Show small pressure changes on a fixed-scale HMI trend

Why does the trend look flat when the pressure really changes?

Because the Y axis must cover the full 20 to 60 psi batch range, and on a chart with 480 vertical pixels that span gives about 0.083 psi per pixel, so a tenth-of-psi difference between parts moves the line about one pixel. The value axis maps pressure onto chart height: psi per pixel = (axis maximum − axis minimum) ÷ plot height in pixels. With a 20 to 60 psi axis over a 480-pixel plot, that is 40 ÷ 480 = 0.083 psi per pixel. A 0.1 psi step is 0.1 ÷ 0.083 = 1.2 pixels. A hundredth-of-psi step is an eighth of a pixel, below what any panel can draw. So the trace reads as a straight line while the logged values move. One pixel is the smallest step the hardware can render; why an HMI trend graph only resolves 1% of full scale works through that limit for a single graph.

Check the log before you touch the display. If the logged values move and only the trace is flat, this is a scale problem and the rest of this page applies. If the logged values are flat too, the fault sits upstream, in the PLC value, the scaling or the logging path, and why the scale value differs from the PLC analog input covers that comparison.

Can the Y axis adjust itself to each batch?

Not on this trend object: its value axis is fixed where you design the screen, with no automatic ranging, so the workable fix is two or more fixed-scale screens and an operator who picks the right one for the batch. Open the trend object in the screen editor and look once at the value-axis settings for an automatic or tag-driven range option. Some trend controls have one; if yours does, write the endpoints from the PLC and you are finished. On the trend objects behind this fix there is no such option, and axis configuration sits behind a login with configuration rights, so plan on the engineering account. Fixed screens then do the same job by hand: one broad screen for context, one narrow screen per batch range. Whichever screen is up, the trigger keeps logging the same values.

What spans should the fixed screens cover?

Use a 40 psi context screen from 20 to 60 psi for batch-to-batch review, and a 2 psi detail screen centered on the calibrated value of the running batch, which keeps the pass/fail region in view while tenths-of-psi steps become obvious. Pick the endpoints from data. Pull the log for several batches, note the lowest and highest values across batches, here 20 and 60 psi, and the spread within one batch, here tenths of a psi. The context screen takes 20 to 60. Each detail screen takes a 2 psi window centered on its calibrated value: a batch calibrated at 42.0 psi gets 41.5 to 43.5 psi.

Keep the pass/fail value inside the detail window. A clipped limit hides the decision boundary, which is the one thing this application cannot lose, since pass and fail sit close together. Do not go narrower than the data needs either: if sensor noise is 0.02 psi, a 0.5 psi window draws that noise 19 pixels tall and operators will chase it. And leave the process alone: do not move the pass/fail limit or the calibration to make a trace look taller. That changes what the process accepts. A scale change changes only the drawing.

How small a difference can one span show?

Divide the span by the plot height in pixels, then give the difference you must read at least 10 pixels: with 480 pixels, keep the span under about 4.8 psi to read a tenth of a psi, and under about 0.5 psi to read a hundredth. The rule in one line: readable span = smallest difference of interest × 48, for a 480-pixel plot. One or two pixels sits at the visual noise floor of a panel read at arm's length; ten pixels is a clear step. The table shows what that means for the three spans this application uses.

The 40 psi span cannot show tenths. The 2 psi span shows tenths well and hundredths barely. The 0.5 psi span shows hundredths. The 480-pixel figure is a full-screen trend on an 800 x 480 panel; measure your own plot area and redo the division, because a shorter plot area shortens every step in proportion.

What do I need before building the screens?

Five things: logged values from representative batches, the calibrated value and pass/fail limit for each batch type, the trend object's value-axis settings in the screen editor, the plot height in pixels, and a login with configuration rights. In detail:

• Logged pressure values from representative batches, with the lowest and highest value per batch written down.
• The calibrated value and the pass/fail limit for every batch type you run.
• The value-axis settings of the trend object in the screen editor, and the account that is allowed to change them.
• The plot height of the trend as drawn on the panel, measured in pixels.
• A labeling plan: every screen states its span in psi, and the operator has buttons or function keys to switch.

How do I build the fixed-scale trend screens?

Duplicate the existing trend screen once per span, set the value-axis minimum and maximum on each copy, label every screen with its span in psi, and give the operator screen-switch buttons so the right scale is up for the batch.

1. Pull the logged values for a few representative batches and write down the lowest and highest value per batch. Across batches here that is 20 and 60 psi.
2. List every batch type with its calibrated value and its pass/fail limit.
3. On the existing trend screen, open the value-axis settings and set the minimum to 20 and the maximum to 60. Label the screen as the 20 to 60 psi batch context view.
4. Copy that screen once per calibrated value. On each copy, set the minimum and maximum 2 psi apart, centered on that value: a 42.0 psi batch gets 41.5 and 43.5.
5. If the pressure tag carries implied decimals, an integer that counts hundredths, enter the axis limits in the tag's own counts; setting trend graph limits for implied-decimal PLC values shows the conversion.
6. Add a fixed horizontal line at the pass/fail value: its height from the bottom of the plot is (limit − minimum) ÷ span × plot height.
7. Label every screen with its span and add buttons or function keys so the operator can switch without leaving the trend.
8. Save and download the project through your normal change-control procedure, then run the checks below.

How does an example batch at 42 psi look on each screen?

On the 20 to 60 psi screen a 0.3 psi spread is 3.6 pixels, under 1 percent of the chart; on the 41.5 to 43.5 psi screen the same spread is 72 pixels, 15 percent of the chart, and a tenth-of-psi step is 24 pixels. Take a batch calibrated at 42.0 psi, with 30 parts reading 41.9 to 42.2 psi and a pass/fail limit of 42.5 psi.

Context screen, 20 to 60 psi: 40 ÷ 480 = 0.083 psi per pixel. The 0.3 psi spread is 0.3 × 480 ÷ 40 = 3.6 pixels, which is 0.75 percent of the chart height. The trace is a line.

Detail screen, 41.5 to 43.5 psi: 2 ÷ 480 = 0.0042 psi per pixel. The spread is 0.3 × 480 ÷ 2 = 72 pixels, 15 percent of the chart. Two parts a tenth of a psi apart stand 0.1 × 480 ÷ 2 = 24 pixels apart. The limit line sits at (42.5 − 41.5) ÷ 2 × 480 = 240 pixels from the bottom, half way up, with the whole spread in view below it.

Same log, same values, two readable pictures. The narrow screen answers part-to-part questions, the broad screen answers batch-to-batch questions.

What about the uneven time between triggered readings?

That is the other axis, and a narrower Y span does nothing to it: triggered samples can look evenly spaced on screen while the logged times are minutes apart, so take elapsed time from the log's timestamps, not from the trace. The horizontal spacing follows the trend object's time-axis behavior. Some trend views plot each point at its timestamp; others space samples evenly by index, which makes irregular triggers look regular. Open the time-axis settings of your trend object and check which behavior you have; some trend views also offer a time-axis mode you can switch between a time span and a tag or a constant. For anything decided by elapsed time, read the timestamps in the log. To show a long run on the screen without scrolling, fitting a multi-hour trend on an HMI without scrolling covers the time-axis side.

How do I check the new screens work?

Run one controlled batch or force known values across the span, then confirm three things: the PLC value and the logged value agree, the detail screen shows a tenth-of-psi step clearly without clipping the batch spread, and switching screens leaves the log recording the same values.

• Force values across the whole detail span, including one just inside and one just outside the pass/fail limit. Both must stay in view on the detail screen.
• Compare the PLC value with the logged value at the tag's resolution.
• Confirm a 0.1 psi step draws at least 10 pixels on the detail screen. At 2 psi over 480 pixels you should measure about 24.
• Confirm the full within-batch spread stays on screen, with no clipping at the top or bottom.
• Flip between screens while a value changes and confirm the log keeps recording.
• Note which scale was active when a screenshot or printout is filed, so a reviewer reading a tall trace knows it was the 2 psi window.

Recommendation: Run one context screen at 20 to 60 psi and one 2 psi detail screen per calibrated value, with the operator choosing the detail screen that matches the batch. Where a decision turns on hundredths of a psi, add a 0.5 psi screen for that check and read the number from the log. Keep the broad screen on the floor for batch-to-batch review, and leave the pass/fail limits and the calibration exactly where they are.

Questions