So, when you know the distortion residuals, can you generate a pre-corrected signal and listen to an "undistorted" one (and compare)? That would make sense to me. I don't quite see a point in listening the distortion alone, as it still says very little about its actual audibility. We know the distortion is there. The question is when it becomes audible.
Last edited:
Yes! The possibilities (and limitations) have not even been explored.
@mikets42 was doing it solo via Matlab until it was incorporated in REW V5.40 beta 32 just 15th June 2024
Download and
for @JohnPM and diyAudio forum if you find this interesting or useful.
I have been wrapping my head around it at the support thread
@mikets42 was doing it solo via Matlab until it was incorporated in REW V5.40 beta 32 just 15th June 2024
Download and
for @JohnPM and diyAudio forum if you find this interesting or useful.I have been wrapping my head around it at the support thread
Last edited:
Tranh, I note from that thread you established in Feb and JuL this year, that Angelo's method gives the same results as Stepped Sine for response & THD ... bearing in mind the windowing and smoothing bla bla issues which are well known and documented.
But has anyone done THD with Angelo's method and compared them with THD using FSAF with the same length of signal? ALL the examples in that thread have FSAF THD results VERY different from 'traditional' methods and/or Angelo's.
Last edited:
You can do this with very short Log Sweeps of 1s or less with Angelo's method. If you vary the length of the sweep, you can easily show how non-time-invariant speakers are .. especially HF units.
If you use the same sweeps to measure distortion as well, you are well on your way to proving speakers are NOT LTI ... like us EVIL speaker designers like to assume 😊
Please use an appropriate xover when testing HF units at high power
Exactly.
During that ‘alignment’ process a lot of things went wrong. You can see sm52 and myself becoming beta testers
I learnt that the mic distortion at 100dB SPL needs to be really low. The background noise floor needs to be really low. The amp, DAC, ADC also matters here…
I recently discovered that my audio interface drops samples on an intermittent basis, despite changing drivers and maxing out on the buffer size.
Going forward, I will probably need to drop the Focusrite. It’s those intermittent problems that are most maddening!!
Member
Joined 2003
FSAF is not a THD result. It's total distortion, includes harmonics, intermodulation, and noise. It also is intended to be used with different stimulus than a sine sweep, so the test signal has a much greater creat factor for better correlation with real audio, or the test can simply be real audio.
But, you can run any comparison you like for free, just download latest REW beta and have at it.
Does this mean no one else has tried to do this? It seems Tranh has problems doing this too.
BTW, these problems are important for Angelo's method too but loadsa people, including AP, Audiomatica, REW and even beach bums in Oz seem to have solved them 😊
The issue is that a new fangled method should give 'similar' results to old fogey methods under the same 'old fogey' conditions. We know FSAF can do this for response. But so can loadsa 'old fogey' methods very efficiently ... eg using music / noise bla bla for response.
What I'm trying to establish is if FSAF can measure "total distortion, includes harmonics, intermodulation, and noise" and give the same results as 'old fogey' methods.
Last edited:
Thanks for this vineethkumar. Are you a modern DSP guru?
If so, can you explain how the subband decimation is done (FIR, IIR, black magid?) and especially how this allows 'perfect' reconstruction? Da free Subband Filtering articles I've managed to find don't make this claim.
Slowly working through da 200+ pages but my single remaining brain cell has problems
I'm only a pseudo wannabe DSP guru from da 80s & 90s and have problems pre10ding to unnerstan dis modern DSP s**t.
Member
Joined 2003
It doesn't sound like you've tried anything to be honest. If you really want to know, just run your own tests and find out.
I'd run some tests of my own today, but the power is out, so I have an excuse.
Peace Dcibel. I've no axe to grind. I'm only interested cos I developed Angelo's method in da 90s.
These days I'm a real beach bum so gain nothing from being proved right or wrong. But Tranh's problems give me a good idea of how difficult it is get sensible answers to my question 😊 Don't forget I've done something similar before. Beach bums are lazy.
But if someone shows FSAF can reproduce 'old fogey' THD results, I might get off my butt. Doesn't seem to be anyone ATM unless you know more.
I'll wait for your tests with bated breathe.
Last edited:
@kgrlee
It's not problems. It's challenges. Like when your soundcard drops samples and gives spurious results.
Once if a blue room it does this (Focusrite's ASIO drivers on my system, (YMMV)
Or even after using third party ASIO drivers, you get a clipped sample (thanks @DcibeL who pointed this out to me in this sample
Or when you don't have your gain settings set correctly, or you can't (yet) compare like for like. (2 drivers should be compared with matched SPL).
So you have to get things correct before you publish spurious results that can't be retracted after 30 minutes.
Yikes!
@DcibeL @fluid
Sonarworks Xref20 not suitable
Thomann PM40c not suitable
ID24 on the way.
Sometimes a leader steps up and courageously leads the way. (Mathematician/DSP expert Mike Zrull)
But did we scare him off?
Doth he darpteth, forsooth, for all inquire mundane queries? Great minds do delight in fellowship.
(Did he leave because he was bored/annoyed here? Genius loves company)
The rest of us just need to have show that's it's not such a scary journey. And perhaps people will consider coming out to have a look around, before venturing out (a.ka. download and try the bloody thing!)
Together We’ll Get There.
It's not problems. It's challenges. Like when your soundcard drops samples and gives spurious results.
Once if a blue room it does this (Focusrite's ASIO drivers on my system, (YMMV)
Or even after using third party ASIO drivers, you get a clipped sample (thanks @DcibeL who pointed this out to me in this sample
Or when you don't have your gain settings set correctly, or you can't (yet) compare like for like. (2 drivers should be compared with matched SPL).
So you have to get things correct before you publish spurious results that can't be retracted after 30 minutes.
Yikes!
@DcibeL @fluid
Sonarworks Xref20 not suitable
Thomann PM40c not suitable
ID24 on the way.
Sometimes a leader steps up and courageously leads the way. (Mathematician/DSP expert Mike Zrull)
But did we scare him off?
Doth he darpteth, forsooth, for all inquire mundane queries? Great minds do delight in fellowship.
(Did he leave because he was bored/annoyed here? Genius loves company)
The rest of us just need to have show that's it's not such a scary journey. And perhaps people will consider coming out to have a look around, before venturing out (a.ka. download and try the bloody thing!)
Together We’ll Get There.
Last edited:
Exactly. You have these problems with Angelo's method too. I hope you solve them for FSAF 😊
Yes, if it wasn't for you, my Ozzy fellow beach bum friend (and @bwaslo ) I now have 4133+2639 & 4191+2669 to take more serious measurements...
Member
Joined 2003
My power is back on, but not much energy left today. Here's some simple loopback measurement at low level, no efforts made to maximize SNR, etc.
Again with a low cost amplifier in the loop. The amp primarily adds some power supply noise. A bit messy on HD chart.
If we plot the harmonics at the harmonic frequency, instead of at the fundamental frequency, the image is much clearer on HD.
FSAF shows this noise clear as day however, and the relative level is quite a bit elevated too.
Again with a low cost amplifier in the loop. The amp primarily adds some power supply noise. A bit messy on HD chart.
If we plot the harmonics at the harmonic frequency, instead of at the fundamental frequency, the image is much clearer on HD.
FSAF shows this noise clear as day however, and the relative level is quite a bit elevated too.
Member
Joined 2003
I had also tested a couple of my own speakers previously. I was focusing on high frequency performance at the time, so the plot is from 800Hz+. Speaker output level is 90dB/1m.
Here's a different speaker, same test conditions and level.
Sorry, I hadn't completed any inter-modulation testing of the above, so it's not clear to me why the second speaker shows a higher level of distortion on FSAF than the first.
Unfortunately since FSAF is a total distortion metric, the chart itself doesn't provide much insight into the cause of the distortion, what order, etc., so it's not going to replace HD or IMD testing in my view, but a useful tool in addition to the traditional tests. Perhaps a much better replacement for a THD+N metric, which was honestly pretty useless to begin with.
Here's a different speaker, same test conditions and level.
Sorry, I hadn't completed any inter-modulation testing of the above, so it's not clear to me why the second speaker shows a higher level of distortion on FSAF than the first.
Unfortunately since FSAF is a total distortion metric, the chart itself doesn't provide much insight into the cause of the distortion, what order, etc., so it's not going to replace HD or IMD testing in my view, but a useful tool in addition to the traditional tests. Perhaps a much better replacement for a THD+N metric, which was honestly pretty useless to begin with.
Last edited:
The closest to FSAF total distortion would be REW's THD(H2..9) curve which I think sums the power of the 2nd to 9th harmonics.
You can estimate quite accurately what this is by visual inspection eg if 3rd is 10dB below 2nd, 2nd+3rd is 1dB above the 2nd curve. If 4th crosses that 2nd + 3rd line, then 2nd+3rd+4th is 3dB above that etc.
I'm not sure how the REW noise curve is done but it should have minimal effect on THD. on the examples here.
BTW, THD is by definition, Distortion + Noise. This is now rather outdated as most music sources are severely band-limited .. which affects the noise measurement
What parameters have you used for your ESS tests? Your FSAF tests?
ATM, I don't think your FSAF THD measurements have any relationship to Distortion measured by Angelo or Stepped Sine ... but neither have any others I've seen. Maybe you, Tranh and everyone else has problems getting FSAF THD to work properly 😊