Troubleshooting circuits is difficult enough when the design domain is pure digital or pure analog, but when the system bridges analog-to-digital the effort that is required to find error sources in the circuit is significantly complicated by virtue of the mixed signal environment.

Microchip Technology has introduced the MXDEV1 Analog Evaluation System (hardware) in combination with the MXLAB Analog Development Environment (software) to assist designers with the evaluation of stand-alone ADCs. The software suite (MXLAB) allows a designer to look at analog-to-digital conversion data from the time, frequency or statistical perspectives. These evaluations can be done real-time as well as at the conclusion of a data collection session. Since the hardware (MXDEV1) includes a well-laid out board with the converter and a breadboard area, this type of evaluation tool can be used to a significant advantage when designing mixed signal circuits.

But the hidden value of this evaluation tool is not only in the hardware. It also resides in the software. The software will work off-line by accepting conversion data from an ASCII file. With this flexibility the designer can troubleshoot his own circuit, provided that the data from his board can be captured.

This article will describe the features of the MXLAB software tool using data that was collected during the development of the MXDEV1 hardware. Details about the MXLAB histogram graph, FFT representation, scope plot and the data display mode will be discussed. Following these troubleshooting examples, you will use the MXLAB tools to identify and describe two different signals that were collected from the MXDEV1 boards.

Although the details of this software will be explained in this article, the most effective way to understand the operation of this tool is to download it to your PC. You can find this evaluation software on Microchip's web site at http://www.microchip.com/10/tools/Analog/Eval/51226a/index.htm. The file that you should download now is:

51226a.zip

After you have downloaded this file, unzip it and install the software by invoking "setup.exe".

Evaluating Conversion Noise Problems
Reference or Input Signal Path?

During the development of the MXDEV1 hardware, it was noticed that Microchip's MCP3208, 12-bit converter was not converting a single code for the dc input from a potentiometer. The circuit that was used when this problem was found is shown in Fig. 1. The data that was collected from this circuit in the development stage is contained in Cap1uf.dat. To download this file (in .zip format) click here.

Fig. 1. This simple circuit has a dc input from the potentiometer configuration. The converter should reliably provide 12-bit accuracy for any potentiometer setting. During hardware development it was noticed that the conversion data that were taken proved there was a problem (illustrated in Figs. 2 through 6.) This prompted capacitor changes in the circuit.

If you take time to look at the "Cap1uf.dat" data file using Notepad you will get a feel for the template that must be used when importing data into this software. You will find:

Microchip Technology
MCP3208
2048 data points
40000 sps
3487
3487
3488
3487
3486
3486
3485
3486, etc

The header for the data has four items, starting with "Microchip Technology". This is followed by the part number MCP3208. The MCP3208 is an 8-channel 12-bit converter. (The four stand-alone 12-bit converters from Microchip are the MCP3201, MCP3202, MCP3204 and MCP3208. Evaluations can also be performed for 10-bit converters. The appropriate names for these devices are the MCP3001, MCP3002, MCP3004 and MCP3008. If you construct your own data file you must use one of these converter names.) Following the device name is the number of data points that are listed in the file. For the Cap1uf.dat file, 2048 samples were taken. This is followed by the sampling speed of the converter at the time of conversion. The listing of the data points start on the fifth line of this file. The number of data points must match the number that is called out in line 3. These data points must all be positive and the maximum value can not exceed 4095 for a 12-bit converter or 1023 for a 10-bit converter.

At this point you should start the MXLAB application (Analog Development Environment.) With the software open, load the first set of data that will be discussed. This can be done by using the menu selection "File | Load Data from File". The data file you should load into the MXLAB environment is "Cap1uf.dat."

With this file loaded, there are a few things to pay attention to on the screen. Notice along the top, below the menu selections, that the sample speed is 40000 sample/s and the number of samples is 2048. This matches the file that we just imported.

Let's look at this data using the MXLAB tool. Open "View | Histogram." When this histogram function is selected, the graph that is shown in Fig. 2 should appear. You will note that the 2048 samples of data taken span across four codes, 3485 to 3488. This data looks "noisy" but it doesn't have a good Gaussian distribution as would be expected if the noise were being generated by RPOT or the amplifier (MCP601.)

click for larger image

Fig. 2. The y-axis of the MXLAB histogram indicates the number of times a particular code was captured. The x-axis span can manually or automatically be adjusted. Basic statistical information is also provided with the plot.

Keeping this observation in mind, open the FFT view with the pull-down menu "View | FFT". The FFT results that appear on your screen indicate that there is no harmonic content to the noise that we saw in the histogram (as shown in Fig. 3.) This observation points back to the original premise that the problem is noise.

click for larger image

Fig. 3. In the FFT graph, various FFT windows can be selected when viewing the data. The magnitude of individual spurs can be quantified by locking the cursor position with a left mouse click.

At this point in the MXDEV1 board development, the engineers at Microchip assumed that the noise was coming from the potentiometer. Given this assumption a 1-µF capacitor was placed across the wiper of RPOT to ground. This added capacitor had no impact on the conversion results. The scope plot provides the missing link that assisted in the final solution.

You can get this screen with the pull-down menu "View | Scope Plot" (see Fig. 4.) When you are in the scope plot screen use your mouse to "Center Envelope" on the left hand side. You may have to toggle this a few times to get a good view. The pattern that you are seeing looks like a random error that occasionally makes the conversion higher than the median. If you change the horizontal scale from 50 samples/div to 10 samples/div at the bottom of the view, you will notice that there is quite a glitch on the first few conversions, as shown in Fig. 5.

click for larger image

Fig. 4. The scope plot visually maps every data point in the time domain.

click for larger image

Fig. 5. If the scope plot x-axis and y-axis zoom capability is used, the signal details can easily be viewed.

The data list can also be viewed by using the same pull-down menu, "View | Data List". The data list for the "Cap1uF.dat" file is shown in Fig. 6. Note that the data list verifies the scope plot results in Figs. 3 and 4.

click for larger image

Fig. 6. A listing of the raw data is shown in this view.

The views of the data that we have looked at suggest that there is some kind of random interference with the conversion process. The possible places for this error to be initiated are with the potentiometer configuration, the amplifier driver to the ADC, the reference voltage source, or digital feed-through. As the evaluation of this circuit came to a conclusion, it was found that the de-coupling capacitor on the 4.096V reference (REF198) was too small. When this capacitor was changed to 100 µF, the conversion consistently produced one bit accurate results. Data from the circuit that has a 100-µF capacitor installed is labeled as Cap100uf.dat. Click here to down load that data file (in .zip format).

Power Supply Noise or Microcontroller Clock?

A second source of noise that was on this board during development came from the power supply. The hardware implementation of the power supply is shown in Fig. 7.

click for larger image

Fig. 7. Hardware implementation of the power supply circuit of the MXDEV1 board. The inductor, L1, was a short for the data taken in Fig. 8.

One design goal for this board set was to provide a way to power the board from the wall. Consequently, Microchip elected to supply a power adapter with the board. This power adapter is basically a switching type that down-converts an ac input (100-240 V ac) down to 9 V dc. This source is then connected to an on-board 9-to-5 V regulator (LM7805.) With this power supply, the circuit (Fig. 1) produced the data that you will find in the power.dat file. To download this file (in .zip format) click here.

The histogram plot of this data suggests that there is a noise problem that has a Gaussian distribution. The FFT plot contradicts this assumption by showing that the data has a small degree of frequency content. The scope plot (shown in Fig. 8) shows that a distinct signal has been coupled into the ADC.

click for larger image

Fig. 8. The harmonic content that is found in the FFT plot of the "power.dat" file is shown in the time domain with the scope plot.

The frequency shown in the scope plot was tracked back to the output of the linear regulator (LM7805). When a choke (L1) was placed in the circuit, the noise disappeared.

We have seen two examples where the MXLAB evaluation tool suite had a significant impact on design process by identifying problems. These problems were found on Microchip's 12-bit MXLAB development kit during the early part of the development cycle. Since the histogram groups the data in terms of code occurrences, this plot showed its usefulness in terms of identifying noise. It also can be used when finding digital communication problems. The FFT plotted frequency information, which identified repetitive occurrences in the data. The scope plot was most useful when identifying un-correlated, cycle problems which the FFT plot missed.

Test Your Evaluation Skills

Given that the fundamental converter problems have been solved there are two more sets of data that have been taken with this board configuration. Both of them have identifiable input signal issues. You decide what these problems are by using the MXLAB suite of tools.

ADC1.dat (download this file in .zip format by clicking here)

Load "ADC1.dat" into MXLAB software and log your results to these questions:

Histogram Results
Mean - ____________
Peak-peak - _______
St Deviation - __________

FFT Results
Fundamental Amplitude __________ Frequency _______
2nd Harmonic Amplitude __________ Frequency _______
3rd Harmonic Amplitude __________ Frequency _______

Scope Plot Results
Validates Histogram and FFT Results? ______________
Type of Signal __________________

Describe Signal _____________________________________________

Click Here for Answer 1

ADC2. dat (download this file in .zip format by clicking here)

Load ADC2.dat into ADCES software

Histogram Results
Mean - ____________
Peak-peak - _______
St Deviation - __________

FFT Results
Fundamental Amplitude __________ Frequency _______
2nd Harmonic Amplitude __________ Frequency _______
3rd Harmonic Amplitude __________ Frequency _______

Scope Plot Results
Validates Histogram and FFT Results? ______________
Type of Signal __________________

Describe Signal _____________________________________________
Why might this signal have so much harmonic content?

Click Here for Answer 2

Conclusion

Although this free software is restricted to analysis of 12-bit and 10-bit conversion data, it has significant worth during the process of hardware troubleshooting and development. The first two examples in this article presented the data the Microchip engineers worked with to troubleshoot the MXDEV1 evaluation kit hardware. The analysis tools that were used in this evaluation were a histogram, scope plot, FFT plot and data list. The histogram groups data in terms of code occurrences. This plot provides useful information about noise problems, but it can also point out digital communication problems. The FFT plot provided frequency information about the signal, which was used to reinforce conclusions reach with the scope plot tool and histogram. The scope plot was most useful for identifying un-correlated signal or cycle problems.

Analog Main | Product of the Week | Columns | Editorial | Tech Notes

		Click here to get your listing up.