How does an Airspy HF+ Discovery compare to a "real" radio?

Question

I was looking for a secondary radio for RX only and I thought a SDR may be a better option. I read about upconverters but then I saw Airspy has a SDR with the upconverter onboard, and claims to have "low-loss preselection filter, high linearity LNA, high linearity tunable RF filter, a polyphase harmonic rejection (HR) mixer that rejects up to the 21st harmonic and multi-stage analog and digital IF filtering", which got me curious, since most "cheap" SDRs don't have preselection filters and are easily overloaded (this is also true for cheap HTs like the BAOFENG which, in my experience, are significantly improved by an external filter).

The specs on this radio are (I have highlighted specs which I discusse further down):

• HF coverage between 0.5 kHz .. 31 MHz
• VHF coverage between 60 .. 260 MHz
• Sensitivity: -140.0 dBm (0.02 µV / 50 ohms at 15MHz) MDS Typ. at 500Hz bandwidth in HF
• Sensitivity: -141.5 dBm MDS Typ. at 500 Hz bandwidth in FM Broadcast Band (64 – 118 MHz)
• Sensitivity: -141.0 dBm MDS Typ. at 500 Hz bandwidth in VHF Aviation Band (118 – 260 MHz)
• Linearity: +15 dBm IIP3 on HF at maximum gain
• Linearity: +13 dBm IIP3 on VHF at maximum gain
• Dynamic Range: 110 dB blocking dynamic range (BDR) in HF
• Dynamic Range: 95 dB blocking dynamic range (BDR) in VHF
• Selectivity: 150+ dB combined selectivity (hardware + software)
• Image Rejection: 120 dB (software)
• Filtering: Polyphase Tracking filters for close range interference suppression
• Filtering: New High Performance Pre-selector for HF with 4 filter banks (corners at ~DC, 5, 10, 17 and 31 MHz)
• Filtering: New High Performance Pre-selector for VHF (Air, HAM, Commercial and Military VHF)
• Filtering: New High Performance Pre-selector for the FM band (Japan, US, EU, OIRT)
• Up to 660 kHz alias and image free output for 768 ksps IQ
• 18 bit Embedded Digital Down Converter (DDC)
• 22 bit Resolution at 3 kHz channel using State of the Art DDC (SDR# and SDR-Console)
• +10 dBm Maximum RF input
• 0.5 ppm high precision, low phase noise clock
• 1 PPB frequency adjustment capability
• Very low phase noise PLL (-110 dBc/Hz @ 1kHz separation @ 100 MHz)
• Best Noise reduction of the market using state of the art algorithms

I tried to compare it to my TS-450S which, in page 6 of the manual states a sensitivity of 2uV for SSB, but it doesn't specify the bandwidth. The claims in the Airspy product specify 0.02uV a 500Hz bandwidth. That's a whopping FORTY dB improvement. Which to me sounds "too good to be true".

My kenwood also states a selectivity of "-6dB: More than 2.2khz, -60dB: Less than 4.4khz". The Airspy HF claims "150dB+", which I don't know how to compare.

In all, would this be a good addition to my shack? I believe even if the analog specs are similar to my almost-30-year-old-radio, it has the advantage of IF DSP, which is something that didn't exist back then, and something I can't afford afford at the moment.

Will it be an improvement over what I have now?

Answer 1

I tried to compare it to my TS-450S which, in page 6 of the manual states a sensitivity of 2uV for SSB, but it doesn't specify the bandwidth.

Yes, because you can't adjust the filtering bandwidth, so there's only one sensitivity here.

Of course, the smaller you make your bandwidth, the lower the amount of noise is you receive, and the less signal power you need for detection.

If you can only design a radio once, you need to pick a middle ground between low noise bandwidth (little noise power, but also cuts off parts of the audio) and high audio bandwidth (which makes for better audio at good SNRs, but for worse sensitivity).

An SDR has its filtering in software or programmable logic – so it's flexible, and you can basically program any bandwidth you like. It hence makes sense for the SDR system (SDR hardware + receiver software) to specify the bandwidth used for a sensitivity measurement.

Generally: SDRs themselves don't have "sensitivity", because they don't detect anything by themselves. Only the software part (hence the S) makes the detection, and that's usually only limited by mathematical bounds. These bounds are actually typically of an information-theoretical nature: you might be familiar with Shannon Capacity, for example, which says that given an SNR, it's impossible to get more than a specific data rate across, no matter what you do. Similar things exist for all kinds of estimators – and remember, no matter which receiver you look at, whether it's analog or digital, whether it's a TV receiver or a radio-controlled clock, they're all nothing but estimators of the signal that was transmitted.

I have replied to very many customers (I work tech support, but for a competitor of the Airspy, in a very broad sense) that the SDR they bought doesn't have a sensitivity. If they can put a number on the SNR their detector mathematically needs to work, and a bandwidth it operates on, giving them the Noise figure is totally sufficient for them to calculate the sensitivity of the system they're building from an SDR device and their own software.

In the Airspy sense, a bit of software seems to come with the hardware (otherwise, the software-based claims make no sense).

The claims in the Airspy product specify 0.02uV a 500Hz bandwidth.

Yeah, that's good,

That's a whopping FORTY dB improvement. Which to me sounds "too good to be true".

but not "incredibly good": It's solid engineering, but look at it this way.

Receiver noise is mostly Johnson-Nyquist thermal noise, which is, at room temperature, -174 dBm/Hz. At 500 Hz (that's 27 dBHz), that's -147 dBm.

So, that's a required SNR of 7 dB, roughly. You'd typically need some 3 to 5 dB of SNR for SSB detection, so that leaves us with a noise figure of 2 to 4 dB – that sounds pretty realistic for good equipment.

My kenwood also states a selectivity of "-6dB: More than 2.2khz, -60dB: Less than 4.4khz". The Airspy HF claims "150dB+", which I don't know how to compare.

Me neither.

The Airspy specs are probably for signals that are aliased into the digitized bandwidth, but not into the channel to be received. But that's speculation, and I must say this spec isn't overly helpful.

The Kenwood spec needs to be read like this:

"For adjacent channels, we can guarantee a suppression of at least 6 dB for anything more than 2.2 kHz from the channel of interest, which increases to -60 dB before you even move 4.4 kHz away from the channel center", I think. The spec doesn't seem to be overly well-written either.

In radio measurement equipment, you'd usually just get a measurement graph, where the x-axis is frequency distance from your channel center, and y is the suppression. Would make more sense, if you asked me, and be easy to measure for the manufacturer.

In all, would this be a good addition to my shack? I believe even if the analog specs are similar to my almost-30-year-old-radio,

It's thoroughly possible that your 30-year-old radio has better amplifiers, and better analog filters - but the analog filters don't matter, when you intend to do the filtering in software, where you can build arbitrarily good filters.

Whether or not it's a good addition to your shack is hard to know – but I honestly think you could do worse :) It's a very flexible thing. You can just use it with a lot of different software, i.e. you can use it to receive SSB just as much as satellite data just as much as digital radio mondiale, just as much as airplane transponders, or…

it has the advantage of IF DSP, which is something that didn't exist back then, and something I can't afford afford at the moment.

Well, didn't exist in consumer radios, yes. Military radio surveillance pretty certainly did it back then; I happen to know a device that was produced in the last years of the former GDR which architecturally looks a lot like a superhet SDR. It's orange, very robust, was probably pretty expensive, and weighs about 20 kg.

But really, basically all low-cost low-bandwidth receivers you can buy these days are based on IF sampling and DSP. The <1€ fully integrated AM/FM broadcast receiver IC in your phone (if you have one where you can use the headphone cable as antenna) is one. Your Zigbee receiver very likely is one. If you still have an old GSM phone lying around: chances are not that bad it has an IF and samples that. It's a pretty common technique, exactly because it makes high low-noise receivers well implementable with the modern technology mix of cheap CMOS ADCs, and logic that doesn't even break a sweat processing a couple hundred kilosamples a second.