Following @DcibeL post about using FSAF measurements across an Rsense of 0.1 Ohm, i open this thread.
First challenge was to find the latest Rew Beta, finally succeeded (only this morning), now i have to get into understanding the set up of Rew in combination with a RME Babyface pro FS.
Somehow my logic and that of Rew do not match very well yet, iow i have difficulty in getting it set up and calibrated.
But my checklist, once it is working, is to check the noise(or garbage ?) not only for drivers but also for the connecting cables.
With this simple circuit...
...and the latest REW beta, the FSAF measurement method allows for one to run this measurement with real audio, and listen to the residual distortion produced. No microphone required, just a soundcard, a resistor and a patch cord. It's not exactly the same as an acoustic measurement, but still provides a lot of valuable insight into driver performance, and enables comparisons between drivers, and various filters or even cabinet arrangements.
For a simple example, a recording of a Wavecor WF120BD04 is attached, tested at 2.8V free air. You can...
...and the latest REW beta, the FSAF measurement method allows for one to run this measurement with real audio, and listen to the residual distortion produced. No microphone required, just a soundcard, a resistor and a patch cord. It's not exactly the same as an acoustic measurement, but still provides a lot of valuable insight into driver performance, and enables comparisons between drivers, and various filters or even cabinet arrangements.
For a simple example, a recording of a Wavecor WF120BD04 is attached, tested at 2.8V free air. You can...
First challenge was to find the latest Rew Beta, finally succeeded (only this morning), now i have to get into understanding the set up of Rew in combination with a RME Babyface pro FS.
Somehow my logic and that of Rew do not match very well yet, iow i have difficulty in getting it set up and calibrated.
But my checklist, once it is working, is to check the noise(or garbage ?) not only for drivers but also for the connecting cables.
Member
Joined 2003
FSAF, harmonic distortion, intermodulation, all are possible with this simple circuit. I will provide some detailed analysis soon, it may take a few days to assemble.
For thee uninitiated, FSAF allows for testing of a loudspeaker with any audio, which results in a "total distortion + noise" result, which includes the harmonic distortion, intermodulation, barkhousen noise and background noise products. It allows for measurement with real audio signal, ensuring test signal crest factor represents real world situations. With a sine sweep measurement, crest factor is very low compared to real audio, and spectrum is not exactly representing real audio either, so usually measurements at multiple levels are required to provide enough data for proper interpretation.
The magic in the FSAF sauce is that it provides an audio file of just the residual distortion products, allowing one to hear the distortion generated by a speaker standing alone without the fundamental audio. No longer do you have to try and carefully pick apart audio details of <1% of the audio, with a single measurement you can listen to just the <1% audio on it's own, which can be quite enlightening. It is also able to produce an accurate frequency response for a speaker, even without exciting the full audio spectrum. It's amazing to test a speaker with a 1kHz high pass and still obtain a perfect frequency response to 20Hz.
Using FSAF as an acoustic measurement with a microphone has some downsides, however. It requires a good mic with low self noise and low distortion at fairly high SPL, a timing locked interface (ie USB mics are out), and it required a very quiet measurement space with good acoustic treatment, so background noise and room reverb doesn't contaminate the results. Not and easily accomplished task for many people.
This electrical measurement provides detailed insight into the distortion products generated by a loudspeaker motor. The small 0.1 ohm resistor allows for a low level voltage pickup that captures all the back EMF captured in the voice coil. It allows for detailed insight into driver performance, allowing for listening comparison of the residual distortion audio without a microphone, and for very low cost. A measurement with better SNR than an acoustic measurement is possible, and without needing the quiet recording studio environment, an average noisey household will do just fine. It enables easy driver comparisons with various filters and even cabinet types. A timing locked interface is still required, I would recommend a USB audio interface so you have an adjustable input gain to maximize SNR. This electrical measurement is not able to capture the driver's frequency response, it will follow the inverse of the driver's impedance, as the 0.1 ohm resistor effectively creates a voltage divider circuit.
More to come, stay tuned.
(Of course, first steps should be a loopback measurement, to ensure a linear frequency response, low distortion and good SNR is obtained without a speaker driver in place. Use a 4-10 ohm resistor in place of the driver for this loopback test)
For thee uninitiated, FSAF allows for testing of a loudspeaker with any audio, which results in a "total distortion + noise" result, which includes the harmonic distortion, intermodulation, barkhousen noise and background noise products. It allows for measurement with real audio signal, ensuring test signal crest factor represents real world situations. With a sine sweep measurement, crest factor is very low compared to real audio, and spectrum is not exactly representing real audio either, so usually measurements at multiple levels are required to provide enough data for proper interpretation.
The magic in the FSAF sauce is that it provides an audio file of just the residual distortion products, allowing one to hear the distortion generated by a speaker standing alone without the fundamental audio. No longer do you have to try and carefully pick apart audio details of <1% of the audio, with a single measurement you can listen to just the <1% audio on it's own, which can be quite enlightening. It is also able to produce an accurate frequency response for a speaker, even without exciting the full audio spectrum. It's amazing to test a speaker with a 1kHz high pass and still obtain a perfect frequency response to 20Hz.
Using FSAF as an acoustic measurement with a microphone has some downsides, however. It requires a good mic with low self noise and low distortion at fairly high SPL, a timing locked interface (ie USB mics are out), and it required a very quiet measurement space with good acoustic treatment, so background noise and room reverb doesn't contaminate the results. Not and easily accomplished task for many people.
This electrical measurement provides detailed insight into the distortion products generated by a loudspeaker motor. The small 0.1 ohm resistor allows for a low level voltage pickup that captures all the back EMF captured in the voice coil. It allows for detailed insight into driver performance, allowing for listening comparison of the residual distortion audio without a microphone, and for very low cost. A measurement with better SNR than an acoustic measurement is possible, and without needing the quiet recording studio environment, an average noisey household will do just fine. It enables easy driver comparisons with various filters and even cabinet types. A timing locked interface is still required, I would recommend a USB audio interface so you have an adjustable input gain to maximize SNR. This electrical measurement is not able to capture the driver's frequency response, it will follow the inverse of the driver's impedance, as the 0.1 ohm resistor effectively creates a voltage divider circuit.
More to come, stay tuned.
(Of course, first steps should be a loopback measurement, to ensure a linear frequency response, low distortion and good SNR is obtained without a speaker driver in place. Use a 4-10 ohm resistor in place of the driver for this loopback test)
Last edited:
To set up REW in combination with an RME Babyface Pro FS, i can only use the Java driver and not Asio for some unknown reason
So continued with Java driver, which appears to be limited to 192kH sampling rate.
So first used Generator to create a sinus and measure first with a true RMS multimeter the Vrms, then reconnect out to in and check with applied voltage. Was off, so i used Calibrate, external signal and applied a known calibrated mic output voltage via genarato and in Calibrate recored it as 94dB SPL.
Now it appears to be ok. Now measure says indeed 94dB.
Still i cannot alter the sampling rate, any suggestions?
So continued with Java driver, which appears to be limited to 192kH sampling rate.
So first used Generator to create a sinus and measure first with a true RMS multimeter the Vrms, then reconnect out to in and check with applied voltage. Was off, so i used Calibrate, external signal and applied a known calibrated mic output voltage via genarato and in Calibrate recored it as 94dB SPL.
Now it appears to be ok. Now measure says indeed 94dB.
Still i cannot alter the sampling rate, any suggestions?
In the upper left part of the Distortion window there is a pull-down menu where you could set the Y axis to show dBv which perhaps is more relevant?
//
//
In my case, to make RME (fireface UCX in my case) work in Asio and recognizable by REW, I just downloaded drivers, and later many other goodies, from RME site, it is all free.
Member
Joined 2003
Start with basics, REW warned you that input level of -56dBFS is very low, so perhaps set input gain appropriately. Output level is set to 1.5mV? Have you calibrated for voltage level? Why so low?
For a loopback test, noise would be more appropriate stimulus for determining successful loopback. How do you interpret SPL scale for an electrical signal? Scale of dBr with full spectrum noise signal, and check the box to show noise floor. Of course, check SPL tab for FR transfer function.
For a loopback test, noise would be more appropriate stimulus for determining successful loopback. How do you interpret SPL scale for an electrical signal? Scale of dBr with full spectrum noise signal, and check the box to show noise floor. Of course, check SPL tab for FR transfer function.
Yes, i was just testing if calibration of the voltage level and the SPL level for given voltange was correct, and that the FSAF worked. The 14.6mV is the voltage value (sensitivity) for 94dB for my calibrated Beyerdynamic MM1 mic. I always verify that the calibration is correct ;-)
The piece of music apparently has a much higher voltage.
Here the 14.6mV with noise ticked:
So mostly noise, with some spikes .
Was hindered in my plans for today, so no more results for now.
Member
Joined 2003
Before going down the FSAF path, let's start with the case for using an electrical measurement as a proxy for an acoustic one. This comparison will use a sine sweep for standard harmonic distortion.
Taking accurate acoustic distortion measurements can be quite an ordeal. It required a good quality microphone to capture low distortion products accurately, and special conditions have to be set up for recording, a quiet space, well treated room, etc. The biggest problem for low distortion measurements is ensuring that the noise floor is below the signal being recorded, which can be difficult acoustically when the signal is approaching the same magnitude of the background noise. Additionally, condenser mics have a rising noise floor towards low frequencies, so low frequency measurements are particularly troublesome. If a similar result can be obtained electrically, we can avoid some of these acoustic pitfalls. I'm sure most people would prefer to not have to lug their test equipment to the room in their house with the best acoustics, and kick everyone out so you can have a nice quiet space to gather data.
A comparison was made using a budget midwoofer - Dayton DA115-8. The driver was placed in a test cabinet, and measured with a Line Audio Omni1 at 315mm distance, and 2.8V level. The acoustic conditions were less than ideal, set up in my office space so there are some early reflections and background noise from the PC fans present, however this is the most convenient location for driver testing.
Here's the acoustic measurement:
This measurement was repeated using the 0.1 ohm sense resistor as the input:
At first glance, you may think that these are wildly different results, however the reality is that they convey much of the same information, it's just that the way they are displayed doesn't make for a good comparison. You may also notice that the noise floor of the electrical measurement is much lower, in the order of 40dB better at 100Hz. The mic measurement displays the acoustic frequency response of the speaker, however the electrical measurement shows an inverse of it's impedance. This is simply due to the fact that the sense resistor effectively creates a voltage divider between the speaker and the resistor.
A traditional HD measurement shows harmonic information at the fundamental frequency that generates the harmonic, and the magnitude of the harmonic is shown relative to the same fundamental frequency. This is a bit problematic, as speakers are anything but linear. If we change the display of the measurement to normalize the frequency response, and show the harmonic levels relative to their harmonic frequency, we get a much different display.
Acoustic measurement:
Sense resistor measurement:
Now things are starting to appear a lot similar. Please forget for a moment about 200-300Hz range, there is evidently something buzzing during the acoustic measurement (highlighting downsides of acoustic measurements).
We can use REW's overlay tool to compare each harmonic directly.
2nd Harmonic: Perhaps not the greatest agreement here, however fairly good through the midrange passband that the driver would be used. Most people seem not that interested in 2nd order anyway.
3rd harmonic, very good agreement.
4th harmonic:
5th harmonic:
From this comparison, I hope to convey that the electrical measurement does in fact provide useful data for loudspeaker comparison without using a microphone.
Taking accurate acoustic distortion measurements can be quite an ordeal. It required a good quality microphone to capture low distortion products accurately, and special conditions have to be set up for recording, a quiet space, well treated room, etc. The biggest problem for low distortion measurements is ensuring that the noise floor is below the signal being recorded, which can be difficult acoustically when the signal is approaching the same magnitude of the background noise. Additionally, condenser mics have a rising noise floor towards low frequencies, so low frequency measurements are particularly troublesome. If a similar result can be obtained electrically, we can avoid some of these acoustic pitfalls. I'm sure most people would prefer to not have to lug their test equipment to the room in their house with the best acoustics, and kick everyone out so you can have a nice quiet space to gather data.
A comparison was made using a budget midwoofer - Dayton DA115-8. The driver was placed in a test cabinet, and measured with a Line Audio Omni1 at 315mm distance, and 2.8V level. The acoustic conditions were less than ideal, set up in my office space so there are some early reflections and background noise from the PC fans present, however this is the most convenient location for driver testing.
Here's the acoustic measurement:
This measurement was repeated using the 0.1 ohm sense resistor as the input:
At first glance, you may think that these are wildly different results, however the reality is that they convey much of the same information, it's just that the way they are displayed doesn't make for a good comparison. You may also notice that the noise floor of the electrical measurement is much lower, in the order of 40dB better at 100Hz. The mic measurement displays the acoustic frequency response of the speaker, however the electrical measurement shows an inverse of it's impedance. This is simply due to the fact that the sense resistor effectively creates a voltage divider between the speaker and the resistor.
A traditional HD measurement shows harmonic information at the fundamental frequency that generates the harmonic, and the magnitude of the harmonic is shown relative to the same fundamental frequency. This is a bit problematic, as speakers are anything but linear. If we change the display of the measurement to normalize the frequency response, and show the harmonic levels relative to their harmonic frequency, we get a much different display.
Acoustic measurement:
Sense resistor measurement:
Now things are starting to appear a lot similar. Please forget for a moment about 200-300Hz range, there is evidently something buzzing during the acoustic measurement (highlighting downsides of acoustic measurements).
We can use REW's overlay tool to compare each harmonic directly.
2nd Harmonic: Perhaps not the greatest agreement here, however fairly good through the midrange passband that the driver would be used. Most people seem not that interested in 2nd order anyway.
3rd harmonic, very good agreement.
4th harmonic:
5th harmonic:
From this comparison, I hope to convey that the electrical measurement does in fact provide useful data for loudspeaker comparison without using a microphone.
Member
Joined 2003
For a similar comparison of a more premium driver. I completed the same test using a Wavecor WF120BD04 - similarly sized to the Dayton above.
Acoustic measurement:
Electrical measurement:
Again, everything is well aligned except for 2nd order. 4th order is difficult to compare on this driver at it approaches the noise floor.
Acoustic measurement:
Electrical measurement:
Again, everything is well aligned except for 2nd order. 4th order is difficult to compare on this driver at it approaches the noise floor.
Member
Joined 2003
Moving on to FSAF, the two drivers were tested using a sample of a popular song. I tested each driver playing free air, at 2.8V there is maybe 1mm of driver movement through the bass, so they are both playing well below their xmax limitations. I then retested using a LR-24 high pass at 200Hz.
the FSAF distortion chart provides a spectrum analysis of the residual audio for the entire 15s of playback. This result is "total distortion + noise" as it includes all harmonic, intermodulation,, barkhousen noise and background noise products, basically any audio recorded that isn't in the original audio stimulus. This is useful information, but it is difficult to determine as there can be many causes for a change in total distortion.
Comparing WF120BD04 with and without a high pass filter, we see about a 5dB reduction in total distortion from 1kHz+
If we were to limit a sine sweep measurement from 200Hz+, we would obtain the exact same distortion plot, so this difference can't be a result of harmonics. What is occurring here is intermodulation, and it is present even at very small driver movement. Let's see what happens when we add a low pass filter at 3kHz.
With a bandpass (LR-24 at 200Hz and 3kHz), total distortion at high frequencies is further reduced, as there is less signal to modulate, however the TD through the passband remains relatively unchanged.
For good measure, the frequency response transfer functions for the above:
Completing the same comparison with the DA115 budget driver, we see less of an improvement, as the midrange also includes high harmonic content.
the FSAF distortion chart provides a spectrum analysis of the residual audio for the entire 15s of playback. This result is "total distortion + noise" as it includes all harmonic, intermodulation,, barkhousen noise and background noise products, basically any audio recorded that isn't in the original audio stimulus. This is useful information, but it is difficult to determine as there can be many causes for a change in total distortion.
Comparing WF120BD04 with and without a high pass filter, we see about a 5dB reduction in total distortion from 1kHz+
If we were to limit a sine sweep measurement from 200Hz+, we would obtain the exact same distortion plot, so this difference can't be a result of harmonics. What is occurring here is intermodulation, and it is present even at very small driver movement. Let's see what happens when we add a low pass filter at 3kHz.
With a bandpass (LR-24 at 200Hz and 3kHz), total distortion at high frequencies is further reduced, as there is less signal to modulate, however the TD through the passband remains relatively unchanged.
For good measure, the frequency response transfer functions for the above:
Completing the same comparison with the DA115 budget driver, we see less of an improvement, as the midrange also includes high harmonic content.
Member
Joined 2003
Okay, so looking at charts and graphs is fun and all, but what does all this really sound like?
Here you go, download my measurements and listen for yourself:
https://drive.google.com/file/d/1HYiV3uP_RX_UQEFkGIesqNXnM-UQ8i_H/view?usp=sharing
To listen, download latest REW beta, open the MDAT file, go to distortion tab, controls, then pick either "Play FSAF Residual" or save file for your own analysis. Recommend using 40dB gain for playback. For analyzing the residual audio, viewing in a spectrogram can be quite useful.
Get ready to declare both of these driver as horrible 😉
Hopefully this has conveyed enough information to get some others testing and listening to their drivers, and comparing variouss real world implementations. Cost of entry is a good soundcard, I recommend a USB audio interface so you have adjustable input gain control, and a resistor and some wire.
Here you go, download my measurements and listen for yourself:
https://drive.google.com/file/d/1HYiV3uP_RX_UQEFkGIesqNXnM-UQ8i_H/view?usp=sharing
To listen, download latest REW beta, open the MDAT file, go to distortion tab, controls, then pick either "Play FSAF Residual" or save file for your own analysis. Recommend using 40dB gain for playback. For analyzing the residual audio, viewing in a spectrogram can be quite useful.
Get ready to declare both of these driver as horrible 😉
Hopefully this has conveyed enough information to get some others testing and listening to their drivers, and comparing variouss real world implementations. Cost of entry is a good soundcard, I recommend a USB audio interface so you have adjustable input gain control, and a resistor and some wire.
Last edited:
Member
Joined 2003
To provide a bit of clarity on the downsides of this "current sense" measurement. Since a driver is not a resistor, the voltage across the 0.1 ohm resistor, or the current through it, will vary slightly over the frequency range. This is why we see an inverse of the driver impedance as the frequency response for this measurement. The small 0.1 ohm value is used to minimize this behaviour, as well as provide a reasonable signal level for direct connection to a soundcard. At low frequency, the result is somewhat impacted by the driver Fs. There may be a better way to provide a true "current sense" function, however it would be much more complex than a single resistor.
I finally got to assembling testrig and run the FSAF sweeps.
Testrig, simplified schema:
Attached the latest mdat file. The 'res' wav file is too large to attach, so a screencopy of the distortion graph:
Be aware this is a SB26ADC tweeter with waveguide. Where the cursor is, is the resonance frequency. Normally i use this with a very steep active hi-pass filter at ~ 3464Hz. Not yet figured out how to include an hi-pass filter in REW.
If i play the 'res' file with a media player at 100% volume it is barely audible, like in the DCibel example i can vagely hear the music, but it is too soft to make sense to me.
Also i used the Arta suite to measure the current distortion with Steps, this example is with 1.7Vrms:
So technically it is fuctioning, also for FSAF, still to figure out a higher level of playback of the 'res' file.
Testrig, simplified schema:
Attached the latest mdat file. The 'res' wav file is too large to attach, so a screencopy of the distortion graph:
Be aware this is a SB26ADC tweeter with waveguide. Where the cursor is, is the resonance frequency. Normally i use this with a very steep active hi-pass filter at ~ 3464Hz. Not yet figured out how to include an hi-pass filter in REW.
If i play the 'res' file with a media player at 100% volume it is barely audible, like in the DCibel example i can vagely hear the music, but it is too soft to make sense to me.
Also i used the Arta suite to measure the current distortion with Steps, this example is with 1.7Vrms:
So technically it is fuctioning, also for FSAF, still to figure out a higher level of playback of the 'res' file.
