Time stamp ruing plots!
in [Matlab]
|
Prev: Zero-Crossings
Next: can someone help me with this
From: Travis Knepp on 17 Jun 2010 16:51 TideMan <mulgor(a)gmail.com> wrote in message <fa89084b-ad05-4906-9362-c7aa64abbc35(a)h13g2000yqm.googlegroups.com>... > On Jun 17, 8:31 am, dpb <n...(a)non.net> wrote: > > TideMan wrote: > > > > ... > > > > > Well, from what I can see, what you've given us is totally irrelevant. > > > How can we possibly discern what you are doing wrong from this? > > > > ... > > > > Err..., crystal ball in shop yet again, is it??? :) > > > > -- > > You're right. > The bloody thing keeps malfunctioning! Ok, ok. My apologies for leaving the details lacking...let's not ruffle our panties. In gps_time I store my date/time in format "HHMMSSddmmyy", the previous guy just stored "HHMMSS". If I plot surf(gps_time,Y,Z) I get something horrendous, but if I remove the date information from gps_time the figure improves, if I just use decimal hour (e.g. 7.45) it gets better yet. Below is how I convert my time to decimal hour, followed by how my predecessor got his. My Time: temp_time = num2str(gps_time(1,oo)); temp_time = datenum(temp_time,'HHMMSSddmmyy'); hour = str2double(datestr(temp_time,'HH')); min = str2double(datestr(temp_time,'MM'))/60; sec = str2double(datestr(temp_time,'SS'))/3600; gps_time(1,oo) = hour + min + sec; I was converting gps_time (again, in 'HHMMSSddmmyy') to a number via datenum(gps_time), which allows me to keep both date and time information, then plotting surf(gps_time,X,Y) with the x-axis units displayed using datetick, but, again, the figures were horrible. His Conversion: Time_UTC{oo}=num2str(gps_time(oo)); temp3 = char(Time_UTC{oo}); gps_time(1,oo)=str2double(temp3(5:6))/3600+str2double(temp3(3:4))/60+... str2double(temp3(1:2)); It seems that when I use less temporal information (e.g. decimal hour vs. date/time via datetick) the figure becomes better (higher graphical resolution), but that makes no sense. I've been through the code a dozen times and can't imagine what is going on to cause the error; I've never encountered anything like it. I'm not terribly experienced, but it seems like everything should be fine regardless of how the time is represented. Again, apologies for lack of detail in original post. I was trying to be brief, and ended up being a pain. Just in case though, for my next post I'll see if I can find some good crystal balls.
From: dpb on 17 Jun 2010 17:02 Travis Knepp wrote: .... > I was converting gps_time (again, in 'HHMMSSddmmyy') to a number via > datenum(gps_time), which allows me to keep both date and time > information, then plotting surf(gps_time,X,Y) with the x-axis units > displayed using datetick, but, again, the figures were horrible. > His Conversion: Time_UTC{oo}=num2str(gps_time(oo)); > temp3 = char(Time_UTC{oo}); > gps_time(1,oo)=str2double(temp3(5:6))/3600+str2double(temp3(3:4))/60+... > str2double(temp3(1:2)); > > It seems that when I use less temporal information (e.g. decimal hour > vs. date/time via datetick) the figure becomes better (higher graphical > resolution), but that makes no sense. ... Think would need an example to cut 'n paste that demonstrates what you don't like vs what you do but... I've not used datetick() much at all, but what I have left me w/ the impression that it generally tries to put too much text on the axis if the number of ticks are very many at all. I'd suggest manually setting the tick values to datenums from beginning to end w/ some number of intermediate but probably not more than 10 or so and work from there. I think from your description it's probably purely a presentation problem of trying to put too much stuff on the tick labels when the date string includes dates as well as time that ends up w/ a lot of text. Oh, reducing font size and going to something like Arial Narrow as a font might help, too. --
From: dpb on 17 Jun 2010 17:09 dpb wrote: .... > I'd suggest manually setting the tick values to datenums from beginning > to end w/ some number of intermediate but probably not more than 10 or > so and work from there. .... Meant "5 or so", more like 10 is where it tends to start when left untended, I think... --
From: TideMan on 17 Jun 2010 17:42
On Jun 18, 8:51 am, "Travis Knepp" <notpartofaddress.travis.n.kn...(a)nasa.gov.removenotpartofaddress> wrote: > TideMan <mul...(a)gmail.com> wrote in message <fa89084b-ad05-4906-9362-c7aa64abb...(a)h13g2000yqm.googlegroups.com>... > > On Jun 17, 8:31 am, dpb <n...(a)non.net> wrote: > > > TideMan wrote: > > > > ... > > > > > Well, from what I can see, what you've given us is totally irrelevant. > > > > How can we possibly discern what you are doing wrong from this? > > > > ... > > > > Err..., crystal ball in shop yet again, is it??? :) > > > > -- > > > You're right. > > The bloody thing keeps malfunctioning! > > Ok, ok. My apologies for leaving the details lacking...let's not ruffle our panties. > In gps_time I store my date/time in format "HHMMSSddmmyy", the previous guy just stored "HHMMSS". If I plot surf(gps_time,Y,Z) I get something horrendous, but if I remove the date information from gps_time the figure improves, if I just use decimal hour (e.g. 7.45) it gets better yet. Below is how I convert my time to decimal hour, followed by how my predecessor got his. > > My Time: > temp_time = num2str(gps_time(1,oo)); > temp_time = datenum(temp_time,'HHMMSSddmmyy'); > hour = str2double(datestr(temp_time,'HH')); > min = str2double(datestr(temp_time,'MM'))/60; > sec = str2double(datestr(temp_time,'SS'))/3600; > > gps_time(1,oo) = hour + min + sec; > > I was converting gps_time (again, in 'HHMMSSddmmyy') to a number via datenum(gps_time), which allows me to keep both date and time information, then plotting surf(gps_time,X,Y) with the x-axis units displayed using datetick, but, again, the figures were horrible. > > His Conversion: > Time_UTC{oo}=num2str(gps_time(oo)); > temp3 = char(Time_UTC{oo}); > gps_time(1,oo)=str2double(temp3(5:6))/3600+str2double(temp3(3:4))/60+... > str2double(temp3(1:2)); > > It seems that when I use less temporal information (e.g. decimal hour vs. date/time via datetick) the figure becomes better (higher graphical resolution), but that makes no sense. I've been through the code a dozen times and can't imagine what is going on to cause the error; I've never encountered anything like it. I'm not terribly experienced, but it seems like everything should be fine regardless of how the time is represented. > > Again, apologies for lack of detail in original post. I was trying to be brief, and ended up being a pain. Just in case though, for my next post I'll see if I can find some good crystal balls. I'm a bit confused.............. In your original post, you implied that you were not happy with the way your algorithm converted to Matlab days, but from this post it appears that's not the problem at all. As dpb says, it looks like it is just a presentation problem. I do lots of plots with time on the x-axis, but I vary the format depending upon the time span. Rather than frigging around with converting hour, min, sec why not just use Matlab days (i.e., gps_time from above) and play around with the format option in datetick? If you type help datetick, you will see lots of options. One thing that I do sometimes after datetick('x',1) which produces labels as dd-mmm-yyyy, is: xlab=get(gca,'XTickLabel'); then remove the -yyyy part from each label except the one at the right hand end, converting the string matrix into a cell array cxlab in the process (necessary because not all the labels are the same size). Then set(gca,'XTickLab',cxlab) There are many ways of solving the problem, these are just some suggestions. |