Markaudio provided a measurement of a driver under development made in an anechoic chamber. The data looks like this:
A user provided this information about the measuring conditions:
I'm wondering where the large response variances of up to 10dB come from. Is this really the driver response, some measurement artifact, room artifact or something else?
Would the swept sine technique have resulted in a different response?
A user provided this information about the measuring conditions:
I'm wondering where the large response variances of up to 10dB come from. Is this really the driver response, some measurement artifact, room artifact or something else?
Would the swept sine technique have resulted in a different response?
Attachments
That is from an anechoic slow sine sweep, as is stated in the quotation. My answer hasn't changed either: yes, you can (as in 'can') get differences with different measurement systems, conditions and techniques. Not exactly news to most people here I imagine. I provided you with an example in the other thread about how differences can show up with varying techniques: slow swept sine conditions can sometimes be rather unflattering of the driver being measured, since they can cause a build-up of mechanical resonances that may not show up as clearly in, for example, a burst (e.g. a logarithmic chirp). This is not invariable: it depends on the driver in question, nor is it automatically a problem under practical conditions, given the constantly varying nature of most musical signals.
As for the driver response, wideband drive units are not like conventional higher mass / drivers designed to operate over a more restricted BW. They employ controlled resonance, either of a secondary radiator (e.g. a whizzer cone) and / or of the main cone & dust-cap substrate to produce a large proportion of their operating BW. Especially as size increases, they generally do not provide as pretty an FR response et al as drivers designed to function over a narrower BW. This is well known, and the vast majority of users of wideband drivers really don't care. They have other priorities in mind. You may not share them, which is fair enough, but they in their turn are entitled to hold them, which is also fair enough.
With that said, the prototype shown above (did you ask Mark's permission to use his graph / data BTW? If not, it seems a trifle discourteous) is indicated at present to be about +/- 5dB across a large proportion of the range, with an increase in the higher registers to help off-axis performance, which is quite common for widebanders. That is no worse than some midbass units out there. To be sure, there are others that will have a somewhat flatter response over their own, rather narrower operating BW, but as noted, the comparison is apples / oranges. They're not designed for the same thing.
As for the driver response, wideband drive units are not like conventional higher mass / drivers designed to operate over a more restricted BW. They employ controlled resonance, either of a secondary radiator (e.g. a whizzer cone) and / or of the main cone & dust-cap substrate to produce a large proportion of their operating BW. Especially as size increases, they generally do not provide as pretty an FR response et al as drivers designed to function over a narrower BW. This is well known, and the vast majority of users of wideband drivers really don't care. They have other priorities in mind. You may not share them, which is fair enough, but they in their turn are entitled to hold them, which is also fair enough.
With that said, the prototype shown above (did you ask Mark's permission to use his graph / data BTW? If not, it seems a trifle discourteous) is indicated at present to be about +/- 5dB across a large proportion of the range, with an increase in the higher registers to help off-axis performance, which is quite common for widebanders. That is no worse than some midbass units out there. To be sure, there are others that will have a somewhat flatter response over their own, rather narrower operating BW, but as noted, the comparison is apples / oranges. They're not designed for the same thing.
Last edited:
Have you ever listened or seen fr plot of Lowther drivers? I have not seen you mentioning them, or anything similar.
It is obvious that for some reason you choose to pick at Mark. Not a fair game. DiyAudio shouldn't be a polygon for releasing personal frustrations 😎
It is obvious that for some reason you choose to pick at Mark. Not a fair game. DiyAudio shouldn't be a polygon for releasing personal frustrations 😎
Just moving the mic a little would result in considerable variation in response as well.
Some take measurements from multiple positions around the system and average them. See Stereophile's reviews.
Last edited:
Indeed. I recommend the OP invests in a copy of these excellent texts:
Testing Loudspeakers: Joseph D'Appolito: 9781882580170: Amazon.com: Books
Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms: Floyd Toole: 9780240520094: Amazon.com: Books
Testing Loudspeakers: Joseph D'Appolito: 9781882580170: Amazon.com: Books
Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms: Floyd Toole: 9780240520094: Amazon.com: Books
I realize you guys were already bickering in the other thread and maybe the OP deserves the hostility given it was off topic and all, I dunno, but I have similar questions about this measurement and it's not about picking on wideband drivers. Isn't the main difference with LMS measurements that the resolution is constant with log freq., while FFT are linear? If so, why wouldn't, for example, the DFT graph from a very long sine sweep in ARTA look similar at the lower end of the spectrum?
Maybe it was an etiquette thing, which, fair enough, but people do ask far, far, far more basic questions about performing driver measurements on here every day without getting a hard time for it.
Maybe it was an etiquette thing, which, fair enough, but people do ask far, far, far more basic questions about performing driver measurements on here every day without getting a hard time for it.
I am very surprised that those are said to be measurements from an anechoic chamber. I would expect anechoic measurements to be fairly smooth, no matter what rate they are taken at. Most anechoic curves are near perfect above 200 or so and then might have some very significant response aberrations below, based on wedge length and outside chamber dimensions.
One exception is big floor grates, that can put a comb filter into the response (reducing in spacing as it goes up in the typical log f plot).
The curve shown has constant "noise" across the range and looks more like an in-room curve with, say, 1/6th Octave smoothing.
The input stimulus or conversion approach shouldn't make much difference. It is true that MLSSA and FFT are inherently linear spaced and LMS is typically log spaced. In the old days with only 512 or 1024 point FFTs then you would see a loss of resolution at the bass end of a log scale. More points means that that isn't much of a factor these days.
I don't see this as a wide band driver problem. That still wouldn't explain the constant amount of "noise" across the frequency band. I have seen wide band measurements in a chamber that had reciprocity correction for chamber inaccuracy, and the low frequency end was so smooth that it looked more like an electrical filter than a speaker. Something isn't right about these measurements.
David
One exception is big floor grates, that can put a comb filter into the response (reducing in spacing as it goes up in the typical log f plot).
The curve shown has constant "noise" across the range and looks more like an in-room curve with, say, 1/6th Octave smoothing.
The input stimulus or conversion approach shouldn't make much difference. It is true that MLSSA and FFT are inherently linear spaced and LMS is typically log spaced. In the old days with only 512 or 1024 point FFTs then you would see a loss of resolution at the bass end of a log scale. More points means that that isn't much of a factor these days.
I don't see this as a wide band driver problem. That still wouldn't explain the constant amount of "noise" across the frequency band. I have seen wide band measurements in a chamber that had reciprocity correction for chamber inaccuracy, and the low frequency end was so smooth that it looked more like an electrical filter than a speaker. Something isn't right about these measurements.
David
Mark's the man to answer that, and he's generally very receptive of interested queries when they're made politely (like Dave's above). I believe he's in Tokyo until the end of the week, which is sod's law as far as timing goes, but assuming things remain civil I imagine he'll be happy to discuss it when he returns.
Last edited:
Understood but would the same driver under the same conditions show the same 10dB variance when measured with the swept sine technique? REW for example can do a 1M log sweep. Would the result look significantly different?
I've measured quite a few full range drivers lately and none of them showed such a response. Even the one Markaudio driver I've bought looks very different.
That's not the topic of the thread. So why do I need to listen to a Lowther driver? I've seen measurements of Lowthers. They look bad but they don't look anything like the graph Mark posted.
I don't pick at Mark and I'm not frustrated. I'm trying to understand the graph he had posted.
Scottmoose said:
You're both talking about a slow, log sine sweep, for goodness sake. The difference between REW (et al.) and LMS is discussed just above, and it's not the sine sweeping swept sine log sine swept sweep technique.
I have (read) both books. That's the reason for my post. I've never seen an anechoic and un-smoothed driver response like the one shown in post 1.
Thanks David. I also suspect a problem with the measurement rather than a driver problem. It looks to me like a snapshot of a steady-state measurement in a room that is not anechoic.
Spekr was disturbed by all the wiggles, and thought there was something wrong.
I felt the same way the first time that I measured speakers in an anechoic chamber.
Most of these raw curves are smoothed or filtered to look nice.
Even the speed of the sine sweep alters the appearance of the curve a lot.
Last edited:
So you're implying my questions have been impolite and that's why he deleted my posts and attacked me?
Who's "he"?
If you see large response variations with small positional changes then the room is probably not anechoic at all or you're measuring in the near field.
Last edited:
Sorry, meant you. It was a good quality ten foot anechoic chamber, and normally I used a distance of 1m, with a B&K mic.
There's a difference between using a sine and reading out the magnitude response directly and using the swept sine technique which calculates the impulse response from which finally the magnitude response is calculated.
My question was if the difference between those two measurement methodologies explains the response shown in post 1 as I'm only familiar with the swept sine technique.
The response in post 1 looks different to anything I've seen and I suspect there are artifacts from the room and/or the measurement that make it look that way.
Care to share one of those measurements so we can get a feeling of how this compares?
- Status
- Not open for further replies.