broadband noise source

anyone have an idea how to generate relatively high levels of broadband noise? Looking for at least -60 dBm/Hz from .4 to 2.5 GHz continuous bandwidth.

Tried a noise diode...way too low an output power for my purpose. I am open to out of the box solutions here.

I calculate at least +33 dBm output power from the noise density. Some overhead for filter with finite slope. I guess the output amplifier needs at least 5 dB more power overhead to handle the crest factor. Ends up with about +40 dBm (10W).

What against using noise diode, bandfilter, pre and power amplifier?

oh I just see my problem... I was doing 10 log 2500, not 10 log 2500,000,000. Yeah that is too high of a noise density.

My intent would be to end up with perhaps 10 miliwatts of integrated power from .4 to 2.5 GHz. So I would need a noised density of -83 dBm/Hz in the noise source. Thanks

57 dB of amplification..that is a little tough on one pcb, but may be possible with a metal shield covering half of the amplifier chain. That could work

was wondering if there was a noise source that starts much higher, like a neon bulb, or some other old school device.

I think only semiconductors will generate noise up to GHz range.

what about an LED driving a microwave photodetector? or maybe an LED driving an optical phase detector with some time delay in one arm?

So far the only reasonable option is to amplify the noise from an avalanche noise diode. Commercial models can be seen at "www.noisecom.com", for instance.
Gas-discharge tubes generate typically ENRs only around 15 dB while avalanche diodes do ~30 dB.
You can try a buzzer gap-discharge noise source but its output is not "white-wideband".It can be tuned to a good output by a coaxial cavity with the discharge gap where the open end is.

I have tried to spatially combine dipole noise radiators with avalanche diodes. A 3x3 matrix generated approx. 37 dB ENR on the axis, so if you need to go without amplifiers in a free space, this way can be useful.

Try this schematic which works for me up to 2.5 GHz:

Yes, I use the same e-b diode in avalanche mode, only into 50 Ohms. If 330 Ohms is used, the output noise spectrum is not flat and at some spots it can peak to 35 dB ENR. The DC current is critical, my best results are for 5..10 mA. Almost any RF transistor e-b diode can be used, it breaks at 6...9 V.

Thanks, especially for the suggestion of a transistor junction...the noise diodes were going to be fairly expensive!