybe 25 cents on a bad day.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
Rob is fond of using delta-sigma modulators instead of PWM. The filter tradeoff is much better. As long as you're making the logic in an FPGA, it's not much more work to do the d-s as compared to PWM. [1]
The EDN guy makes the common mistake if fiddling around with circuits, and not backing off to appreciate that the entire thing is a linear circuit that has a transfer function. And this guy works for Linear Technology!
John
[1] but I confess that I really don't understand how one makes a precise d-s dac or adc. Seems like the edge timing would have to be accurate to femtoseconds or something like that.
I did my first software D-S for an auxiliary temperature controller a few months ago, at Tim's suggestion. Worked great--I'll probably never use PWM+filtering again. PWM is great for controlling RC airplane servos, though.
(That's actually a serious comment--$150 will get you a beautiful servo with titanium gears and boatloads of torque, good for many sorts of jobs. A 50-Hz PWM is all you need, perhaps extended to higher resolution via D-S tricks in the timer interrupt handler.)
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
You must be blind as well as stupid. Maybe it will sink in this time.
BWuaahahahahahahahahahahahahahahahahahahahahahahahahahahahahaa! You are such a funny example of stupidity that it makes my belly muscles ache from laughing... at you.
Gated integrator's.. aka boxcar averagers. They seem to have gone out of style*. I was thinking I could use one to pick off the tail of the photodiode pulse response. Hit the edge of the PD with a laser pulse. (sorry dragging this thread back to the start.)
So here's a question,. (I'm not expecting any definitive answers.) Is the slow response from the edge of the PD due to a diffusion effect. The edge absorption is in a doped region and carriers must diffuse from there before entering the depletion region, where they are swept away... and give a sharp pulse. ('full shot noise') or, Is the slow response from the sides due to edge defects in the undoped intrinsic layer. Charge carrier near the edge get trapped 'in the middle' for a micro second or so. (And thus show less than full shot noise.)
George H.
*(I guess it's easier to digitize everything and let 'god' decide what to keep.)
This starts to get a little subtle, I think. ISTM that you have to draw a conceptual line between processes that contribute randomness, i.e. primary photodetection and diffusion, and ones that don't, e.g. the RC rolloff due to the photodiode capacitance. So let's model the PD as two noisy current sources in parallel--a slow one and a fast one--with a capacitor in parallel.
I'd be willing to bet that both the fast and slow current sources produce shot noise equal to sqrt(2*e*I_N), where I_N is the instantaneous current arriving by the Nth mechanism.
(One has to be a bit careful about the statistics of time-varying things, of course.)
Then you apply the transfer function of the RC rolloff in order to figure out what that gets you at the output.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
Of historical interest: (Two different Anders(s)ons, both papers good)
=========================================
Anderson, Larry K. ;
Measurement of the microwave modulation frequency response of junction photodiodes [using microwave noise measurements]
This paper appears in: Proceedings of the IEEE Issue Date : May 1963 Volume : 51 , Issue:5 On page(s): 846--847
=========================================
Torbjörn Andersson, Alan R. Johnston, and Hans Eklund
Temporal and frequency response of avalanche photodiodes from noise measurements
Applied Optics, Vol. 19, Issue 20, pp. 3496-3499 (1980)
Abstract
This paper describes a method of obtaining the temporal and frequency response of avalanche photodiodes (APD) by performing simple noise measurements. From the measured noise spectrum and by using the Hilbert transformation technique, the complex transfer function of the detector is determined. response can then easily be calculated by means of fast Fourier transforming. The method has been applied on a high speed APD, with a bandwidth of ~2 GHz, and on a relatively slow APD, with a bandwidth of 0.2 GHz, to calculate the pulse response from a short optical pulse. The calculated pulse width for the fast APD was 215 psec, and the corresponding measured width was 210 psec, while for the slow APD the calculated and the measured widths both were 3.1 nsec. Also the shapes of the pulse responses showed excellent agreement. The method depends on the essentially identical frequency response of an APD and associated circuits for noise due to steady-state illumination and for a signal.
All very true. I'm suggesting it's a bandwidth question. If one channel (the slow current source) has a longer transit time (across the depletion region). Then the band width is less. So my picture for that would be some edge inducted trap states. The charge carriers are photo-excited in the intrinsic region, but one falls into an edge trap and gets stuck there for a few microseconds. (Is that a reasonable trapping time?) . Then the qv (charge times velocity) current pulse gets broken into two. (That last for a shorter time, in this 'simple' picture.) Mind you, I'm pretty much making up the idea of such edge defects.
Hmm, this is not really a bandwidth issue.... It's a splitting of the current pulse in two. Each, on average, only 1/2 an electron long. The qv pulse integrated over time has to add up e. The charge of the electron. (That is kinda subtle.)
The other possibility is that the slow response is due to absorption of the photon in an edge doped region (p or n). The slow response is just waiting for them to diffuse into the depletion layer, where each gives one nice current pulse. This seems much more likely, since diffusion times are in the microsecond range and that's what I see.
Yeah, I was trying to separate the current pulses in the PD from how my external circiut responds. (It seems to all connect at the PD capacitance.) As you said all these current pulses are much faster than the other RC times in the external circuit...
The second one is my favourite method for characterizing the linear response of PD/TIA combinations. You don't need a VNA, just a flashlight and a spectrum analyzer.
Cheers
Phil Hobbs
(in sunny Rancho Bernardo by myself for a few days--any locals want to hoist a cool one and talk about circuits?)
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
Well, there's an easy way to find out--shine a laser on the centre of the PD and then the edge, adjust for the same DC photocurrent, and measure the noise spectrum. My bet is that the noise spectrum is exactly the same in the two cases.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
Are you staying at the RB golf course spa place? How is it?
Not too far away is the Stone Brewery
formatting link
where the beer is good [1], the food is excellent, and the outdoor setting is beautiful. Try it. Or wait for me... I'll be in the neighborhood on Monday and Tuesday.
John
[1] like most gourmet brews, most of their beer is a bit hoppy for my taste.
There was a wedding there this weekend, so I'm at the Hilton Garden tonight, but then at the golf course place Sunday through Tuesday. (My flight leaves at 7:30 AM on Wednesday, and the airport is a half hour away. That gets me to LGA by about 4, so if I'm lucky I'll get out of Queens with a whole skin.)
I'm going to go to Palm Sunday Mass (Divine Liturgy actually) at St. John of Damascus Orthodox Church
formatting link
which I'm looking forward to very much. I've never been to an Orthodox liturgy--I live close to one of the two main Orthodox seminaries in the US, but they don't have a parish attached, just the school's chapel.
And my son Simon is due to come back from his 8 months in Africa on Thursday. One of the girls he's travelling with just came down with malaria, and he thinks he may have a touch of it too. (All good things to light candles for tomorrow.)
So sure, let's hoist a few then.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
BF862 is *the* magical jfet. There's nothing else like it.
If you use it as a source follower, and bootstrap the drain, Cin will be tiny. So I figure, given a photodiode, the lowest noise situation is to have it do nothing but charge its own capacitance. The fet follower is the next best thing to connecting nothing at all to the photodiode.
So the PD, hit with a pulse of light, makes a voltage step into its own capacitance. The BF862 and downstream stuff amplify the step, and then you can pseudo-differentiate it back to a pulse, or just digitize the step.
Of course, every pulse, plus any dark current, raises the DC voltage at the gate of the fet, so it eventually crashes. But you can occasionally yank down the fet source, forward-bias the gate diode, and DC reset the whole system. Turns out I have time do that in one potential application.
No resistor, no Johnson noise, no loop stability/noise peaking issues, hardly any extra capacitance, no shot noise much beyond the photons themselves.
If all you're looking for is fast pulses, a multi-gigohm resistor to ground, or a sorta TIA feedback config, will do the DC reset function continuously, and the Johnson noise will be lowpass filtered by the diode capacitance such that it doesn't matter much. But not as cute.
A variation for low frequency illumination might be to reset it periodically, integrate-and-dump style.
This may be silly of course, but playing with silly circuits is still a good thing to do. You never know where it might lead.
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.