Is a low power broadband transmitter possible? say 0-2ghz?

[rwgast_lowlevellogicdesin]

I want to generate noise on the whole spectrum up to 2ghz, the point being is to give my SDR a flat DB response at any frequency, this way I can test the response of preamps, filters, etc.... Im only trying to get enough power to where I set it right next to my SDR and can analyze the whole spectrum without spikes in the FM and UHF bands. I also dont need to modulate ot do anything special I just need noise, kind of like an oscillator based jammer but with a huge range and very little power to keep it legal.

Im sure this is possible but is there a practical and cheap way to do it?

[jt]

Taking advantage of noise generated by a diode in avalanche breakdown is commonly used for broadband white noise generation.  You might find the following .pdf of interest.

http://www.rf-microwave.com/datasheets/2581_generic_KIT-NS_01.pdf

It will be a lot easier if you can connect your noise source to your system-under-test via a coax cable rather than an over-the-air broadcast.  That way you don't need to worry about broad-band antenna tuning and you will be less sensitive to local noise sources.  And because you won't be broadcasting, you also won't need to worry about the FCC knocking on your door. 

[SteveLy]

Digital is easiest on paper. Uncorrelated random values give you white noise up to the Nyquist frequency = 1/(2*sampling interval). The probability distribution of the random values does not matter, only that they're uncorrelated/independent.

[T3sl4co1l]

Digital is easiest on paper. Uncorrelated random values give you white noise up to the Nyquist frequency = 1/(2*sampling interval). The probability distribution of the random values does not matter, only that they're uncorrelated/independent.

Yeah, but you'd need a LFSR (or similar method) running at 10GHz+*, to achieve the stated bandwidth.  (On the upside, an LVDS or ECL output would be quite simple to couple into a 50 ohm transmission line, and deliver a modest number of dBm's.)

*As I recall, it's not flat up near the sample rate, due to time correlation (it's only random every t = n*Ts, not T ~= 2*pi/BW).  The result is a sinc(f) weighting, which flattens out at lower frequencies (say < Fs/5 or 10) to a quite reasonable white noise signal.


A smaller bandwidth is probably a smart idea, anyway.  A quite reasonable bit of noise can be made easily (say, 100s of MHz), which could then be modulated up to higher frequencies using a diode balanced mixer and a variable frequency generator (as local oscillator).  This, hopefully, will give a flat band with a small peak in the middle (due to LO bleed-through).

Any amplification following the mixer will be rather expensive, because 2GHz (with a lower limit of ~DC??) isn't the cheapest thing.  Not with calibrated gain anyway.  (MMICs are cheap, sure, but they aren't very flat either.)

Tim

[tggzzz]

Answer: yes.

Next question: is it legal?

Next question: if illegal, why is it illegal?

Next question: if illegal, what are the penalties (legal and otherwise)?


(Of course, if you don't transmit there are no such issues to consider)

[German_EE]

http://www.rf-microwave.com/datasheets/2581_generic_KIT-NS_01.pdf

plus

http://www.minicircuits.com/pdfs/MAR-2SM+.pdf

[rwgast_lowlevellogicdesin]

Ok thanks, this really helps I knew you could use diodes to generate noise but I didnt realize how broadband it was. I have no problem directly plugging in to the DUT then the radio, and ditching the airwaves for test purposes.

what is an LFSR?

[helius]

https://en.wikipedia.org/wiki/Linear_feedback_shift_register

It's a shift register with feedback. The simplest kind of PRNG to build in hardware.

[German_EE]

A LFSR is good for generating noise at audio frequencies but for RF you would need a seriously high clock frequency, hence the use of noise diodes. Another source of noise is a neon bulb.