44 kHz Doppler radar air speed meterr phase shift in wind tunnel

Unless you use two mutually prime carrier frequencies, and use the Chinese Remainder Theorem (or the Trunk and Brockett improvement) to determine which rollover cycle one is in).

Good. They probably use Peebles as the textbook.

I'm thinking of the traditional F-S detector with two LC tanks stagger-tuned. While phase is certainly involved, because frequency is the rate of change of phase, the F-S output does not change if one changes the phase of the carrier being tested without changing the frequency, so a F-S detector is no good at detecting a steady phase shift. So, after a short time, the detector output will drift to zero, in effect claiming that a steady windspeed is zero.

Nor are F-S detectors especially precise.

The purpose of the algorithm is to allow one to use multiple mutually-prime carrier frequencies, to resolve the rollover problem. Sounds like two frequencies would suffice.

If one arranges the circuit correctly, the two frequencies can be simultaneous. Failing that, they are used alternately, switching fast enough to follow short-term variations in windspeed such as gusts.

To follow turbulence, the simultaneous approach may be required.

If the beams entering and leaving the flow are perpendicular to the flow, no correction is needed. The transducers are big, and will disturb the flow

The standard alternative is to have the beam cross the flowtube at a slight angle, directly from transducer to transducer without mirrors, with transducers mounted in recesses in the tube wall.

And clarifies the thinking, in my experience.

On a sunny day (Sun, 01 Dec 2013 11:39:27 -0500) it happened Joe Gwinn wrote in :

Yes you mentioned that, but here I have only one carrier frequency,

I have read it now.

Well depends how you look at it, I look at the phase comparator part, and would only use that:

In fact, if you look at this youtube video I made,

Right, good enough for audiophiles though :-)

I like the sampling way, generate a pulse on start positive going wave of Tx. Reset a ramp on start positive going wave of Rx, With an analog switch sample the ramp. Almost 360 degrees, linear,. Digital: Use the 2 PIC comparators to detect pos going wave, start counter one interrupt comparator 1, read counter on positive going wave of 2, Something like that, 18F14K22 PICs have 2 build in analog comparators that have < 10mv offset. No extra hardware needed. And I already have the PICs,

I just watched the Chinese moon lander launch live on CCTV news via satellite, recorded it too, already set to go to BluRay, 1.5 GB. Now that is nice to see all those young people do such a nice job, X band communications. directly into moon earth orbit, everything working 100 %, Landing gear and solar panels on moon lander already activated, plutonium RTG power source. Last were the Russians in 1977. China is getting ahead.

I don't visualize the proposed circuit.

Yep. But no longer used even there, these days. State of the art is a phase lock loop.

Well, this sounds a bit noise-sensitive.

At RF the standard way is to feed the two 44 KHz signals to a double balanced ring diode mixer (from MicroCircuits) followed by a RC low pass filter to eliminate the 88 KHz sum signal to yield a DC voltage that varies with phase. These are cheap and small, and provide near-ideal conversion properties.

However, 44 KHz is a bit low for purchasing such things, the transformers being the usual problem. Not to mention lack of sufficient market. So, one rolls one's own.

One approach (which I used in the 1970s or 1980s) is to switch the received carrier using the transmitted carrier to drive a inverted bipolar transistor. Now days, I'd use a FET of some kind.

Or, I would combine amplifier and switch into an AD630 Balanced Modulator/Demodulator from Analog Devices.

They will succeed, because they have decided that they will do what it takes, and now have the industrial base to do it.

And it's a really good way to train an entire generation of engineers et al.

As for turbulence in the pipe, you will need screens covering both openings, to keep junk and critters out, and the screens should flatten the flow out, the main issue being to ensure stability and predictability.

Joe Gwinn

On a sunny day (Sun, 01 Dec 2013 14:59:24 -0500) it happened Joe Gwinn wrote in :

Rip apart the coupled coils (bandfilter). Now one is the coil in the transmitter, and the other in the receiver.

State of the art is digital,....

There is very little noise because of the high Q and large amplitudes.

I have a ??? chip somewhere tha thas single inputs, not diffrential, upto soem GHz, but I think that is not needed, a PIC should do.

Yep

Maybe I won't use no pipe...

Yes, but it's still a PLL. Works much the same, but cost is orders of magnitude less.

When you use it outside, turbulence alone will give you multiplicative noise.

Yes, one can do it that way. Best to use matched schottky diodes, to keep the offset low. Balance of both transformer and diodes is the key, and is easier said than done.

I suppose, but I would go about it differently. I'd use the PIC to synthesize the two mutually-prime carrier frequencies, and to do the bookkeeping to track the various measurements, and to do the math.

Use of some analog components can greatly reduce the load on the PIC, one example being going from handling two carrier signals at 44 KHz versus one phase signal at roughly DC.

More generally, one does the high-frequency but simple stuff in analog hardware, and the slow but complex stuff in the PIC.

Actually, what PIC chip are you using?

Then, a crossing-beam arrangement, which allows computation of wind direction as well. One can use two mutually prime carrier frequencies, and alternate them between E-W and N-S. One PIC would manage the whole show.

Joe Gwinn

An alternative that starts out by measuring true airspeed has to be augmented by a system that takes into account the air density before it's of use for aircraft control purposes. To calculate the density requires a measurement of the static pressure (and temperature). People have been killed in airliners as a result of the static pressure vents getting blocked (in the particular instance I'm thinking of, it was covered by adhesive tape).

There was no good reason for AF447 to have crashed, but crews will crash aircraft, despite every technological defence, if they don't do what they're trained to do.

Sylvia.

On a sunny day (Sun, 01 Dec 2013 16:26:29 -0500) it happened Joe Gwinn wrote in :

No no no no, strange you did not see it as it was you who mentioned IQ. In digital (software defined radio) the instantaneous phase (vector direction) is derived form the IQ signal. In the next sample (of IQ) the vector points elsewhere. The atangent gives the 'amount of phase change', and that is a measure of the frequency deviation. Well I have put it a bit simple, but if you can program in C, look for rtl_fm,c, IIRC it is part of the rtl_sdr package, maybe look for that, It does pretty decent narrow and wide band (like broadcast) FM. I experimented a bit with that code, it is interesting as it has a fast atan2 too. Look for atan2 in the source. Software defined radio is the current 'state of the art'.

Well this is why I am testing things, yes that is the interesting part,

In my amateur radio days I made a great 4 diode balanced mixer at about what was it 4 or 7 MHz, driven by tubes you know, very nice to make a double sideband signal, Also did it at 9 MHz, and then followed by a 9 MHz crystal filter to make single sideband (exciter).

More generally just do as much into software as possible, the PIC 18F14K22 is a 64 MHz clocked and has 2 _analog_ comparators, that can interrupt. It also has hardware frequency counters, PWM unit, SPI, I2C, RS232, DAC, internal voltage references, clock multiplier PLL, internal oscillators, what not.

Yes I gave a link to that wind speed meter earlier on in the thread. There are also a lot of PIC projects on m website:

And other projects:

Everything works, but now there are so many that I will have to think of better names so I can remember what it did once I see some directory with code. There are many many more not on the website, some for obvious reasons.

Umm. Pure analog PLLs were used in the 1960s, but for very specialized applications, because of the expense. For instance, the US Space program used PLLs for the voice links used to hear astronauts speaking from the Moon's surface.

All the tricks you mention were known then, but were too expensive until semiconductor technology had advanced to the point that a PLL could be integrated. First as analog, later as digital. Analog was first because digital requires far faster transistors than analog to handle a given signal frequency.

For lower frequencies, possibly. For higher frequencies, the RF stuff needs to be analog, because current semiconductors just are not fast enough (or quiet enough).

Yes, for audio purposes, this works well. But a phase detector cares about DC and offset stability.

All for the better. How much code space?

Sounds able to run a crossing signal dual-frequency phase-measurement setup.

Yes, I recall the crossing-beam setup.

Are the transducers waterproof?

Joe Gwinn

Mmmm, turned into a food thread already? I made french toast with walnuts and sliced banana's sauteed in butter for Saturday breakfast. (smothered in maple syrup.) My wife said she'd keep me for another week.. which takes the pressure off. :^)

Grin.. I started the turkey broth yesterday.... I don't know what I'll do with it? But it's a great way to clean out the veggie bin in the frig.

George H.

On a sunny day (Mon, 02 Dec 2013 09:48:59 -0500) it happened Joe Gwinn wrote in :

I have used 4046 PLLs for FM audio generation and demodulation, narrow band too.

You need to read up a bit, my rtl USB stick goes to 2.2 GHz, PLL tuner, zero IF, quadrature ADC, it will decode any modulation system, inclusive DVB-T, AM, FM, pulse whatever, dump1090 shows all the air traffic.

Think it was 16 kB (8 k words) 512 bytes RAM, 256 bytes EEPROM, and of course a multi channel 10 bit ADC. I have only filled the code space once:

It can do a lot, depends on the programmer though, and the language, I program it only in asm. I have the complete auto-pilot plus sensors reading in it, plus downlink for HUD display, plus some other stuff.

You can get those water proof, 12 Euro or so at conrad.nl.

n.

al from the opposite side.

sajous xy display

s.

le,

n know that?

ratio detector could do the phase

Hi Jan, You've gotten my interest. It looks like most people do a time of flight measurement. (which strikes me as error prone... how do you 'defin e' the start and stop times.)

But this does a phase comparison

I'm a bit confused as to where you are getting your two signals from. If y ou are just looking at the Tx and Rx signal. Then it seems to me there sho uld always be some phase shift between these two. Or did you "set that to zero" at zero wind speed? You could also adjust the reflecting surface til l you got a zero. But if you are inside a tube then aren't you getting ref lections from all over the place?

I'm not too sure about you calculation. I would have guessed that it only depended on the distance between the Rx and Tx in the direction of the wind .. and not also on the distance perpendicular to the wind direction.

Say are the Rx and Tx the same.. so that you can switch roles? Use Rx for Tx and so on.

George H.

On a sunny day (Mon, 2 Dec 2013 12:37:44 -0800 (PST)) it happened George Herold wrote in :

I have done the time of flight, see my other postings, but at these low air speeds the difference is hard to measure. For time of flight I just wrote some C code for Raspberry to switch the transmitter 10 ms on, 10 ms off. The tx rising edge is quite sharp, and so is the received signal's rising edge, when the tx goes off, then the amplitude decays within about 1 ms, but for time of flight you only would use the rising edge. Because I could not detect the low air speed and IIRC it was Glenn who suggested using phase, So I tried xy on the scope to see fif that was measurable, and it is.

Ch1 is from the collector of the tx driver, ch2 is directly on the rx transducer that is tuned in resonance with a big L.

I was lucky :-)

The transducers are directional, yes you get reflections from more than one place, but those are weaker, the shortest path is the strongest,

Yes there needs to be a sine(angle) somewhere I think.

Well this was with receiver 'upwind', The other way around should work too,

In the radar world, 100 MHz is considered audio.

I work in the GHz range.

I was an assembly-language programmer in the 1970s and a bit into the

The transition to Fortran came as soon as the computers became strong enough, because it takes on average four lines of assembly code per line of Fortran, and human effort is roughly proportional to lines of code (regardless of language (except for APL)).

Good.

By the way, there is an article that may be useful: "An analysis of the classical Doppler effect" by C Neipp et al, Eur. J Phys. 24 (2003)

Joe Gwinn

On a sunny day (Tue, 03 Dec 2013 09:41:15 -0500) it happened Joe Gwinn wrote in :

When working in TV a somebody gave me a real nice book on radar, cannot remember the name, I learned a lot from that, but never really worked in radar.

Yep, I am up to 12 GHz (x-band) now.

This is a 1.6 GHz TV DVB-S transmitter experiment, works very well:

We were required to learn C in the eighties, before that everything I did was asm on micros, and asm and BASIC on the first IBM PC. And BASIC on a Commodore PET:

I remember playing moon landing game on that.

Am I imagining or has google changed search algo? I was looking for the Chinese moon landing, basically orbital parameters, and where it is now, and found noting useful, not even up to date news with google, tried bing, and it got interesting really fast (but no parameters either). Maybe it is secret.

I really do not now all that stuff about code lines, and programmers doing 3 lines without errors a day or whatever :-) Once you have good asm libraries and a decent editor much is cut and paste, It was harder with a line editor... But yes, C is much faster, but C is not suitable IMO on small micros and it does not even make sense if 90% of your routines is talking to peripherals.. That is all register work, counting processor cycles, timing.

I never used Fortan, I did buy a book on ADA in that time, never used ADA, and I had a book on ARM.

Yes, found it Neipp2003.pdf To me these things are extremely simple, other things are not... probably because to much is parroted in literature too.

In fact every thing is simple one you have done it... :-)

Well, WW2 ground based "Chain Home" radars operated below 100 MHz. It's a start.

Receivers are typically the hard part, especially if one needs low noise figure and a wide dynamic range in the presence of strong interference. Digital can achieve such things only at frequencies that are a small fraction of the digital logic clock rate.

I don't know how much the Chinese have released, but I imagine that the orbital elements are nonetheless known, as there are many radars and telescopes that can track it. Eventually, someone will publish the data. Perhaps one of the astronomy magazines.

The usual dodge is a mix of C (not C++) and assembly-coded C subroutines or functions. This allows one to get the advantages of both languages.

I did a lot of Fortran, and did some Ada83. Most of what I did with Ada was to find ways to evade Ada83's limitations for embedded realtime applications. Ada95 solved many or most of those problems, but too late to save Ada from being eclipsed by C/C++ in the market.

Yes. I prefer math done by physicists versus physics done by mathematicians.

Yep.

Joe Gwinn

Oh and Physics done by engineer's, rather than engineering done by s physicist. I always thought that in college Math should be taught by Physicists, Physics by the Engineers, I don't know who would teach the engineering.. maybe a *working* engineer. And what do the Mathematicians do?

George H.

College math, at least for engineer, should be taught by engineers, as should Physics. I had one math prof (diffeq) who used EE problems. It helped a lot.

My father would have agreed with you about working engineers teaching engineering and he *was* an EE prof. ;-)

Teach philosophy?

The two best math classes I ever took were one on Fourier transforms taught by Bracewell and one on asymptotic methods taught by the estimable Professor Stefanos Venakides, who started out as an EE and switched to math.

Pure mathematicians really don't seem to like teaching math classes that are a means to an end.

Cheers

Phil Hobbs

An afterthought. One reason that the Doppler literature is confusing is that there are two kinds of Doppler effect, Classical and Relativistic, and people often neglect to say which one they are writing about.

Before Einstein, there was only the Classical Doppler effect, and Relativity had not been invented, so no distinction was made - it was just the Doppler effect.

The key issue is that the the wave traveled at a fixed speed relative to the medium (say air or water or metal) in which the wave propagated, so if the medium was moving, the wave was carried along (convected).

People tried to apply this theory to light, and it didn't go well. The Michelson­Morley experiment was an attempt to detect this convection in the Aether due to the motion of the Earth. No such thing was detected.

Einstein sorted it out by realizing that light traveled at the same speed in all reference frames, which requires that there be no convection phenomena for light (and thus for all electromagnetic phenomena).

Circling back to the question about ultrasonic windspeed measurement, the correct theory to use is classical, and if the article doesn't say that it is in regard to the classical effect, keep looking. The exception being stuff written before 1905.

Joe Gwinn