cheap design for a fast keying transmitter (microwave)

[CopperCone]

So in order to test the PlutoSDR POI, I think that what I need is something like an oscillator hooked up to a switch.

I wanted to key a transmitter at various intervals for various duty cycle (easy to control with a pulse generator), but I need some kind of fast RF switch.

I'm under the impression that I should be looking at PIN diodes but I am not sure. I would say a power handling capability of -40dBm is fine.

Are these a good choice?

I have a old spectrum analyzer that's way beyond economical repair, I thought I might be able to harvest some parts from it. I am however reluctant to take apart the attenuator. It has a YIG oscillator too. Are there any parts in a typical SA that could handle this requirement? (it goes to 3GHz)

I wanted to try keying duration's of 10nS, 100nS, 1uS, 10uS and 100uS. I would like to use a signal of at least 1GHz, but being able to do multiple frequencies up to 6GHz would be good to see how uniform it is.

Or, are there designs for some kind of oscillators that lock really quick, so I can just use a regular transistor to apply power? The problem with this is that I wouldent actually know how long it oscillates for, because I don't have a fast oscilloscope.. so the design would have to be well proven.

[randsl]

What sort of frequency accuracy you need?

If you need really accurate frequency, I don't think switching the oscillator is not a good solution. Normally, if you use a PLL to generate a really accurate frequency, the lock in time is typically us or ms range. So switching the oscillator is not a solution if you need ns, 10 us pulse with very accurate frequency.

However, if you don't care about the frequency (and transients) and just need a RF pulse which is closer to a certain GHz value, you can use a single transistor oscillator with turning on/off with short pulses with the required duration. I've done it with 24 GHz frequency and 2-3 nS pulse on time.

[CopperCone]

so these oscillators start up very fast?

I was worried because once you flip the switch, the oscillator needs some time to start up before its truly on... I was not sure what that would be.

I guess it needs to be within 30MHz or so.

Any particular circuits in mind?

[randsl]

if you use a single transistor oscillator, with no varicap or DROs, yes it does, but the frequency is not accurate as it is a free running oscillator & it's subject to voltage transients during the start-up, die-down... There's a small delay typically in ns scale. but these ones are very fast ...

If you need to be within 30 MHz closer, then this is not an option I guess...

I'm unable to provide you the exact one which I used for 24 GHz as that one is related to a my workplace. However, it's vary similar to the topology used in this research paper : https://www.researchgate.net/publication/269040235_Multifrequency_Oscillator-Type_Active_Printed_Antenna_Using_Chaotic_Colpitts_Oscillator

If you switch on/off with a microcontroller pin (if you need < 10 ns, you may need special pulse gen. circuit), which can provide sufficient current typically (10 - 30 mA) it works well.

[randsl]

If you use a vericap or dielectric resonator, then the frequency accuracy is getting bit better than this method. However, startup time slows down..  but again, that depends on what you need. If you need slower keying, and no need of very accurate freq. could be a solution as well..

If you need faster keying, and accurate frequency need to consider a different solution...

[CopperCone]

ok, actually, is it repeatable? All I care about is repeat-ability within 20MHz or so so i can test it according to the real time span of this thing. The exact frequency I don't care. I realize this caused some confusion. Like, I can find the signal, and hone in on it.. but I can't be chasing it around because the whole object is to determine probability of intercept statistics.

I don't know much about single transistor amplifiers, but this book has a design here:

https://books.google.com/books?id=4Op56QdHFPUC&pg=PA275&lpg=PA275&dq=2n5109+oscillator+circuit+ghz&source=bl&ots=HKTdlFwyy1&sig=D8kmy9w5jN5IEKKwt8x86ebpdv0&hl=en&sa=X&ved=0ahUKEwi3x82J2JTWAhXK8YMKHUz8DrcQ6AEINTAF#v=onepage&q=2n5109%20oscillator%20circuit%20ghz&f=false

I think I can replace the t-line with a inductor (i am winding air inductors on a magnet wire with a drill bit), using an online calculator. I only have a DE-5000, but it looked like the calculator was pretty much right for a 22nH inductor I wound. I chose that because thats the only RF transistor I found right now.

Not sure what to do about that transmission line. Suggestions would be welcome. I can deadbug it on a piece of copper. I was just gonna try a bunch of air inductors between 10 and 1000nH. Hopefully it won't burn out (the book says he replaced the inductor with a t-line).

Should I just use strait pieces of wire instead?

[randsl]

As long as you apply the same pulse (amplitude, width and rise, fall times), and you are not disturbing the oscillator surroundings by any means (such as putting hand, objects closer to it) or not changing the load, there's no obvious reason to change the frequency. I'm getting pretty good results...

If you mean repeating is rebuilding the circuit, my answer is this one is not repeatable, you have to hand tune it by changing component values or cutting the tx lines.

If you need a good accuracy in frequency, there's another solution.

Use a good stable frequency source, such as a vco controlled by a PLL (may be a single chip ones with integrated VCOs : http://www.ebay.com.au/itm/35M-4-4GHz-RF-Signal-Generator-Module-Synthesizer-ADF4351-Development-Board-AF/292181749179?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2060353.m1438.l2649)

Set the signal source to the required frequency, and keep it running. Then put an RF amplifier chip at the output. Switch on the amplifier power with the pulse you need to keying in.
Normally amplifier has a 20 - 30 dB attenuation with no supply power condition. So you can have at least, 30 dB amplitude difference between on/off states if you use a 10 dB amp with 20 dB attenuation at power off satate. So it's not a true on/off keying. It's actually ASK. Depending on your requirement, might be an easy option.

                                                                |----switch the power line
                                                                |
                    Stable signal source -----> RF amp -----> (pulse out)

Obviously, the output pulse has amplitude transients (trapezoidal shape due to slow response of amplifier startup), but has a stable frequency. With this method you can't generate very narrow pulses. But I've achieved ~10 ns by switching RF amp chip power line (obviously I didn't use filter caps in the power line).

If you don't have an access to a spectrum analyzer, frequency counter or high speed scope  to measure the frequency, the easiest way is using a PLL chip as it gurantees the set frequency using the digital interface SPI or I2C.

[CopperCone]

yea if i get the signal under the noise floor it might as well be off


I don't have a IC preamp MMIC on hand but its a candidate for being ordered. 10nS is more then enough. I did not think they were so fast.. it does seem like the neatest solution

I always assumed that the power stages of chips don't like being modulated like that.. does it cause faster wear? I thought the quiescent current of those is pretty substantial

I normally don't think of fast repetitive power cycling of complex devices (i.e. more then a transistor), its a design no-no for me, but I may be wrong. And its perfectly fine for test equipment I guess.

I once used PWM control of a LM317 in CL mode for a LED string, and it did not live too long, but the cause of death is really unknown, it could be cheap ebay parts (back when I was really new). I did have a ESD mat however, proper soldering station , etc.

I assume you had the decoupling capacitor right behind your switch? or were you worried about inrush?

[randsl]

Yes, all of the points you've mentioned are valid for high power devices. You can't switch so fast I guess and I haven't tried to do so as well...

It's my bad, forgot to mention which type of RF amp... It has to be a medium power amp (I've used GaAs MMIC) with 10 - 15 dB gain and draws may be ~100 mA.

You can use a LNA chip for this purpose as well. I guess it's easy to use LNA MMICs in this case as they draw much lesser currents (20 - 60 mA typ) and easy to switch with a BJT/logic gate or with a couple of microcontroller pins.

These MMICs can be switched with pulses as narrow as 10 nS and believe me they last years & years with a couple of millions of switching cycles per day.

[CopperCone]

the oscillator in the google books does not appear to work. It drew like 80mA and got hot enough to smoke the flux on it with only RF output during power up (broadband stuff).

I think I will do it the modern way.

[randsl]

yeah, getting oscillators working cab be bit tricky... my experience is, if this kind of oscillator draws more than 40-50 mA, then it's not oscillating. You have to adjust the L,C values and resonator sections.
If you simulate well before construction, may be bit easier to get going... but why bother simulating if if you need only one oscillator...  go in the easy way... 

[AF6WL]

Suggest you keep the oscillator and amplifiers powered up and feed the signal through a double balanced mixer LO to RF port.
Apply the pulse generator to the IF port ( DC coupled )

[G0HZU]

You can make a  fairly stable 1GHz oscillator in about 2 minutes using a handful of parts assuming you have a decent UHF transistor. Eg something like a classic BFR91. Just build it ugly style on a sheet of PCB. No need even to use SMD parts. Two capacitors, two resistors, the transistor and a couple of short bits of wire. I can show you how to efficiently couple to it to minimise loading. I can show you how to do all this if it helps?

This is probably as easy to make as a schoolboy crystal receiver. This method will work OK up to maybe 2GHz especially if you swap over to SMD parts.

Suggest you keep the oscillator and amplifiers powered up and feed the signal through a double balanced mixer LO to RF port.
Apply the pulse generator to the IF port ( DC coupled )

Yes, that would work well. It would need a mixer that can still work reasonable well at 1GHz on the RF and LO ports but it would be a quick and easy way to do it

[CopperCone]

yea I gotta find another RF transistor. Any idea why the circuit I found in the book wont work with the 2n5109 part?

Draws 60mA at the start, then runs away to drawing like 100mA, hot enough to start fuming the flux on it. Won't oscillate, only has some detectable wideband noise when I connect the power wire, looks like a noise burst across 400MHz-1GHz or so..

I used 22nH inductors for the legs (5 turns of the wire diameter I had around the drill bit I used which is like 0.05inches), and I replaced the transmission line with a ~40nH inductor (to achieve the ~120 ohm transmission line impedance), also wound around the drillbit with twice the turns of the other two inductors. (these are air wired). I measured the inductor wires and the winding cores using a micrometer and got the dimensions online using a calculator, then verified at 100KHz with a LCR meter. I also tried replacing the t-line with a 22nH inductor (same as the other two) and it did the same thing.

I think I might be able to just use the noise burst to quantify the SDR, I will try to find an RF transistor if you want to help me model the circuit around it.

I did point to point wiring on a piece of 2x2cm copper sheet. I am measuring the current draw with a HP fluxgate magnetometer around a piece of braid going from the resistor leg to ground. The circuit is powered by -24V. The longest lead distance I have is approximately 2mm, which I think should be fine for 1GHz. , and the ground of the transistor is 1.2 inches of braid (so the probe can hook around it).

I will try to shorten it, perhaps its the problem. The braid is gootwick solder wick.

Also, stupid question, but http://www.pa1m.nl/pa1m/wp-content/uploads/2012/09/2n5109.jpg does the pin out here (same as the datasheet for the 5109) show t from the top with a xray or bottom like the part is upside down? this things driving me nuts, I might be doing something stupid

[randsl]

Well, if the oscillator is not starting most probably the resonator section is not strong/sharp enough or the gain is too low to keep the oscillation...

I think you would agree that if we don't use a proper PCB it's bit difficult to 'simulate' the setup. 

Why don't you go for a cheapie like this one : http://www.ebay.com.au/itm/like/372013834932?chn=ps&dispItem=1 since you need only something to test your SDR.
With the vericap, I've keyed in 50 kHz pulse before (not this one), if you need faster switching, you can remove the vericap and put a fixed cap.

Here are 3 more well documented and more practical schematics than the google books' one :
https://www.infineon.com/dgdl/AN061.pdf?fileId=db3a30431400ef68011426c13f560683
https://www.infineon.com/dgdl/AN002.pdf?fileId=db3a30431400ef6801142771e61c0770
https://www.nxp.com/docs/en/application-note/AN11225.pdf

I've tried the 10 GHz one few yrs ago, and I was able to get it going without much hassle. So I guess the 1 GHz one may also work well.

[randsl]

Regarding 2n5109, yeah agree, datasheets sometimes making things unnecessarily  complex. my guess is it's what you can see from the bottom (pins side). https://www.centralsemi.com/get_document.php?cmp=1&mergetype=pd&mergepath=pd&pdf_id=2n5109.PDF

Also noticed that fT of this transistor is 1.2 GHz, so most probably there might not be enough gain to keep the oscillation. If you have some another transistors in hand, try a one with bit larger fmax and ft.

[CopperCone]

on a sidenote, is 100KHz enough to test a inductor operating at such frequencies (air coil)? I am not familiar with how the parasitics will scale.

I can do up to ~10MHz in this jig I made up, rather precisely, using a true rms thermal meter and an oscilloscope to measure phase shift... unfortunatly the phase shift is my least accurate measurement, only good to 1 degree or so.

Should I manually test these inductors ?

I can potentially test it at a higher frequency, but I am unable to measure voltage at that point, as my fastest oscilloscope is 200MHz, and  my true RMS meter (which can offer more resolution then the oscilloscope) is only rated to 100MHz. I would also need to make a new jig

I wonder if the PLUTOSDR can measure these things, or I suppose I can use my network analyzer thats good for 300MHz.

[CopperCone]

does anyone know how the nature of the 'transient' seen on startup of this oscillator looks like?

Does it 'sweep' really fast, or is it real broad band noise? I would be interesting if it was infact a fast sweep circuit, which swept power across a band quickly, rather then increasing total average noise density.

[RoGeorge]

in order to test the PlutoSDR POI

If by testing POI you mean Probability Of Intercept, then it should be 100%.

AFAIK, AdalmPluto is not a spectrum analyzer, but a SDR. It should stream ALL IQ points to the USB, and the ADC/DAC is running continuously, so once it is tuned and locked to a frequency, it won't miss any sample, not even one. Of course, the PC OS or the PC application might miss some samples if it is not able to cope with the speed of the USB stream coming from the AdalmPluto, but then the PC might be the one with POI less than 100%, not the AdalmPluto itself.

Am I missing something, please? I'm asking because I am a beginner in SDR.

[CopperCone]

typically the specification is quoted that for some keying duration, the probability of intercept will be 100%. if the signal appears for a shorter duration then the minimum time for 100% probability of intercept, the signal must be repetitive and captured by statistical luck, like waveforms per second on an oscilloscope.

[RoGeorge]

That is exactly what I was trying to say, there are no periods of "pause" expected, like in waveform per seconds of an oscilloscope. Once the AD9363 is tuned and locked on a frequency, then the Rx data stream is continuous, without any periods lost.

The minimum pulse that can be viewed is given by the ADC frequency. AD9363 have a maximum of 20MHz Rx bandwidth: http://www.analog.com/media/en/technical-documentation/data-sheets/AD9363.pdf
The 20 MHz limit comes from the maximum clock of the ADC (40 MHz), and not from the analog parts of the AD9363, so any pulse longer than 1/20MHz = 50ns will have a POI of 100% guaranteed. Anything shorter than 50ns will be based on luck only, so it's not relevant to test it with pulses shorter than 50ns.

One more thing to take into account here is that the ADC clock can be changed in the range of 400KHz-40MHz (this is how the 200KHz-20MHz tunable channel bandwidth is achieved). So, if we set the Rx bandwidth to 1 MHz, the minimum pulse with 100% POI guaranteed must be at least 1 microsecond long. Anything shorter than 1 us will be based on luck.

Of course, all that stands only if the whole chain works perfectly, and without any software bugs, so a test for POI totally makes sense, even when we expect 100% POI.


Now, assuming that the whole receiving chain is working properly, and that the above logic is correct, this is how I propose to test the POI: instead of preparing a pulsed carrier signal (a GHz range switch is not trivial), send just the pulse without any carrier. A single pulse of 50ns can be produced much easier with some digital counters or with logic gates.

For a narrowed Rx bandwidth, the DC pulse must be longer in order to have a 100% guaranteed POI.
E.g. for 500KHz Rx bandwidth, the minimum pulse with 100% POI guaranteed is 2 microseconds.

If the single DC pulse is seen on the output with 100% POI, then most probably the whole Rx chain is working correctly, so there will be no need to prepare a modulated pulse with variable length, which will be much harder to DIY than a non-modulated DC pulse.

[G0HZU]

Here's the circuit for the quick and basic 1GHz+ oscillator. This is a very simple negative resistance oscillator. You do need to use a fast transistor like a BFR91 and the 1.8pF (COG ceramic disk) capacitor at the emitter needs to be fitted snugly between the collector ground node and the emitter with very short legs. Eg legs just long enough to solder the cap snugly across collector and emitter. Or you can use a SMD cap here. The 2.2pF cap in the resonator section needs to have short legs as well. If they are left long then the frequency could go below 1GHz.

The 22nH inductor is just a piece of wire about 20mm long. It could be a cropped resistor leg for example. The resistors can be classic 0.25W leaded metal film resistors, eg MFR series or they can be SMD. But it should work fine even if you don't use SMD parts.

Note that the power supply is negative 9V (-9V). Also, this isn't a high performance oscillator in terms of phase noise or stability but it will be fine for what you are doing. To couple a signal from it use a grounded pickup wire about 1cm from the 22nH inductor. Then feed this into a 10dB attenuator. it should then give out something like -15dBm to -10dBm but this depends on how close you place the pickup wire to the 22nH inductor.
There's no need to make a PCB for this. Just build it point to point in dead bug fashion on a small piece of copper PCB ground plane. But try and keep all connections short!

[G0HZU]

Any idea why the circuit I found in the book wont work with the 2n5109 part?

I had a look and that circuit was designed by Randy Rhea who founded Eagleware and produced the Genesys simulator. When I first looked at his circuit on a simulator I couldn't see any negative resistance looking into the 2N5109 transistor using a recent s-parameter model for the 2N5109 at 15V and 50mA. It was the nearest model I could find online.

But then I tried looking in an old version of Genesys because this comes bundled with lots of s-parameter models of older transistors. I did find an s2p file for the 2N5109 and when I used this particular model then the Rhea oscillator circuit does generate a useful amount of negative resistance up at 1GHz. However, I did wonder if the 22nH inductor L2 at the base should be a slightly bigger value as it might begin to affect the circuit if it is just 22nH. But maybe it is OK if it has a reasonable amount of self capacitance. This will make it appear a bit bigger than 22nH at 1GHz.

So it could be that this circuit is marginal, but I can only guess. One 2N5109 model says it works and one says it doesn't.