[Elecraft] K3 TX delay

[W2RU - Bud Hippisley]

One thing I haven't seen mentioned in this discussion is the sending speed at which the benefits of QSK "disappear" in the K3.

Speed in words per minute (wpm) is very roughly equal to the number of dashes in a 5-second interval.  Thus, at 25 wpm, there are 25 dashes in 5 seconds, or 5 dashes per second.  But a dash consists of 4 "periods":  three consecutive "mark" periods, followed by a single "space" period. So...in one second there are 5 x 4, or 20 such periods in 25 wpm CW, and therefore a single period is theoretically 50 milliseconds (ms) long.  Of course, if keying "weight" is set above 50% (or 1.0-to-1.0 mark / space ratio for dots), the key-up period is LESS than 50 ms.  A similar calculation at, say, 35 wpm, gives us a key-up (or "space") period of about 35 ms. 

But to avoid generating excessive clicks, the leading and trailing edges of RF output during each mark element ramp up or down gradually, with the total interval at either end of the element having been measured at 4 ms (onset of RF output to 100%) in one specific "clickless" K3.  If we assume we have adjusted the weight using the 50% RF output level as the reference (an assumption that is, itself, worthy of a completely separate posting), there's another 4 ms (1/2 x 2 x 4) that we have to subtract from each space period.     

Of course, for QSK to truly be QSK, we have to be able to hear receiver audio output for at least part of the "space" period between two consecutive "mark" elements (either dot or dash).  The minimum K3 TX Delay that can be set via the CONFIG menu is 8 ms.  It's my understanding that at least part of that delay is for "covering" the time it takes all the various synthesizer and DSP circuits in the K3 to switch from RX to TX, especially since the RX and TX frequencies can be different when in QSK mode.  It's also my understanding that there's a similar delay at the END of a transmitted element, when all the circuits have to return to their RX states.  Because frequencies and other circuit parameters are actively changing during those two periods, it's unlikely that useful RX audio is available at any time during either 8 ms interval. 

Clearly, then, there is a MINIMUM total interval of (2 x 8) + 4, or 20, ms when no received signals can be heard by the K3 user between mark elements.  In fact, even at the 8-ms setting, it's very possible that MORE than 16 ms of each space period is "blanked".   I say that because a number of us have found that to get proper weight from the K3 in QSK mode, we have to artificially adjust the length of its "mark" periods (with, for instance, the "Compensation" parameter of a WinKey or other suitable outboard keyer) by adding 6 or 7 ms to each key-down interval.  (Note that this a fixed interval for a given K3 regardless of sending speed, and is not "weight" in the conventional sense.  Good keyers, like the WinKey, allow weight, ratio, compensation, and other parameters to be independently adjusted.

So now, with the K3 CONFIG menu set to 8 ms (its minimum value), we have 20 + 7, or 27, ms of each "space" when the receiver is blanked.  Interestingly, that's just about the length of an entire space at 35 wpm!  So, for all practical purposes, QSK is simply not possible with a K3 at CW speeds above 35 wpm.  Below that speed, the ability of QSK to help you detect the presence of the other station is statistical, depending on whether enough of the character elements sent by the other operator happen to coincide with some part of the narrow usable window within your own transmitted "space" intervals to cause you to conclude he's really sending again.

But it gets even worse:  That same K3 above was measured as having 15 ms total delay from initial closure of its KEY OUT line to the start of RF output when using the "new" QSK mode with TX Delay set to 8 ms, the minimum possible.  Further, there was an additional 15 ms delay between the cessation of RF output and the lifting (opening) of the KEY OUT line.  Since RX audio is muted the entire time the KEY OUT line is closed, then we have to add another 14 ms [2 x (15-8)] to the 27 ms; that gets us to 41 ms, the entire "space" interval between consecutive "mark" elements at 24 wpm or higher! 

Conclusion:  Modern transceivers filled with synthesizers and DSP circuitry are no match for properly designed analog receivers such as the old Hallicrafters and Collins units when it comes to QSK performance.  For the K3 and most other modern transceivers,  at common CW contest speeds "QSK" doesn't mean what it used to ("able to hear the other station between our own transmitted DOTS"); the best we can do above 25 wpm or so is to hear the other station between CHARACTERS.

The K3's QRQ mode may be better in this respect than the figures discussed above but, since QRQ mode currently does not allow the use of RIT, XIT, or Splits of ANY delta-f, no matter how small, it's of limited interest to me since my primary interest in QSK is knowing when a DX station has come back to someone while I'm still sending. 

Finally, with respect to amplifier QSK:  I am aware of five ways of enjoying QSK capability when using a separate amplifier:
     1.  PIN diode circuits (Alpha 87A, some Ameritron models, some outboard PIN-diode "boxes) triggered by the KEY OUT line from the K3
     2.  High-speed reed and vacuum relay combinations (current Alpha models, AG6K circuitry, etc.) triggered by the KEY OUT line from the K3
     3.  Electronic bias switching (EBS) in conjunction with a shunt-type vacuum tube TR Switch (E. F. Johnson, e.g.) and requiring NO trigger signal
     4.  EBS in conjunction with a separate RX antenna (always with suitable RX front-end protection, of course) and requiring NO trigger signal
     5.  Diplexing schemes which don't care if / when the transmitter is keyed

Most important, the clunky, noisy, typically open-frame relays used for manual, PTT T/R switching in many amplifiers should NEVER be considered capable of being used in QSK mode!  In fact, many of these relays are sufficiently slow that they shouldn't even be used in VOX mode because they'll be "hot-switching" RF, even with the delays in the K3.  

PIN diodes and EBS can be VERY fast — with total switching times under 1 or 2 ms.  Inexpensive reed and vacuum relay circuits typically perform in 2 or 3 ms, thanks to the use of speed-up circuit techniques.  With respect to #3 above, the maximum voltage out of the RX ANT terminal of a Johnson TR Switch during transmitter key-down periods is high enough to trigger the internal RX front-end protection circuitry in the K3, so it is not a viable approach, even though I have successfully used Johnson TR Switches with Kenwood solid-state transceivers for over a quarter century.   

And FAR more than a "quarter century" ago I began my own ham "career" with a form of #4.  I used a separate receiving antenna and the EBS was very simple:  all my CW rigs and amplifiers back then were operated in Class C mode, and fully cut off during key-up intervals!  

Bud, W2RU