Pure tone testing is valid basis for gain optimization, but is not revealing. Two tone is little better.
E-MU 0404 USB is soundcard used at fs 48kHz:
Here is 19kHz+20kHz:
Much more revealing are results comparing swept sine v MLS. With full spectrum, IR returned by each looks nearly identical; HP behavior of coupling capacitors is revealed, as is FIR brickwall filtering used in this soundcard. This is not revealing either.
Swept sine of 131072 sample length and inverse are generated, and turned into 3x tandem repeat.
MLS 131071 sample sequence is turned into 3x tandem repeat.
Each signal has approximately one octave notch filter centered about 1kHz applied via FFT.
The two signals are captured by loopback using same output and input levels used for above two tone measurement.
IR is recovered for swept sine by convolution of inverse with test signal and loopback result.
Likewise, time reversed copy of 131071 sample MLS is convolved with test signal and loopback results for IR recorvery
Overlay of spectra for the four conditions:
The notches of the test signals are deep, with nearly identical spectral envelopes. For loopback results, notches become partially filled with harmonics and IMD components. Floor for swept sine conditions is close to floor seen with unmodified sweep's total harmonic distortion level. MLS doesn't show harmonics/IMD, but distributes these effects across time domain of result. The notch for MLS result is filled further, revealing demands of reproducing continuous broad band signal.
Signals at output of DAC may exceed values represented by sample data. With sine, swept sine this never happens. With 19kHz + 20kHz two tone test, it can, but in limited periodic fashion.
For used MLS, normalized sample values are all 0.5 and -0.5. Swept sine was used with same value range.
E-MU 0404 USB is soundcard used at fs 48kHz:
Here is 19kHz+20kHz:
Much more revealing are results comparing swept sine v MLS. With full spectrum, IR returned by each looks nearly identical; HP behavior of coupling capacitors is revealed, as is FIR brickwall filtering used in this soundcard. This is not revealing either.
Swept sine of 131072 sample length and inverse are generated, and turned into 3x tandem repeat.
MLS 131071 sample sequence is turned into 3x tandem repeat.
Each signal has approximately one octave notch filter centered about 1kHz applied via FFT.
The two signals are captured by loopback using same output and input levels used for above two tone measurement.
IR is recovered for swept sine by convolution of inverse with test signal and loopback result.
Likewise, time reversed copy of 131071 sample MLS is convolved with test signal and loopback results for IR recorvery
Overlay of spectra for the four conditions:
The notches of the test signals are deep, with nearly identical spectral envelopes. For loopback results, notches become partially filled with harmonics and IMD components. Floor for swept sine conditions is close to floor seen with unmodified sweep's total harmonic distortion level. MLS doesn't show harmonics/IMD, but distributes these effects across time domain of result. The notch for MLS result is filled further, revealing demands of reproducing continuous broad band signal.
Signals at output of DAC may exceed values represented by sample data. With sine, swept sine this never happens. With 19kHz + 20kHz two tone test, it can, but in limited periodic fashion.
For used MLS, normalized sample values are all 0.5 and -0.5. Swept sine was used with same value range.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.