Tcalibration FAQ

Understanding the Terms

How do I convert percentage error to voltage error? Multiply the percentage error by the span and then divide by 100. Multiplying by the span means it is no longer a fraction of the span. Dividing by 100 means it is no longer a percentage. Note, the span is twice the peak value.

How do I convert from percentage of span to percentage for a particular voltage? Multiply by span and divide by actual voltage. Multiplying by the span means it is not longer a fraction of the span. Dividing by the actual voltage then makes it a fraction of the actual voltage.

What can be positive or negative? Adjustment values (c and m), Spot Errors and the DC Offset can be positive or negative. Other features such as Maximum Error, Linearity and Inaccuracy are forced to be positive (the absolute value is used) to make comparison easier

Are DC Offset and c the same? DC Offset is an error so has RMS Noise error and experimental error (during calibration) added. c is an adjustment value so does not have errors added to it because it is not appropriate (the errors relevant to c are instead built into the Maximum Error). Otherwise DC Offset and c are the same.

How can the Maximum Error be converted to an absolute voltage automatically by the software without worrying about the number of significant figures the calculation is known to? Because the Maximum Error has already had calibration error added to it. So, error has been incorporated in an absolute way already.

Why is the formula for adjustments (reading -c)/m? The c and m values describe the relationship between the reading (on the y axis) and the input voltage (on the x axis). Therefore, to recover the original input voltage from the reading it is necessary to undo the equation of the line of best fit. For example, if the c value is +0.02V then obviously you need to take that away from the reading to get closer the the input voltage. The same logic applies to the m value meaning we divide. And it is necessary to minus c first because to undo it (using the inverse function) the operations have to be applied in the opposite order.

What kind of lead should I use? I look to use RG-58 cables for testing due to the low-noise I find they deliver.

What is a 5-point calibration? What other ones are there? A 1-point calibration typically tests with a positive voltage equal to 80% of the maximum possible to record. A 2-point calibration typically also tests the response to a zero voltage input. A 3-point typically also tests with a negative value: The whole range will have now been tested and a multiplier and offset can be calculated. A 5-point adds two extra points at 40% and -40% (again, picking typical values): this might seem to be a waste of time but crucially can pick up on the surprisingly common situation where the response is an S-shape that is accurate at both ends and in the middle of the range but inaccurate away from there. Sometimes a 9-point calibration is done but this would only pick up on some very rare errors and in my experience is more often used for calibrating based on a series of manual inputs such as temperature or pressure where the inputs themselves are a bit unpredictable.

Is there a training option available for the Tcalibration sofware? If you or your organisation have been supplied with the Tcalibration software then you can ask for the sample data set which is for an instrument (with some problems to make the data interesting) with +/-10V, +/-1V, +/-0.1V and +/-0.01V ranges and 8 channels. Voltage response and timebase data are included so you can create a full calibration certificate. You can then compare your results with how the calibration certificate should look. Make sure you have both the sample data and the model calibration certificates and reports. Note, the spreadsheets used for analysing the sample data were version 1.27 with a couple of minor text differences in some of the operator instructions and some of the cells with numbers relating to the granularity error are one positioned one row higher than in the published version. If you are using a later version the Linearity, DC Offset, Inaccuracy, RMS Noise and timebase calculations are all backwards-compatible so you can still compare your answers. From the software versions 1.28 onwards there are updated example error values so they will be different.

A time-efficient approach to the DC response spreadsheet. Depending on your instrument this can take up to 5 minutes to fill out. This is a typical time to spend creating a thorough audit trail for a calibration. You can speed this aspect up by filling out all the details for one range and then just copying this spreadsheet, renaming it and making minimal adjustments for each of the other ranges.

Why doesn't the software allow each channel to be potentially loaded from a different file? Because it would add a big layer of complexity to the range spreadsheet. It was a judgement call based on the amount it would be needed and the complexity it would add. If it is an issue for you do let me know. In my experience new instruments are increasingly making it easier to save all channels at once so I believe this is less and less of an issue.

If you have more than 64 channels. (Simple answer.) This is relatively uncommon in these instruments so to make the spreadsheets smaller and run faster the maximum in one calibration certificate spreadsheet is 64. If you have more than 64 channels you can use additional calibration certificate spreadsheets. The "Transfer from Range Files" has yellow boxes near the top that override the default channel names and will enable you to renumber them. A time-saving approach is to set this up for a given instrument and then keep using (a copy of) the same preprepared spreadsheet every time you calibrate one.

If you have more than 64 channels (or channels split between files). (Advanced answer.) There are two ways of combining data if you find the data has to initially be loaded into more than one spreadsheet per range. You can either manually copy data into one spreadsheet (instructions in the range spreadsheet) or you can load more than one file into the overview (this is generally easier with just one downside being that the average adjustment values will not be done over the whole instrument).

How to find out more about a cell. Many cells, especially those for operator input, are laid out vertically. By reading from top-to-bottom from cell-to-cell it is presented in a logical order, with extra information, advice and even some calculations are shown unnecessarily just to make it clearer to the user what is happening. To understand an input cell: the writing above it and the writing to the right-hand-side offer help with deciding what to enter.

Why are different things on the spreadsheet yellow, green, grey and purple? The yellow cells are for the operator to fill in. The green cells are calculated automatically and should not be changed. The grey (often italic) text is calculated automatically and part of more intelligent aspects of the code such as filtering results so it is meant to be ignored except perhaps by some advanced users. The green cells are more obvious because they have numbers or text that can make sense even to a non-programmer. Some things are purple and this is just to highlight something to make it stand-out but has no special meaning.

Cell A31 onwards in the "Summary Data" tab of the DC voltage spreadsheet is on the topic of: "Error to add due to RMS Noise preventing accurate measurement." Other errors are added too. Is this correct? A single error bar is added to each result to reflect the fact the error may be larger as the calibration process cannot be 100% accurate. The error bar is the unreliability in the measurement made during the calibration and in many cases calibration will be done precisely enough that this will effectively be zero (as it is lost in the rounding).

Why do I see an error of 0.0034 in one box and then 0.00 in the next box? Is the number missing? This can happen due to rounding. This happens most often with the bright green boxes that act as totals or final figures based on calculations above it. This final figure will be rounded to two decimal places and so may seem to be zero. The actual number transferred to later calculations is the non-zero value you just can't see it. The decision to do this was made because the final results are to two decimal places: this was always going to look strange so it was decided to have this confusion somewhere that it was easier to see rather than at the moment it gets transferred to another tab.

Why are the RMS Noise for val1, val2, val4 and val5 by default removed from the analysis? This is so that the RMS Noise is from the zero volt/earth/ground measurements. For reasons discussed in the RMS Noise and Ground article this is a good way of representing the RMS Noise of the instrument.

Linearity, Inaccuracy, DC Offset and the Maximum Errors in the calculation tabs do not include experimental (calibration) error. This is why you can find the same thing with a different answer in the final results. The benefit of only adding the experimental (calibration) error later is that comparisons of the results during calculations are easier and there is a clarity that all results get experimental (calibration) error added at the same time. (c and m do not get experimental error added at all.)

Printing

Printable areas. These spreadsheets have some tabs with areas pre-formatted for printing. Options have been set to make printing an easy process on most computers (but Excel can sometimes mess with the formatting, for example trying to print onto more than one page). Your computer may override some of these settings. Each tab should print to a single page with the content centred. If this does not work for you then find the "page setup" options (often accessible from the page that appears when you start printing) and confirm that it is set to "fit to 1 page(s) wide by 1 page(s) tall" and the margins are set to "Center on page Horizonally Vertically". It is set to have no header nor footer. Additionally, I have set the spreadsheets up printing to A4 so you may need to change that or your computer may adapt automatically.

Analysis/Report Writing

Adding value to your calibration report. Only so much can be automated in the software. Here are some examples of higher-level analysis I like to do for my calibration reports. (1) My preferred approach is to exclude any channels with errors from the overall analysis sheet. I then add a cover sheet with details about the performance of those excluded channels. (2) An approach I use where there are a number of instruments of the same type for one customer is to do analysis of each instrument separately. Then combine them all into overall results sheet for them all with a full list of all excluded channels from all instruments. (3) If a channel is faulty you may choose not to input a signal to it, particularly if it is causing problems on the other channels. In this case any results shown by Tcalibration for that channel are just based on an open circuit and not real. It is recommended you manually delete them from the output sheet before printing. (4) When doing analysis for a calibration report remember the tab on the voltage response spreadsheet that allows you to look visually at the data points for a given channel.

What if you've tried everything and the results look wrong? There are obviously millions of possible mistakes. Here are some of the more common ones. (1) Is the problem only with one (or some) of the channels? In this case the channel may be faulty or the lead you used for that channel may be faulty. I recommend retesting (or at least looking at) either the whole instrument with leads switched-over/changed or retesting the one channel. (2) Are you sure you are not mixing up the peak value and the span? These are relevant in a number of places during the calibration process. If you have made a mistake you may need to recollect the data entering the correct test voltages. (3) Sometimes an instrument automatically changes the units it uses to save the data depending on the range/values being collected. In this case you may need to change the peak value to the correct units on the "LoadData" tab of the range spreadsheet. Looking directly at the raw data file with a text reader (or maybe opening it directly into Excel) can help you diagnose this issue. (4) There are a lot of yellow cells to enter data into. Many do not need changing but perhaps you accidentally changed one of them without realising it or something similar. It does not take very long to open a fresh spreadsheet and try again. This will often fix a problem. (5) Sometimes the operator accidentally collects data before changing the input voltage meaning a series of voltages such as -8V, -4V, 0V, 4V, 8V is entered where one of the voltages is a repeat of the previous one. You can look at the raw data to look for this issue. Generally this creates a high Inaccuracy error for all channels. (6) I find that if I am using this all day I can go into auto-pilot and one of the first things I forget to do is change the root directory for the data. This is worth checking if the data looks wrong. (7) Some instruments will automatically save a new file every time you even open a preview of the data in the control software. In this case sometimes you will hit preview early, perhaps before the voltage source has stabilised or before you remembered to change the voltage. In this case the filename or folder name may not be sequential. Sometimes in this case I forget that for one or more ranges there was extra data stored that I need to ignore during analysis.

What if there is a high error on a channel and/or range? You might choose to retest and see if the same error appears. This is what I usually do if it is a new error I have not seen during a previous calibration with the same instrument. Alternatively it might be a predictable error, such as steadily higher error as the voltage range decreases and this might not need re-checking. The "valPlot" tab on the range spreadsheet can be useful for investigating high errors: using the numbers below the graphs can help and note that some of these are percentages of the span and some are absolute voltage which relates better to the graph. It should be possible to pick out the absolute voltage values on the graph, as shown by the blue and red markers.

How did you combine all the calibration certificate pages into one file? I use PDFMerge. This is software for Windows. It enables you to drag-and-drop PDF files into a list and then produce a new PDF made up of all the files in the list.

Can the Maximum Error (using only the c adjustment) be greater than the Maximum Error (using both c and m adjustments)? Yes it can. The Maximum Error finds the differences between the measured values and the input values. The c and m adjustments are based on minimising the squares of the differences (a least squares approach is used). You might ask why use least squares if this happens: the overall trend is more important than choosing the largest error and trying to make it as small as possible.