Wirkungsweise und Messanleitung eines Audio Notches
Operating Principle and Measurement Guide for an Audio Notch Filter
Im Elektor 9/2022 stellte ich ein Audio Notch vor das von 2 hintereinander geschalteten Fliege Filtern gebildet wurde.
Hier folgt nun eine etwas ausführlichere Erklärung zur Wirkungsweise eines Audio Notches und eine Messanleitung wie man damit die Messung von Oberwellen in einem Audio Spektrums mit Hilfe eines Audio Analysers verbessern kann.
Um die Qualität eines Sinus Signals, zumeist 1 kHz, zu beurteilen schaut man sich am besten das Spektrum mit Hilde eines Audio Analysers an. Die Messgrößen THD (Total Harmonic Distortion), SNR (Signl to Noise Ratio) und THD+N (SINAD, Signal to Noise and distortion) geben zwar Hinweise wie gut ein Sinus ist, erst das Spektrum zeigt die ganze Wahrheit. Hier kann man unterscheiden ob es Störungen verursacht durch „Netzbrumm“, Oberwellen, Mischprodukte oder andere Störsignale (DCDCs etc) sind.
THD: Dieser Wert gibt an wie groß die Summe der Oberwellen ist im Vergleich zum Pegel des Messtons. Wie viele Oberwellen berücksichtigt werden ist Definitionssache. Der Wert gibt keinen Hinweis auf die Verteilung der Oberwellen. Rausch- und Störsignale werden nicht berücksichtigt.
SNR: Hier wird das Verhältnis des Rauschens, inklusive Störsignale, im betrachteten Frequenzbereich zum Signal angegeben. Oberwellen werden nicht berücksichtigt.
THD+N: Bei diesem Wert werden alle Signale (Oberwellen, Rauschen, Störsignale) berücksichtigt und ins Verhältnis zum Messsignal gesetzt.
SFDR: (Spurious Free Dynamic Range) Dieser Wert gibt den Abstand des Messignals zur größten Störfrequenz an. Dabei ist es egal ob es eine Oberwelle, Rauschen oder ein sonstiges Störsignal ist.
Die folgenden Messungen und Bilder wurden mit dem analogen Generator eines Audio Precision 2422 und dem Analyser eines Quantasylum QA403 gemacht.
Um bestmögliche Messergebnisse zu bekommen muss man die Aussteuerung des ADCs im Analyser optimieren. Übersteuert man ihn so wird das Signal oben und unten abgeschnitten und die Folge sind jede Menge große Oberwellen. Bild 1 zeigt das Spektrum eines heftig übersteuerten Signals.
Trotz optimalem Arbeitspunkt produziert der ADC und sein Ansteuerverstärker selbst Oberwellen, die man nicht von denen des Eingangssignals unterscheiden kann.
Messanleitung:
Im zweiten schließt man das Signal an den Eingang des Notches an und verbindet den umschaltbaren Ausgang des Notches mit dem Eingang des Analysers. Der Schalter am Notch muss dabei auf „Flat“ stehen. Mit Hilfe des im Notch verbauten Attenuators bringt man die Amplitude in den Bereich von -5 bis -10 dBV ( optimaler Arbeitsbereich des Notches) und merkt sich die Einstellung des Abschwächers. (das Notch muss natürlich eingeschaltet sein).
Im dritten Schritt wird der Schalter in die Stellung „Notch“ gebracht, die Amplitude des Testsignals hat sich von 7 dBV auf -52 dBV verringert, also um 59 dB. Bild 8 zeigt nun das Spektrum. Eventuell kann man die Frequenz des Testsignals verstellen um die bestmögliche Dämpfung zu erhalten, bzw. die Notchfrequenz einstellen, falls möglich. Das Notch ist sehr steil und selbst kleine Frequenzänderungen haben große Auswirkungen.
Am anderen Ausgang des Gerätes ist das Notch Signal um den Faktor 10 (+20 dB) verstärkt.
Bei korrekter Einstellung kann man so etwa 10 dB bessere THD Werte messen.
Hier zum Vergleich die Messungen mit dem Analyser des AP2422.
Die folgende Tabelle zeigt die Messwerte der beiden Analyser mit und ohne Notch. Quelle ist in allen Fällen der analoge Generator des AP2422.
English Text
In the September 2022 issue of *Elektor*, I presented an audio notch filter formed by two cascaded Fliege filters.
This was followed in September 2024 by an adjustable version featuring three frequency ranges and a state-variable topology.
However, the articles may not have fully covered the operating principle and practical application of the circuit.
Presented here is a more detailed explanation of how an audio notch filter works, along with instructions on using it to improve harmonic measurements within an audio spectrum using an audio analyzer.
To assess the quality of a sine wave signal—typically at 1 kHz—the best approach is to examine its spectrum using an audio analyzer. While metrics such as THD (Total Harmonic Distortion), SNR (Signal-to-Noise Ratio), and THD+N (SINAD/Signal-to-Noise and Distortion) provide indications of signal quality, only the spectrum reveals the full picture. It allows one to distinguish between disturbances caused by mains hum, harmonics, intermodulation products, or other interference sources (such as DC-DC converters).
The following is a brief explanation of these metrics:
THD: This value represents the sum of the harmonics relative to the level of the test tone. The number of harmonics included in the calculation depends on the specific definition used. This value provides no information regarding the distribution of the harmonics; noise and other interference signals are excluded.
With the QA403, the levels of individual harmonics can be displayed in a graph.
SNR: This indicates the ratio of noise—including interference signals—to the signal itself within a specific frequency range. Harmonics are excluded from this calculation.
An SNR value is of little significance without specifying the frequency range.
THD+N: This value takes all signals—harmonics, noise, and interference—into account and expresses them as a ratio relative to the test signal. In older audio analyzers, this value was determined using an analog method: the measurement signal was heavily attenuated, and the remaining signal was measured.
In modern analyzers, this value is also determined digitally from the spectrum. The bandwidth must also be specified alongside this value.
SFDR (Spurious-Free Dynamic Range): This value indicates the difference in level between the measurement signal and the strongest spurious frequency—regardless of whether that spurious signal is a harmonic, noise, or some other form of interference.
The following measurements and images were obtained using the analog generator of an Audio Precision 2422 and the analyzer of a Quantasylum QA403.
To achieve the best possible measurement results, the ADC drive level within the analyzer must be optimized. If the ADC is overdriven, the signal is clipped at the top and bottom, resulting in numerous high-amplitude harmonics. Figure 1 shows the spectrum of a severely overdriven signal.
Even without overdrive, an ADC's distortion characteristics change with the magnitude of the input signal. The optimal operating point should be determined by adjusting the signal in 1 dB steps (or finer increments) while monitoring the THD value.
If the input signal becomes too small, harmonics and other disturbances may disappear into the noise floor.
Figure 2 shows the spectrum of a 7 dBV sine wave measured with my QA403, with its attenuator set to 12 dB. The displayed SNR is -114.31 dB and the THD is -119.11 dB.
Changing the attenuator to 18 dB alters the values as shown in Figure 3. THD improved by nearly 4 dB to -122.9 dB—clearly visible in the level of the harmonic at 3 kHz. However, the SNR worsened to -110.63 dB, likely because the analyzer's own noise floor now has a greater impact on the measurement result.
This demonstrates that the optimal signal level for THD and SNR can indeed differ. For a given amplitude, the analyzer's input attenuator is used; on the QA403, this offers 6 dB steps ranging from 0 to 42 dB.
An averaging factor of 8 was used to reduce noise in the display, though this does not alter the measurement itself.
The QA403 allows for the display of a bar chart showing harmonic levels relative to the fundamental frequency (Figure 4).
Even at the optimal operating point, the ADC and its input driver amplifier generate harmonics that are indistinguishable from those present in the input signal.
This is where the notch filter comes into play.
When correctly tuned to the signal frequency, it attenuates the sine wave signal by 50 to 60 dB. This way, the ADC itself cannot generate its own harmonics from this now-small signal.
The exact amount of signal attenuation is not critical; the signal simply needs to be small enough.
Naturally, the notch circuit itself must be of sufficient quality to avoid introducing distortion.
Since distortion can occur if the input level is too high, it may be necessary to reduce the amplitude before the signal reaches the notch. Harmonics already present in the input signal are attenuated just as the input signal itself is.
Figure 5 shows the frequency response of a notch filter. The signal at 1 kHz is attenuated by approximately 50 dB.
However, the harmonics of the input signal (2 kHz, 3 kHz, etc.) are not attenuated by the notch. The filter must be "sharp" enough to suppress the signal as much as possible without altering the amplitude of the harmonics.
Measurement procedure:
First, measure the amplitude of the signal directly using the analyzer and note the level. If the signal exceeds 0 dBV, the analyzer's attenuator must be used. We continue to use the 7 dBV signal (see Figures 2 and 3).
Next, connect the signal to the notch filter's input and connect the notch's switchable output to the analyzer's input. Ensure the switch on the notch filter is set to "Flat." Using the notch's built-in attenuator, adjust the amplitude to the -5 to -10 dBV range (the notch's optimal operating range) and note the attenuator setting (the notch filter must, of course, be powered on).
In our example, the notch's attenuator is set to 12 dB. This reduces the original 7 dBV signal to -5 dBV. The QA403's attenuator can now be set to its minimum (0 dB). Figure 6 shows the reduced signal. THD remains in the -122 dB range. With the QA403, you can set the "external gain" to -12 dB to correct the level display.
Figure 7 shows the menu; the level once again displays the original 7 dBV, and there is an indication of the -12 dB external gain.
The automatic THD and SNR readings should be interpreted with caution, as the level attenuated by the notch filter can easily shift due to temperature drift.
At the device's other output, the notched signal is amplified by a factor of 10 (+20 dB).
To display correct levels, this 20 dB gain must be accounted for in the external amplification settings.
Figure 9 shows the spectrum without correction (Input Gain = -12 dB), while Figure 10 shows it with correction (Input Gain = +8 dB).
With the correct settings, you can measure THD values that are approximately 10 dB better.
Naturally, this requires a sufficiently pure input signal. Digital signal generators rarely meet this standard; an analog generator is a better choice here.
For comparison, here are measurements taken using the AP2422 analyzer.
Figure 11 shows the spectrum without using the AP2422's built-in notch filter.
The spectrum looks better with the notch filter engaged (Figure 12).
Comparing Figure 8 with Figure 12 reveals that the QA403, when paired with an external notch filter, compares favorably to an Audio Precision unit—albeit an older model.
The following table lists the measurements from both analyzers, with and without the notch filter. In all cases, the signal source is the AP2422's analog generator.
Admittedly, this procedure is quite complicated just to achieve a few decibels of improvement in THD readings.
You can easily bypass this effort by purchasing an Audio Precision APx555B for over €40,000. It contains both an analog generator and the notch filter, and is currently the best thing you can buy.
Want to build a project?
Bring your design to life with the Elektor PCB Service, powered by Eurocircuits. Upload the project files and order professionally manufactured PCBs or assembled boards through a proven European production platform.
Supporting KiCad, Eagle, Gerber, and ODB++ formats, the service is suitable for everything from prototypes and validation builds to series production and volume manufacturing.
Made in Europe. Fast. Reliable. Professional.

Diskussion (0 Kommentare)