Call Search
     

New to Ham Radio?
My Profile

Community
Articles
Forums
News
Reviews
Friends Remembered
Strays
Survey Question

Operating
Contesting
DX Cluster Spots
Propagation

Resources
Calendar
Classifieds
Ham Exams
Ham Links
List Archives
News Articles
Product Reviews
QSL Managers

Site Info
eHam Help (FAQ)
Support the site
The eHam Team
Advertising Info
Vision Statement
About eHam.net

   Home   Help Search  
Pages: Prev 1 2 3 [4] 5 Next   Go Down
  Print  
Author Topic: CW issues  (Read 17949 times)
M0HCN
Member

Posts: 473




Ignore
« Reply #45 on: August 01, 2011, 06:24:08 PM »

Sigh!
I don't have a horse in this race (If I want an SDR I am quite capable of rolling my own), but I thought this was about to get interesting when someone posted a link to actual measurements, you know, as in real data, but it appears that the ongoing pissing contest is more entertaining.....
Leaving the err "personalities" out of this can we have a conversation about what acceptable CW  performance means?

Can we start by actually trying to quantify what we mean by acceptable CW performance (I am working on a (mostly non SDR) rig of my own, so this would be useful information for when I pen the DSP code for it)?

Sidetone latency: < 10ms.
Sidetone latency jitter: <2ms.  This is I suspect the big one for having a radio feel right to a CW operator (Based on what gives music keyboard players fits).

These are trivially met in an embedded micro or DSP  as you do all the work on a sample by sample basis in the ISR, but are somewhere between harder and MUCH harder to meet on a general purpose OS (you really need to timestamp the key up and down events with the corresponding audio sample position, possibly do it in the micro running the AD and data link over USB or whatever (Hell, generate the sidetone there and pass it to the PC instead of the RX audio!)).
Note that neither of these say all that much about any latency from the key going down to the rig making RF, just about the key -> sidetone timing, so presumably we need to specify how accurately the RF envelope should track the key down time.

RF Pulse duration accuracy: +-1ms (seems reasonable and is clearly doable, note that this is really time from the start of the keying envelope rise to the start of the decay, the actual envelope will of course be longer).

It is highly architecture dependent but for full break in a time for the key up to RX back at nominal sensitivity is surely important (This includes the decay part of the keying envelope and the time taken for relays to switch/pin diode minority carriers to decay and the AGC to recover).  A dot is generally taken as 1200/wpm ms, so for 60 wpm a dot is 20ms, if the rig recovers in say 10ms then full break in is just about possible.

What have I missed, and do these numbers look sane to the more serious CW operators out there?

Now the contentious bit is that given we are working on a non RT capable platform (At least for FLEX, I am using a proper DSP on bare metal), none of these numbers can really be guaranteed 100%, so what does the failure histogram look like, four nines, five nines, six nines? The software gets harder to write at an exponential rate and there is only so much development time.  

Having these numbers (or ones like them) for the Flex hardware/software would go a long way to actually quantifying the performance for the doubters.

Regards, Dan.

Logged
KE5JPP
Member

Posts: 0




Ignore
« Reply #46 on: August 01, 2011, 06:29:20 PM »


....you tried to use a piece of one of my post as an excuse for another one of your ridiculous childish Flex bashing post
Michael, W4HIJ


No Mikey, I am responding to the parts of your post that I can decipher and make any sense of.  Most of it is gibberish.  I remove or shorten those parts so that others don't have to wade through your painful gibberish all over again to discern some kind of meaning or point.

Funny you call it "Flex bashing" since I only re-posted what was reported by another Flex customer on Flex's mailing list.   Roll Eyes 

Gene
Logged
WB6RQN
Member

Posts: 484




Ignore
« Reply #47 on: August 01, 2011, 09:15:54 PM »

Sigh!
I don't have a horse in this race (If I want an SDR I am quite capable of rolling my own), but I thought this was about to get interesting when someone posted a link to actual measurements, you know, as in real data, but it appears that the ongoing pissing contest is more entertaining.....
Leaving the err "personalities" out of this can we have a conversation about what acceptable CW  performance means?

PLEASE! I would LOVE to know what constitutes good CW performance so we can measure it.

Quote
Can we start by actually trying to quantify what we mean by acceptable CW performance (I am working on a (mostly non SDR) rig of my own, so this would be useful information for when I pen the DSP code for it)?

Sidetone latency: < 10ms.
Sidetone latency jitter: <2ms.  This is I suspect the big one for having a radio feel right to a CW operator (Based on what gives music keyboard players fits).

That sounds sensible so far. I wonder if it would be possible to roll a simple keyer with the ability to fiddle with these parameters so that we can do statistical analysis of what people like and don't like?

Quote
These are trivially met in an embedded micro or DSP  as you do all the work on a sample by sample basis in the ISR, but are somewhere between harder and MUCH harder to meet on a general purpose OS (you really need to timestamp the key up and down events with the corresponding audio sample position, possibly do it in the micro running the AD and data link over USB or whatever (Hell, generate the sidetone there and pass it to the PC instead of the RX audio!)).
Note that neither of these say all that much about any latency from the key going down to the rig making RF, just about the key -> sidetone timing, so presumably we need to specify how accurately the RF envelope should track the key down time.

That should be possible to do by using a DSO triggering off the keying event and then measuring the envelope from the transmitter in order to avoid any latency in a receiver. Besides, the pulse would get stretched in the receiver IF filter.

Quote
RF Pulse duration accuracy: +-1ms (seems reasonable and is clearly doable, note that this is really time from the start of the keying envelope rise to the start of the decay, the actual envelope will of course be longer).

It is highly architecture dependent but for full break in a time for the key up to RX back at nominal sensitivity is surely important (This includes the decay part of the keying envelope and the time taken for relays to switch/pin diode minority carriers to decay and the AGC to recover).  A dot is generally taken as 1200/wpm ms, so for 60 wpm a dot is 20ms, if the rig recovers in say 10ms then full break in is just about possible.

What about keying latency? How long between the start of a key transition to a change in the RF envelope?

Quote
What have I missed, and do these numbers look sane to the more serious CW operators out there?

Now the contentious bit is that given we are working on a non RT capable platform (At least for FLEX, I am using a proper DSP on bare metal), none of these numbers can really be guaranteed 100%, so what does the failure histogram look like, four nines, five nines, six nines? The software gets harder to write at an exponential rate and there is only so much development time.  

Having these numbers (or ones like them) for the Flex hardware/software would go a long way to actually quantifying the performance for the doubters.

I agree 100%. Find out what the real numbers are and then let the chips fall where they may. I will be happy to run the tests for the various bits of Flex hardware. As of Wednesday I will have all three Flex radios at home, the 5000 with all the bells and whistles.

73 de Brian, WB6RQN/J79BPL
Logged
AK7V
Member

Posts: 250




Ignore
« Reply #48 on: August 02, 2011, 09:05:32 AM »

Seems to me you need a real, high speed, hardcore CW op to evaluate whether these radios "do" CW properly.  But those ops won't buy the radio unless they're convinced they can do CW on them. 

The failure histogram would be interesting to see.  I don't have an SDR, not really interested in getting one, but could imagine that a radio running on Windows could preform flawlessly for a while, but then stutter a bit and make sending jittery for even just a second.  Disk caching, antivirus starting, some transient OS bottleneck or delay... Someone would need to spend a decent chunk of time operating with it.  Someone who would notice and be bothered by those types of problems, while simultaneously being able to diagnose them.  Does anyone like that exist?

Also, I'm not sure there's much difference between what CW ops like and don't like.  Minimum time between key down and side tone, key down and RF, predictability, etc.  A "real time" experience.  Measure a standard, well regarded CW rig and aim for that. 
Logged
K0OD
Member

Posts: 2557




Ignore
« Reply #49 on: August 02, 2011, 10:29:01 AM »

I use my 5000 at 30+ wpm. Sometimes in contests. Once for a full weekend in the CQWW.

Only problem is the side tone "squelch tail" which is noticeable mostly on rarely-used six-meter CW. I was calling the ST0R last night on 40 and couldn't hear the side tone problem at all.

The squelch tail is usually very weak, sometimes non existent. I have a TS-850 on the desk next to the 5000 and I never use the old Kenwood anymore.
Logged
M0HCN
Member

Posts: 473




Ignore
« Reply #50 on: August 02, 2011, 10:50:19 AM »

The real question is, where is the upper limit on things before they become a noticeable issue to more then the very occasional CW operator, measuring a K2 or whatever does not tell us that.

Just figuring out which numbers matter for CW seems to be a hard problem, as in we do not seem to be sure what to measure, much less what the limits should be!

This is far more of an issue for the PC based SDR crowd then it is for people doing the DSP in an ADSPxxxx or TMS320 or whatever, as Windows just does not easily do very low latency, except under very specific conditions (Some pro audio manages reasonably low numbers, but it is still hardly sample by sample).  Knowing where the limits for acceptable performance for (90,95,98,99.5) % of operators are would really help the PC SDR crowd to know where to throw the development effort.

I strongly suspect that the latency jitter is FAR more of an issue then the actual latency is (provided the sidetone latency is in low 10ms range), but I only have evidence based on what musicians report for midi latency jitter, which while similar is hardly definitive for CW.

Interestingly Linux with the RT patches applied gives a scheduling latency of below 100us in some configurations, and around 400us for a stock kernel, so if Windows (where I could not find the figures) is not too far off this sort of number then scheduling latency should be below 1ms which implies that most of the latency problem stems from the implementation of the SDR not from the underlying OS.      

I think I am going to take this over to the CW forum, and see what those guys have to say.

Thought for the SDR software crowd: Increasing the amount of overlap between the FFT buffers will increase the CPU load, but cause the IO buffers to become smaller allowing lower control latency without compromising the FFT length, it might be a sane tradeoff.

Regards, Dan.

Logged
W4HIJ
Member

Posts: 367




Ignore
« Reply #51 on: August 02, 2011, 12:49:10 PM »


Interesting that when you quote me that you leave things out to suit your needs. Talk about spinning things!  In this case the very important part you left out was where I said "took me all of 20 minutes to do so" OK, the be honest maybe it was 10 minutes.  Cheesy So if I can do all the things you just listed in 10 to 20 minutes, I must be Superman!
But see, I didn't have to do any of that. Just adjusted some buffers. As usual you come up looking like you don't know jack about what you're talking about because you don't. BTW, never once put ferrites on anything with my 5K and now not with my 1500. Perhaps you guys need to check your ground systems??? Thanks for the laughs, keep the ridiculous post coming.... I need more humor in my life!!
Michael, W4HIJ

Mikey, these were all suggestions collected from the Flex mailing list, not my suggestions.  You really need to learn how to read carefully before posting and making a fool of yourself.   Roll Eyes Roll Eyes Roll Eyes

Gene

And Gene is right Michael. That is sometimes the way it is. OTOH, the difference is, most people don't have to do that. And when people do, most of the time it takes only one or two of those things to nail the problem. It is all a matter of scale. As I read it, Gene wants everyone to believe that is how it is for everyone, not just the occasional problem. My experience is very different -- that for most people, it just works and works well.

73 de Brian, WB6RQN/J79BPL

Yes Brian, SOMETIMES  But my experience and I suspect the majority of Flex owners experience is completely different. I'm not saying the problems don't exist for some folks, If they didn't we wouldn't see the complaints.  But, I suspect these problems affect a small number of the overall population of Flex users. Very few people take the time to post that they are having no problems. I do, if only to counter act  all the BS spewing from the keyboards of our two resident two year olds.  I've downgraded my opinion of their emotional maturity level because I realized what a disservice I was doing to my nine year old daughter by calling Stan a ten year old! As for Gene, he's demonstrated what level he is on by referring to me as "Mikey" another childish attempt to  try and get under my skin.  
 Anyway, the two village idiots aside,  I would be very interested to see what the test you propose bear out but I can only go by my personal real world experience with my 1500 in the last couple of weeks that I've been working CW with it. As already stated, I'm using a straight key and I have not had any of the issues that have been being discussed. Sidetone is fine, there is no discernible delay that I can hear.  Just like my RX skills, my sending skills were a bit rusty so I would think that if there were a delay that it would tend to mess me up more so than it would a more seasoned CW op.
Also as already stated, I don't use paddles and a keyer so I can't comment there and I don't have any antennas that will load up on six meters at the moment so I can't comment on the problem people are reportedly experiencing there. What I do know is that on 40 and 20 meters, things have been fine.
 I guess I must just be lucky that I'm not experiencing all these other issues. Perhaps I should be playing the Lotto more often! Cheesy
 I'll just follow along with the discussion of the technical aspects from now on as I've grown weary of sparring with the two year old twins. You and Dan continue on please....
73,
Michael, W4HIJ
 
« Last Edit: August 02, 2011, 12:56:30 PM by W4HIJ » Logged
M0HCN
Member

Posts: 473




Ignore
« Reply #52 on: August 02, 2011, 01:19:17 PM »

Apart from the rather "self selected" nature of the reports you get for problems, there is another issue with establishing what is going on here...

People differ, and what may just not be a problem for one can make something completely unusable for another, just the way it is, sensitivity to things like delay and delay jitter can (and do) vary by whole orders of magnitude between individuals (There was work done on this by the musical keyboard crowd), so actually finding values for this stuff is hard (And satisfying everyone needs massive over engineering for the 95% case). 

Now if I had more time, more money and more energy I might consider pulling a "sherwood engineering" on CW keying performance, complete with data tables and calibrated tools, which is really what needs doing here to see what is really going on, as it is my interest is mainly in transmitter performance as the low hanging engineering fruit on the bands these days (30dB below PEP is NOT a respectable 2 tone result, Motorola app notes were managing -35dB below ONE tone 20 years back (Equates to ~41dB below PEP)).

The only folks who really know what the frequency with which technical support is needed for Flex radios are Flex, and even if they were publishing that data it would not tell the whole story, but they are far from the only manufacturer to have major support pain with software systems (You dont want to see what MUSICIANS sometimes manage to perpetrate in terms of ways to break computer based hardware).

Regards, Dan.
Logged
KE5JPP
Member

Posts: 0




Ignore
« Reply #53 on: August 02, 2011, 05:08:12 PM »

As for Gene, he's demonstrated what level he is on by referring to me as "Mikey" another childish attempt to  try and get under my skin.  
 Anyway, the two village idiots aside,  
73,
Michael, W4HIJ
 

Mikey, I have removed most of your blathering and nonsensical reply.  Since you are now resorting to name calling it seems I have in fact gotten under your skin.  If you are so bothered by my responses to you, then please do both of us a favor and put me on ignore like you have claimed to do with Stan.  Roll Eyes Cheesy Roll Eyes

Gene
Logged
WB6RQN
Member

Posts: 484




Ignore
« Reply #54 on: August 02, 2011, 06:08:40 PM »

This is far more of an issue for the PC based SDR crowd then it is for people doing the DSP in an ADSPxxxx or TMS320 or whatever, as Windows just does not easily do very low latency, except under very specific conditions (Some pro audio manages reasonably low numbers, but it is still hardly sample by sample).  Knowing where the limits for acceptable performance for (90,95,98,99.5) % of operators are would really help the PC SDR crowd to know where to throw the development effort.

By default, Windows runs many things that are unnecessary. Most of these services can be turned off, e.g. automatic software update, network time update, etc. In addition I disable paging/swapping. Once you get rid of the unnecessary cruft, it appears that Windows actually is a credible OS. The best source of information for this are the gamers. They are trying to eke the last percent of processing power from a Windows system and are amazingly successful.

Quote
I strongly suspect that the latency jitter is FAR more of an issue then the actual latency is (provided the sidetone latency is in low 10ms range), but I only have evidence based on what musicians report for midi latency jitter, which while similar is hardly definitive for CW.

Interestingly Linux with the RT patches applied gives a scheduling latency of below 100us in some configurations, and around 400us for a stock kernel, so if Windows (where I could not find the figures) is not too far off this sort of number then scheduling latency should be below 1ms which implies that most of the latency problem stems from the implementation of the SDR not from the underlying OS.      

I think I am going to take this over to the CW forum, and see what those guys have to say.

Good idea, but I bet they haven't quantified things.

Quote
Thought for the SDR software crowd: Increasing the amount of overlap between the FFT buffers will increase the CPU load, but cause the IO buffers to become smaller allowing lower control latency without compromising the FFT length, it might be a sane tradeoff.

That is pretty standard for configuring PowerSDR to run CW. It is what I did when I was trying to do my QSK test on the bench. It worked.

73 de Brian, WB6RQN/J79BPL
Logged
KE5JPP
Member

Posts: 0




Ignore
« Reply #55 on: August 03, 2011, 03:32:47 AM »


By default, Windows runs many things that are unnecessary. Most of these services can be turned off, e.g. automatic software update, network time update, etc. In addition I disable paging/swapping. Once you get rid of the unnecessary cruft, it appears that Windows actually is a credible OS. The best source of information for this are the gamers. They are trying to eke the last percent of processing power from a Windows system and are amazingly successful.


If guys are going to dedicate a computer for use with their Flex and they are going to lock it down by removing or avoiding using many Windows applications and services, they need to take into account the cost of the dedicated computer and the time required to fiddle with stripping Windows down into the total cost of ownership of a Flex.  I bought a K3 and I just plugged it in and started operating.  I did not have to buy a dedicated computer, then go through the process of locking it down and stripping out unneeded services.

When you look at the $600-$700 cost of a Flex-1500 it seems cheap at first glance.  But if you are going to dedicate a computer to running your Flex-1500 in an attempt to get the best performance in CW, for example, you are actually spending more like $1000-$1200.  There are other good performing radios in that price range to look at too.

Before someone jumps in and says "I just hooked it up to a computer I already have and I use the computer for everything", please realize that you are lucky that is all you had to do.  Some guys aren't so lucky and end up having to fiddle a lot or purchase a dedicated computer.  Just keep that additional cost in mind when comparing total costs to other manufacturer's offerings.

Gene
Logged
KE5JPP
Member

Posts: 0




Ignore
« Reply #56 on: August 03, 2011, 03:44:01 AM »


Now if I had more time, more money and more energy I might consider pulling a "sherwood engineering" on CW keying performance, complete with data tables and calibrated tools, which is really what needs doing here to see what is really going on, as it is my interest is mainly in transmitter performance as the low hanging engineering fruit on the bands these days (30dB below PEP is NOT a respectable 2 tone result, Motorola app notes were managing -35dB below ONE tone 20 years back (Equates to ~41dB below PEP)).


While the Sherwood table is interesting, there is a danger of falling into "specsmanship" where a few numbers that most people are not going to ever be able to discern in everyday use are taken as the determining factor in which radio you are going to buy.  Sherwood has said that even though some radios are at the top of that list as far as numbers, some of those radios are ergonomic or user interface disasters and he would not want to own one.  The same thing can happen with a Sherwood-like table of CW performance numbers.    I am not sure that the numbers would tell you enough about the actual CW operator's experience with a particular radio.  I am guessing that what is subjective acceptable performance for one guy is going to be unacceptable to another.  I am not saying tables of measured numbers are not useful.  I think they are more useful to the designers of the hardware than the end users though.

Gene
Logged
M0HCN
Member

Posts: 473




Ignore
« Reply #57 on: August 03, 2011, 04:56:15 AM »

Indeed, no table of numbers catches things like critical controls buried 5 layers down in a menu (HT makers you know who you are (all of you))!

Any yea, optimizing for a single parameter is almost always a mistake, but I am mainly interested in trying to find out which numbers actually matter as clearly some make far more difference then others.

Buying a radio because it is at the top of that table (which is mainly  concerned with close in IMD) is poor decision making but that table is a very useful benchmark for those parameters for those of us designing and building our own rigs.

The actual point at which any of the numbers become 'good enough' will vary widely between individuals of course, that is a given, so  there is no absolute line in the sand (Hell, I have been using an old FT 101-DZ while I build my dream radio, the thing is pathetic by modern standards, but you can still have fun).

You don't necessarily have to agree with Mr Sherwoods particular set of priorities (I find his Drake obsession a little strange), but the data is undeniably real at least for the samples he tested, and is useful even if it cannot be used as a sole decider of the 'which rig' question.

Regards, Dan.


Logged
W4HIJ
Member

Posts: 367




Ignore
« Reply #58 on: August 03, 2011, 12:49:10 PM »

Just as a point of information, there is an application out there that the gamers use called G-Boost that makes it possible to shut down a lot of the unnecessary stuff that Windows runs. Just takes a click of the mouse button and another when one wants to restore the PC to "normal" use. Neal recommends it and I have been using it before running PowerSDR. I guess I'm one of the "lucky" guys though in that I really don't even have to do that to use my Flex with my "do everything" PC. I run G-Boost more as a precautionary measure than anything else.
Michael, W4HIJ
Logged
WB6RQN
Member

Posts: 484




Ignore
« Reply #59 on: August 03, 2011, 02:58:34 PM »

If guys are going to dedicate a computer for use with their Flex and they are going to lock it down by removing or avoiding using many Windows applications and services, they need to take into account the cost of the dedicated computer and the time required to fiddle with stripping Windows down into the total cost of ownership of a Flex.  I bought a K3 and I just plugged it in and started operating.  I did not have to buy a dedicated computer, then go through the process of locking it down and stripping out unneeded services.

I would say that is a fair generalization. In my case, not so much because I run a dual-boot machine. On the windows side I do strip out the unnecessary stuff. In fact, I have found that to be a generally good course of action for all the windows systems I use and/or maintain.

And I like the performance I get from the Flex vs. the K3, dedicated computer or no. The trade off seems quite reasonable to me and I am not sorry I made it.

Quote
When you look at the $600-$700 cost of a Flex-1500 it seems cheap at first glance.  But if you are going to dedicate a computer to running your Flex-1500 in an attempt to get the best performance in CW, for example, you are actually spending more like $1000-$1200.  There are other good performing radios in that price range to look at too.

What you say is true. It ends up being a matter of choice, I think.

Quote
Before someone jumps in and says "I just hooked it up to a computer I already have and I use the computer for everything", please realize that you are lucky that is all you had to do.  Some guys aren't so lucky and end up having to fiddle a lot or purchase a dedicated computer.  Just keep that additional cost in mind when comparing total costs to other manufacturer's offerings.

The surprising thing, (or maybe not so surprising) is that just hooking up a computer and having it work while you use it for everything, seems to work more often than not. But it does require the owner/operator to know that it is possible that a problem can occur. It is one of the things a user needs to know in order to make an informed decision.

73 de Brian, WB6RQN/J79BPL
Logged
Pages: Prev 1 2 3 [4] 5 Next   Go Up
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.11 | SMF © 2006-2009, Simple Machines LLC Valid XHTML 1.0! Valid CSS!