DAX vs USB - Real World Test

  • 7
  • Praise
  • Updated 4 years ago
Well, KY6LA, you drove me to it....

I did a somewhat "scientific" look at JT65 decodes using DAX vs a Signalink USB. The results are interesting.

Setup:

Flex 6500 tuned to 28.076 MHz
WSJT-X version 1.4 beta - 2 simultaneous instances
AGC OFF
Filter bandwidth 3900 Hz
Windows 8.1 in a Parallels VM
WSJT-X #1 using DAX RX1
WSJT-X #2 using USB Codec
Sampled for 60 minutes

DAX connected WSJT-X #1 decoded 247 transmissions, with an average -9.6 signal. Minimum received signal was -22.

USB connected WSJT-X #2 decoded 253 transmissions, with an average -9.4 signal.
Minimum received signal was -22.

So, the SignaLink USB actually decoded 6 more transmissions, although the reception levels were about the same. Not sure where the "extra" decodes came from. Couldn't find a good reason for them. I suspect the difference in levels may just be a cumulative rounding error between decodes on each package that tilted slightly in favor of the USB run.

The DAX audio was noticeably cleaner on the waterfall at the very lowest edge (below about 120 Hz), where the USB device had noise present. 

Here's the USB waterfall:




Compared to DAX:


So, what does it mean? Probably that when set up properly, WSJT-X does a bang-up job of decoding signals under most situations. The humble SignaLink USB does a pretty good job, but is not as "clean" as DAX. I did not test transmit today, but know from experience that DAX TX is substantially better looking on the scope than the sound card output. 

I'd use DAX based on the "cleanliness" of the signal, especially on transmit. The extra little bit of signal-to-noise ratio could be helpful in a weak signal environment. But, in a pinch, the sound card will work fine.

I did not test in the presence of large, interfering signals, or on modes other than JT65. In those cases, your mileage may vary. California estimates lower. Must be 18 years or older to enter. Void where prohibited.


73,

Geo
Photo of George Molnar, KF2T

George Molnar, KF2T, Elmer

  • 1538 Posts
  • 546 Reply Likes
  • Scientific, sorta

Posted 4 years ago

  • 7
Photo of Guy G4DWV/4X1LT

Guy G4DWV/4X1LT

  • 1687 Posts
  • 387 Reply Likes
Thanks George for that experiment. It's answered a few questions here.

best 73
Photo of 'IB' - LB6B

'IB' - LB6B

  • 8 Posts
  • 1 Reply Like
That is surprising if the extra decodes are valid? Is that measured with a "modified" Signalink or off the shelf?
http://www.frenning.dk/OZ1PIF_HOMEPAGE/SignaLinkUSB-mods.html
Photo of George Molnar, KF2T

George Molnar, KF2T, Elmer

  • 1537 Posts
  • 546 Reply Likes
Stock Signalink. The decodes seemed to be valid, but I didn't look too deeply into them. Might re-run and see what happens.
Photo of 'IB' - LB6B

'IB' - LB6B

  • 8 Posts
  • 1 Reply Like
Brilliant George! An unmodified Signalink should at least be a couple of dB down from a DAX simply because of the design flaws described in the article in the above link. Something if definately broken with the DAX if it comes out worse.. I hope and trust this is not the case?!