Beispiel einer kompletten Labormessung mit unserem Audioanalyzer QuantAsylum QA401



This method has been around since the late 1930’s. This test involves subjecting the device under test (DUT) to a low frequency tone (usually 60 Hz) and a mid-high tone (usually 7 KHz) and looking at the output. The ratio is also important, and the lower frequency tone is usually the stronger stimulus signal, and is usually 12 dB higher than the higher frequency tone.


The ITU method might be even older than the SMPTE measurement, though it was originally known as the CCIF method. While the SMPTE was created by motion picture engineers, the CCIF/ITU test was created by communication engineers. In this test, two tones are generated 1 KHz apart and at the same amplitude. Usually, this is 19 KHz and 20 KHz, and then the resulting 1 KHz tone level is measured and referenced to the combined amplitude of the 19 and 20 KHz tones. But there are other variants of the test where the second and third order products are considered, or the 2nd through 5th order products are considered. The 1 KHz measurement is easy to make on the QA400.

In setting up to make an SMPTE IMD measurement, we’ll use both generators on the QA400. Set the first generator to 60 Hz and 0 dBV, and the second generator at 7 Khz and –12 dBV (recall the SMPTE spec is that the higher frequency tone was 1/4 the level of the lower frequency tone, and 1/4th is the same as -12 dB).

In the plot below, we can see the frequency and amplitude of the tones is correct. In the green box, you can see the tone at 7 KHz has some mixing products appearing. If we turn off the 60 Hz tone, the spectrum around 7 KHz gets clear. These are indeed mixing products.



If we zoom in on the 7 KHz tone, we can see the products more clearly. And indeed, measuring with markers reveals these are 60 Hz modulation products. In computing the IMD products, we’re interested in comparing the modulation products relative to the carrier, or 7 KHz tone. In the graph below, we can see a single modulation product is –106.7 dB below the carrier. There’s another on the other side at the same level. These two would sum and increase the overall level by 6 dB of this particular pair. How far out you go generally depends on how quickly the harmonics fall off. The SMPTE spec is interested in ALL of the energy outside the 7 KHz tone, and thus it’d be correct to sum them all together. But you can get fairly accurate by simply considering the first pair or two. In this case, the harmonics are 100.7 dB below the carrier, and thus the SMPTE IMD figure would be –100.7 dB or 10^(-100.7/20) * 100 = 0.000922%




To make the basic ITU measurement, we set the two generators to 19 KHz at –6 dBV and 20 KHz at –6 dBV, respectively. These two tones will sum to 0 dBV, which will also be the reference stimulus for this measurement. Remember, for the ITU test, the levels will almost always be equal.



Above, we can see the resulting 1 KHz mixing product. Remember, this is called a mixing product because it is the difference between the two stimulus frequencies (19 KHz and 20 KHz). There are other mixing products too. And which ones you consider are a function of the flavor you have chosen for the IMD test. In more advanced instances of this test, you might also considering the following:


1 * F1 +/- 1 * F2 = 19 +/- 20 = 39KHz and 1 KHz (2nd order IMD)

2 * F1 +/- 1 * F2 = 38 +/- 20 = 58 KHz, 18 KHz (3rd order)

1 * F1 +/- 2* F2  & 1 = 19 +/- 40 = 59 KHz and 21 KHz (3rd order)

2 * F1 +/- 2* F2 = 38 +/- 40 = 78 KHz and 2 KHz (4th order)

3 * F1 +- 2 * F2 = 57 +/- 40 = 97 KHz and 17 KHz (5th order)

2 * F1 +/- 3 * F2 = 38 +/- 60 = 98 KHz and 22 KHz (5th order)


For the simple case, we’ll consider just the 1 KHz resultant 2nd order IMD product. Here, with a 0 dB reference and a 1 KHz product at –113.4 dB, the resulting ITU IMD is –113.4 dB, or 10^(-113.4/20) * 100 = 0.000213%. To properly put this measurement in context when sharing the result, it’s very important to specify the stimulus frequencies and that only the resultant 1 KHz tone was considered.

In the plot above, we can clearly see the other frequencies around the ~20 KHz signals. Zooming in shows more detail. We can see the 18 KHz product (3rd order) and 17 KHz product (5th order). A reasonable measurement to make would be a 2nd and 3rd order measurement. In this case, we’d want to sum resulting 2nd and 3rd order harmonics and compare those to the stimulus.



Beispiel einer THD Messung mit unserem AudioAnalyzer QuantAsylum QA401



Next, we're going to make a graph of the amplifier's THD versus output Power. We need to turn off the A-Weighting that was enabled in the noise test above.

To sweep THD versus power, we'll use a plugin selected as shown below:



The configuration screen for that appears as follows. 

The settings for the various entries will change depending on the amp. What this test will do as configured above is as follows:The amp will sweep a 1 kHz tone from -30 dBV to 0 dBV. The power output will be calculated based on the specified load impedance and specified external gain. These settings are very important to have correct!

The attenuator will be managed automatically. What this means is the test will automatically engage the attenuator when needed, and where that occurs depends on whether you are measuring THD or THD+N. 

The test will also automatically stop IF the measured power exeeds 10W AND the THD exceeds -40 dB. This will protect the amp and load, and should be used anytime you are dealing with an unfamiliar amp. 

Running the test as configured above yields the following graph. The test was first run at volume midpoint, and then repeated at max volume. 



From the plot above, we can see there is reasonably continuity between the two volume settings. In other words, you aren't penalized from a THD perspective by leaving the volume at max all the time (although you will pay a penalty in noise floor). 

Finally, let's take a look at the spectrum at max volume and 192 kSps so that we can look out to nearly 100 kHz. The output level was adjusted until just before the THD started to degrade quickly, which is around the 350W point.



Note the level of gain provided by the amp: about 37.6 dB. Overall the spectrum above looks well behaved out to 100 kHz, and is absent any spurious tones that might harm high frequency drivers. Given that the amp is rated at 1000W in bridge mode, and that our measurements showed the 10% distortion point being hit at 665W, it seems a stretch to claim this is a 1000W amp. For commercial audio, an amp should achieve the rated power at 1% THD 24x7, with peaks beyond that hitting 10%. 

This post walked through the concerns and considerations of measuring a high-power/high-gain class D amplifier, and how you can ensure your system is setup correctly to give correct measurements.

The Behringer NX1000 is a discrete realization of a class D amp. It has a clever feature that was tested but not documented in this post where it adapts its frequency response to the load. And it works. That is a shortcoming shown previously in the TPA3255 amp post. But considering TI's TPA3255 hits 600W at 10% and this NX1000 hits about 670W at 10%, it's tough to justify the NX1000's discrete effort at this power level. Probably the reason a Behringer engineer might give is that their current discrete architecture provides them a path up to thousands of watts (side note: there's a good write-up here on amp power levels required for live music) . That allows them to tier their amplifiers by the number of MOSFETs in the power stage. Something that couldn't be done with current class D chipamp.

When working with high power amplifiers, it's easy to overstress the DUT or the load. The most common cause for this is not understanding the gains of each component in the signal chain, or incorrectly entering a gain as an attenuation (or vice versa). Move slowly when you are bringing up a new DUT, and sanity check your first measurements with a DVM. It takes 60 seconds, and is a sure-fire way to diagnose incorrect setup parameters in your measurements. 


Beispiel einer THD Messung mit unserem AudioAnalyzer QuantAsylum QA401

The THD spec on the QA401 is -108.5 dB, measured in loopback mode, single ended, L- shorted to ground, attenuator off, 32K FFT, Hann windowing, 48 KSPS and -10 dBV. 

Because the measurement is made in loopback mode, it's hard to distinguish where the DAC performance ends and the ADC performance begins. In a previous blog post, we took a look distortion limits of the QA401 when using a notch filter. 

In this post, we'll take a look at some sweeps to see where the sweet spot is for the QA401 in loopback mode. 

In this first plot, you can see a THD sweep from -22 dBV to -8 dBV. Best case THD appears around -14 dBV and 2 KHz



Focusing on that operating point, we see the graph below. Using markers, we can see the first harmonic at 4 KHz is -120 dB below the fundamental, which is exceptional. The measured THD at this point is -115.7 dB--considerably better than the spec'd value for the QA401 of -108.5 dB. Of course, some QA401 may be slightly better and some may be slightly worse than this -115.7 dB figure.



At 1 KHz, the optimal point for 1 KHz operation we can read off the swept plot is at -16 dBV. Below is a plot of that measurement:




While not as impressive as the 2 KHz point, the above shows that the second and third harmonics are both about 120 dB below the fundamental.



Texas Instruments TPA3225 EVM

We're using the QA450 in manual mode in these tests. In manual mode, you change loads (4 or 8 ohms) by pressing on-screen buttons. The QA450 also supports automated control via REST. 


The supply used for all of the tests below is a 48V/10A ACDC brick, with the trimpot dialed up to 51V. This limits some of the testing because it cannot supply enough current for drive both the left and right channels to maximum power. For most plots below, the data was collected on the left channel only.The noise of the amp at 51V was verified with a lab supply to ensure the brick wasn't degrading the amp performance with its own noise. It was not. TI's EVM doesn't care whether you are using a lab supply or an economical fixed ACDC supply. 


A good first test when first trying to understand an amplifier is a multi-tone test, because it provides a quick way to look at the gain, gain balance, the flatness of the spectrum and the cross-talk all at once.The total RMS power from 20 to 20 KHz using a multi-tone stimulus is about 17.3W into 8 ohms in the plot below. From the text display at the bottom of the graph we can see the amp has a gain of roughly 27.5 dB, exceptional (0.01 dB) matching between the channels, about -75 dB of cross-talk, and just a bit of roll-off (about 0.4 dB) up near 20 KHz. Generally, we might state the gain of the amp is 27.5 dB +0/-0.4 dB. And no worrisome spurious products are observed.Here the Multitone QA401 output level is -6 dBV (total RMS in 20 to 20 KHz). For inputs lower than -6 dBV into the TPA3255, the spectrum appears very similar. 



Nudging the QA401 multitone output up 1 dB to -5 dBV input shows a picket fence of products suddenly appearing--growing about 30 dB in the process (see plot below). The CLIP_OTWZ LED also flickers on the EVM. This LED by itself suggests that the junction temperature is higher than 125C according to Table 2 in the EVM. But the total RMS of all the tones is 22.4 DBV which is about 22W. What is going on? How can a 300W amp be putting out just 22W and increasing the input by 1 dB cause its performance to collapse?
The answer is the the crest factor: While the RMS of the 12 tones is 22.4 dBV = 13.18 Vrms, the momentary peak of the signal can result in a short excursion on a single terminal of +/-25V across the load as verified with a scope. This means that's a total of 50V across the speaker--and at that point the amp has run out of supply rail and/or is current limiting and/or is also fighting on-die thermal limits due to the instantaneous dissipations encountered. In short, multi-tone is a great tool, but make sure you understand that is incredibly demanding on the equipment and it's easy to push things too hard (Some quick math: A single sine has a crest factor of 20 * Log10(1/0.707) = 3 dB. Worst case, the 12 tones in the multi-tone test sequence could potentially peak at 36 dB above the RMS. Be careful!)Dialing the output back to -7 dBV and switching to a 4 ohm load we see the following:
This time we see the Peak- is reported as -1.75 dB. That is the minimum amplitude of one tone is about 1.75 dB below that of the mid-band tone. And visually we can see that is happening at the 20 KHz frequency. So, the act of changing from an 8 to a 4 ohm load has resulted in the frequency response of the amp dropping at 20 KHz by about 1.4 dB. And this shouldn't be a surprise: Class D amps require the output filter be designed for a specific load. If you change the load, the output response will change. OK, thus far we have some import information: Nominal gain is around 27.5 dB and there's a bit of droop at the higher end with lower impedances. 

Frequency Response

Multitone looks across the band with a dozen frequencies. But let's take a more detailed look at the frequency response at 4 and 8 ohms. For this, we'll sweep from 20 to 20 KHz at -18 dBV into the TPA3255. With an 8 ohm load this will give about 1W of output. For this test, the sampling rate on the QA401 was set to 192 Ksps and Flat Top windowing was used. 
After running the sweep, we're presented with the following graph. This is a new window introduced in the last month or so. Notice the ability on the left side of the dialog to customize trace colors, thickness, names, etc. For now, we'll do nothing and just take the graph as is. 
Next, we switch the load to "4 ohms" on the QA450 programmable load:
And then we run the exact same sweep again, just as we did above. This time, after the sweep completes, we're presented with a new option:
The Graph Selection dialog box is saying "I notice you have at least one graph window already open. Do you want to add this new sweep to an existing graph OR do you want add it to a new graph?"The "QA401 Graph 112" matches the title bar in the first graph we generated. The "112" is just a random 3 digit number to disambiguate the different graph windows. Let's opt to add it to the existing window. And now we're greeted with:
The graph above needs tweaking. The trace colors are the same, the title is ambiguous and trace names aren't helpful. Let's use the features of the graph tool to fix this (changing colors, adding notes, changing titles) and then copy 800x480 sized bitmap to the clipboard for pasting into this post:
From the plot above, we can readily see the response at 20 KHz is about 1.6 dB down with a 4 ohm load versus an 8 ohm load. The gain of the amp is also reduced a bit due to the amp output impedance, which we'll look at in more detail below. But in short, this is a big improvement over previous versions of the application. No import/export, finding traces in directories, remembering the names, etc. In the section below looking at output impedance we discuss this load sensitivity a bit more. Next, we'll move on to look at amp distortion.


Like the frequency response sweep, capturing a THD+N versus power takes just a few seconds to run. The setup for this test is shown below, with some new options highlighted.
The first option is that the QA401 application can manage the attenuator automatically for you. This is helpful because at lower levels of power out, having the attenuator off will lower the noise floor. If you elect to have the plug-in manage the attenuator, then the measurement will always start with the attenuator active. From there, the plug-in will decide when it makes sense to enable the attenuator. Generally, this will occur at around -20 to -10 dBFS, depending on whether you are measuring THD or THD+N.Switching the attenuator automatically has a downside in that it introduces a bump in your output display, which usually will need some explanation if you are presenting the data to others. So, if your application isn't showing a difference at lower levels with attenuator on and off, then leave the attenuator on and take the glitch-free trace. The TPA3255 is an exceptional performer, and for that reason the data presented below lets the plug-in manage the attenuator.The second highlighted checkbox is the ability to enable an early abort of the test sequence. This is helpful if you don't know the limits of the amp, but you know you want to stop as soon as the THD (or THD+N) exceeds a certain threshold. In the example above, we see the Abort Minimum Power Level is 10W, and the Abort Maximum THD is -85 dB. With these settings, if enabled, the plug-in would automatically stop at any measurement where the measured power was greater than 10W AND the THD had exceeded -85 dB. The resulting output from the THD+N sweep is below. In the above settings, note that we elected to have the plug-in manage the attenuator. You can see where the plug-in engaged the attenuator. Also we now have THD+N displayed in both dB and as a %. 
How does this compare with TI's sweep of the part? That graph is below. TI has a lot of additional filtering that was used on their measurement, including an AES17 and AUX-0025. They also ran the test at 75C. The distance between power steps looks very small (1 dB or less versus the 2.5 dB steps taken above).But in any case, the two plots compare quite favorably, and it took just a few minutes to setup, run and prepare the graph.



A sweep of IMD is important on Class D amps as it's difficult to measure linearity at higher frequencies using a conventional THD techniques because the world beyond 50 KHz in Class D amps is usually fairly harsh. Below is the setup for the IMD measurement. That this time we are enabling early termination: If the power level exceeds 1W AND the IMD exceeds -60 dB, then the sequence will terminate early. 
There isn't an option for the type of IMD measurement right now in the above configuration (ie, SMPTE or ITU), but there will be at some point. The test above will measure the level at 1 KHz while applying a 19 KHz and 20 KHz tone at equal levels (commonly known as the ITU-R or CCIF/ITU). If you specify an output level of 0 dBV, then the level of each tone will be -3 dB, and the combined RMS of the two non-coherent tones will be 0 dBV. The response at 1 KHz is referenced to the combined RMS of the two tones. In the future will be the ability to specify the max order of distortion. Currently, a single second-order product is considered (F1 = 19K, F2 = 20K, and  ABS(1* 19K - 1* 20K) = 1 KHz.).A previous post talks more about IMD measurements (including background) on the QA400.The resulting plot of the IMD measurement is as follows. TI doesn't publish IMD figures on the TPA3255. 
A plot of about 10W into 8 ohms shows the following activity. We can see the total RMS of the 19 and 20 KHz tones are 19 dBV, and the amplitude of the resulting 1 KHz tone product is about -84 dBV, resulting in measurement of about -103 dB.



Measuring output impedance of an amplifier is done by taking measurements of the output across two loads, usually an open and a typical speaker load, or in this case, a 4 ohm (instead of an open) and 8 ohm load. And then through some algebra we can compute the output impedance of the amplifier. The setup screen for this plug-in is below. We'll first sweep at 8 ohms, and then sweep at 4 ohms. We'll also do the left and right channel together:
When the test starts, we're instructed to first connect the 8 ohm load. With the QA450, we make sure the 8 ohm button is pressed.
That sweep will take a few seconds, and the we see the second prompt. There we connect the 4 ohm load:
Upon completion of the second sweep, we get the following graph of output impedance:
At 1 KHz with a 4 ohm load, this is a damping factor of roughly 30. At 20 KHz, the output impedance is nearly 2 ohms. This is overwhelmingly coming from the output LC filter. The output L is 15 uH, which has an impedance of 1.8 ohms at 20 KHz. This highlights a common complaint with class D. Historically, class D amps have had three big problems: Poor PSRR, marginal to poor output impedance (especially at higher frequencies), and poor tolerance to load changes (as we saw above in the frequency response of 4 versus 8 ohms). TI has addressed the first problem of PSRR by closing the loop on the TPA3255. But the second and third problems are due to the output LC. And TI only closed the loop prior to the LC in the TPA3255. In the future, look towards class D chipsets than can close the loop using feedback taken post-LC. The math and processing to do so must be very difficult. But you can bet the big class D chipset companies are working hard to solve this. 


A noise measurement was made with the inputs shorted and with A-Weighting enabled. TI's spec indicates a typical 85 uVrms noise (20 to 20 KHz), and what was measured below was 95 uVrms. This measurement was into 4 ohms.



The gain of an amplifier should be linear across it's operating region. That is, if the amp gain varies based on the incoming signal level, it can result in strange amplitude modulations. The graph below is the setup for the gain linearity test. The set up indicates we'll sweep the QA401 output from -120 to +5 dBV, in 5 dBV steps, and measure the gain at each point. What we'd like to see is a single horizontal line.
The plot from the test is shown below. There are 3 regions to note: The first is left side of the graph below -110 dBV or so input level. Here we can see that if a very small signal is input to the amp, then the gain is around 28.5 dB. And on the far right side, with very high input levels around +5 dBV, the gain drops to about 26.5 dBV.
What we see above on the far left side of the graph and on the far right are to be expected. The key is the steller performance in the middle region from -110 dBV to just over a few dBV of input.On the left side, the deviation can be explained by noise. The signal is so small that its difficult to separate the noise from the signal. But it doesn't matter, because we're not able to hear the signal at this level either. And on the right side, the deviation can be explained by amplifier compression. But at this level, the output is so large (200 to 300W) it just doesn't matter. In summary, the gain linearity of this part is very good. 

A Side Note on the QA450 to Power Measurement

In many of the above tests, that the power into the QA450 was exceeding 300W in some cases (the manual has more analysis on this). The QA450 cannot handle these levels of power for very long. But if you are doing swept tests with smaller FFT sizes, it can handle them for long enough periods to make the measurement. And this is an important point to make again: The QA450 isn't designed for sinking hundreds of watts for several minutes. It's designed for making very fast measurements at several hundred watts under automated control.