Receive Latency

Barry N1EU

I just made some measurements on cw receive latency and wanted to share them.  I used a simple methodology of recording the tx sidetone in my K3S while transmitting a single dit and recording the subsequent cw receive in a Flex 6500, Orion II, and ANAN-100D all set for 400hz receive and no dsp functions engaged.  The latency figures measured  were 175msec, 45msec, and 70msec respectively.  I hope going forward the Flex number can be improved.  175msec is just long enough that it's getting to the point of me being slightly delayed when trying to quickly respond in a dx pileup.

73, Barry N1EU

Sergey R5AU

Hi Barry, this is not new and you well confirm what we have on hands for now with F6000 UM:

however such mesurements are verys depends on the filters shape factor you know what was discussed ex. here:  https://community.flexradio.com/flexradio/topics/cw_filter_shape_factors?topic-reply-list%5Bsettings%5D%5Bfilter_by%5D=all

i am not againts with you notes but necessary to have apple-to-apple compare
or we need to talk concerning introduction of the  Shape Factor "tuning knob" into the F6000

P.S. personally i have no issue with Latency, usually my setup is CW 400Hz and 32WPM and higher

K0UNX

My Apple Macs give you a choice of analog output or FIBRE output.  So does my TV, and my HI-FI equipment.  I had gone the FIBRE route to eliminate physical connections, ground loops, and the hum associated with connecting things together.  It didn't take long to realize that whenever you have AD and DA converters, you introduce latency with each step.  The result is "LIP-SYNC" problems watching TV and ECHO effect with audio.

SDR's have to take analog signals and digitize them to send them over a network or to process them.  On the other end of the pipe, the computer or terminal (Maestro, etc.) has to convert the digital stream back to analog for our ears.

The TV in our living room was purchased in 1974, and has a "converter box" to receive OTA HDTV and present it on that old set.  We also have a modern HDTV in another room.  The latency between the two when you're standing halfway between them is on the order of FIVE SECONDS.

Yup!  Latency is with us.

I've gone back to analog wires connecting my stuff together, and put away all the fibre gear.

Jim
K0UNX
Littleton, Colorado

Barry N1EU

Thanks Sergey, but I'm not seeing 85msec, I'm seeing twice that much latency - big difference that concerns me.

Andrew O'Brien

Just to stir things up... is there any evidence that a 175 Msec delay makes ANY difference in whether the DX station picks you out or not ?  That delay might actually help.
Andy K3UK

Mark Griffin

Barry,
This is nothing new. If you were to take a Flex User poll and ask them to pick between lower latency versus unlimited filter width, the unlimited filter width would win out. I myself have a K3 and also a Kenwood TS-990S.

The K3 has hardware filters and so does the Kenwood. And of course you can adjust the width of the filter with the limits of the firmware that is installed. Do you use your Flex for contesting?

The contest radio club I belong to was to have N4HY, who was one of the architects of Flex Radio at our club meeting tonight. But due to illness he is not able to make it. I was going to ask him some questions related to the Flex after the meeting. But that will have to wait for another time as they say.

I was going to purchase a Flex. I was very impressed with the hardware performance. But with all the issues that are software related, I purchased the 990S. I just see too many issues when it comes to software. That is why I was very apprehensive. Mark Griffin, KB3Z

Barry N1EU

If you don't do a lot of contesting, this is probably a non-issue.  But I often engage in trying to bust a pileup for a needed multiplier.  Precise timing of my calls is crucial in the ebb and flow of the pileup.  175msec is enough to throw my timing off a bit.  

I'm under the impression that Flex is seriously going after the contester market, so I think this IS an issue.

Maybe the filter slopes need to be slightly shallower but that's not currently a configurable parameter.

Walt - KZ1F

Mark and Barry, you both mention you have multiple radios. Something I've never quite understood is how does that work? For instance, I have 3, the 530 and 1500 are collecting dust. Someone, a yr or two ago, said their k3s was their everyday radio but then when is the 6000 used? So I was curious in your two cases.

Mark Griffin

Walt,
I was going to do a SO2R situation when I considered adding the flexradio. I had a Top Ten DX Doubler all ready to go. But after getting my 990S I have a main receiver as well as a sub receiver.

That is one option I did not have with my K3. So when I use my 990S with N1MM Plus for contesting  I have the software set for S02R and I can see a bandmap not only for spots on the main receiver but also the Sub Receiver.

Currently my K3 is sitting on a couch. I have it as a backup rig just in case the 990S falls to pieces. And I hope that doesn't happen! I just used my 990S for the CQWW SSB and CW contests, and I prefer it over my K3.
Just personal preference I guess.

Hopefully this gives you some insight on how I operate when it comes to contests.

With day to day operating I will probably be on 40 meter CW which is what I have the main receiver on. The sub receiver I usually have on 40 meter SSB. I have 2 speakers so I can listen to both.

Mark Griffin, KB3Z

Sergey R5AU

Barry, i assume you mesured two way signal cirucaltion, however in the table mentioned one direction ex. from  the Mic to Ant

Barry N1EU

No Sergey, this is 175msec ONE way, from antenna to speakers/headphones.

Sergey R5AU

Got it, very strange, I'll remesure at the next weekend

Andreas Junge

Barry,

are you measuring the delay at the speaker output in the back of the radio or are you using the "Remote" audio? I am sure the latency table from the manual is based on output at the speaker port of the radio.

Routing the audio through the "Remote" seems to add a considerable amount of latency.

73, Andreas, N6NU

Barry N1EU

It's measured at the rear panel speaker output.

I suspect that the current 6K steep filter slopes is a double negative for cw contesting - it results in greater latency and it stops you from hearing off-frequency callers when you're running.  Not-too-shallow, not-too-steep is what's needed - like the PowerSDR "1024 buffers" slope that matches the Orion slope:

Walt - KZ1F

That's great, thank you Mark. A few months back I had to take a road trip to HRO in NH to return a rotor. The gas it took was way less than shipping and I got to see, feel, and play with their radios. Absent were Flex, and Elecraft. But the 990s was there as were Icoms and Yaesu's. I liked the 990, aside from it's size and weight, it seemed pretty full featured. Now that is a candy shop. This is meant as a rhetorical and I am certain FRS has already 'done the math' but I wonder if selling retail would ultimately sell more units with a positive margin for FRS. It's gotta be deemed less profitable for whatever reason. I didn't see Alinco's there either, not that I was looking for them,

Sergey R5AU

Barry with F6K in use FIR filters what a different from the regular one what you explained, but latency measurement results what you get are very strange 

Mark Griffin

Walt,
You make an excellent point that neither Flex or Elecraft were at HRO. Elecraft is basically either a kit or assembled. And you pick the options that you want. Flexradio is pretty basic except for adding a GPS or Antenna Tuner on the 6300. Both Elecraft and Flexradio seem to be quite successful with their marketing strategy.

My first major radio was a TS-940 and I was going to reactivate that one. But decided to get the 990S instead. Plus the deal that HRO had where it was about 1500 dollars less was a major reason why I purchased it. It is a large, heavy rig, but with all the features I really like it. There is a learning curve, but that will happen with any piece of equipment you buy today.
Mark Griffin, KB3Z

Craig K9CT

Barry

My opinion is that this latency is real but not having any effect on the ability to DX or Contest. I have been grinding away using SO2R and have not seen any negative effect in SSCW or ARRL 160. I can listen to my second harmonic and get a feel for the latency. I don't change my operating at all.

Being on the DXpedition side of things...some delay before calling is a GOOD thing. Almost everyone calls at the same time. The DX station is usually tuning away from the last station worked in a split and it takes a half second before he can hear the next call. Notice how many times in a phone pileup that the DX station will ask for the Echo Uniform or cw pileup EU? That is because you were the last station he heard. 

If you are running on cw, two or three stations on the same freq is uncopyable but you will notice the last couple of letters and work that guy. 

So far, other than knowing there is some latency, it has not had any impact on my operating. Should it be improved as much as possible? Yes, I would agree. 

I was more concerned about turnaround time. The time it takes to go from TX to RX. In SO2R, I am very impressed how quickly the 6700 switches. The second I CQ on one slice, I am listening to the second slice. Once done CQing, I am now instantly listening to the first slice. Is there a real delay? Yes, but not perceptible or meaningful in real time operating.

You can actually listen full duplex in SO2R, so no actual noticed delay unless you are near your harmonic. 

73, Craig K9CT

 

Barry N1EU

First off, big congrats Craig to what I believe is the best showing ever (at least in USA) in a contest using a DDC/DUC xcvr.  I agree with everything you said but there is still a real downside to the receive latency for my contesting.  When I'm trying to break a hellacious pileup on a needed multiplier, and I'm hearing the dx well, I feel that precise timing of my calls is a big advantage to punch through.  That's the edge I would lose with 165msec receive delay. 

I just ran some more experiments.  Latency drops from 165msec to about 100msec when I increase to 422Hz bandwidth.  I then ran some passband plots to look at the slopes and compared them with my Orion II (roofing filter plus dsp at their steepest setting of "199 taps").  Yes, the Flex filter slope is slightly shallower at 422Hz than 400Hz, accounting for the lower latency.  BUT the Orion filter slope is MUCH shallower still than even the 422Hz filter slope.  I would post some plots but the only software I have available produces logarithmic traces and they look a little strange.

This data seems to indicate that Flex has really skewed the filtering to extremely steep filter slopes without offering the option (yet?) of shallower slopes. Steep filter slopes are great when I'm tuning a crowded band or I have major QRM on the side when running.  But otherwise, shallower filter slopes comparable to what's being provided in rigs like the Orion II and K3 will result in MUCH lower latency and would also allow me to hear off-frequency callers, which I generally want.  On another note, when I'm just cw ragchewing, I really prefer the tone of a shallower sloped filter.

73, Barry N1EU

km9r.mike

Please do not take this the wrong way and I hope you do the same so you will agree. While working a contest pileup I go from loudest to weakest. I think it is just human nature to be lazy and work the easiest stations first.. Even if I get a quick prefix that is subsequently covered by a loud full call , I will default to the loud full call unless the prefix copied was a multiplier. I am not saying that your calls are not loud, at the worst I am just saying there may be louder stations calling at the same time and the station being called also goes from loudest to weakest as well.

At the same time, I think a leading edge technology machine like the flex SS rigs should be the envy of other manufacturers when it comes to measurements like this and am disappointed that they currently are not in this particular measurement. : (

Andreas Junge

Barry,

thanks for measuring this! BTW, would it be possible to get the numbers for the "Remote" setting? To me it feels like it adds  another 150 to 200ms.
 
73, Andreas, N6NU

M0GVZ

The delay is irrelevant, really truly it is. In fact it may actually help you as I've always found that as a low powered station and operating mobile I get a lot more success being the "tail end Charlie" as it were so that the tail end of my transmission gets heard after the cacophony of noise from everyone braying away has subsided. "Station with victor zulu in the call"  or "station with golf victor zulu in the call" is often what I hear from the station running the frequency.

Look through my videos on Youtube. You'll see I get through pile ups usually in the first couple of calls. You'll also notice I repeat my callsign twice to ensure I'm tail end charlie. One of them, the one with the President Lincoln where I get through a pileup to AU3NIAR I'm only running 21W into a Hexbeam and manage to get through second call. 

At the end of the day though the biggest delay is actually in the operator. 

W7NGA

Too funny .. whoda thunk the day would come where adjustable latency becomes a feature?

Andreas Junge

I think this is besides the point. For one, I like to decide if I am tail-ending when calling and not leave it to the radio. It's about what can be done vs how it is implemented right now.  Barry has shown that if the filters are set to 422 hz width the latency drops by 65ms. The steep filter curves come at a certain price and I am willing to pay most of the time, but not always. If I get an option to select a filter shape of steep vs shallow with the resulting effect of high vs low latency then I can decide.  

I am also comparing a remote K3 with RemoteRig to the "Remote" audio setup of a Flex 6700 and the delay on the Flex is WAY higher than the RemoteRig.  It feels like there is a 150 to 200 ms added on top of the necessary delay dictated by the filtering. I think there is room for improvement. RemoteRig can do it well....

73, Andreas, N6NU

Barry N1EU

The delay is irrelevant

It's irrelevant to YOU.  Several cw contesters have contacted me off-line and told me it's not irrelevant to them.

73, Barry N1EU

Lewis Cheek

I have a 6500, and the more I use it in cw while contesting the less I like it. The total delay is not something I like. My hope is after Maestro is rolled out ( I NEED KNOBS ) Flex will work on reducing latency, if not I'll be moving back to Elecraft. Many times the station I'm working is well into the exchange or calling cq before my receiver is switched. I know a couple of BIG time contest stations are using Flex and perhaps their input will help resolve.

Lew
N4CO

W7NGA

Lew, could you expound further? I am trying to follow your example. You are transmitting, stop, and after the Flex switches to receive the other station is already mid-CQ? It seems to me that the time to go from transmit to receive is essentially instantaneous (~7ms) and you receive almost immediately. The latency contributes a delay before you hear the signal only. Reversing this line of thought, you are hearing the other stations CQ ~165ms after they stop transmitting. They should hear your reply about ~170ms after they stop and listen (ignoring their own latencies). I measured this at one time but there is essentially zero latency between key-down and RF at the antenna.

W7NGA  dan

Lewis Cheek

Dan, guess I need to work on this a bit . When I stop xmit and start hearing the other station they are well into the exchange before I hear them. Now I do understand some guys are quick, and perhaps they don't wait until I have finished . However, this occurs very often. I usually run using a 400 hz bandwidth.Wish I had a way to test this myself. I'll really be disappointed if Maestro exacerbates this. For my operating gaining some knob control is all I need. I'm looking to be able to use my left hand with Maestro and keep my right hand on keyboard, etc.

Lew

W7NGA

Lewis, I removed my comment about Maestro because I realize now that to be a contesting machine you must be hard-wired to the Flex and not be Remote. If latency is a problem now, you sure don't want to add networking delays.  I was enthused about Maestro to use it as an expensive FlexControl but I have been told that it cannot control SSDR on the PC. Perhaps in the future. I have become far too accustomed to my beautiful 24" displays, and 4 slices, to fall backwards with the Maestro.

W7NGA

Let's back up some more. Your signal travels at 186 miles-per-millisecond. About 1300 miles during your T/R switching interval. About 15,000 miles in the difference between a very low latency receiver and the Flex. Hmm .. not much. 

So, the issue would be when the transmitting station finishes calling CQ, you will not be able to know this until ~165ms after the fact (should we include transit times?). You hear him stop, your brain processes this fact and introduces its own latency, and your arm eventually presses the CW key to transmit his call. 

Instead of buying a new radio, wouldn't one be almost better off to simply hone their reflexes? Reflex exercises. See the light, press the buzzer. Or maybe just move closer to the DX station(s) and buy yourself a millisecond? Hehehehe .... 

Lewis Cheek

You're on to something here! Just need to also account for rotation of earth and if signal is coming from west to east, east to west, north to south or south to north. My xyl says I don't hear well, so maybe a good wax job prior to contest is in order.