How big are the FFTs? What devices are they implemented in?
I'm curious for SDR use. The Hermes project has what they call Direct Fourier Conversion DFC where they interface 120MSPS 16-bit ADCs via an FPGA to a PCI-e port and inject data almost straight into a GPU, that does 1-million point FFTs on it. They can then extract up to 80 channels from the frequency data and iFFT them back to time domain to demodulate the audio. Useful to have that many channels in a WebSDR.
But I'm interested to know how much FPGA resource (or other, DSP chips maybe?) it would take to do the same without a honking computer_ & GPU... How would you tackle this problem in hardware?
On Saturday, 8 January 2022 at 15:46:10 UTC-8, Clifford Heath wrote: ...
A mixture of 128, 256 and 512 point complex FFTs. Each axis has different number of samples.
NXP, TI and Infineon have DSPs with FFT accelerators and other special purpose accelerators for radar processing. ...
The FPGA solution we did first required a $30,000 FPGA from Xilinx, then we used a $3,000 Altera/Intel Stratix 10. The special purpose DSPs with accelerators take much less power (and are orders of magnitude cheaper). Even a "honking computer and GPU" would be hard pushed to provide the performance needed.
John Larkin <jlarkin@highland_atwork_technology.com> wrote in news: snipped-for-privacy@4ax.com:
Ever herd of diversity. Far better to array 8 antennas and then let some hardware pick the best signal. Not the strongest signal, the best signal. That was analog NTSC broadcast days.
If the guy is in the US he is talking about digital signals on the old UHF TV band, now referred to in the US as Broadcast HDTV.
In getting the signal most HDTV recievers usually either get ALL of the signal, or simply mute that channel when not able to gather the perfect packet data. There is no snow and there is no ghosting.
Anyway... diversity receivers never added channels because strength was not the issue. It was about quality. So around a race track they would have diversity recievers and antenna arrays to get the best audio signals from the drivers as they circled the track. The hardware did not add channels it switched to the best channel. It is similar to what cell service does as you leave one cell and enter another.
That's a common misconception based AFAIK on a statement by Kraus concerning wire antennas specifically. There's a lot of life after wire antennas. ;)
If half the power is reflected, the antenna has a VSWR (on the business end) of no better than
1 + 0.707 VSWR = ---------- = 5.8 : 1. 1 - 0.707
Your average HP waveguide-to-coax transition is a 1/4 wave antenna stuck through the H face, 1/4 wavelength from a waveguide short. Its efficiency is way over 90%.
In free space, there's more than one mode to worry about, of course, so things generally aren't that good, but you can make adiabatic waveguide horns that have VSWRs near 1:1. Here's a small one that's specified at
1:1.15:
formatting link
By reciprocity, the coupling is the same in both directions, considering only the antenna mode. Of course it'll reflect a lot more if you come in with an orthogonal mode--ideally 100%.
As I pointed out way upthread, a single pair of wires can interrogate only one optical mode, corresponding to an etendue of lambda**2/2.
That's a super useful fact when considering how various detection and transmission schemes scale with wavelength. If you want the RF to come out of one pair of wires, the projected solid angle Omega' subtended by the antenna pattern has to obey
Omega' <= lambda**2 / (2 * collection area).
By building sufficiently bad antennas, you might think you can interrogate a wider Omega', but that's illusory--it just means that the matched mode is more complicated, so that it doesn't quite match the incoming field anymore. (You can derive the etendue limit from thermodynamics, so arguing otherwise eventually amounts to asserting that one can make a perpetual motion machine.)
Incoherent combination techniques, e.g. a photodiode or (at RF) N antenna/receiver sets with their outputs summed, can interrogate many modes at once, at the price of squaring the required dynamic range.
Nice, Jan. Finding a hotspot can be necessary. As someone who lives a few km away from the actual objects shown in your wallpaper (source of many ATSC signals), even my TV has trouble now & then, with a simple antenna at the window (...is not on a high floor) cheers, RS
On a sunny day (Sat, 15 Jan 2022 15:41:51 -0800 (PST)) it happened Rich S snipped-for-privacy@gmail.com wrote in snipped-for-privacy@googlegroups.com:
Yes, terrestrial, but I am so glad I have a satellite dish... Reception here from all over the world, Germany, UK, Russia, China, even Cuba is 100% ! Only problem I ever had was when a big thunder cloud passed between satellite and dish. And now we have interference from the new local radar station.. but most channels are transmitted both in normal and HD resolution and somehow HD works anyways. There is also the issue of that national TV networks only show you what they want you to know / think, while the world via satellite shows all sides of the political / economic spectrum. I have a choice of more than 900 free to view satellite channels with my movable dish.
As to ATSC it was developed after DVB-T in Europe to deal with multi path reflections IIRC:
formatting link
We went completely digital a bit earlier in 2006 with DVB-T:
formatting link
days we have DVB-T2 here.
Satellite uses DVB-S and DVB-S2 S2 is close to the Shannon limit:
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.