Testing QDX receiver performance for FT8 signals

I recently built a QDX digital transceiver from a QRP Lab kit. QDX Digital Transceiver

I was intrigued by a technical feature of the device (direct synthesis of the RF waveform), its low cost, and its compact, uncluttered physical layout. It looked attractive for portable SOTA operations when compared to my FT817 with associated interface, CAT cable, soundcard dongle and USB hub.

To gain quantitative information about the performance of the unit, I tested the receiver against a Flexradio 5000a (home station) and compared it to an FT817 (SOTA station).

Equipment: I set up my 20m centre-fed dipole as an inverted-V in my back-garden in Calgary and used a splitter to provide identical signals to my home station, a Flex5000a and to the QDX. The Flex was operating PowerSDR v2.7.2 running on an old Lenovo computer under Windows 7.

The QDX was connected to an HP Stream laptop computer running Windows 10. Both computers were running WSJT-X v2.5.4

The FT817 (with TCXO module) was connected by a home-brewed audio interface to an external soundcard and thence to the same HP Stream laptop computer used with the QDX.

WSJT-X records a file “ALL.txt” that contains the decoded output of the program. This file was imported into Excel for analysis.

Procedure: The Flex and QDX systems were run simultaneously for 1 min 45 seconds (seven decoding periods of 15 seconds each for the FT8 signals) during which time the Flex decoded 218 messages, and the QDX decoded 142 messages. Every message decoded by the QDX had a corresponding Flex decode. There were however, 76 decodes by the Flex that had no corresponding decodes by the QDX. For the 142 messages decoded by both radios the reported S/N ratios were very similar, differing by only 1 dB on average with little variability.

I did look at the S/N ratio reported by the Flex for the 76 signals not decoded by the QDX and found that these had lower S/N ratios when compared to the signals decoded by both systems. It appears my QDX is slightly deaf compared to the Flex5000a.

The Flex and my FT817 (portable) were run simultaneously for 24 mins 45 seconds (99 decoding periods). A total of 3671 decodes were received, 3138 by both Flex and FT817, 250 by the Flex only and 283 by the FT817 only. Comparing the S/N reported for each signal decoded by both radios, again there was little difference, the Flex values averaged 1 dB higher, but there was more variability, a standard deviation of 2 dB. The Flex5000a and the FT817 portable system have essentially the same receiver performance in this test.

I tried one further experiment. My portable system uses a cheap, no-name soundcard dongle. I swapped this for a Sound Blaster X1, reported to have superior specifications. To my surprise, the system performed quite poorly, decoding only a fraction of the decodes reported by the Flex. I am still investigating this observation.


The use of a splitter to present identical signals to two radio systems is a useful method to compare the performance of the radio receivers.

Receiving FT8 signals is a more variable process than I expected. The QDX appeared to be working just fine but was missing about 35% of the signals decoded by my Flex5000a home station (and by inference, my FT817 SOTA station).

While my FT817 portable station was working as effectively as the Flex5000a, it appears I was lucky in my selection of the soundcard I was using. I plan to investigate the performance of various soundcards, and to determine what makes a good card in this role.

I would be interested to hear of experiments by other people to determine the effectiveness of various portable digital radio systems.



Hi Ian,

Interesting. What do you think was the reason for the 817 decoding some signals missed by the Flex and vice versa?

Some of these minimal component count rigs lack the dynamic range of better ones. I think that’s a problem that a switchable preamp in the Rx path could solve, but it means more weight and complexity.

Poorly filtered smps have been implicated in less than optimum sound card performance by masking the lower levels.

A 16 bit a/d should have enough dynamic range but maybe some chips with problems have made their way onto the market.


1 Like