Comparing a HackRF Clone against the Original

Over on the Great Scott Gadgets blog Michael Ossmann, the lead creator of the original HackRF has put out a post comparing his original HackRF with one of the many clones on the market. The HackRF is a low cost wideband transmit capable SDR that was released via Kickstarter crowd funding back in 2014. Even up until today it is one of the most popular SDRs for radio experimenters due to it's versatility, open source nature, and low cost.

Within the past few years Chinese clones of most SDRs including the HackRF have appeared on the market often at substantially reduced pricing. As the HackRF is fully open source hardware, copies are legally allowed, however buying a clone does not support the original developer and can put strain on their support services. The general consensus amongst clone purchasers is that they work fine, but when there are problems you take the risk of not being able to expect any sort of support or warranty from the the cloner. Also while the clones work fine, up until now we have not yet seen any performance comparisons yet.

In his post Michael Ossmann tests a clone which is even advertised to have improved upon the original design. Michaels post goes into more detail, but long story short, the clone has clear transmit performance issues above 1 GHz, and at the worst point produces 22 dB (150x) less power out compared to the original. In terms of receive performance the clone performs even worse, showing very poor sensitivity when compared to the original. Michael notes that this clone would not have passed the QC procedure used for the original.

We believe that the original HackRF has created significant value to the RF community through software, tutorials and their hardware. Over the years countless projects and research/conference papers have been enabled by the HackRF. So even regardless of potential performance and warranty issues we think it is ethical to support the original creators if your budget allows it.

HackRF Receive Performance Test. Above 5 GHz the test signal was below the noise floor.

42 comments

      • jim

        thank you for prompt reply clifford. is this still the best bang per buck or is there better sub $200 rx/tx sdrs on the scene?

        • Clifford Heath

          There are many different SDRs on the market, fewer with Tx as well as Rx, but you will need to understand the differences and decide what matters for your application. The HackRF isn’t excellent at anything. It excels only in the very broad range of things at which it is mediocre. It’s a good buy if you want a little bit of everything, but nothing in particular.

      • Andy

        I placed an order, and Aliexpress auto-canceled it a couple weeks later because
        the vendor failed to ship it. I tried to ask the vendor about it, and got no reply. Do you have another source?

        I’m in the United States.

        Urgently need a half-dozen of these now, and more later, to make drone detection
        and tracking equipment to protect Ukrainians from Russian grenades falling from the sky in contested territories.

        • Clifford Heath

          This is not the right place to ask about suppliers. AliExpress is an umbrella for tens of thousands of businesses – which one did you order from? Because the manufacturer shipping my design has stock and has been shipping.

          • Andy

            I ordered on July 2nd from the supplier you linked to in your message above:
            https://www.aliexpress.com/store/group/HackRF-One/5892199_10000002152810.html

            Aliexpress auto-canceled on July 13th.

            I want to get the boards with your design due to the improvements discussed on this forum page. Chose that particular supplier on Aliexpress because it’s the one you mentioned. Posted message here after supplier was non-responsive because I didn’t know where else to ask if that’s even still the correct supplier (your message was more than 4 months ago).

            Have other people been buying from the supplier with no problem for the past month?

            Thanks.

            • Clifford Heath

              Yes, others have. I even contacted the supplier to confirm, after one contact said their unit hadn’t shipped (after 2 days). Supplier told me they had stock and were shipping. But *please* stop talking about it here. This is not the place for it and I won’t answer you again here. Contact me privately if you want to. I’m easy to find.

  1. cya

    Is there a hi res jpg with correctly mounted parts? I originally sent email to Carl with this “gsg news” and he published it here. I told him that was just few days before I received my HackRF clone and I was bit disappointed why I bought it.
    I tested new board, it seems that is working fine but I don’t have tools to check actual operation. So opening shield and comparing it with old design wouldn’t help.
    So, again is there schematic or hi resolution images of new design?

    • Clifford Heath

      I still have no confirmation that I correctly guessed the manufacturing fault – we only know of two faulty units out of thousands made – but my working hypothesis is that the baluns around the RFC5072 mixer may have been fitted rotated 180 degrees. If you snap off the RF shield can, you should see the baluns mounted as in this picture.

      The design files have always been published on my github.

      • cya

        Hi Clifford, many thanks for image and link.
        Please correct me if I am wrong, I guess that we can distinguish them by silk, your version has “CKT 08 mar 2021”
        Now confirmed, my board is same as on your picture, and my friend brought original design clone HackRF, we compared them briefly 1 pc with 2 sdr# and seems to me that noise floor is lower and when amp is activated it has more gain on your version so this is big advantage as far as I am concerned.

    • alan tracey

      i got a Chinese clone from ebay back in may 2020 i can send you some high res pics just let me know and you can email my business alan dot tracey at reconosint dot net.
      I dont know if its the exact same model but id like to know how good mine is compared to others or the original, I was never able to get anything other then fm public radio you know the 97fm – 108fm to reliably work on it. I bought it to use with projects like IMSI catcher script and Im a DragonOS user as well, Some other stuff works but I also have a LimeSDR and ive had a couple RTL-SDR’s the NuEllect that type I believe and those had better results. I seen Aaron come along ways and highly support hi Project on DragonOS by the way.

      • Clifford Heath

        Alan, I tried to contact you, but your reconosint domain has no MX record. Do you have an email address that works?

  2. Dunno

    To be fair to GSG – the USRPs I use have twice suffered from a blown RF front end. And its SO expensive to fix too. Amost two of your full-priced Hack RFs to resolve. The first time was completely my fault. I knew immediately when I keyed up what I had done.

    The second time I was like WTF. I did nothing wrong. Must have zapped it connecting FM-Bandstop with open pins to the device. Can’t think of anything else.

    And to be fair to Ettus / NI, many of their cards can withstand 0 dBm, some up to +10 I think… -5 dBm seems a bit harsh.

    Just my 2c.

  3. John

    It’s difficult to understand the motivation to clone an existing design knowing that it will undercut the work of the original creator and create support headaches for them. Getting two cheap Chinese units doesn’t seem sufficient motivation.

    Perhaps the original intent was simply to damage GSG.

    • Bob

      And low quality assurance (QA), boards either ship totally untested or with minimal testing to keep the price as low as possible. Fitting parts the wrong way around as stated by Clifford Heath is a perfect example of bad QA.

    • willmore

      Read the comments before commenting unless you want to sound very foolish. This design is from 2014 and is done as open hardware. That means that people are free to reproduce or modify the design. Since the design is so old, some of the parts included are no longer available or the best performing for the task. So this ‘clone’ is a redesign using better and more modern (and still availabe) parts.

      • Clifford Heath

        Quite right willmore. I had hoped that Michael would adopt my changes, but he still doesn’t understand the need. He seemed amenable to a change, so I designed it and offered a Github PR, but he still rejects it. So I contacted a clone manufacturer who had done other work for me, and they asked for further revisions to avoid unobtainable parts, which I did and again offered to GSG. You can read the whole history here . Full credit for getting the HackRF to work, but he’s not an RF designer, and no serious radio person takes the HackRF very seriously. But now GSG has their fingers firmly in their ears, and I am blocked from every forum they control – even email.

        The only people who are acting to hurt GSG here are GSG staff themselves. And given the way they continue to abuse customers with this faulty design that is not of merchantable quality, perhaps some self-flagellation is warranted?

          • snn47

            I calculated the received power for a 17 kW output power (sounds like a pulse compressed Solid-State PSR), a 0 dBi receive antenna facing the PSR dish at the same height for 34 dB antenna gain (not clear if the gain is referenced to dBi or dBd) a signal of around 8 dBm in 2km distance would be received.
            But the actual signal would have to be much lower, because in order to avoid ground clutter any PSR antenna will use an uptilt of the antenna pattern to reduce the antenna gain by several dBi to eliminate ground clutter. If the antenna is well below the horizontal plane of the PSR dish the antenna is lower again.
            So are you certain it was the referenced S-Band PSR that damaged the circuit, since S-Band PSR transmit between 1 µs up to ~120 µs up to ~1200 times a second? There are normally a lot of other strong transmitter in operation at and around an airport, other pulsed systems can be L-, C- or X-Band PSR, SSR, MLAT.

            • Clifford Heath

              You are probably right. The radar (40MW EIRP) was only a surmisal. The actual cause could be almost anything – and that’s the point. It doesn’t take much to exceed 15dBm, and very many people actually damage their HackRF in situations where it should survive.

              • snn47

                Radar is always the system that everyone thinks/rembembers of due to the antenna-form/-size or its rotation. Other non rotating systems that also produce high Peak Power and EIRP are less obvious/remembered like
                – BC/TV transmitter (100 kW transmitter power or more without antenna gain ) or
                – OFDM based cell phone systems, which can combine high gain (>26 dBi and/or beam stearing) antenna and have a with Peak to Average Power Ratios (PAPR) of up to +17 dB.
                Was it really a fast death of the components, that were caused by short pulses from a High Intensity Radition Field or is it not more likely
                that it was a continious exposure to/need to process all strong signals over the large receiver bandwith, that caused the damage in the medium term time, when no filter are used to reduce the combined energy of all frequencies. 0 to 6 GHz includes a lot of strong signals, (non-)pulsed and/or large PAPR?

        • elmood

          I read the GitHub thread and thought that Clifford was quite reasonable in the detail of his fix and its justification. My designs have ESD protection on even the most mundane low frequency ports so I’m a bit mystified why this wouldn’t be the #1 priority for an RF design like this. I would be embarrassed if anything went out with such a shortcoming.

          I’ve ordered a clone unit (it’s in the mail) and feel even more okay with it after seeing the attitude of the GSG folks. I’ll probably add the protection mod. However before I waste a lot of time, I’d like to characterize the unit so I can know the overall performance and also compare later in case any damage is suspected.

          Can anyone recommend some basic tests to perform to verify the sensitivity, noise level, output power, etc.? Are there any benchmarks out there to compare against? I have a spectrum analyzer with a tracking generator so I can take measurements and make tones at various levels and frequencies.

          • Clifford Heath

            Thanks for your support. Note that ESD protection isn’t the main issue (though the TVS is badly chosen at 15V) – it just can’t handle excessive RF power.

            I can’t help with an existing test plan. If your spectrum analyser has a remote interface you could script a test sequence using hackrf_transfer or a similar console-mode tool for spectrum analysis. I’d be happy help design an automated test – please contact me by email.

      • John

        Any share of the revenue being given back to the original designer? If so great! If not, why not?

        I fully understand open source, but just because someone leaves the window open shouldn’t mean someone else is free to loot the building.

  4. Gérald

    The current design is from 2014, almost 8 years old already! I hope that a refresh will come soon. But chips shortage could be a problem nowdays…

    • Clifford Heath

      The existing design is compromised in so many ways that only a completely new design can provide substantial improvement. I’m sketching one out at present, but don’t hold your breath waiting for it. Suffice to say that there is now no excuse for an underpowered CPU or an 8bit digital path these days

  5. snn47

    @ Clifford Can you explain why there is this sudden step-like in-/de-crease in output power and sensitivity of >10 dBm between 2.1/2.7 GHz?

    • Clifford Heath

      Between 2.1/2.7 GHz the RF path goes straight to the WiFi chip, bypassing the RFFC5072 frequency converter. It still has all 3 protection devices in circuit, and in this test that appears to result in 2dB loss. Below 2.1GHz it up-converts, and above 2.7 it down-converts (on receive, opposite on Tx). The apparent 12dB of additional loss in both directions and across all frequencies can only be explained by the frequency converter (an original GSG design) being broken. The loss is more marked above 5GHz, which is possibly explained by the protection devices, but is compensated by the additional power and gain of the newer MMIC amplifiers that replace the (now unavailable) MGA devices. Except you can’t see that in this unfair test.

      • snn47

        Thank you, for explaining the difference between the receive range of the WiFi Chip and the up-/down-converter.
        My critique to the writeup is that it does not describe the test setup and how the results were obtained e.g.
        – type of Spectrum Analyzer, which settings like Resolution Bandwidth
        – type of cable and length between the devices
        – and if the settings and connections were identical during SDR tests.

        If a single cable and/or connector is damaged or was damaged during dis-/connection the following results are be flawed and therefore are not comparable.

        While a continious high power across the operational frequency range is desireable, it is also important to know the bandwidth, noise and spurious emissions of the generated signal differes. between the two SDR.
        For the receive tests an additional question would be the type of signal generator that was used, and could the generator provide -5 dBm from >0 to 6 GHz during the sweep.

        While we have thanks to affordable SDR today the option to easily generate and receive signal over a wide bandwith and have an idea but not a valid measurement of what is going on. Unfortunately many tend to take any displayed value as fact without questioning the seen values and curves can be true or just wishfull thinking. There are reasons why comercial wideband test equipment and cables are expensive, why they need to be constantly calibrated and when necessary replaced, and why any test setup needs to be tested for function and calibrated before measurements are taken.

  6. Fazal Majid

    Since the design is open-source, the clone vendor must have substituted inferior parts in the bill of materials in the not-so-grand old tradition of chabuduo. It would be interesting to compare the two and see which corners were cut.

      • snn47

        You accept the tests and conclusions from Michael as valid, but do not question, e.g. if there is also a methodical error since the test setup used for measuring of both SDR is not described, please refer to my post from today in response to the additional information Clifford provided.
        The figures between ~2.1 and 2.7 GHz, that do not use the additional up-/down-conversion circuits, differ only by a few dB, which supports Clifford conclusion that the components involved with up-/down-conversion may be defective.
        If the performance is so inferior why did Michael not ask for a replacement in order to have a functioning clone?

    • Clifford Heath

      I am the designer of the “clone” version that Michael tested. I can categorically state that he has a broken unit. The RFFC5072 frequency converter in his unit is not functioning correctly, probably because the baluns have been fitted incorrectly (rotated 180). My design actually outperforms the original on a number of metrics, as well as not being super fragile like the original, and the manufacturer has shipped several thousand units to very satisfied customers. Of course, that has caused a problem for GSG, which no doubt motivated the current response. In any case, the malfunctioning converter is of an identical design to the original (I made no changes there), so assuming Michael has not damaged it, it must have a manufacturing defect. I utterly reject his implication that it has a design defect. He simply doesn’t know how to interpret his own measurements.

    • Clifford Heath

      On the contrary, the vendor has no engineer on staff who knows how to make substitutions, nor to test the effect (unfortunately this also means they have minimal QA, but I haven’t been able to help them improve that). The several units I have (my entire compensation for my work) have all genuine parts. Before the chip crisis, the entire cost of production, including all genuine parts, was around $US50.

  7. David

    $339.95 for a real HackRF One versus $105.00 for a clone is quite a stretch cash-wise; too much to simply dismiss based on the testing of a single unnamed clone by someone who is not independent. If Michael Ossmann wants to make a convincing enough argument for most people to shun the far less expensive clones, he should commission an independent tester to compare a statistically significant number of different clones against the baseline performance requirements of real HackRF Ones during final inspection. And by the way, narrowing that eye wattering 324% real versus clone price gap might help sell more real HackRF One units regardless.

Post a comment

You may use the following HTML:
<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>