CW latency measurements ...

W7NGA

I thought I would take a few minutes this morning and measure CW latency based on N1EU's thread and investigations. Thankfully, I find this quite reasonable and not an impediment to my operating in the least. Receive/Transmit switching time is ridiculously low and also a consideration in the overall scheme of all things contesting.

Here are three plots of latency:

68.5ms   @ 1khz
101.5ms @ 800hz
165ms    @ 400hz

Essentially the same 165ms latency for filter widths less than 400hz. 

Hope this helps ...

W7NGA  dan
San Juan Island, Wa.

Charles - K5UA

Hi Dan, can you run these for the 50, 100, and 150 hz filters? Thanks for your time.

Charles  K5UA

W7NGA

Charles,

I did. Latencies for CW filters are all essentially the same for 400hz and lower.

Mark Griffin

So in a nutshell, the less degree of filter you use, the less the delay between transmit and receive. Does that mean that it takes a lot more for the SDR software to use heavy filtering than not? Please excuse me, if I didn't phrase that correctly. Mark Griffin, KB3Z

Barry N1EU

Thanks Dan!!!  You confirmed my earlier measurement/posting  ;-)

W7NGA

Yep ... I thought the visuals would add credence to your thesis Barry!

Sergey R5AU

Dan thank you for clear statement 

Barry N1EU

Mark, just to be clear, we're measuring the delay between someone else transmitting and you hearing it through the Flex 6K receiver.

Mark Erbaugh

I believe that with narrower filters (or at least steeper filter skirts), more samples must be processed, it's not necessarily the processing time, but the time the processor must wait for the samples.

Charles - K5UA

Thanks so much

Charles  K5UA

Ken - NM9P

RE:  the "ridiculously low" transmit switching time:
If I remember correctly, someone, (I think you, or Paul or Lee), tested and found only 8 ms between key-down and beginning of the CW signal.  I don't remember which software version that was, and I don't know if that figure still holds in the latest version.  But I assume it still does. I haven't noticed any significant changes in CW performances in many versions.  Do you confirm this?

Ken - NM9P

Sergey R5AU

Ken, here we are talking with some other one - full signal delay through received or backward through transmitter (in case ex.: SSB ) what reflect filter parameters. I don't assume anything changed in the CW TX time in 1.5.1

Ken - NM9P

I understand.  That is a separate measurement, and a separate (but related) issue.

But he referred to the transmit switching time in relation to that issue.  I assume it was because it may moderate the receiver latency in actual operation by reducing the total turn-around time between transmit and receive (and back.).

Ken - NM9P

Ken - NM9P

I know this has been discussed before, but it would be interesting to know how much latency savings could be made with a user-selectable option with a "softer" CW filter with a little gentler slope factor.  (especially at narrower bandwidths.)

I believe Steve addressed this once, but I don't remember the figures, or if any figures were actually published.

I also do not know what OTHER things might be "broken" in the process of adding such a feature for changing the slope factor of the filters.....

Ken - NM9P

Sergey R5AU

Yeah it close each to other but different, our filters are FIR based and have fixed group delay time means very limited relation between filter sharp and filter ring. But in any way right now a good time to measure everything again.

W7NGA

Ken, you are correct. It is the aggregate times that ultimately define how fast one could hear a signal end and get RF to the antenna. That is, my Collins 75A4 has essentially zero latency ... but it takes me ~500ms (or more) to switch everything around. Still, that is considerably faster than when I was a Novice and had a knife-blade switch to go from transmit to receive. As I recall, I first had to reduce the gain on the BC-348, flip the knife-blade antenna switch to switch the antenna from receiver to transmitter, and then hit the transmit switch on my homebrew transmitter! Five seconds later (or longer) I was ready to send CW at a blazing 5 wpm. 7 wpm on a really good day!
 

Barry N1EU

Ken, I think the latency improvement by going to a shallower slope would be considerable.  When I tick the cw BW up to 420Hz from 400Hz, the latency goes down from 165msec to about 100msec.  I believe that improvement is due to the firmware using a shallower slope.  (and I wish I could save that as a filter preset)

The transmit latency seems minimal - its so small that I can easily key using a second receiver as my sidetone while I key the 6500.