testing the CW program

testing the CW program

Postby kz1x » Sat Jan 24, 2015 5:09 am

I have spent some quality time with the new program and will do more over the weekend.

I think I will 'move' the test from my Atom powered Gateway netbook to an i5 Lenovo T420.
This way I can tell if the performance issues I experienced are related to the relatively slow netbook.
The Lenovo isn't new but it's still reasonably snappy.
kz1x
 
Posts: 232
Joined: Wed Aug 07, 2013 1:28 am

Re: testing the CW program

Postby AE9RB » Sat Jan 24, 2015 10:10 am

Maybe the performance issues can be fixed.
73 David AE9RB
AE9RB
Site Admin
 
Posts: 511
Joined: Thu Nov 15, 2012 5:27 am
Location: Portland, Oregon, USA (CN85ql)

Re: testing the CW program

Postby kz1x » Sat Jan 24, 2015 5:05 pm

More specifically, on both of the computers, I experienced:

RX: all signals sound phase 'watery' ... AGC action also seems to have a high and low frequency component? I do not know if I should adjust some I/Q parameter or what else can be done. This happens at any bandwidth; I would like to have one wider-than-500 Hz setting also. The screen graphic should show the filter skirt, not just the passband width. A manual notch and shift-from-passband-center would be nice future features.

TX: the keying vs sidetone felt like some random number of large-hundreds (or small-thousands) of microseconds were 'inserted' between each key action and the sidetone. Since there's no TX I could not say if this also happens on the RF. No amount of weight change could affect the phenomenon; the weight adjustment does work of course but the range is quite broad. The 'center' 50% of the current adjustment range should be spread out over the linear range of the sliding control. I tried sending at several speeds and with multiple different paddles to assure myself the experience wasn't the wetware. The Ultimatic and bug emulation modes are clever. Dot and dash memories work well and as expected. A text-in-CW-out window would be a nice add. The spacing adjustment is unique and I didn't really understand it. I hope that the RF TX when enabled is a raised-cosine function and that the TX can operate QSK including an option to move the amp key to the envelope edge.

Other observations:

I saw no discernible difference in performance between the two computers, both Win7, but with vastly different CPU.

A Debian/Ubuntu 14+ port would be great, especially if I don't have to compile it. Quisk /= CW.
kz1x
 
Posts: 232
Joined: Wed Aug 07, 2013 1:28 am

Re: testing the CW program

Postby AE9RB » Sat Jan 24, 2015 8:39 pm

There will be a new AGC in the next version. The current one works mimics hardware AGC but only works great on SSB.

There's always going to be a tiny bit of keyer lag in a non-realtime OS. I want to know what WPM speed it becomes intolerable at.

The raised cosine shaping is in there already. Try changing it to zero. The sidetone your hear is what will be transmitted.

The only reason a classic radio has a filter 500Hz big is that you'd have a hard time finding signals with something smaller. With SDR you can look at the spectrum scope. Why do you want a bigger filter than 500Hz?

What's the use case for a notch filter? If QRM is getting through the 50Hz filter then a notch isn't going to help you.

I don't know what shift-from-passband-center means.

How well does Quisk work with CW?

Cool that it runs on Atom! I'm betting it "just barely" does. All the heavy stuff is in there already so it should continue to run on Atom.
73 David AE9RB
AE9RB
Site Admin
 
Posts: 511
Joined: Thu Nov 15, 2012 5:27 am
Location: Portland, Oregon, USA (CN85ql)

Re: testing the CW program

Postby kz1x » Sat Jan 24, 2015 9:21 pm

There will be a new AGC in the next version. The current one works mimics hardware AGC but only works great on SSB.

>> OK

There's always going to be a tiny bit of keyer lag in a non-realtime OS. I want to know what WPM speed it becomes intolerable at.

>> Well, intolerable is a word filled with pejorative. It's also too subjective. I would use a different word.
>> Also, it's not a function of keyer speed in this case, since the problem I found happens at all speeds.
>> The lag is not constant. It seems to be different for almost every character.
>> An objective standard to your "intolerable" point in this case is instead the answer to a binary question:
>> "Would you use the Peaberry V2 for CW using this program?" and for that my answer is (right now) non-binary.
>> My answer right now is: "Not if I had another radio to use instead."

The raised cosine shaping is in there already. Try changing it to zero. The sidetone your hear is what will be transmitted.

>> OK. I was primarily referring to the transmitted RF envelope that results, which may or may not be the same.

The only reason a classic radio has a filter 500Hz big is that you'd have a hard time finding signals with something smaller. With SDR you can look at the spectrum scope. Why do you want a bigger filter than 500Hz?

>> Because I like to tune around and listen to multiple CW signals in a passband of perhaps 1 kHz. It's not a matter of eyes.
>> As a "more traditional" ham, most of my personal radio interface is auditory.
>> You will also need to be able to support at least XIT/RIT, or A/B VFO TX switching. Can't work DX if you don't have some kind of split.

What's the use case for a notch filter? If QRM is getting through the 50Hz filter then a notch isn't going to help you.

>> This is a fair point. However, the preference could readily be to listen wider and clobber monotonic interferers.

I don't know what shift-from-passband-center means.

>> It means you leave the tuned frequency put and move the edges of the passband asymmetrically around it.

How well does Quisk work with CW?

>> It's unusable in that mode as far as I can tell.

Cool that it runs on Atom! I'm betting it "just barely" does. All the heavy stuff is in there already so it should continue to run on Atom.

>> OK. It'd really be cool if it ran on my Galaxy S5, if you're looking for environment challenges. ;-)
kz1x
 
Posts: 232
Joined: Wed Aug 07, 2013 1:28 am

Re: testing the CW program

Postby AE9RB » Sat Jan 24, 2015 10:39 pm

If you can't stand the 10-13ms required by the operating system for audio buffering then you'll never be happy with any software keyer. Or maybe both audio systems on both your computers have extra buffering which can only be fixed by whoever wrote the audio driver. You can always turn off the software keyer and use hardware like my Arduino keyer.

Is +/- 10kHz sufficient for split operation? More than that could affect QSK since the Si570 would have to be retuned every tx/rx switch. Tune too far and you have to wait for the PLL to relock.
73 David AE9RB
AE9RB
Site Admin
 
Posts: 511
Joined: Thu Nov 15, 2012 5:27 am
Location: Portland, Oregon, USA (CN85ql)

Re: testing the CW program

Postby kz1x » Sun Jan 25, 2015 2:34 am

I could accept a fixed amount of offset latency from my finger to the sound. The problem is the varying amount of latency.

I have the Arduino keyer shield and a Leonardo to go with it. But that's a sandwich of stuff to assemble to send CW. CW is supposed to be the simplest of modes. Part of the charm; the computers should be an accessory, you shouldn't need two in tandem to sample a contact closure. Once the setup is made complicated, you might as well do something more powerful with it. N2APB thought that all out pretty well http://www.sdr-cube.com/ordering.html and I agree with his conclusions.

Perhaps a PB V3 PSOC could have an optional $1 MSP430 in front of it handling the key. I know a guy who would write the firmware for it gladly. Or, maybe the latencies I am describing are distributed across the PSoC, USB, PC, etc. in which case it doesn't matter. The Flex rigs I have tried have 'challenged' CW mode operation too.

It's rare to have split operation for CW more than ~ 5 kHz so 10 would be fine, as long as the control is not too coarse.
kz1x
 
Posts: 232
Joined: Wed Aug 07, 2013 1:28 am

Re: testing the CW program

Postby AE9RB » Sat Jan 31, 2015 9:16 pm

I wonder if your spacing is set to something other than element. The others enforce proper character and word spacing which may seem like random delays if you aren't expecting this behavior.

The next version will have a significantly improved sidetone generator. I changed to WASAPI, which is the new audio API introduced in Vista. There are zero buffers in the application now, so any jitter or lag will be from the drivers and operating system.

FWIW, the Mac sidetone in the next version is indistinguishable from a hardware keyer. The OS X audio system is fantastic. There's only one API to choose from and it's lower latency than anything in Windows.
73 David AE9RB
AE9RB
Site Admin
 
Posts: 511
Joined: Thu Nov 15, 2012 5:27 am
Location: Portland, Oregon, USA (CN85ql)

Re: testing the CW program

Postby AE9RB » Sat Feb 07, 2015 12:08 am

Please try the keyer in version 0.2 and let me know what you think.
73 David AE9RB
AE9RB
Site Admin
 
Posts: 511
Joined: Thu Nov 15, 2012 5:27 am
Location: Portland, Oregon, USA (CN85ql)

Re: testing the CW program

Postby kz1x » Sat Feb 07, 2015 7:38 pm

Massive improvement to the keying. Like, night and day.

Of course I will need to see the RF envelope to match it, but, as an improvement, just wow. I could not replicate my annoyance.
The weight scaling is just about perfect.

Now ... regarding the AGC ... there was some improvement. Still hard to listen to.

The under-the-water effect is now reduced but still present, and multiple signals in the passband are discernible instead of being turned to mush.
However a noticeable and unpleasant effect still exists. Signals have phase distortion.

Loud signals (indicated S9+10 and up) get clicky sounding where they did not before.

AGC should have an 'off' button option.

I cannot see any circumstance where the keyer sidetone should not equal the RF offset.

The extra bandwidth setting is very helpful.

The spectrum should be able to autozoom at 250H or narrower selection; the VFO should autoselect the appropriate digit position.

Any chance at point-and-click for the VFO?

The app needs a mute button. Tap to mute output; tap again to unmute. Flash the icon red when muted.

The S meter should be a dBm indicator (option?) and it should have an accessible calibration constant.

The v0.2 app routinely starts on the Atom netbook without the spectrum display, but the Lenovo T420 does not have this symptom.
kz1x
 
Posts: 232
Joined: Wed Aug 07, 2013 1:28 am

Next

Return to Peaberry CW

Who is online

Users browsing this forum: No registered users and 1 guest