Eric Wallin's Internet Homepage

2. Passive Component Measurements

Introduction

Now that our output level has been set to ~1W in Speaker Workshop and our input level has been set via the appropriate recording slider, we are ready to move on to doing the channel calibration. Finally we will do something real and measure the values of several types of passive components. But first we need to perform one final input level check.

MLS Signals & Input Clipping Levels

Speaker Workshop uses a special class of signals rather than sine waves to perform most of its measurements. This type of signal is known as an MLS, or Maximal Length Sequence, and it is generated from a relatively simple algorithm. When you look at the spectra of an MLS signal it looks like white noise (random) but in fact it is deterministic (non-random) because it is produced by a very deterministic algorithm. The MLS is used for all types of acoustic testing since it is noise-like and posseses a low peak to average power ratio (low crest factor). This produces a good SNR (Signal to Noise Ratio) for the recorded signal, and so subsequent calculations performed on the recorded signal can be more exact and informative. The MLS has some rather magical properties due to its "acts random but really isn't" behavior, but that's another story for another day. For now we need to make certain that the soundcard inputs don't clip when Speaker Workshop uses this type of signal.

Figures 1 & 2. MLS signal used by Speaker Workshop in the frequency domain (left) and in the time domain (right).

We need to establish a baseline as to where the soundcard input clips. In the basic level setting of the previous section, you should have seen input clipping when setting the correct input level. Go ahead now and do some recording of 1W sine waves through the Jig II set to DIRECT mode (SW1 center, SW2 right, SW3 down) and adjust the appropriate input level slider until you see mild clipping on the tops and bottoms of the recorded sine wave. Note the min and max values in Speaker Workshop's VU meters. With my modified AudioPCI this point is around 27.5K.

Figure 3. Mild input clipping when recording a sine wave.

Figure 4. VU meters associated with the mild input clipping above seen.

After you have determined the input clipping min/max levels, put the input slider back to where you get a ~14K peak signal as we did in the prevous section.

Test Signal Preferences

Now we need to set the preferences of the MLS test signal. Before we do, however, we need to get some basic information about our soundcard. Open the soundcard wizard again by going to the menu item Options | Wizard | Check Sound Card... and once the checking is complete, click on the "More" buttons for the input and the output. Scroll down to the end of the "Formats" box and see what your soundcard is capable of in terms of sample rate.

Figures 5 & 6. Extended information about the soundcard input (left) and output (right).

As per these dialogs, my soundcard is stereo and capable of a 44.1kHz sample rate at the input and output - all good things to know.

- NOTE - I am using version 0.91 of the software and have noticed that wizard does not report back the true maximum sampling rate of my soundcard. For the Ensoniq Soundscape, the max sampling rate is 48 kHz.

Now close this dialog and go to the menu item Options | Preferences... and click on the Measurements tab. You should get a dialog that looks like this:

Figure 7. Test signal preferences.

Here is an explanation of each control and how you should set it for now:

Table 1. Test Signal Preferences.

Control	Description	Set To
Sample Rate	This slider sets the sample rate your soundcard will use for input and output.	Set it to the maximum your soundcard will support as per the documentation that came with it, and not the max as reported by the wizard.
Sample Size	This slider sets the length of the MLS signal and thus the duration. Larger numbers give you longer measurement periods, but the with improved frequency domain resolution. For calculation purposes, you can divide the sample size by the sample rate to get the sample duration in seconds. A 32.768k sample at 44.1kHz sample rate yields 32.768 / 44.1 = 0.743 seconds of sample time. Frequency domain resolution is the inverse of this number, or 1 / 0.743 sec = 1.346Hz, which is reported in the Precision Results section below this slider.	I usually use the 32,768 or 65,536 setting.
Volume	This entry area automatically sets the Wave slider on Window's Volume Control to whatever you enter here. This sets the output level.	Set this the same as whatever produced a 1W sine wave output when you determined the basic output level.
Reverse Channels	This checkbox will reverse the left and right input channels for impedance testing if checked.	Leave this unchecked for now, you will know soon enough if you need to check it.
Use Preemphasis	This checkbox will apply a low-pass filter to the MLS output signal, and will apply a similar high-boost filter to the input signal. Using preemphasis lowers the energy in the highest frequency ranges, giving the lower frequency ranges more relative power.	Leave this unchecked for now.
Repeat Count	This entry area allows you to repeat the MLS signal as many times as you like. The sample time will increase by whatever you set this to as a multiplicative factor. For example, if your sample time is 0.743 seconds and you set the repeat count to 5, the sample time will be 0.743 x 5 = 3.715 seconds. Be aware that lengthening the sample size this way will not increase the frequency domain resolution. But repeating the MLS multiple times can increase the SNR, and thus the accuracy, of your measurements. Using repeated MLS signals with acoustic measuremnts like frequency response requires some additional consideration of the MLS length. Echos from a given section of the sequence should have time to die out before that section is repeated. We'll cover this in more depth later.	Set this to '1' for now if it isn't already set to that.

Channel Difference Calibration

The channel difference calibration measures the differences between the soundcard inputs and adjusts for it. With the JIG II still in DIRECT mode (SW1 center, SW2 right, SW3 down) go to the menu item Options | Calibrate... and up will pop the calibration dialog:

Figure 8. Calibration dialog.

Click on the "Test" button in the "Channel Difference" section, then the "Next" button, then the "Next" button again, and finally the "Finish" button when it is not grayed out. This brings you back to the calibration dialog. Click on the "OK" button to close it.

Now inspect the VU meter in Speaker Workshop:

Figure 9. Post channel calibration VU meter readings.

These are the readings I get after performing the channel difference calibration. As you can see, the min and max values are less than the clipping levels for my soundcard input (27.5K) which we measured earlier. So the inputs aren't clipping, which is a good thing, and we can move on to inspect the calibration data itself. Open the "System" folder on the design tree, and double click on the icon named "Measurement.Calib". A window named "Measurement.Calib (dataset)" should open.

Figure 10. Calibration frequency domain graph.

To make this graph look neater, right-click in the chart area and select Chart Properties... from the pop-up dialog. Click on the X Axis tab and in the scale section set the Maximum to "20k" and the Minimum to "10". Next click on the Y Axis tab and put the numbers "12" and "-12" in the Maximum and Minimum entry areas, respectively. Set the Major gridlines distance to "3", and make sure all Auto Minmax boxes are unchecked. Finally, Click the click on the Y2 Axis tab and put the numbers "180" and "-180" in the Maximum and Minimum entry areas, respectively. Set the Major gridlines distance to "45", and make sure all Auto Minmax boxes are unchecked. Click the OK button. You should something like the figure above. Right-click again in the chart area and select Make Chart Default. This will force all future frequency domain graphs to use this format.

If the left and right inputs are perfectly matched, you should ideally get two straight lines at 0 dB and 0 degrees for the magnitude and phase response, respectively. How does my card look? The dark magnitude line on the graph above is almost ideal, with very little deviation from the 0 dB level above 20 Hz. This indicates no significant level differences between the left and right channels. So far so good.

The phase data on the graph above, which is the gray upwardly curving line, is another story. The phase error is minimal at the lower frequencies, but increases to around +180 degrees at 1/2 the sample frequency (also known as the Nyquist rate). Phase error like this is revealing evidence that the Ensoniq AudioPCI employs the trick of utilizing a single multiplexed analog to digital (A/D) converter to handle both of its stereo inputs. The single converter is switched quickly between the two inputs, first doing a conversion on the left channel, then on the right, and so on, performing two conversions within a single sample period. Since the conversion for both channels isn't happening simultaneously, the 1/2 sample delay between the left and right channels shows up as an increasing phase error with frequency. Soundcards are designed this way since a single 2x speed converter is usually cheaper than two separate 1x converters. The designer can eliminate almost half of the conversion hardware this way.

Is this 1/2 sample delay between the left and right channels bad? As the graph shows, since the same A/D converter is used for both channels we get identical magnitude response, which is a wonderful thing. And the phase error caused by the 1/2 sample delay can be effectively subtracted from our measurements by performing the channel difference calibration. Look at it this way: when given the choice when comparing two measurements, you always want to use the same equipment to make those measurements in order to minimize systematic measurement error. So I guess I'm actually happy that my soundcard uses the same A/D converter for both channels, even if this introduces some conversion delay between the channels.

Impedance Calibration

Now that the channel differences have been taken care of, we can do the impedance calibration. Place the JIG II in the IMPCAL16 mode (SW1 up, SW2 left, SW3 down) and open the meue item Options | Preferences... and click on the Impedance tab.

Figure 11. Impedance calibration tab of the Preferanced dialog.

If you know the input impedance of your soundcard, enter this here. If you don't know, use the defaults recommended in the SW help. After that, press the upper Test... button and then press the Next > button on the dialog that pops up. Enter the value of your 16 ohm calibration resistor and press the Next> button. Wait until the recording has stopped, then flick SW1 to the down position. Enter the value of your 4 ohm calibration resistor and press the Next > button. Now take a look at the ending dialog:

Figure 12. Impedance calibration results.

If you don't get something reasonable for the reference resistor (7 to 9 ohms) and something reasonable for the series resistance (less than 1/2 ohm, positive or negative) then you should check your jig connections and repeat the calibration. Note that you can hit the < Back button a couple of times to get to the beginning of the calibration procedure, and then you won't have to re-enter the resistor values again since SW remembers them for you. When your numbers are good, smash the Finish button then the OK button to close the preferenced dialog.

Impedance Measurements

One way to verify the impedance calibration is to measure the calibration resistors. To do this, place the jig in IMPCAL16 mode (SW1 up, SW2 left, SW3 down) then click on the "sine_test (signal)" window and select the menu item Measure | Passive component and the measurement should start. Next place the jig in the IMPCAL4 mode (SW1 down, SW2 left, SW3 down) and do another passive measurement. Here is what I get:

Figures 13 & 14. Calibration check: 4 ohm measurement (left) and 16 ohm measurement (right).

Go ahead and do these measurements several times and confirm that they are consistent. Next, measure a capacitor and an inductor. Put the jig in IMPMEAS mode (SW1 center, SW2 left, SW3 down) and install a capacitor across BP3 and BP4. Here is what I get when measuring a 1uF cap (which BTW measures 0.963uF with my Fluke meter in capacitance mode):

Figure 15. The result of measuring a nominal 1uF capacitor.

Back home