Notes from the Test Bench
By Bruce Hofer, Chairman & Co-Founder, Audio Precision
August was another busy month, highlighted by my annual visit to the Danish Technical University (DTU) in Denmark to assist Professor Michael Andersen with his very popular one week PhD course on class‑D amplifier design. The students asked many excellent questions, like “why are we NOT using the AUX‑0025/0100 measurement filter when making measurements” and “what’s the difference between THD+N and THD?” I was also quite impressed with the proficiency they all showed in using our audio analyzers to measure several prototype amplifiers.
This year, Audio Precision donated an APx525 analyzer with the BW-52 high bandwidth option to the university to help Dr. Andersen and his students. The over 1 MHz FFT capability of the instrument will allow them to see the actual switching frequencies of their class-D designs, which until now they could only see in the time domain on relatively low resolution oscilloscopes. The gleam in Dr. Andersen’s eyes and the smile on his face says it all...
AP Chairman Bruce Hofer presenting an APx525 analyzer (with BW-52 option) to Danish Technical University Professor Michael Andersen.
Output: Measuring Microphone Preamplifier Noise
by Adam Liberman, Editor Audio.TST
In Audio.TST’s March edition, we discussed how to measure phantom power with APx. This month, we again look at microphone preamplifiers, this time focusing on measuring noise with both the APx and 2700 Series analyzers.
Making noise measurements of microphone preamplifiers (and other high-gain circuits) involves special techniques that differ from those used to measure low-gain devices. When gain is set very high, the noise in the input stage gets greatly amplified and becomes the primary noise source. A signal-to-noise ratio measurement doesn’t do a very good job of describing this noise, because the signal level is unknown—it will depend on the sensitivity and self-noise of the microphone that’s used with the preamp. That’s why we use equivalent input noise (EIN) and noise figure (NF) to describe noise at high gains. This makes it much easier to recognize good results and to compare devices.
Equivalent Input Noise (EIN) is a measurement of noise generated in the input stage of a device, and is made by looking at the noise on a device’s output and then subtracting the gain. We measure input noise at the output for two reasons: it would be very difficult to measure such a tiny signal, and the noise is generated within the device’s gain stage—there isn’t an actual place where you can measure it directly. For this reason, it is called “equivalent” input noise.
When measuring input noise, the mic preamp’s input must be shunted with the specified source impedance or the results will be invalid. This is because the value of the source impedance affects the amount of noise generated within the input stage, and because all resistors themselves cause noise. This noise, called Johnson noise or thermal noise, is caused by the thermal agitation of electrons. The thermal noise generated by a 150 Ω resistor at 20° C and 20 kHz bandwidth is –130.9 dBu. The resistor should be a high quality metal film type so as not to introduce additional types of noise.
Chart 1 Thermal noise for various source impedances, 20° C and 20 kHz bandwidth.
To illustrate these concepts, we’ve measured noise on a preamp and have shown the results in the graph below. This preamp has an EIN of –127.2 dBu with a 150 Ω source. Since the thermal noise of the source impedance is –130.9 dBu, we can conclude that the preamp has added 3.7 dB of its own noise to the resistor noise. It is therefore said to have a noise figure of 3.7 dB. Stating in terms of noise figure makes it easy to see how close to noise-free a mic preamplifier’s input stage is. If it had added no noise, its noise figure would be zero. The quietest mic preamplifiers can achieve a noise figure of around 1.0 dB with a 150 Ω source. Note that although noise figure is defined as the x/y ratio noise factor expressed in decibels, the two terms are often used interchangeably.
Figure 1 Microphone preamplifier noise graph (see larger image).
The standard source impedance with which balanced microphone preamplifiers are tested is 150 Ω (200 Ω in Europe), so at a minimum, this source impedance should be used to produce or verify specifications. However, an increasing number of condenser microphones, especially those with ultra‑low self-noise (3 – 14 dBSPL), have very low output impedances on the order of 25 to 50 Ω. Therefore, testing preamplifiers using a 40 or 50 Ω source impedance can give additional real-world performance data.
Unbalanced microphone inputs are tested with either a 150, 200, or 600 Ω source. With semi-pro gear, such as portable digital recorders, unbalanced inputs are normally tested at 150 Ω (200 Ω in Europe), while consumer grade gear such as computer sound cards are often tested at 600 Ω. If the preamplifier under test is to be used with a dedicated built-in microphone, then the source impedance should be set to match the actual operating conditions.
Keep in mind that these input noise measurements are useful for comparing mic preamplifiers against each other and against the theoretical minimum noise for a given source impedance. In the real world, microphones and mic preamps work as a team, and noise performance of the system is also affected by the self-noise and sensitivity of the microphone. As a system, we can measure or calculate signal-to-noise ratio at a given acoustic level (dBSPL). However, when measuring a preamp in isolation, we don’t talk in terms of signal-to-noise because we don’t know what the signal level will be in actual use.
Making the Measurements
Microphone preamp noise can be easily measured with any AP analyzer. Make sure to set the output impedance of the generator as required (see the discussion above). Since the APx585 doesn’t have selectable output impedance, you can add additional resistance in series with the signal conductors inside of an XLR connector.
Proper cabling is extremely important when measuring low signal levels. On the 2700 Series, try setting the output to both Balanced Floating and Balanced Grounded to see which is quieter. Lifting pin 1 (ground) at the male end of the XLR cables may also be helpful. Use the FFT Spectrum Monitor in APx500 or the FFT Spectrum Analyzer in AP2700 to check for extraneous hum or interference. In noisy environments, it may be best to terminate the preamplifier input with a terminator plug fitted with the correct resistor instead of back-terminating at the analyzer.
Measurements should be made over the entire gain range of the preamplifier, starting either at minimum or maximum gain. At each measurement point, the gain of the preamplifier is set and then a noise measurement is taken.
To set gain, turn on the audio generator and adjust the preamplifier gain control while reading the signal level coming back into the analyzer. If the preamplifier has an input stage gain control, then this control should be used to vary the gain, and the output gain control or slider should be fixed at full or reference level. If only an output attenuator is provided, then this control will have to be used to vary the overall gain. If there is a passive input stage attenuator, then it may be necessary to plot the noise several times using different combinations of settings.
The generator level used to set the gain is not critical, as long as the output is not noisy or distorted. As the gain is increased, it will be necessary to reduce the generator signal level to prevent output distortion.
Figure 2 Connecting the APx525 to eliminate the effect of source resistance on gain measurement. Remove the extra cables before measuring noise and terminate the preamp inputs as required.
In APx500, there is a convenient Level and Gain measurement. However, to eliminate gain error due to the slight voltage drop across the generator's source resistance, it is necessary to take some extra steps: Attach an XLR splitter to Analog Output 1, and connect one of the splits to the preamplifier, and the other to Analog Input 1. Then, connect the output of the preamp to Analog Input 2. Now, using the APx500 derived result Compare (Ratio), you can measure the exact gain. With this method, the generator output level reference is taken after the internal source impedance and not before.
Eliminating this error can significantly improve gain measurement accuracy, especially with preamps that have a relatively low input impedance. For a preamp with an input impedance of 2 kΩ, the error is 0.7 dB with a 150 Ω generator output impedance. Even if you were to temporarily lower the generator output impedance to 40 Ω, the error would still be 0.2 dB.
Figure 3 Setting preamp gain in APx500 (see larger image).
To measure gain in AP2700, you can set the analyzer to read amplitude in dBgA (dB relative to the generator). However, to eliminate the source impedance gain error as discussed above, set Analog Output A to GenMon and do a 2-Ch. Ratio measurement.
Figure 4 Connecting the 2700 Series analyzers to eliminate the effect of source resistance on gain measurement.
Figure 5 Setting preamp gain in AP2700.
After setting the gain, the generator is turned off. The noise level can now be read using the Noise measurement in APx500, or by setting the amplitude meter to read in dBu in AP2700. Noise measurements are normally taken over a 20 Hz to 20 kHz bandwidth, both unfiltered and A-weighted. Make sure that you have removed the Y-cable from the APx if you used it to set gain, and that the preamp input is again terminated either by the analyzer or by a terminating plug of the correct impedance.
Figure 6 Reading noise in APx500 (see larger image).
Figure 7 Reading noise in AP2700.
Reading the Graph
Looking at the example in the graph above, you’ll see that at high gains, input noise completely dominates and the noise figure becomes constant. The minimum noise factor is 3.7 dB, measured at 60, 70, and 80 dB of gain.
As the gain is reduced, the output noise would ideally be reduced by the same amount. In reality though, as we reduce the gain below 60 dB, it fails to drop at the same rate, due to both the rise in impedance of the amplifier’s feedback loop as well as the increasing influence of output noise.
As the gain is further reduced, the constant output noise completely obscures the input noise. You can see that we’ve continued to calculate EIN down to 0 dB gain in order to draw a complete curve, although at this point we really aren’t looking at input noise at all. Even though this preamp’s gain can’t be reduced below 10 dB, we’ve extrapolated the curves to illustrate how the difference between the output noise and EIN curves is always equal to gain, and therefore at 0 dB gain they converge.
Sound Advice: MLS vs. Chirp Speaker Measurements
Question: I currently do acoustic measurements using the MLS method on an AP 2700 Series analyzer. I’m thinking of acquiring an APx500 Series instrument, which uses Continuous Sweep (Chirp) instead. Will I get the same results with both analyzers?
Answer: Yes, you will get the same results, as you can see in the graphs below, as long as the speaker has low harmonic distortion.
If the speaker is being over-driven or has a high amount of harmonic distortion, the Continuous Sweep results will be superior. Continuous sweep only looks at the fundamental frequencies and therefore is unaffected by harmonic distortion. MLS, on the other hand, does not exclude the harmonics, so frequency response and phase measurements will be affected.
Continuous Sweep also has the advantage that it returns 14 results at once (Energy Time Curve, Impulse Response, Level, Relative Level, Deviation, Delay, Phase, Group Delay, Level and Distortion, THD Ratio, THD Level, Distortion Product Ratio, Distortion Product Level, Acquired Waveform), instead of only two (level and phase) as with MLS.
When comparing, make sure that you use the same settings for time delay (to delay the acquisition until the first sound reaches the microphone) and acquisition length (to stop the acquisition just before the first reflection arrives).
Turn off phase wrapping, as the wrapping reference point is sensitive to the pickup of extraneous subsonic noise in the room and may vary between acquisitions. In AP2700, set the Phase meter to Auto, and in APx500, choose “Absolute” instead of "Absolute Wrapped" in the phase view.
Continuous Sweep (Chirp) Level, APx500 Series.
MLS Level, 2700 Series.
Continuous Sweep (Chirp) Phase, APx500 Series.
MLS Phase, 2700 Series.
Test Results: AP News & Events
Audio Precision has two open engineering positions (more details online):
Senior Software Engineer: Senior software development of extremely high performance digital audio test and measurement instrumentation and applications. Individual contributor as well as providing technical leadership to the software team.
Engineering Manager: Manage software and hardware engineering department for development of extremely high performance digital audio test and measurement instrumentation and applications. Leads cross-functional project teams to deliver quality products and features to the market on time. In conjunction with the Vice President of Engineering, develop long-range positions on programs or technologies for consideration of the Product Planning Team.
Both jobs are at the factory in Beaverton, Oregon. Competitive salary and benefits. Audio Precision is an Equal Opportunity Employer. See our Careers page for more details.
©2011 Audio Precision, Inc.