Audio Measurement Techniques - cont

Distortion & SNR

Distortion is the measurement of the alteration of the original shape of a waveform after it has undergone amplification or conversion from electrical to acoustical output.  We typically measure Total Harmonic Distortion + Noise or THD +N.  Often you see A/V publications displaying an FFT distortion analysis which shows the fundamental test frequency and the subsequent distortion byproducts or harmonic content.  But are their measurements meaningful?

FFT Distortion Analysis of Yamaha RX-Z11 at Full Rated Power

The fundamental test frequency of 1kHz was applied to the receiver driven to near full rated power of 140wpc into an 8 ohm load for this test.   You can tell this by taking the level at 1kHz from the graph which is 30.393dbV and converting it to watts as follows:

[10^(dBv#) / 20)]^2 / Rload or in our case 10^{30.393 / 20) = 33.1Vrms  => 33.1^2 / 8 = 137 watts. 

I often see other review publications publish an FFT distortion measurement they claim to have taken at full power for a 200 watt power amp whose fundamental test frequency is at a level much lower than the rated power. 200 watts into 8 ohms corresponds to about 32dBv, yet the measurement shows a level of 9dBv which is one watt.  

For a quick and easy way of converting volts to dBV, use this calculator. 

When looking at distortion measurements, it's important to know how they are conducted per the following variables:

• Output level
• Frequency Bandwidth
• Test Signal Used
• Load Impedance (for amplifier measurements)
• Room noise conditions (for acoustical measurements of loudspeakers)

We have seen manufacturer specs on preamps that make a high output voltage claim without specifying load impedance.  While the preamp may be able to deliver 2Vrms into a 100kohm load, what about when it's trying to drive a high quality headphone system that is rated at 100 ohms and trying to deliver 1000 times the current?  Unless the preamp has output devices capable of delivering current into a 100 ohm load, the subsequent output voltage will most definitely be much less than 2Vrms.

The same holds true when looking at amplifier Signal to Noise Ratio (SNR).  Manufacturers typically spec SNR at full power instead of 1 watt.  1 - 3 watts is where the amp spends most of its time when averaged with a long time constant and where low noise is critical for high resolution playback.  So if Amp A is rated at 200wpc with an SNR of 105dB while Amp B is rated at 400 watts with 106dB SNR, which amp actually has a lower SNR?  To answer this, we must compare apples to apples by scaling back the SNR to the same power level.  Let's use 1 watt.

Using V = sqrt(P*R) we can calculate our voltage

Using 20 log (V/2.82)  we can scale to 1 watt

Amp A: SNR 105dB @ 200 watts   105dB - 20*log [sqrt(200*8) / 2.82] = 82dB @ 1 watt

Amp B:  SNR 106dB @ 400 watts   106 - 20*log[sqrt(400*8)/2.82] = 80dB @ 1 watt

So in reality Amp A has a 2dB advantage in SNR when compared to Amp B at the same power level.  This of course also assumes SNR was measured in the same method and with or without weighing. 

Output Level

You probably won't believe this but what if I told you a 1" dome tweeter can produce 20Hz?  Yes it most definitely can but there is a caveat; its output level.  At 20Hz, the output level is lower than that of a gnats fart but that’s all too often an overlooked point in manufacturers marketing literature claims. 

Quite often we see subwoofer manufacturers claim their sub is flat down to 12Hz (a frequency humans can't even hear).  Some go so far as to publish a frequency response graph on their website to prove their point.  What they don't tell you is that they are using DSP processing to linearize the response of the subwoofer system and its 12Hz extension is only good at vanishingly low output levels which most certainly cannot be felt. If you look closely at their measurement it's typically done at levels between 80-90dB at 1 or 2 meters groundplane.  This is especially true with most floorstanding speakers that have a claimed frequency extension of 20Hz yet only have  a pair of 6" woofers to reproduce bass frequencies.

Once the levels are driven up however, compression kicks in and severely limits the output abilities of the loudspeaker/subwoofer, especially at the lowest frequencies.  So in many cases, the manufacturer is trading SPL for extension solely for marketing purposes as discussed in our article below.

See: Trading SPL for Extension A new Trend in Subwoofers?

We've read many subwoofer reviews where a reviewer is touting the max output of a subwoofer they measured in-room and with no mention of distortion levels or whether or not the sub was producing audible mechanical noise or distress.

This reminds us of a recent subwoofer review from another A/V publication that measured a 10" sub in a small enclosure where they claimed it was outputting 100dB at 18Hz.  The caveat was the measurement was taken in-room and at only one foot distance from the driver.  If you scale that measurement back to 2 meters (without even removing room gain), that 100dB becomes 84dB which is barely audible, and certainly not as stellar as the reviewer would have its readers believe.  84 dB at 20 Hz is only slightly louder than the background noise found in most living rooms because of the loudness effect of the human hearing mechanism.  If you factor out room gain of that measurement, the reality is that sub is likely producing less than 80dB at 18Hz 2 meters away.  This is certainly nothing to brag about in a review or subsequent marketing literature but such is the case when dealing with the laws of physics and reality. 

The same holds true with power output for amplifiers.  Most review publications and manufacturers will provide amplifier output power at 1kHz instead of full bandwidth.  They typically run instantaneous power vs distortion sweeps at a single frequency instead of full bandwidth testing into 8 and 4 ohm load impedances and almost never run continuous or dynamic burst power tests to get a true picture of performance.  We have found some of the newer Class D A/V receivers simply fell apart when driving a 4 ohm load above 3kHz which doesn't show up with a 1khz power sweep but can surface in real world listening tests using 4 ohm speakers.  Several loudspeaker manufacturers informed us they had to do alterations to their 4 ohm speakers in order for these receivers to drive them without shutting down. 

We have covered amplifier measurements in detail in prior articles. For information on how we measure amplifiers see: 

Audioholics Basic Amplifier Measurement Techniques

The All Channels Driven (ACD) Amplifier Test

Conclusion

As we learned in this article, product performance data can very easily be manipulated to look good and more often than not inappropriate measurement techniques lead to erroneous conclusions about actual product performance. Manufacturers understand the role specifications play in the purchasing decisions made by consumers, and in some cases tell us half-truths for what are purely financial motives.

It's very easy to manipulate loudspeaker response curves and data by:

• Increasing the dynamic range (vertical axis)
• Apply excessive smoothing (ie. 1/3 octave smoothing)
• Incorrectly summing nearfield measurements
• Using low resolution during measurement taking (collecting an insufficient number of data points and then “connecting the dots” to make the product look better than it really is).
• Displaying max output levels taken in-room instead of in a 4pi or 2pi environment to eliminate associated boundary gain and room gain in real rooms
• Displaying max output levels without specifying distortion levels or whether or not the loudspeaker is suffering from audible mechanical noises and/or distress under such testing levels

It's very easy to manipulate amplifier data by:

• Measuring power into 1 discrete frequency only instead of full frequency bandwidth (20Hz to 20kHz)
• Furnishing SNR data without specifying power level or methodology used for testing
• Furnishing preamp output level data without specifying load impedance
• Providing FFT distortion measurements without knowing the true output level the test was conducted at

Consumers need to not only understand what measurements represent about product performance, but how those measurements were taken and how they were displayed.  Take the time to know these facts before allowing a manufacturers spec sheet or a so called "professional review" sway your option about product performance on a product you haven’t experienced first hand for yourself.  Measurements are only as useful as measurer's ability to correctly take and publish them. 

Never underestimate the role of politics and marketing plays in a company whose main goal is in achieving sales and boosting the bottom line.  Few consumers are willing to spend the time necessary to fully educate themselves on the topic, and the reality is manufacturers know this, and often jury rig the measurements for what they know will be the numbers most likely to be viewed by consumers prior to making a product purchasing decision.  Only by understanding the underlying measurements, and asking hard questions will the consumer find the whole truth.  Last but not least, do not forget one's own ears are a more reliable arbiter of truth, than salesmen on commission.  Make Jack happy by showing him you can handle the truth and you want more than marketing graphs from manufacturers and A/V review publications.

I hope this article sheds some light on this very often overlooked topic.

Acknowledgements

Special thanks to Dr. Floyd Toole, Paul Apollonio and Mark Sanfilipo for their peer review and additions in making this article more complete.

Recommended reading: "Sound Reproduction Loudspeakers & Rooms" by Dr. Floyd Toole