Flex 6700 CW Keying v. Elecraft K3

  • 2
  • Praise
  • Updated 4 years ago
Last night, I compared transmit CW keying characteristics between my Flex 6700 and Elecraft K3 in "QRQ" mode. Keyed waveform images attached. The shorter keyed waveform time from the left edge is the image captured from the Flex 6700.

Test conditions:

- Agilent DSOX-3024A Digital Storage Oscilloscope for data capture;
- RF sampled from a Bird 4723 RF variable coupler;
- 100 watts transmit power from each transceiver on 80m into a 50-ohm load;
- External keying from a modified WinkeyerUSB. The keyer is modified with a port that goes to logic high when the PTT line goes low. This was done to ensure correct scope triggering;
- Screen capture is one "dit" at 40 WPM;
- Horizonal axis is 5 ms./div.

Notice that the Flex 6700 generates full power 6-7 ms. after key closure. By contrast the Elecraft K3 in "QRQ" mode produces full power at approximately 13 ms after key closure.

Also note from key closure to key-up, the Flex 6700's waveform goes to zero at roughly 7 ms and approximately 10 ms. for the K3. Both transceivers produce excellent keyed waveform shaping at the leading and trailing edges.

I should also point out that the Flex 6700's keyed waveform ramp time has two steps. What's shown is the faster keyed envelope time. The transition occurs when the 6700's internal keyer goes from 19WPm to 20WPM. But wait, I just told you that I'm using an external keyer. That's right. But the keyed ramp time is still a function of the internal keyer speed. I'll attach another image showing that effect.

The concept is a good one in that at lower speeds, the ramp time can be slowed while at faster speeds, it can be shortened up. As long as the raised cosine characteristic is preserved, it's a good idea IMO. The bottom line is that if you're using an external keyer, you may elect to slide the internal keyer to match that of the external keyer. Again,that transition occurs at exactly 19-20 WPM on the slider. I am not sure of Flex's plans to automatically adapt that transition when external keying is used. I see no reason to change it as long as the op is cognizant of that transition. No harm is done by a mismatch. But die-hard CW fans may want to keep this effect in mind.

Paul, W9AC

First image is the Flex 6700:

-
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes

Posted 4 years ago

  • 2
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
And, the Flex 6700 keying shown below when the internal slider is set to 19 WPM or lower. Keep in mind that the 6700 is still being keyed by an external keyer. Only the internal keyer slider value in SSDR is changing.

Paul, W9AC

Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
I failed to mention that the green-colored trace is the PTT line. Key closure is signified by the higher DC level, then falls to zero when the key is opened.

Paul, W9AC
Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1678 Posts
  • 573 Reply Likes
Paul,
Thanks for the captures and comments. Great work!

I don't think I've seen any previous comments about the ramp up change or ever heard of it on another rig so that is interesting.

Regards, Al / NN4ZZ
Photo of Ken - NM9P

Ken - NM9P, Elmer

  • 3875 Posts
  • 1181 Reply Likes
But that does explain the extra thump on receive on slower keying speeds requiring a slight increase of delay to about 18-20.
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
Ken,

There's likely a relationship to what you and I are hearing (when the slider is at 19 WPM or less). I have to believe that artifact can be ironed out since it nearly goes away at 20 WPM and above. I never send my character sets below 20 WPM even during slow QSOs so it's pretty much a non-issue for me. Sending is sorta' done in a Farnsworth fashion when I'm asked to QRS.

I'm still impressed with the overall QSK accomplishment at this early stage of SSDR development. Most folks probably cannot appreciate the software engineering effort required to achieve this level of T/R switching performance from an SDR transceiver, especially when one considers wide splits, RIT/XIT, multiple slices and pan adapters running, etc. What may seem like a simple process in the analog world can be incredibly complex in a SDR.

Paul, W9AC
Photo of Ken - NM9P

Ken - NM9P, Elmer

  • 3863 Posts
  • 1176 Reply Likes
I started out with my 6500 sending at 16-18 WPM since my code speed had deteriorated from several years' absence from CW work. The "thump" was annoying at first until I discovered that a slight adjustment of the delay muted it completely. I still set it about 50 on a noisy or busy band because my brain gets confused hearing "between the dits." But I have the delay set a lot faster than I EVER did on any other rig because it can be made to be so smooth.

Now after a few CW contests and DX pileups, I am sending at 20 again. I think I need to do some maintenance on my 25 year-old Bencher paddles because I am getting some skipped characters and odd timing. It seems as if It is missing a contact once in a while. Before 1.0 I could blame it on the timing quirks that Flex was still working out. But now it is all on me! (Or my paddles) I have cleaned them, but I may have a bad contact, or need some higher power cleaner and adjustment.

In any case, I have found the 6500 to be smoother in function and sound than my TS-850 was. And that is saying something, because I thought the 850 was a superb CW rig which I loved for 20 years before replacing it with my Flexes.

The 6500 has made me a CW op (part time) again! And an SSB Contester, and a DX chaser, and an audiophile, and a Digital Mode experimenter.......

(Did I say that I love this rig!)

73 Ken - NM9P
Photo of Stan - VA7NF

Stan - VA7NF

  • 401 Posts
  • 86 Reply Likes
Paul,
Thanks for the informative analysis.

Something in the keying feedback path starts making extra characters at 30WPM. In your investigation did you happen to compare key down to sidetone generation? Does it start before/at/after RF start?

Between fingers, 6700, sidetone speaker out, brain and back to fingers there seems to be a delay in moving the fingers. Probably in the vintage brain but I notice better keying on an external dedicated keyer with sidetone.
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
A couple folks have requested the same. The DSO has 4 channels so I can add 6700 headphone audio to ch. 3. This will show timing between keyer PTT, RF waveform, and Rx audio with sidetone. I've gotten lazy with DAX, so I'll first need to wire up a 1/4" to BNC adapter.

Paul, W9AC
Photo of Sergey, R5AU

Sergey, R5AU

  • 802 Posts
  • 96 Reply Likes
Paul, thank you for such report in details, i am with Stan regarding his request, it will be quite interesting to look on this
Photo of Ken - NM9P

Ken - NM9P, Elmer

  • 3875 Posts
  • 1181 Reply Likes
I would suggest that the test be repeated after 1.1 is released to see if there were any changes / improvements / glitches introduced.
Photo of Alan - KA4B

Alan - KA4B

  • 107 Posts
  • 11 Reply Likes
Paul, if I am perceiving things correctly it looks like on the plot of the K3 signal, once the initial ramp up has been completed, there seems to be a gradual increase of signal strength, probably of 2 or 3 percent. I don't see that on the Flex plot. Any explanation?
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
Good eye. That's due to ALC stabilization.

Paul, W9AC
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
I've now added a third trace, colored in blue. This trace represents audio with CW sidetone from the Flex 6700 headphone jack. Three screen captures were taken. All other test conditions remain the same as in the case shown above.

Display #1: CW speed set to 20 WPM in SSDR;

Display #2: CW speed set to 19 WPM in SSDR;

Display #3: CW speed set to 19 WPM with the horizonal scale set to 20 ms./division. Notice the "glitch" at about 190 ms on the blue colored trace. This shows the "thump" that's present when the internal keyer speed is set to 19 WPM or less.

Probably the most important take-away here is that the CW sidetone ramps "lock step" with the actual keyed RF envelope. Not only that, but the sidetone ramping is completely independent of the internal speed slider. Notice that although the keyed RF waveform changes from 20 to 19 WPM, the sidetone doesn't. This means that you're hearing the sidetone start about 8 ms. after key closure. I'll take that any day.

Flex gets an A+ for CW timing. Still the audible "glitch" heard in the headphones at ~ 190 ms. is a bit of an annoyance for those ops sending at 19 WPM or less from the internal keyer. For those using external keyers, simply set the internal speed slider to 20 WPM or more, then dial in any speed you like on the external keyer.

Paul, W9AC

Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 969 Reply Likes
Paul, we're working hard to complete v1.1, but we'll come back and look at this for v1.2. I'm not sure what might be causing it off the top of my head.
Photo of Barry N1EU

Barry N1EU

  • 495 Posts
  • 121 Reply Likes
Great stuff Paul! Can you please tune the receiver to band noise with sufficient RF/AF gain to create some scope deflection and repeat the trace and show the receiver recovery after transmission?
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
Have a look below. Keying is 20 WPM. A few interesting things going on...

Notice that the band audio is actually mixing with the CW sidetone for half of both keyed elements. I suspect this is due to latency.

Second, notice the receive band audio recovering in between "dits" about 50 ms after the fall of the CW sidetone and keyed RF envelope.

Paul, W9AC
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 969 Reply Likes
Let me explain what it going on here -- it's real subtle. There is latency in any receiver, most of it from filtering in the multiple stages of downconversion and demodulation (whether digital or analog). Let's say that the receiver is delayed in time by 40ms. If you are doing QSK keying and your dots are 60ms long, it will take 40ms after your dot for the RX to recover and then you will only hear 20ms of actual receiver noise in between the dits. As speeds increase and filters on CW get tighter, this problem gets worse making QRQ QSK impossible if you don't mix audio from the receiver and the side tone. We thought a lot about this and decided that the ear and brain are perfectly capable of filtering this out and by doing this we can provide QRQ QSK up to 100WPM. So this is how we implemented it.

If you use a separate TX and RX antenna, the FLEX-6000 does a fine job of handing you RX audio and TX at the same time. To prevent any overload in the RX during the precise moment that a tone goes out, we mute the receiver. But this muting occurs in the digital domain and is timed precisely so that there is no TX audio present in the receiver. Since there is no AGC in the RF section of a FLEX receiver, there is no recovery period for the receiver to "come back down to earth" after hearing its own transmit signal.
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
Steve,

Great explanation and it makes perfect sense.

Paul, W9AC
Photo of Stan - VA7NF

Stan - VA7NF

  • 401 Posts
  • 86 Reply Likes
Thanks again Paul,

The 8ms delay in output is fine, especially since it is shutting down the RX and notifying any amplifiers etc, before starting the RF.

I would still like to see the sidetone match the key down because at 30WPM and more the activator and sidetone are being more out of sync. At 60 WPM, I used to do message handling at that speed, it is over 50% shifted. As the WPM decreases the audible delay becomes insignificant.

With remote CW mode network delay and jitter 6000 key up/down will become erratic. In this mode the PC must supply the sidetone concurrent with key action and the flex will need around 100ms jitter buffer as in VoIP.
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 969 Reply Likes
Exactly. This is why the CW engine in the FLEX-6000 is built for full remote with separate keying, sidetone synthesis and RF keying components.
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
Last set of screen captures. Same test with the Elecraft K3. Again, the transmitted RF waveform is the yellow trace; the key line is the green trace; and blue represents receiver headphone audio with sidetone. The top display shows Rx recovery time in "QRQ" mode. The bottom trace shows Rx recovery time with QRQ deactivated.

In both instances: (1) the onset of the CW sidetone is delayed to approximately the same degree as the Flex 6700. The K3 does not exhibit latency and I wouldn't expect much, if any. (2) Recovered Rx audio is faster on the K3 when compared to the Flex.

Paul, W9AC

Photo of W5UN_Dave

W5UN_Dave

  • 315 Posts
  • 30 Reply Likes
I run full break-in on the K3 at all time, usually well above 20 wpm. What I notice all the time in rapid fire QSO's such as working DXpeditions, is that when I key up on the final character, the responding station will have already begun sending, and I will miss the first couple of dits/dashes. I find this lag latency very frustrating.

I cannot comment on the 6700 full break-in yet, as I have not used it enough. I expect better. Anyone else want to comment who has experience with full break-in with both radios?
Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1678 Posts
  • 573 Reply Likes
Paul,
Excellent information, thanks for taking the time to do all of the tests and analysis!

Regards, Al
Photo of Sergey, R5AU

Sergey, R5AU

  • 802 Posts
  • 96 Reply Likes
Hi Paul,
thanks a lot again for figures, i wonna to compare with ver 1.0, you know such "glitch" as mixing band audio an side tone + side tone "lock step" and not remember in 1.0, can you compare it ? I think CW was better in 1.0.
Could be interesting regarding FRS comments, i guess Agilent oscilloscope is Independent reviewer

Paul, just to be clear, in case 6700 you utilized only ANT1 circuit for RX/TX or two slices with separate RX and TX circuits(ex: ANT1 and ANT2) ?
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
Yes, only ANT1. I did not try more complicated use with more slices, panadapters, splits, etc.
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 969 Reply Likes
Incidentally we broke QSK CW in the Beta v1.0.x. Performance in v1.1 will be the same as the production v1.0 software.
Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1678 Posts
  • 573 Reply Likes
Paul,
Two Questions
--------------------
First, about the keying ramp up difference based on the CW speed setting. When using a straight key (i.e. IAMBIC OFF) does the speed setting still have the same effect on the signal? If so, and assume you are keying at about 15 WPM, would one or the other speed settings (and associated ramp ups) sound better on the other end? This is just a curiosity and you have spent a lot of time testing already so I'm not asking you to do any more but would like your thoughts.

Second, do you mind if i download some of your screen captures and comments for future reference and to share? I'd be sure to credit you for the work.

Regards, Al / NN4ZZ
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
Al,

1) Same effect with the straight key or an external keyer since iambic is turned off. Recall that all my testing is with an external keyer. It's possible to test the internal keyer by modifying the configuration of the trigger input. For anyone wishing to test timing of the internal keyer, a "Y" adapter is needed at the key jack. The key jack will have a positive key voltage until keyed to circuit ground. One side of the Y adapter goes to the Flex, the other to the scope's triggered input. The scope would would then be set to trigger on the trailing edge and not the leading edge.

2) By all means, go ahead and use the images.

Paul, W9AC
Photo of Al / NN4ZZ

Al / NN4ZZ

  • 1678 Posts
  • 573 Reply Likes
Paul,
Thanks, I compiled into a Word document. Again, nice job!
Regards, Al / NN4ZZ
Photo of Barry N1EU

Barry N1EU

  • 495 Posts
  • 121 Reply Likes
Paul, I'm looking at your last Flex trace (Jan 28 21:31:35 - not the K3 trace). What I'm seeing there is 50msec latency for rx recovery after xmsn, correct?

I think the mixing of rx audio with the sidetone is problematic and is probably a bug - agree? Someone sent me a soundclip a few days ago of their 6500 and I saw the same thing and was very confused by it.
Photo of Paul Christensen, W9AC

Paul Christensen, W9AC, Elmer

  • 313 Posts
  • 132 Reply Likes
Yes, but it's not occasional -- it's at the beginning of each keyed element when using full break-in. It's of such a short duration that I can't even tell the band noise + sidetone mix is even happening. It's an interesting visual effect, but operationally, I think it's meaningless.
Photo of Barry N1EU

Barry N1EU

  • 495 Posts
  • 121 Reply Likes
Thanks Paul. What was confusing me is that we're dealing with two factors at the same time: latency and TR turnaround. Because the turnaround is so much faster than latency, the rx is actually turned on and buffers are filling before the audio actually pops out when latency expires.
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 969 Reply Likes
It is done on purpose to make QRQ QSK possible -- see my note above for the details
Photo of Barry N1EU

Barry N1EU

  • 495 Posts
  • 121 Reply Likes
Wow, as an occasionally serious cw contester, I take my hat off to what you guys have accomplished!
Photo of Steve Ellington

Steve Ellington

  • 1 Post
  • 0 Reply Likes
I'm 5 months late on this. The band noise at the onset of the sidetone can be annoying but I've done a few things to help reduce the effect. 1. Installed a separate ethernet card just for the 6300. This seems to have reduced latency vs. the router connection. 2. Turn the master volume up high and control volume level with the slice volume control. I'm not sure why this help. 3. Use an external audio filter for headphones seems to help but of course filters how their downsides. 4. Slide the L - R balance control partially to one side or entirely. Sidetone from one speaker will be pure and the other speaker will contain both. My ears found the switching effect less distracting. 5. Keep the bandwidth as narrow as practical. Obviously this reduces atmospheric noise. 6. Use APF and or peak the EQ to the sidetone frequency. Keep in mind what N5AC said about high speed QSK. If the band noise were to be muted during the beginning of the sidetone, QSK speed would be effectively reduced and you could miss a transmission. I do think that once a certain amount of "Delay" is inserted by the user we could do without the sidetone noise. This would make operating a bit more pleasant for those not interested in QSK. 73 Steve N4LQ
(Edited)