Newbie: How this simple 10/24GHz oscillator works? (no DRO, two resistors+microstrip)

volker_muehlhaus said:

But doesn't that require a linear frequency sweep?

Click to expand...

I do not see why everybody wants the linear frequency sweep. The resulting distance scale may be not exactly linear,. so you have to calibrate it. Try to use a ruler for the distance, etc.

jiripolivka said:

I do not see why everybody wants the linear frequency sweep.

Click to expand...

You misunderstood the problem.

The problem is not a nonlinear curve between IF frequency and distance. That would be trivial to solve indeed.

In FMCW, the IF frequency depends on the dinstance to the target.
For an ideal linear sweep, that IF frequency is constant during the sweep, so that we have a nice clean signal for processing.

The problem is this: If the sweep is not linear, the IF frequency will change during the sweep, for a given fixed distance. Instead of a single peak after FFT, which indicates the distance, you now get a smeared signal without clear maximum. How to interpret that signal now? In IEEE Explorer, you can find some papers about error modelling for non-linear sweep and approaches to handle that error in postprocessing. I've not looked into those recents papers, but know that linear sweep was considered most important when I worked a bit on FMCW in the 1990's.

jiripolivka said:

The article shows not a 10 GHz oscillator but one that operates at ~1 GHz, and harmonics could be detected by a car radar.
As you have no data on the particular transistor, you cannot design an oscillator with it. I have made similar transistor oscillators; the trick is that you have to make blind experiments with a number of transistors. Only some would oscillate and mathematics cannot help but experience. If you wish, look in an old cavity of an UHF TV tuner. There the frequency coverage of 470-890 MHz was achieved also by experiments.

Click to expand...

I am a little late making this post, but, inspection of the image linked in the lead post shows the device to be an MRF901, used to be fabricated by Motorola.

Jim

---------- Post added at 11:31 ---------- Previous post was at 11:24 ----------

RF_Jim said:

I am a little making this post, but, inspection of the image linked in the lead post shows the device to be an MRF901, used to be fabricated by Motorola.

Jim

Click to expand...

Datasheet, incl S-parameters: **broken link removed** -- Case style 317-01 Style 2


Osc design - mentions MRF901 - Discrete oscillator design: linear, nonlinear, transient, and noise domains - Randall W. Rhea - Google Books


PG 19 addresses MRF901 designs - **broken link removed**

Jim

ok ok, i agree that linearity of FMCW is important, as volker_muehlhaus said. But from my previous experiments with collector-voltage modulated DRO i saw that some distances can be "extracted" from what i hear in earphones. Even with wrong modulation and some nonlinear frequency curve. I am almost sure it is not applicable for many serious applications. The main problem is that all reflecting objects contribute to the beat spectrum, not only the biggest one in the beam. In case of good linear frequency curve each object gives us some peak on the spectrum related with the distance. In case of nonlinear frequency sweep each object give us some small "mountains", that interfere in spectrum. Indoor measurement also adds many double reflections. But for amateur use it have benefits: it is simple, but we still can get some distances from some points; we get a working high frequency oscillator.

After reading some on FMCW technique i see that i do not want fully implement it. For getting linear frequency sweep we must do some nonlinear voltage sweep. In this case our VCO must have some digital output for zero-crossing counter of local oscillator... implemented in microstrips and some elements i do not know =) Then this output must be fed to some high-speed ICs for counting and accumulating. After all DSP must read this counter several times to get frequency curve, alter voltage sweep to make next period readings more linear. Also need some equipment to measure and tune all that zero-crossing stuff circuity. There are too many parts i can stuck with. So i still love my dumb idea of wrong vco method =)

Here in the attachment image i show you some real 24GHz oscillation circut. Actually it was a very tiny motion sensor (5 x 4 x 1 cm). Two patch-antennas on the front side. Each antenna had a sealed cavity on the back side. My drawing shows only TX antenna cavity. All dimensions are not precise, taken from drawings i made before. That was a bad idea, because now I can't do proper simulation i CST Sonnet lite program, and it is difficult to understand how it works.

I will be happy if somebody help with my question, so i can "work it out":
I want make the same scheme in lower frequency, for example 5GHz..10GHz. To get some ideas about how it works I try to simulate the original 24GHz oscillator filter (3 microstrips with 1.5 and 3mm length). But when I start simulation, i see strong reflection of emwave and sometimes weak current leaks through that "filter". I think that it is wrong. What should I see in simulation? I thought that it must be some bandpass filter, that passes 24GHz and do not passes anything around that frequency in some range.


p.s. I downloaded too-many-too-read articles on Active Integrated Antennas.

Yes, that is a band pass filter. You did not like the DRO band pass, so a microstrip band pass is the alternative.

As mentioned a million times before, this is a design for a fixed frequency osciallator (motion sensor -> doppler radar with fixed frequency). For your needs (wideband tunable VCO), this makes no sense. You should really start looking into VCO design if you need a VCO.

I read about aia-antennas can be modulated with or without varactor. Also it is pretty acceptable scheme because i do not need long frequency sweep. For example automobile parking radar or something like that, collision avoidance. 50MHz sweep can be enough. I know i think like a kid and want to use this dumb bandpass filter he he. Sorry if it is annoying.
P.s. I want etch some simple board, just need some example of real microstrips.

volker_muehlhaus, i still do not understand, why i need any bandwith more than 50MHz? I just read some articles on range measurements, for example http://www.libsou.com/pdf/01610381.pdf - it says that 100MHz frequency sweep is used for maximum range of 100 meters. I know that wide frequency sweep gives more opportunities for long distances and better range resolution. But what kind of problem with small bandwith i can have?

The second question is about microstip parallel coupled bandpass filters.
For example: need microstrip resonator for 10GHz frequency:
1. Calculate wavelength λ=299792458/10000000000 = 0.030 meters = 30 mm
2. Find out PCB substrate dielectric constant. For example i found some "relative dielectric constant" of good laminate: ε=1.7
3. Correct the wavelength λ2=λ/sqrt(ε)=23 mm <- is that right?
I think i can use calculator placed here: **broken link removed**
ε=1.7 w=0.04 h=0.8 then i get ε of 1.5 and λ2=24.5mm

update: h=0.032, (0.8mm to inches)

4. Use "λ/4" parallel coupling bandpass filter where overlapping is λ/4 and the resonator length is λ/2.
So i get lonely microstrip with length of about 12 mm, that fed with 6mm microstrip at the top and em wave jumps to the 6mm microstrip at the bottom (almost as on this picture:
View attachment 69994
)
5. okay, it is time to etch the board and cut bandpass feedback with the knife until it works. And of course more calculations when i study more things =)

i still do not understand, why i need any bandwith more than 50MHz?

Click to expand...

I suggest to create a radar target model (one or multiple objects with a time-of-flight and reflection coefficient each) on pencil and paper and check the bandwidth requirements for unambiguous detection. You'll notice that minimal detectable object distance as well as resolution of multiple objects has to do with a c/2Δf parameter.

If you don't intend multiple object detection and low minimal object distance, 50 MHz can be O.K.

Terminator3 said:

volker_muehlhaus, i still do not understand, why i need any bandwith more than 50MHz? I just read some articles on range measurements, for example http://www.libsou.com/pdf/01610381.pdf - it says that 100MHz frequency sweep is used for maximum range of 100 meters.

Click to expand...

The link does not work.

Terminator3 said:

I know that wide frequency sweep gives more opportunities for long distances and better range resolution. But what kind of problem with small bandwith i can have?

Click to expand...

Wider frequency range gives more accuracy, not more range.
In your example above, if they have 100mm range and low accuracy/low resolution, a 100MHz sweep might work.

With narrow sweep and poor sweep linearity, you will reduce accuracy/resolution.

Terminator3 said:

For example i found some "relative dielectric constant" of good laminate: ε=1.7

Click to expand...

Not sure if that value makes sense. The lowest εr found in "normal" substrates is usually 2.2 for teflon based substrates.
For your microstrip line, some field will be in the substrate and some in air, so that you get an effective ε (somewhere between εr=1 for air and the substrate εr)

To design your line, use a transmission line calculator like AWR TXLine (free download from http://web.awrcorp.com/Usa/Products/Optional-Products/TX-Line/ )

Some new questions:
1. What's the name of microstrip building block shown on the image in the attachment? I want to find out what is it used for.
2. Is it good idea to use diode mixer to do VCO? Usually microwave mixer used to get IF output from LO and RF input. But also there is a way to do amplitude modulation of LO. I can't understand how it's shematic would look like. Also there is some miximg issue: after multiplication we get two frequencies: Fa=F1+F2 and Fb=F1-F2, but i only need Fa. Local oscillator is F1 (some Ghz), modulating it with low-frequency F2. How this problem can be solved? Both resulting Fa,b frequencies would be high, so no way to do pass filters. Some guy said i can try IQ-modulation (make kind of balanced mixer).

It's a directional coupler.

Terminator3 said:

2. Is it good idea to use diode mixer to do VCO? Usually microwave mixer used to get IF output from LO and RF input.

Click to expand...

Do you realize that VCO means voltage controlled oscillator? This is can be implemented with an L-C oscillator where the C is tuned (varactor diode). Some design examples are here: https://www.qsl.net/va3iul/High_Fre.../High_Frequency_VCO_Design_and_Schematics.htm

A mixer is a multiplier. You can multiply LO and RF, to get the sum/difference frequencies. You can multiply a carrier with a low frequency signal, to obtain amplitude modulation.

Terminator3 said:

Also there is some miximg issue: after multiplication we get two frequencies: Fa=F1+F2 and Fb=F1-F2, but i only need Fa.

Click to expand...

That is why your "mixer as a VCO" concept is not used. You try to replace a simple design (varactor tuned VCO) by something much more complicated (baseband sweep + IQ modulator). Makes no sense.

thanks for directional coupler, FvM!

volker_muehlhaus, thanks for fast reply: =)

That is why your "mixer as a VCO" concept is not used. You try to replace a simple design (varactor tuned VCO) by something much more complicated (baseband sweep + IQ modulator). Makes no sense.

Click to expand...

yes i agree.. It's because i just realized how mixers work from trigonometric point of view. Mixers is the only microwave "thing" I etched by myself on pcb.

Btw, i tried some trigonometry:
first mixer gives: cos(f1*t)*cos(f2*t)=0.5*cos((f1-f2)*t))+0.5*cos((f1+f2)*t))
second mixer gives cos(f1*t-90)*cos(f2*t-90)=0.5*cos((f1-f2)*t))-0.5*cos((f1+f2)*t)) - i am not sure about -90 shift for f1.
f1 - local 5..10GHz oscilator frequency
f2 - modulation frequency, for example from sound card (he-he, seems that maximum range is 22kHz?)
this two things must go away somehow: +0.5*cos((f1+f2)*t)) -0.5*cos((f1+f2)*t)) = 0.
and finally we have cos((f1-f2)*t)). Maybe can be used for tuning stable LO frequency obtained from DRO oscillation.

Ok, going to see https://www.qsl.net/va3iul/High_Fre.../High_Frequency_VCO_Design_and_Schematics.htm link.

Hello! Do not want to create new topic. Here is my question:

this time is not about VCO.

When using this scheme as downconverter it looks pretty clear and documented in many sources.
What if we fed low-frequency into IF pin? Then RF recieving antenna becomes transmitting antenna. I see some problem, because that antenna still recieves something "wrong", and it leaks to mixer?

Is it good idea to put directional coupler at RF pin, then put FET amplifier, and finally the antenna?

Dear All, what good books on practical microwave design you can suggest? With some talkthrough examples, when some frequency is chosen, laminate, LO-pass filters, microstrips, R&C's calculated. I saw a chapter from Sayre's "Complete Wireless Design", and want to find something like that.