New Airspy HF+ Pics and Sensitivity Test
The Airspy HF+ is an upcoming product from the Airspy team that is intended to be a high performance HF/VHF receiver at a low price. Its frequency range will be DC to 31 MHz, and 60 to 260 MHz and the bandwidth will be about 660 kHz. So why choose the HF+ over the Airspy R2, Mini or SDRplay which all have larger frequency ranges and bandwidths? It seems the focus of the HF+ is to be an extremely high dynamic range receiver. This means that strong signals should almost never overload the receiver making it very good for DXing weak signals (listening to weak signals from very far away). On other receivers once you turn the gain up strong signals can block reception of the weaker ones.
Recently we saw the release of some of the first 3D renderings of the product. Now finally we have a photo of the actual PCB which is shown below. The RF sensitive innards are hidden away within a shielding can, but we know from the product page that inside are the switches, filters, tuner, ADC and 18-bit DDC.
Also, over on Twitter, @lambdaprog, lead creator of SDR# and of the Airspy HF+ has uploaded some sensitivity tests. It seems that sensitivity will be at least -136 dBM at 20 meters, as a -136 dBm signal still comes in with 21 dB of SNR. Similar sensitivity results are obtained on the FM Band.
Sensitivity tests of the prod revision of the Airspy HF+. It looks like -136 dBm still gives a good readable signal on 20m. cc @ITeadstudio pic.twitter.com/n4KaKi0eAB
— prog’ (@lambdaprog) July 17, 2017
More sensitivity tests of the Airspy HF+… FM DXers will like this one 🙂 pic.twitter.com/2H2CZJSgt9
— prog’ (@lambdaprog) July 17, 2017
More tests of the Airspy HF+ production. Here's a plot of the MDS (500 Hz) in VHF. pic.twitter.com/ilte8Hz7sQ
— prog' (@lambdaprog) July 20, 2017
More tests of the Airspy HF+ production. Here's a plot of the MDS (500 Hz) in HF. pic.twitter.com/LkD7mCIgmB
— prog' (@lambdaprog) July 20, 2017
The Airspy team have sent us a sample unit from an early manufacturing test and we hope to have a full review available a few weeks after we receive it.
http://airspy.com/downloads/changelog.txt
Dear All
Do you think that this new sdr receiver will be in line in terms of performance and sensitivity with at least the Elad sdr fdm 2?
Are the sdr perseus still the best around? Even after 9 years?
Thanks again
CLAUDIO
Current, not the newest, NFM demodulators in an all-in-one chip, have sensitivity of -160dBM. Also GPS devices run at -160dBM – see older Holux for example. So I wonder why such low spec devices are coming out? For HF it is okay in QRM rich regions, but any receiver for VHF+ should be much better than -136dBM.
I was wondering the same! Why don’t we have more better radios than we have today?!
The answer, alas, is that we have more idiots using radios without understanding the laws of physics than radio engineers who care about explaining what they are doing.
For example, the average appliance operator or radio blog poster can’t tell if -160 dBm is greater or lower than the natural background noise at the typical NFM channel bandwidth at a given frequency.
PS: My post is only funny for people who know something about RF and can make fun of the credulity of people who don’t.
how can you go to -160dBm when the mathematics say that with a perfect receiver with 0dB of NF in a bandwith of 500Hz you get -174 + 27 = -147dBm !!!
to get -160dBm with a perfect receiver you have only 25Hz of bandwith. good luck with you NFM signal 🙂
not counting that the noise in HF/VHF is much higher -174dBm/Hz.
see this ITU report:
https://www.itu.int/dms_pubrec/itu-r/rec/p/R-REC-P.372-13-201609-I!!PDF-E.pdf
Lugi, do not be upset with the “digital” guys. They do understand very well the 0 and 1. The analog guys, they understand the rest, above, bellow and in between 🙂 Explaining them 10 x log(BW) is a rocket science 🙂
R-REC-P.372-13-201609-I!!PDF-E.pdf does not account for the unnecessary increase in man mad noise due to poor designs or quality of parts e.g. in cheap circuits in LED-Lamps.
For those that are unable to open the link
https://www.itu.int/dms_pubrec/itu-r/rec/p/R-REC-P.372-13-201609-I!!PDF-E.pdf
type R-REC-P.372-13-201609-I!!PDF-E.pdf into the ITU search to download it.
The link has to be written in the proper way
Also you run into the phase noise of the tx, how clean does an rx have to be?
First of all, you will run into your receiver LO or reference oscillator phase noise, problem known as reciprocal mixing. Then we can talk about others TX phase noise quality. To be honest, I do not see the high quality reference oscillator on the presented unit. I really do not know why we are running the double oven reference oscillators in our devices where they are just consuming a lot of current. Obviously this can be done some other way…. 🙂
-136dBm “okay on HF for a QRM rich region”?
I would die for such a low noise floor on HF !
-136dBm is 0,0355uV making a noise floor of 9dB below S -“zero” 🙂
Moreover, I would be deeply worried about our galaxy, which apparantly stopped making noise 🙂
In a 2.4KHz filter, that sensitivity should work out just fine at 10m and above, some HAM rigs only have a -125 sensitivity in a voice bandwidth on hf and they get along great. In a cw bandwidth that thing should be even more sensitive.
Can it run on HDSDR ??????????????????? This thing is got my interest .
why would you use HDSDR?
Anonymous , I like it better than SDR Sharp .
When I manage to buy one of such radio, I will attempt to create an extio dll , as I did in the past for Airspy.
The source code for the access library is already on github, so it should not require much time.
Andrea
Send us a short mail with your details. We’ll get you a complementary unit dispatched when the first batch is ready. Your time and effort are very much appreciated.
Cheers
Ian
Adam,
Where do you see a leakage problem in the pictures? I don not see any connectors,leads or connections.
A lack of shielding with weak signals would cause noise ingress on measurements, and lead to worse SNR than you see here.
Looking forward to definitive figures on this little DX machine!
No doubts, the 18 ADC will bring another dimension to the SDR but the presented figures are too nice to be true 🙂
If you have S/n 21db with the -136dBm input signal it does mean you have the MDS -157dBm. we can all agree that this far to futuristic. On the other hand, it this is true, the same receiver will saturate quickly with a descent antenna on the HF. As a rule of the thumb the 18 bit can give as dynamic range of 110dB roughly, so with the MDS of -157dBm the same receiver will run in troubles with a signal of -47dBm. If you attach descent HF antenna to the spectrum analyzer you may notice a much stronger signals on the HF during the evening hours in Europe on 7.3 mhz, 5 MHz etc…. I can be wrong, but the problem in the presented photo is the leakage from the signal generator. It is not easy to preserve such a good signal shielding with such a small signal levels. I do have R&S SME03 signal generator where the signal can be attenuated up to -144dBm but the leakage is present already at -130dBm. I am surprised that a guy running such a nice Agilent device is not aware of that problem.
If you look at the bandwidth it is 46.512kHz, so I would assume that they are sampling at 744192 SPS and decimating by 16 for an extra 4 bits of dynamic range (so 22-bits). 22-bits would be a dynamic range ~134dB.
Decimation does NOT reduce KTB, it reduces the quantisation noise of the ADC, not the noise from the source. The indicated receiver bandwidth is 2.4 kHz, and if the signal level and SNR were correct, then the noise floor in this bandwidth would be -136 – 21 – 10.Log(2400) = -190.8 dBm/Hz. This is simply IMPOSSIBLE given that KTB is -173.8 dBm/Hz at 300 kelvin.
Also a true sensitivity test would use a modulated signal of defined modulation parameters.
Don’t forget that this device contains an LNA, Tuner etc. The noise figure will be the noise figure and no amount of decimation will help this.
I think Adam is right, this is simply signal generator leakage giving a misleading result.
I’m surprised how so many experts (combined) could not interpret what they see. What is shown is the carrier over averaged noise ratio. It is independent from the filter bandwidth used.
To convert that into actual SNR you need the number of FFT bins and the sample rate, and probably the window function as well.
Sure, this looks better than most other SDR’s, but these aren’t 500 Hz or even 2.4 kHz measurements.
I am not sure how SDR# measures SNR, but what you suggest seems entirely logical. But it also suggests that the claim of a 21 dB SNR with a -136 dBm signal is misleading (to put it mildly) and that this is not a meaningful sensitivity measurement unless (as you say) the FFT bin size is stipulated. I also reiterate the point that for sensitivity to be meaningful, it needs to be defined for a modulated signal with a given level of modulation parameters. Anyone can pull a CW tone out of the noise by increasing the FFT size and zooming in on the FFT until you hit the numerical dynamic range of the FFT engine itself. That on its own says nothing about the receiver performance.
“Decimation does NOT reduce KTB” – yes, there is a hard floor. I agree with “-136 – 21” but not “– 10.Log(2400)”. The source bandwidth could be 50 Hz or less, why does the source bandwidth have to be 2400 Hz ?
You are right, unless the bandwidth is clearly stated, the SNR for an CW tone is meaningless. I do not know how SDR# calculates SNR, but as the demodulator filter is set to 2.4 kHz, it seemed logical that this would be the bandwidth that the noise power is calculated in. If this is not the case, for all we know, it might even be 1 Hz, in which case the NF would be 16.8 dB, which would actually be more than good enough for HF, but not an indication that the receiver would be capable of receiving -136 dBm signals either in the 20m or VHF bands
I was looking at the numbers and thinking that something was wrong, I was thinking about Johnson-Nyquist noise:
Johnson-Nyquist noise @ 300K for 46512 Hz is ~ -127 dBm
Johnson-Nyquist noise @ 300K for 744192 Hz is ~ -115 dBm
Johnson-Nyquist noise @ 300K for 36000000 Hz is ~ -98 dBm
But if I assume that they are using a pair of 13-bit ~36 MSPS Sigma Delta ADC’s and decimating in the digital domain …
Then decimating 36MSPS by 48 is ~750000 SPS @ 18.58 bits
Then decimating 36MSPS by 768 is ~46875 SPS @ 22.58 bits
So the Johnson-Nyquist noise floor in the analog domain is at -98 dBm @ 36MHz.
Which due to processing in the digital domain is lowered by ~57.67dB to -155.67dBm (@22.58 bit).
And at full bandwidth of around ~750000 SPS that would be lowered by ~33.59dB to -132dBm (@18.58 bit).
With the above assumptions in mind when I looked at the graphs, I see that my thinking may have been flawed.
I’m still making a lot of assumptions, I don’t know, I’m reading between the lines. I’m ignoring ENOB, because I don’t know what it is.
I should clarify, they have not provided the ENOB, so I can’t factor that in.
Then there is all the FFT processing gain to factor also.
So you would gain 6.02dB every time you quadrupled the number of bins in the FFT ?
FFT gain = 10*log(M/2)
Sauce: https://www.designnews.com/aerospace/where-does-fft-process-gain-come/100022666833951
Which is 6.02dB every time you quadrupled the number of bins in the FFT.
Exactly right. reducing the FFT bin size makes a signal appear more visible by reducing the integrated noise power within each FFT bin, but it does not improve the fidelity of the signal in a given demodulator bandwidth. This is one of the tricks used to make decimation appear more effective than it actually is. For a given FFT size (say 32k), decimating by a factor of 4 will reduce the bin size by a factor of 4 and reduce the apparent noise power on the display by 6 dB. This is certainly not increasing the SNR by 6 dB
That Microchip+ATMEL, 100-pin QFP, is that a SAM ARM MCU ?