Mate! Doing outdoors measurements can be bad for ye health!.⚠️ what with the cane toads, snakes, bats ruining your fun important R&D.
If you’re on typical Darwin “acreage” like me mates, you don’t need no stinkin’ no echo room.
But if ya got two adjoinin rooms, one bigger than the other- you can make ya own no echo room- get some wedges and turn thatspare room lab into an small no echo room, just like me mates over at Purifi:
Your next speaker? @Dave Zan
“It’s going straight to the poolroom!”
Actually I do most of me thinkin in what the kids call mad dad lab.
PS. I’m on the case! In coming weeks I’ll let you know about this log swine sweep vs stepped sine thingymabob. Thanks to me mates @Hörnli @IamJF
Let me see if I can measure that 8” Purifi like me mate @lrisbo
Ubeaut Ushindi ideas
If you’re on typical Darwin “acreage” like me mates, you don’t need no stinkin’ no echo room.
But if ya got two adjoinin rooms, one bigger than the other- you can make ya own no echo room- get some wedges and turn that
Your next speaker? @Dave Zan
“It’s going straight to the poolroom!”
Actually I do most of me thinkin in what the kids call mad dad lab.
PS. I’m on the case! In coming weeks I’ll let you know about this log swine sweep vs stepped sine thingymabob. Thanks to me mates @Hörnli @IamJF
Let me see if I can measure that 8” Purifi like me mate @lrisbo
- Ultra Low HD: e.g., -60/-70dB (HD2/HD3) @ 1kHz
- Ultra Low IMD: e.g., -50dB @ 30Hz/255Hz
Last edited:
The current project is the classic (imperial measurement units) 2 way 15" driver + horn loaded mid :
15" JBL 2226 woofers in 4 cu ft closed box.
4" VC JBL 245x compression drivers on substantial horn.
Plus JBL 18" 2445 subs in 10 cu ft vented box.
The subs and woofer boxes are finished, need to build the mid horns or buy JBL 2384.
I have JBL 2405 slot horn drivers for super tweeters if I need them.
I like your pic, I was very impressed with what you had done to your place, until I read the small print.🙂
It did make me realise that I could actually do that with my spare rooms.
Best wishes
David
Last edited:
No it's not, that won't run without an entire environment, software, libraries etc.
What are the details, versions, set up options?
Perhaps more than you seem to expect, a lot of statistics is concerned with time series and my prior consultancy was analysis of employment time series, we considered ARIMA and SARIMA.
I even did my introductory presentation to the team about the similarities of the stats we needed for our econometric analysis and the statistics of loudspeaker analysis, that was fun.
"[F]ast methods", not so much, our data volumes didn't need it, at least not once I cleaned up the old SAS and SQL code.
But I know they exist and have some limited familiarity, please share your expertise.
Best wishes
David
Last edited:
D
Deleted member 375592
I must apologize for the confusing explanations ahead of time. I started talking at 15 and till now think in pictures. Their verbalization is a bit hard for me.
There are traditional methods of System Identification which are Best Linear Unbiased Estimators but they are horribly slow. Processing 15 seconds of response to an arbitrary input takes ~40 hours on i7-11700K. There are faster methods that work only on certain inputs, like Exponential Sine Sweep (ESS). To find an Impulse Response (IR) all you need to do is to feed this ESS to a spk and record mic. Then you convolve the mic signal with the ESS - and voila, Room IR (RIR) is here. Thanks to the fast methods of convolution developed in the 60s, convolution is blindingly fast.
If you convolve the original ESS with mic, you get RIR. If you double the frequency of ESS, you convolve the "double-frequency" ESS with mic and get 2nd harmonic time domain H2IR. If you triple the frequency - 3rd, H3IR. etc. Of course, you need to avoid frequency wrapping and cut the "double / triple /etc frequency ESS" when the instantaneous frequency crosses Nyquist. Then you convert the time-domain HxIR-s to the frequency domain and - voila! You get the "best" matched filter estimated harmonics. You can de-noise HxIR by standard methods like Ephraim-Malah or whatever else you prefer, and get the harmonic info from under the noise. It is way better than tracking the top of the ridge as Angelo Farina suggested originally.
Although I am not aware of previous publications of this method - you never know... well maybe I have [shamefully] reinvented a bicycle. If so, please excuse me.
Historic note: The first of such inputs was linearly swept sine, aka Chirp, whose autocorrelation function is Dirac's Delta function. It was discovered by Darlington in 1947 for cold-war radar "pulse spreading". Since then, such functions and sequences have been called "perfect". MLS is near-perfect. They are widely used, from GPS to mobile phones, behind the scenes. Since then, many other near-perfect functions have been discovered.
Thanks,
Michael
OFFTOPIC ON
The equation is simple: either you are rich and have access to AC, or you die.
Not a good time to close your eyes, burn fuel and be sarcastic about it.
OFFTOPIC OFF
I am the first to enjoy sarcasm. But this is no fun, and too many people don't get it. Temperatures are breaking high records around the globe. Remember India last summer? Rio de Janeiro above 60°C?
The equation is simple: either you are rich and have access to AC, or you die.
Not a good time to close your eyes, burn fuel and be sarcastic about it.
OFFTOPIC OFF
Political campaigning is not allowed on this forum. The public policy debate regarding climate change is very much political. I am sure if I started quoting Bjorn Lomborg or Steven Koonin on the subject, I would be flagged. So let's keep it about DIY speakers...
Just to be sure I get this right: is your idea to repeat the same measurement several times (using the same test signal), and then determine the mean to reduce the (random!) noise? Sure, that works.
More or less.
Partly inspired by the fact that @IamJF mentioned the noise floor of the mike without discussion.
In the old days (pre DAC and DSP) the noise floor of the mike was essentially an absolute limit, nowadays it is not.
I am very sure he knows that but the discussion was incomplete without a mention of the possibilities.
Secondly, I was interested to discuss those possibilities.
How does "repeat the same measurement and determine the mean" compare with make the sweep slower?
A slower sweep takes more time and should reduce the noise but also have potentially better frequency resolution.
So "repeat the same measurement several times and determine the mean" will work but may be sub-optimal?
I suspect so but not quite sure, discussion of this stuff makes it clearer.
Best wishes
David
@mbrennwa
An opinion is as common as an rectum- everyone has one. And in the era of misinformation, disinformation as well as increasing distrust in authority figures from individuals to institutions, everything including what we currently know according to best available science can, and is disputed. And thus almost everything in life can, and IS debated, and solution proposed, and whether is can be done at community level or whether it needs regulation. And thus everything can politicized and polarized.
Exhibit A:
https://www.ftc.gov/news-events/new...plifier-rule-make-testing-methods-more-useful
An opinion is as common as an rectum- everyone has one. And in the era of misinformation, disinformation as well as increasing distrust in authority figures from individuals to institutions, everything including what we currently know according to best available science can, and is disputed. And thus almost everything in life can, and IS debated, and solution proposed, and whether is can be done at community level or whether it needs regulation. And thus everything can politicized and polarized.
Exhibit A:
https://www.ftc.gov/news-events/new...plifier-rule-make-testing-methods-more-useful
Think about how sad this statement is….and that folks would be so beaten down by political rhetoric to believe that the natural phenomenon of climate change is a political topic? This is called ‘scientism’……and anyone with a brain should decry it whenever possible…..and that includes here where fairly intelligent people often gather to discuss and debate topics.
There still exists the foundation of all applied science and that’s the scientific method……best not to forget.
D
Deleted member 375592
The one-long vs. many shorts has been Discussed by A. Farina in https://www.researchgate.net/public...ues_and_their_Applicability_in_the_Real_World
Off-topic: the world is much more complicated than we humans can comprehend. Therefore, for any problem, however complicated, there is always at least one simple, effective, easy-to-understand, politically correct wrong solution.
Off-topic: the world is much more complicated than we humans can comprehend. Therefore, for any problem, however complicated, there is always at least one simple, effective, easy-to-understand, politically correct wrong solution.
Back to the topic of microphones, here is a library of them:
https://micpedia.com/library/?pp=omni
Anyone have experience with the following professional (I use it for my job) microphones:
1. Sennheiser mkh 8020 condenser-microphone
2. Schoeps cmc-6xt + mk2:
https://micpedia.com/library/?pp=omni
Anyone have experience with the following professional (I use it for my job) microphones:
1. Sennheiser mkh 8020 condenser-microphone
2. Schoeps cmc-6xt + mk2:
Last edited:
The link isn't to a discussion by Farina, is this a typo?
The paper is decent, I think I read it a few years back, thanks for the reminder anyway.
But doesn't address the extended sweep vs. many shorts, at least not explicitly.
So I did search for the Farina paper and yes, he recommends extended sweep over multiple synchronous addition of shorter sweeps.
Chuffed that my suspicion was confirmed.
Fixed this typo.
Now, as tktran said, "back to the topic of microphones"
Best wishes
David
D
Deleted member 375592
On pages 15-17:
"The usage of averaging several impulse responses for improving the signal-to-noise ratio is a deprecated technology when working with the ESS method.
Synchronous time averaging works only if the whole system is perfectly time-invariant. This is never the case when the system involves propagation of the sound in air, due to air movement and change of the air temperature. So, the preferred way for improving the signal to noise ratio is not to average a number of distinct measurements, but instead to perform a single, very long sweep measurement, as clearly recommended in the ISO 18233/2006 standard.
However, in some cases the usage of long sweeps is not allowed...
... a number of independent sweeps can provide almost acceptable results, provided that the deconvolution and averaging of the impulse response are performed in reversed order (first averaging, then deconvolution), and in the frequency domain."
I may add that a sweep longer than Ttemp may affect the results a bit.
"The usage of averaging several impulse responses for improving the signal-to-noise ratio is a deprecated technology when working with the ESS method.
Synchronous time averaging works only if the whole system is perfectly time-invariant. This is never the case when the system involves propagation of the sound in air, due to air movement and change of the air temperature. So, the preferred way for improving the signal to noise ratio is not to average a number of distinct measurements, but instead to perform a single, very long sweep measurement, as clearly recommended in the ISO 18233/2006 standard.
However, in some cases the usage of long sweeps is not allowed...
... a number of independent sweeps can provide almost acceptable results, provided that the deconvolution and averaging of the impulse response are performed in reversed order (first averaging, then deconvolution), and in the frequency domain."
I may add that a sweep longer than Ttemp may affect the results a bit.
The link you posted doesn't have pp. 15-17. So I suspect you mean a different document, and the link isn't correct, as I already asked.
But the quote is excellent, what document is it from?
It's not the seminal Farina paper, IIRC.
Best wishes
David
While true, let’s not exaggerate those influences. Speed of sound changes roughly with 0,6m/s per K. At 1m distance, the difference would be 0,0000085s per K. A phase shift of 60 degrees at 20kHz. How likely it is that temperature changes with 1 K between measurements?