Is it possible using tiledlayout to have one set of tiles with shared tight y-limits and other tiles with much larger y-limits without losing the zoomed view of tiles with tight y-limits?

I am trying to create a figure with one set of time-series using

ylimit = [-100,100];

and another few time-series with the automatically-chosen y-limits, which are different by several orders of magnitude. Each plot shares the same x-axis limits. I want to really zoom in on the dynamics in the first set of time-series. That is the reason for the static y-limits. However, when I add the time-series which have much larger y-limits, the first set of time-series appears to zoom out and the dynamics are very difficult to see.

hf=figure; clf
chans = 16;
t = tiledlayout(chans,1,'TileSpacing','Compact','padding','compact'); %setup tile for all subplots
cmap = colormap; %get current colormap
allColors = cmap(chans:chans:chans*chans,:); %split colormap into diff colors per chan
ylimit = [-100,100]; %set y-limit for zoomed-in view of dynamics
for subplotN=1:chans-3 %plot first set of time-series
    nexttile
    x = 1:2841;
    y = randn(1,2841)*50;
    plot(x,y,'LineWidth',2,'Color',allColors(subplotN,:)) 
    ylim(ylimit)
end

This code yields the following figure. Looks decent, I can see the dynamics:

Now, I add the next time-series with a ylimit that is several orders of magnitude different

nexttile
y1 = randn(1,2841)*1e6;
plot(x,y1,'LineWidth',2,'Color','b');

This causes the previous traces to zoom out so it's way harder to see their dynamics:

How can I fix this? I want the previous time-series to stay zoomed-in as I add the remaining tiles with significantly larger ylimits.

Thank you in advance for any help.

Answer 1

[Cris LaPierre]

To me, it appears the previous timeseries YAxis Limits have not changed (i.e. it is still "zoomed in"). I think what has happened is that, by adding a plot requiring an exponent to capture the scale of the Y values (x10^6), the actual height of each plot has also shrunk to to accomodate an exponent. The fix is to find a way to plot the data without adding the exponent for the larger scale variables.

One approach may be to divide by whatever the explonent is.

nexttile
y1 = randn(1,2841)*1e6;
% find exponent automatically and divide y1 by it when plotting
x10=floor(log10(max(y1)))
plot(x,y1/10^x10,'LineWidth',2,'Color','b')

I wonder if stackedplot might be a better option for you. You'd have to modify your code a little, and wouldn't be able to color each line (at least not easily), but it does a better job keeping the y-axes as tall as possible.

figure;
chans = 16;
for subplotN=1:chans-3 %plot first set of time-series
    x = 1:2841;
    y(:,subplotN) = randn(2841,1)*50;
end

y(:,subplotN+1) = randn(1,2841)*1e6;

stackedplot(x,y)

[Bryan Conklin]

Thanks Cris. I was able to get what I needed using your suggestion about dividing to scale the image down.

The stackedplot render actually provides a much higher resolution view of the dynamics. However, I don't know if I can get it to look the way I need. See image below for what I was able to do with subplots. The key at the bottom seems the trickiest. I used primitive lines and text commands to make it with subplot. Do you think this figure is possible with stackedplot?

[Cris LaPierre]

I think there is probably a way to do it, but I'm also sure it would take some effort to figure it all out. This is definitely not an area of expertise of mine, though.

[Bryan Conklin]

Thanks Cris. I'll start another thread trying to figure out how to configure stackedplot to my specification.

[Cris LaPierre]

Another thought. Another althernative might be forcing the exponent to be 0. Perhaps that is more desireable than dividing by 10e^10, especially if you actually don't know what the maximum magnitude to be encountered will be.

ax = gca;
ax.YAxis.Exponent = 0;

You can then use ytickformat to set the formatting if you want.