That's what people live by in many countries, in the end, they get accused of dumping because the products are too cheap.ronc said:
(subthread about Dr Geddes' speakers split off to here http://www.diyaudio.com/forums/showthread.php?s=&threadid=121618
) (subthread about Noise Control split off to here http://www.diyaudio.com/forums/showthread.php?s=&threadid=121622
)
I am getting confused becasue the thread was opened in an effort to share some distortion related studies, yet the above statement seems to state that those distortions are not so rellevent to good sound reproduction. So that seems to end discussion.gedlee said:
So now are we actually trying to see if these distortion figures are still meaningful? Are we trying to find other means of distortion measurment methods and metrics?gedlee said:
That's what people live by in many countries, in the end, they get accused of dumping because the products are too cheap
\ If America or any other country cannot produce a competive product then its their problem.
Back to the subject. You can measure distortion, but can you actually hear it. The human ear is not a spectrum analysis scope.
ron
\ If America or any other country cannot produce a competive product then its their problem.
Back to the subject. You can measure distortion, but can you actually hear it. The human ear is not a spectrum analysis scope.
ron
The distortion figures of THD and IMD are not meaningful. Period. And they cannot be made to be meaningful. There are no readily available measures that are. Even a spectrum of the distortion ala ucla88 are not complete because they ignore a very important characteristic and that is WHERE the nonlinearity occurs. This cannot be seen in the spectrum alone. As Howard and I have both noted one needs to consider the phase of the nonlinearity - which is synonamous with "where".
I analyzed our data considering only weighting the harmonics and this was a vast improvement over THD and IMD, but it still fell short of the GedLee metric which also takes into account the phase, or where the nonlinearity occurs. The problem is not simple and attempts to simplify it inevitably fall short of the mark.
I analyzed our data considering only weighting the harmonics and this was a vast improvement over THD and IMD, but it still fell short of the GedLee metric which also takes into account the phase, or where the nonlinearity occurs. The problem is not simple and attempts to simplify it inevitably fall short of the mark.
gedlee said:
No doubt the 3 toneburst spectrum is not the be all and end all. However, I am somewhat limited to available and reasonably priced testing hardware/software. I look at my distortion numbers a little bit like a baffled FR curve. Not the final say in distortion but a good starting point.
Off to have a glass of wine...
ucla88 said:
I think that there are better ways to measure these things. I did a lot of work on this. But in the end I concluded that no one was interested. The feeling seemed to be "sure THD is meaningless, but its so easy to do!".
The equipment requirement is not elaborate, a sound card, mic and some software, but the post processing is. Since there are no automated systems to make good distortion measurements you have to basically do the processing yourself. Doable, but a pain. One CAN plot of the nonlinear curve from a single tone spectrum, which is valid at that level and that frequency. If things don't change with frequency then your pretty much done - (for that level!), but loudspeakers do change with frequency and so you have to do this at multiple frequencies.
And what about thermal issues? Do you address those at all?
I would like to ask some questions about the information in this thread.
If the distortion figures of THD and IMD are not meaningful, is there something else which could tell us about the distortion a loudspeaker produces?
You said, that a system can be made so that those figures are relevant or irrelevant. What is the difference between those systems?
If the distortion figures of THD and IMD are not meaningful, is there something else which could tell us about the distortion a loudspeaker produces?
You said, that a system can be made so that those figures are relevant or irrelevant. What is the difference between those systems?
Currently I am working in the path with the assumption that if I can control the stored energy problem such that the CSD show fast enough decay, then the distortion issues will become more significant. But I really have not spent much time into distortion yet. Additionally, when we measure distortion acoustically, the source of distortion is not clear. Perhaps with a properly focussed lazer measurement system like Klippel, it is possible to get closer to the motor distorsion, then little by little find out what part is creating more significant small signal distortion. If it is the acoustic aspects that show more significant distortion, then the diaphragm would need to be studied further.ronc said:
MaVo said:
I would say that there is no practical way to ***** the nonlinear distortion in a loudspeaker that is meaningful to its perception. I think that I have laid the groundwork to efficient techniques that could be used, but they have not ben implimented.
In a loudspeaker system, bigger is better. Larger drivers have less nonlinearity for a given SPL. Compression drivers have virtually inaudible nonlinear distortion. Using a large woofer and a compression driver makes a system in which nonlinear distortion is not going to be a consideration.
As the system gets smaller nonlinearity becomes more of an issue and fast. Unless one plays the smaller drivers at lower levels they will reach a point where they are nonlinear and this will be the limiting factor in their usage. As a rule of thumb, for every halfing of the drivers size you would lower the MaxSPL by about 12 dB. Why 12 dB and not 6 dB? Thats because the efficiency will drop by about 6 dB and the nonlinearity will grow by about 6 dB, hence the SPL at which the nonlinearity is an audible factor will drop by about 12 dB. These are "rules of thumb" but you get the idea that going smaller is a big hit in SPL performance.
At a low enough SPL level nonlinearity is not a factor in any loudspeaker.
Now there is also the thermal modulation factors, which also tend to favor bigger. But these are not the topic here.
Something occured to me today :
We would need a computer model of the human hearing system, with psychoacoustics and all.
Turns out we have it : all lossy encoders (MP3, Vorbis...) have such models.
So, would it be possible to extract the phychoacoustic model from, say, the Vorbis encoder, and use it to determine distortion masking ?
If I have the time, I might try that.
We would need a computer model of the human hearing system, with psychoacoustics and all.
Turns out we have it : all lossy encoders (MP3, Vorbis...) have such models.
So, would it be possible to extract the phychoacoustic model from, say, the Vorbis encoder, and use it to determine distortion masking ?
If I have the time, I might try that.
peufeu said:
Opera, I think its called, has been tried and it worked reasonably, but its design was not intended to measure loudspeakers.
I was leading a group within the AES to find ways to quantify the subjective aspects of nonlinearity - thats where the first papers came from. Opera was one of the considerations. I recently resigned from that group because of my belief that there really wasn't any interest in doing this and I had concluded that it wasn't very important. I wanted to do something that mattered.
The work that Lidia and I did was based in psychoacoustic models of masking, etc. the exact same things that Perceptual Coding is based on. But we were after a simple approach that worked "well" but need not be "precise". We were trying to replace THD and precise was hardly necessary. But that work didn't go anywhere because of a lack of interest. The more people look at the problem the more they conclude its not worth the effort.
As I say, its not that nonlinearities aren't important, they are, but I don't need an elaborate measurement to tell me what to do to make them irrelavent. Simple engineering gets the job done and its time to move on to something that makes more of a difference.
If you ask me to design a loudspeaker where nonlinear distortion does not play a role in its perception its pretty easy. You may not like the solution, but its not that hard to lay it out. Now is when the tradeoffs come into place. In the marketplace those tradeoffs are virtually always made to the detriment of nonlinear distortion, because, quite simply, its not always viewed as negative. And do they care to know about the distortion that they are creating - no not really. They measure THD and call it a day. They got a pointless result, but it didn't really matter in the first place. Everybody knows from experince that THD doesn't work so we've all ignored for years. But everybody still asks for it and its still measured on a regular bases, but for all practical purposes it is ignored.
Making that nonlinear-distortion-free loudspeaker sound good (accurate) in a small room, now thats where the challenge is.
peufeu said:
Electronics is a whole lot easier to do because the nonlinearities tend to be fairly independent of frequency.
What I have found works extremely well is to look at the spectrum of a sine waave as you drop the input signal level. If the higher orders of the spectrum rise, as they often do, then this would be a highly audible distortion - ie. crossover. The distortion that occurs at levels near clipping is pretty much irrelavent. And you can't look at THD + noise with level, because at the low levels all you will see is the noise. You have to note the levels of the harmonics themselves.
There was a post some time back on how I make this measurement. I have found that it tells me all that I need to know about electronics.
