Ok, I'll see if I can dig them out of my files. They were somewhat smoother, I think, and used a rather slow sweep
I'm truly confused by this thread. Why are we attempting to understand a pre-released driver measurement of a full range driver in the multi-way forum? Is it to keep the discussion away from the one person who could shed any real accurate light on the subject (the developer himself)? The graph in question was a swept sine. The room was anechoic, so we can safely assume it wasn't a room artifact.
If you're truly interested in discussing differences in measurement technique, you should pick a released driver that has been shown to have good QC and thus similar T/S numbers across production runs. As it sits this thread looks like nothing but a dirty internet troll.
* Moderator note: Thread moved to Full Range forum.
a) That person actively suppresse the discussion by deleting my posts. -> I had to go somewhere else
b) It was suggested to me not to "de-rail" threads and go somewhere else. -> I had to go somewhere else
c) It was suggested to me "to do some more research and study." -> I had to talk to someone else
d) I'm genuinely interested in the measuring conditions the led to the results shown in post 1
The only thing left would be the driver. I don't believe it's that bad.
Do you have such data? I don't. That's why this thread exists. I can only provide my own data and measuring conditions which I'm happy to do. And as I've said before my data doesn't show anything close to the graph in post 1. Even a Markaudio driver I own doesn't look anything like that.
Last edited:
Oh I see. Scott and I both thought you were just referring to the input signal, repeatedly. A Google search for "swept sine technique" results in all sorts of pro-level papers and products using the term to refer to specifically impulse FFT-based measurements and specifically to non-FFT measurements. Words, what good are they anyway eh?
That was my question too (see post 6, and 7).
I dunno, I only bothered to post because I saw the measurement, thought it seemed very strange for anechoic, and then saw the reactions to spekr asking about it and wondered if maybe there was something major I don't understand about LMS measurements, because it seemed strongly implied by a whole pile of people that, unless you're totally clueless, it looks perfectly normal.
I don't have any particular interest in the specific driver and whether the measurement makes it look good or bad or whatever.
This is like tasting raw beef and wondering why it doesn't taste like hamburger. The measurement that you're doubting is a driver still in development. I assume (yes this is an assumption) the development process would be to shape the general FR curve (looks pretty nice) then use klippel or a similar tool to isolate and remove the worst resonances in the design to get rid of the peakiness. You're comparing it to finished products, noticing it's different, then looking for fault in the measurement process. No wonder you were asked to quit commenting, as if it was my measurement process I would be offended in your logic.
Sorry if I wasn't clear enough. I was referring to the swept sine technique by Farina. All major measurement software these days use it.
Well, if the ragged response is just some artefact of the driver being a prototype then why didn't Markaudio say so? I wouldn't have made any additional post. Instead he wrote a lot of posts but never said anything remotely connected to what you're suggesting.
The entire measurement is an artifact of the driver being a prototype. That goes without saying and is why your logic of it being a faulty measurement process is so affronting. Mark has a very professional measurement process that he has well documented.
Isn't that something that happens primarily at higher frequencies whereas at lower frequencies the driver still shows a more textbook-like pistonic behavior with little or no cone break-up even if it's a wide band driver?
No, I don't know who speaker dave is. In general, the logic of debating speculations of a pre-release response curve's measurement process with anyone other than the people involved in the measurement process is absurd. The particular driver in question is something I've never seen before - an interchangeable power train, basket, etc... - and that alone could explain all of your concerns. Can you find a driver example that has interchangeable parts with a smoother frequency response?
You've seen Mark document his measurement process and yet you still question it.
It's also very bad forum etiquette to post multiple replies in succession. This practice of yours could be the main reason why many people including Mark seem to be getting frustrated with you.
You've seen Mark document his measurement process and yet you still question it.
It's also very bad forum etiquette to post multiple replies in succession. This practice of yours could be the main reason why many people including Mark seem to be getting frustrated with you.
1) FWIW the curve that worries you so much actually looks quite good to me, if it's really unsmoothed.
Welcome_to_the_real_World [tm]
In fact, if actually raw it should be fuzzier, which makes me think that LMS signal detection and conversion does have some kind of built-in/implied smoothing, maybe not on purpose but the result of Math signal processing/conversion/you name it.
2) that said, I don't know what is this thread doing in this specific subforum.
Yes, I read your "explanation".
If so, and following its Logic, why not post it in the Drag Racers Forum or in the New England Gardeners association page?
3) I don't know you (at all) nor Markaudio (in that case only that he has made respected speakers for a long time) but in the lack of further data, in general I tend to trust more the "has done" people than the "never did" type.
4) and yes, this thread has a faint smell of bad blood in it .
Might be wrong, of course.
Anybody can be wrong ..... even you 😱
Welcome_to_the_real_World [tm]
In fact, if actually raw it should be fuzzier, which makes me think that LMS signal detection and conversion does have some kind of built-in/implied smoothing, maybe not on purpose but the result of Math signal processing/conversion/you name it.
2) that said, I don't know what is this thread doing in this specific subforum.
Yes, I read your "explanation".
If so, and following its Logic, why not post it in the Drag Racers Forum or in the New England Gardeners association page?
3) I don't know you (at all) nor Markaudio (in that case only that he has made respected speakers for a long time) but in the lack of further data, in general I tend to trust more the "has done" people than the "never did" type.
4) and yes, this thread has a faint smell of bad blood in it .
Might be wrong, of course.
Anybody can be wrong ..... even you 😱
Typically, +/- 3dB smoothed curves have hitorically been considered good. +/- 5dB raw data isnot that far off.
Now a more important question to me, what is the linearity, or even the ability to reproduce, information 40 dB down from this?
dave
Please research. What is your response to his post?
Again, it was suggested to me that the process would be industry standard and post 1 is a general question about measurement software/methodology so there's nothing absurd about it.
If this would be the case why doesn't Markaudio simply say so?
No, I've NOT "seen Mark document his measurement process". Got a link?
People talk to me and I reply. You consider this "very bad forum etiquette"? In my neck of the woods it's called having a conversation.
Could you please provide examples of raw responses? Raw responses of drivers in an anechoic chamber that is.
There are very knowledgabe people from within the industry in the multi-way form. They probably have seen a lot of anechoic driver measurements. That's why I started the thread here.
Me too. My hope is that more knowledgeable people like speaker dave post their comments.
True, that's why I started this thread.
Hey spekr, earlier today I got flustered with the tone in this thread and Mark's repeated posts in other threads expressing his discontent with people who are doubting his words on measurement. However throughout this thread you seem to have taken everyone who attempts to look from a different angle at the problem you've presented as being against you, and have had a tone of 'with me or against me' that continues to off-put me and sours my posts too. This will be my last post in this thread.
Here's a link to a post Mark himself made roughly an hour before you made this thread, in response to your similar queries - or rather accusations, depending on how they're read - where he clearly tells you it was a fully anechoic chamber on an IEC baffle at 1m: http://www.diyaudio.com/forums/full...io-alpair-7p-vs-alpair-7-3-a.html#post4089718
He also asks you to do a quick search to see where he's detailed his measurement process in the past. I did that and within ten minutes found the following:
Here's a post where Mark has detailed production tolerances as well as measurement process: http://www.diyaudio.com/forums/markaudio/138092-new-4-driver-markaudio-2.html#post1779186
Here's a thread where Mark discusses with members of the forum possible future changes to his measurement and reporting process: http://www.diyaudio.com/forums/markaudio/251911-future-testing-data-analysis-presentation.html
Here's a thread where other users debate Mark's anechoic results: http://www.diyaudio.com/forums/full-range/253108-ma-driver-measurements.html
Overall, this seems to be a source of much frustration for Mark because he's so entwined in the community and we're all skeptics at heart, but there comes a point were it seems like the horse has been beaten to death on this discussion (and I'm new to it).
I stand by the fact that it's absurd to use a driver that nobody except Mark has in hand, to make any sort of conclusions about Mark's measurement process. Pick one we can take to labs near us and perform ideal tests on, otherwise it's all just fear, uncertainty, and doubt being spread as we go around in circles wondering and Mark's well built reputation is burned at the steak of FUD.
Can you point to any other manufacturer who has detailed their production control and measurement process as thoroughly as Mark has?
I say this because I'm very interested in learning more about ideal measurement processes too, but your logic on the matter is hurting my brain. Really, it is. I don't mean this as a personal attack. I'm merely trying to actually understand why you seem fixated on this particular singular response curve and why you've posted in no less than three threads about this one preliminary test curve? Find a driver you can hold in your hand and measure. When the Alpair 8 is released, then you can rack your brain on the intricacies of its response and compare your personal anechoic baffled 1m response to his final published version. Until then, lets all refrain.
Here's a link to a post Mark himself made roughly an hour before you made this thread, in response to your similar queries - or rather accusations, depending on how they're read - where he clearly tells you it was a fully anechoic chamber on an IEC baffle at 1m: http://www.diyaudio.com/forums/full...io-alpair-7p-vs-alpair-7-3-a.html#post4089718
He also asks you to do a quick search to see where he's detailed his measurement process in the past. I did that and within ten minutes found the following:
Here's a post where Mark has detailed production tolerances as well as measurement process: http://www.diyaudio.com/forums/markaudio/138092-new-4-driver-markaudio-2.html#post1779186
Here's a thread where Mark discusses with members of the forum possible future changes to his measurement and reporting process: http://www.diyaudio.com/forums/markaudio/251911-future-testing-data-analysis-presentation.html
Here's a thread where other users debate Mark's anechoic results: http://www.diyaudio.com/forums/full-range/253108-ma-driver-measurements.html
Overall, this seems to be a source of much frustration for Mark because he's so entwined in the community and we're all skeptics at heart, but there comes a point were it seems like the horse has been beaten to death on this discussion (and I'm new to it).
I stand by the fact that it's absurd to use a driver that nobody except Mark has in hand, to make any sort of conclusions about Mark's measurement process. Pick one we can take to labs near us and perform ideal tests on, otherwise it's all just fear, uncertainty, and doubt being spread as we go around in circles wondering and Mark's well built reputation is burned at the steak of FUD.
Can you point to any other manufacturer who has detailed their production control and measurement process as thoroughly as Mark has?
I say this because I'm very interested in learning more about ideal measurement processes too, but your logic on the matter is hurting my brain. Really, it is. I don't mean this as a personal attack. I'm merely trying to actually understand why you seem fixated on this particular singular response curve and why you've posted in no less than three threads about this one preliminary test curve? Find a driver you can hold in your hand and measure. When the Alpair 8 is released, then you can rack your brain on the intricacies of its response and compare your personal anechoic baffled 1m response to his final published version. Until then, lets all refrain.
I don't understand where folks are seeing this as an attack or bad blood or frustration on the part of the OP. Read his first post carefully, it is objective and neutral and asking for possible explanations of a why a measurement is the way it is. It seems that anytime anyone posts a frequency response plot from MA, the troops start circling their wagons for battle. I don't get it,
and I keep seeing this time and time again. I think if you want objective non flaming responses at an honest question regarding MA drivers, you need to redact the plots and all supporting info to show the facts anonymously - that is not tied to any manufacturer or model number to protect the innocent. I went to this thread based on the title of wanting to learn more about anechoic measurements and keep seeing this drawn out saga that has been repeated many times before until a moderator comes in and closes it.
FWIW, the freq response plot shown in Post 1 looks great to me - I would say that is fantastic performance for a 4in full range driver.
Last edited:
I have absolutely no axe to grind with regard to MarkAudio, but I also feel that those curves are unusual to the point of being suspect.
I have been measuring speaker professionally for over 30 years and used a number of anechoic chambers as well as ground plane and impulse test approaches to anechoic measurement.
Although most chambers have significant low frequency flaws, essentially the standing waves of a typical room show through below the effective range of their wedges, they also usually exhibit a clean smoothness that no in-room curve can ever show. The curve shown in the OP looks more similar to an in-room with frequency smoothing (say 1/6th Octave) type of curve. It has the randomness of a noise measurement, although that randomness would typically increase for lowest frequencies.
It doesn't help that the curve uses a 120 dB range (50 is more typical) so it will look way flatter than otherwise. Still, the 10dB narrow band variation is considerable for an anechoic measurement.
I also wonder about IEC measurements. I don't know if that is and open back IEC baffle or an IEC dimensioned cabinet, but either is likely to put some response signature into the curve due to diffraction effects (1 m measurements will help/reduce that, but should also knock down the random wiggles).
There is a concept in measurements called "quefrency". This is an intentional flipping of the word "frequency". Basically it is to look at the wiggles of a response curve and interpret them as a wiggle frequency. (A log magnitude conversion of the frequency response curve as if it were a time waveform, which it isn't.) The understanding with quefrency is that a reflection close in time to the sound's origin can only give low quefrency wiggles. High quefrency wiggles require much later reflections.
As the curve shown has high quefrency wiggles all across the curve it implies that many diverse and late arriving reflections are contained in the measurement and at a level sufficient to give 10dB variation. That is unlike any anechoic measurement that I have ever seen. Even a bad chamber with a heavy floor grate will show a single periodic ripple, (Scrunching together at high frequencies on a log f plot) rather than the random variation.
Anyhow, the curves look suspicious to me. (The Zaph Audio curves mentioned look quite respectable.) I make no comment/have no opinion on the product itself.
David
I have been measuring speaker professionally for over 30 years and used a number of anechoic chambers as well as ground plane and impulse test approaches to anechoic measurement.
Although most chambers have significant low frequency flaws, essentially the standing waves of a typical room show through below the effective range of their wedges, they also usually exhibit a clean smoothness that no in-room curve can ever show. The curve shown in the OP looks more similar to an in-room with frequency smoothing (say 1/6th Octave) type of curve. It has the randomness of a noise measurement, although that randomness would typically increase for lowest frequencies.
It doesn't help that the curve uses a 120 dB range (50 is more typical) so it will look way flatter than otherwise. Still, the 10dB narrow band variation is considerable for an anechoic measurement.
I also wonder about IEC measurements. I don't know if that is and open back IEC baffle or an IEC dimensioned cabinet, but either is likely to put some response signature into the curve due to diffraction effects (1 m measurements will help/reduce that, but should also knock down the random wiggles).
There is a concept in measurements called "quefrency". This is an intentional flipping of the word "frequency". Basically it is to look at the wiggles of a response curve and interpret them as a wiggle frequency. (A log magnitude conversion of the frequency response curve as if it were a time waveform, which it isn't.) The understanding with quefrency is that a reflection close in time to the sound's origin can only give low quefrency wiggles. High quefrency wiggles require much later reflections.
As the curve shown has high quefrency wiggles all across the curve it implies that many diverse and late arriving reflections are contained in the measurement and at a level sufficient to give 10dB variation. That is unlike any anechoic measurement that I have ever seen. Even a bad chamber with a heavy floor grate will show a single periodic ripple, (Scrunching together at high frequencies on a log f plot) rather than the random variation.
Anyhow, the curves look suspicious to me. (The Zaph Audio curves mentioned look quite respectable.) I make no comment/have no opinion on the product itself.
David
- Status
- Not open for further replies.