Also level of directivity index has big role in this. Large speakers with DI~5 dB at LF and rising towards HF don't need very linear DI spectrum. Small conventional speakers can't move you "there" without listening at near field so they're better try to emulate features of original sound sources; to move instruments and vocalist into your room. This requires some assumptions/statistics about recordings; how close/far mics typically are etc. Original sound sources such as human head, cello or piano don't have linear DI spectrum so reproduction in conventional room with speakers having very linear or too steep diving DI at mid-range sound retracted/distant, less natural and presence (=dead). Respectively, speakers with more constant/flat DI at mid up to ca. 1 kHz and then smooth steepen towards HF sound more alive, natural and present.
Directivity is not the only feature having effect on naturalness. Timing down to bass range and dynamics should be good too.
I guess experience and controlling this kind of details is not very common. Generic studies focusing to mediocre consumer products with random samples can't detect all nuances in a controlled way. It would be sad if such basic science develops into a religion on any forum.
I prefer much more Erin's Audio Corner measurements. His "Globe" Plots are great (not only). 
In case of Amir's reviews I see problem in his verbal expression.
In case of Amir's reviews I see problem in his verbal expression.
This happens when people are not involved in all aspects and over simplifying things. The majority of those people haven't been involved in all aspects of speaker building. You can already see and hear that in the language they use and what aspects they skip over way to easily.
Directivity is an important parameter, but that doesn't mean that other parameters aren't important anymore all of a sudden.
I think Erin's experiences and descriptions are very much in line with my own. He at least tries to correlate what he's hearing with what he's measuring.
In some cases even little resonances can be pretty audible, even when the directivity is pretty constant. But there are many other things, some that Erin also seems to miss.
That being said, any third party measurements are welcome.
I often don't even bother reading (or listening to) the verbal description.
It's extremely clear at this point that some brands fall flat on their faces.
Some obviously don't know what they are doing and some actually do.
Goes to show the power of marketing.
It also goes to show how much in this field is covered up by a whole bunch of fancy talk and complete lack of transparency.
Btw, it wouldn't be to difficult to calculate the error of the directivity.
A constant directivity would show a straight line in the directivity index (DI) vs frequency plot.
One only has to calculate between this ideal straight line and the actual data.
In the end it's possible to get the standard deviation of all those points.
The lower this number (the standard deviation) , the more it follows a perfect constant directivity.
One can also calculate the angle of this line to get a measure of how strong the speaker is beaming.
I have done this myself a few times but doing it "manually " (well in Excel) was a pretty tedious process.
But in a program like this it can be fully automated, especially because the data is already there.
The difficulty is only where to choose the starting point of this line as well as the ending point.
I did it by eyeballing it, but haven't thought about doing that mathematically yet.
Also the DI above a certain frequency isn't that important anymore because of the very high local absorption of those frequencies. I would say roughly above 8kHz or so.
Below a certain frequency there is the opposite when standing waves will be the dominant factor.
A constant directivity would show a straight line in the directivity index (DI) vs frequency plot.
One only has to calculate between this ideal straight line and the actual data.
In the end it's possible to get the standard deviation of all those points.
The lower this number (the standard deviation) , the more it follows a perfect constant directivity.
One can also calculate the angle of this line to get a measure of how strong the speaker is beaming.
I have done this myself a few times but doing it "manually " (well in Excel) was a pretty tedious process.
But in a program like this it can be fully automated, especially because the data is already there.
The difficulty is only where to choose the starting point of this line as well as the ending point.
I did it by eyeballing it, but haven't thought about doing that mathematically yet.
Also the DI above a certain frequency isn't that important anymore because of the very high local absorption of those frequencies. I would say roughly above 8kHz or so.
Below a certain frequency there is the opposite when standing waves will be the dominant factor.
Last edited:
Of course. I look back on a review I posted a week or two ago and think "doh, I wish I had remembered to say that" or "man, I need to make another video explaining that so people don't gloss over it". That's why I made this center channel discussion video where I spend 35 minutes using data, ripple tank simulations, in-room measurements and even using binaural recordings of various center channels in my room to help people understand what comb filtering is and why horizontal MTM speakers are so problematic.
I was really proud of that video. Yet, I wish I had time to leave in some bits that I had to remove or just did not include. I didn't talk about reflections in that video and I even said I wasn't going to because of time. I didn't talk about linearity of response. I focused solely on horizontal radiation. And I edited a good bit out because no one wants to watch an hour long video these days. Except for me. If it's a good podcast/interview.
So, when you see things that you think I missed, also please keep in mind that I am limited on time. Maybe I didn't talk about it because I was focused on something else in the data. Maybe I didn't talk about it because I plain old forgot. Maybe I left it out. Maybe I thought "hey, the hardcore guys don't need this explained and it's way past the boundaries of actual usefulness for the average Joe, at least right now". I don't always include an entire suite of data such as step response or group delay because, again, I don't know how useful that is to the community - at large. Then there's the fact that I have to pay for storage space, generate the data, etc, etc.
As an analyst (as my day job), I could stew over measurements and do the "but, if" stuff all day. There are also many, many tradeoffs to consider not just in loudspeakers but in practically everything that is worth evaluating. I'm reviewing for a large audience. An audience that consists of 2-channel purists, home theater junkies and some are "even full-range single-point-source only" folks. It makes the task of trying to boil down quantifiers and application limits of various designs in 15 minutes or less (otherwise, you lose them and thus their connection/appreciation for the data).
IOW, it's easy to gather the data. It's hard to make it relevant and entertaining enough to keep them engaged and be able to explain all the things I see or find interesting all in a short time span. In my written reviews, I often leave out a LOT of commentary because I assume (as I say in the reviews) that people looking at the data know how to read it. But for videos, that's a different task.
If I were making a video about a DIY design for DIY'rs it would be completely different than me making a store-bought studio monitor review for mixing engineers (side note: I may soon be partnering with a famous producer in future videos). Because you go into each with the expectation that the audience not only understands what you're about to say but also understands themselves. But that's not the case. The people who read/watch my reviews come from all sorts of backgrounds. Some are there to learn, some are there to pass the time. And that's what makes providing data and reviews challenging. If I'm being honest, I think I'm now doing a pretty good job of what I set out to do. It has taken me a while to find my groove but I'm happy with the response that I'm getting; people saying that I helped them understand a topic or an attribute more than they did before. That's how we all win. At least in my humble opinion. My underlying goal in all of this is to help people see data as useful; useful to their purchasing decisions. Doing so will bring the community - as a whole - up. That may be naïve of me. But, dammit, I'm trying.
- Erin
Last edited:
There are quite some things I would like to reply to, on a constructive and positive way that is.Of course. I look back on a review I posted a week or two ago and think "doh, I wish I had remembered to say that" or "man, I need to make another video explaining that so people don't gloss over it". That's why I made this center channel discussion video where I spend 35 minutes using data, ripple tank simulations, in-room measurements and even using binaural recordings of various center channels in my room to help people understand what comb filtering is and why horizontal MTM speakers are so problematic.
I was really proud of that video. Yet, I wish I had time to leave in some bits that I had to remove or just did not include. I didn't talk about reflections in that video and I even said I wasn't going to because of time. I didn't talk about linearity of response. I focused solely on horizontal radiation. And I edited a good bit out because no one wants to watch an hour long video these days. Except for me. If it's a good podcast/interview.
So, when you see things that you think I missed, also please keep in mind that I am limited on time. As an analyst (as my day job), I could stew over measurements and do the "but, if" stuff all day. There are also many, many tradeoffs to consider not just in loudspeakers but in practically everything that is worth evaluating. I'm reviewing for a large audience. An audience that consists of 2-channel purists, home theater junkies and some are "even full-range single-point-source only" folks. It makes the task of trying to boil down quantifiers and application limits of various designs in 15 minutes or less (otherwise, you lose them and thus their connection/appreciation for the data).
IOW, it's easy to gather the data. It's hard to make it relevant and entertaining enough to keep them engaged and be able to explain all the things I see or find interesting all in a short time span. In my written reviews, I often leave out a LOT of commentary because I assume (as I say in the reviews) that people looking at the data know how to read it. But for videos, that's a different task.
If I were making a video about a DIY design for DIY'rs it would be completely different than me making a store-bought studio monitor for engineers. Because you go into each with the expectation that the audience not only understands what you're about to say but also understands themselves. But that's not the case. And that's what makes providing data and reviews challenging. If I'm being honest, I think I'm now doing a pretty good job of what I set out to do. It has taken me a while to find my groove but I'm happy with the response that I'm getting; people saying that I helped them understand a topic or an attribute more than they did before. That's how we all win. At least in my humble opinion. My underlying goal in all of this is to help people see data as useful; useful to their purchasing decisions. Doing so will bring the community - as a whole - up. That may be naïve of me. But, dammit, I'm trying.
- Erin
But that's also going heavily off-topic, so maybe it's a good idea to move that to a separate topic?
You are free to PM me.
Just keep in mind the reasons I gave above. You may request x/y/z but there are limits. This is something I do on the side. It isn't my day job. And balancing my hobby with my family and my job is priority.
So, sure, PM me and we can talk there.
I thought more about just an open brainstorm topic
Personally I found that always more useful, more people know more than one (or two)
I am aware about the limited time you have, really appreciated everything you do!
I thought more about just an open brainstorm topic
Personally I found that always more useful, more people know more than one (or two)
I am aware about the limited time you have, really appreciated everything you do!
Edit for conciseness:
If you have a specific request or question then feel free to PM. I won't have time nor the desire to keep track of a thread with a multitude of posts/requests. I'm doing about as much as I can as it is these days. Just being honest.
I'll leave it at that. No need to keep going OT in this thread, as you already mentioned.
Last edited:
ASR appears to systematically and intentionally maintain a lie that timing has no perceptible effect on sound reproduction. What we hear, sense and measure is not flying magnitude responses generated at 2.83 Vrms constant voltage. A strong brand and resale value are more significant than genuinely good performance and sound reproduction, so fan boys can be fooled easily and endlessly by focusing the most popular measurable quantities. Smooth directivity and frequency responses sell the best at the moment so reviews overflow duplicate useless data such as normalized polarmaps or waterfalls and balloons or polar charts at few individual frequencies while practical dynamics and timing analyzes are totally ignored.
Fortunately, there are more knowledgeable and open communities and individuals. Manufacturers have ended up replacing or upgrading lagging non-dynamic products despite of their commercial success and perfect spinorama.
Shouldn't care a sh*t about that but somehow I do.
Fortunately, there are more knowledgeable and open communities and individuals. Manufacturers have ended up replacing or upgrading lagging non-dynamic products despite of their commercial success and perfect spinorama.
Shouldn't care a sh*t about that but somehow I do.
Member
Joined 2003
Your sample is actually about twice better than mine having 8th order slopes. This shows that IR+SR+ETC graph alone is difficult to interpret and not adequate though is able to show some features GD can't. Group delay and possible excess group delay (assuming that minimum phase calculation is reliable enough) fills information holes and make graphs comparable.
It depends on woofer and mid driver. It's quite popular to use 2nd order acoustic slopes (instead of 4th) for the lowest XO to drop GD to about half. That may be impossible with open baffle with small cone area and low XO without active FIR. Boxed designs are usually possible to make with 1st order electrical HP.
You could also read appendix 1 in VCAD manual. Transient perfect design is one option especially with coaxial driver arrangements.
High XO works okay with 4th order acoustical slopes because that does not cause so much GD. High XO also is also easy for FIR.
Are we talking about relative group delay or absolute group delay?
Relative group delay can also be seen in the phase response as well as how the frequency response connects.
Or just plotting the GD graph itself obviously.
Often these graphs give a better insight than looking at the impulse/step responses. The exception is looking at and comparing the start of the impulse itself.
In (mechanical and room) acoustics the impulse is sometimes used a bit more often to estimate durations etc.
In the end it doesn't have to be textbook perfect as long as the total performance doesn't suffer from it.
It's all about orders of magnitude and significance (something a lot of people don't seem to think about or even know the existence of)
Relative group delay can also be seen in the phase response as well as how the frequency response connects.
Or just plotting the GD graph itself obviously.
Often these graphs give a better insight than looking at the impulse/step responses. The exception is looking at and comparing the start of the impulse itself.
In (mechanical and room) acoustics the impulse is sometimes used a bit more often to estimate durations etc.
In the end it doesn't have to be textbook perfect as long as the total performance doesn't suffer from it.
It's all about orders of magnitude and significance (something a lot of people don't seem to think about or even know the existence of)
I guess this thread is a good one.
I notice, ARTA and several others suggest a L-Pad in front of the sound card. Usually a 910 and 8200 Ohm with a pair of back to back 4.1 Zieners.
OK, considering a sound card is usually 10K input, and a power amp would not notice a 900 Ohm load, I am sure it works fine.
But I was wondering, to be more versatile, not just ARTA, but REW, RightMark etc, where you may want to measure lower level signals, that 900 Ohm input impedance seems a bit low. Would it not make more sense if it was a tad higher? Say 5K?
Then, they usually show the zieners on the input side where they would get the full current of a higher voltage input. Would it not make more sense if they were on the output side, i.e. input to the sound card. That way a miss-probe or transient would not as likely fry them. ( and then your sound card)
I think I'll make a little box with a 0, 10, 20 attenuation selectable.
FWIW, my old Behringer mic got flakey, so I bought the MiniDSP mic. Turns out it is LEVEL calibrated not just frequency, so one does not need a calibration capsule or guessing with a radio Shack SPL meter for a reference. I to me raw driver electrical tests with WooferTester which is far easier than the sound card jig method.
I notice, ARTA and several others suggest a L-Pad in front of the sound card. Usually a 910 and 8200 Ohm with a pair of back to back 4.1 Zieners.
OK, considering a sound card is usually 10K input, and a power amp would not notice a 900 Ohm load, I am sure it works fine.
But I was wondering, to be more versatile, not just ARTA, but REW, RightMark etc, where you may want to measure lower level signals, that 900 Ohm input impedance seems a bit low. Would it not make more sense if it was a tad higher? Say 5K?
Then, they usually show the zieners on the input side where they would get the full current of a higher voltage input. Would it not make more sense if they were on the output side, i.e. input to the sound card. That way a miss-probe or transient would not as likely fry them. ( and then your sound card)
I think I'll make a little box with a 0, 10, 20 attenuation selectable.
FWIW, my old Behringer mic got flakey, so I bought the MiniDSP mic. Turns out it is LEVEL calibrated not just frequency, so one does not need a calibration capsule or guessing with a radio Shack SPL meter for a reference. I to me raw driver electrical tests with WooferTester which is far easier than the sound card jig method.
Member
Joined 2003
Of course the resistor values are not set in stone, I use 22k and 47k on my jig. The ARTA jig application note includes the calculations so you can determine whatever set of resistors you like. I put those calculations in a spreadsheet that I’ll post here a little later on.
Low signal level is not the issue, but rather probe impedance vs load impedance, and how much output you can reliably measure before clipping the sound card input.
I personally don’t use the zener diodes, IMO they are only required if you are prone to doing silly things. I find the clipping indicators on my USB audio interface to be more than adequate. If you have built the jig properly it should be no more complicated than the woofer tester, but much more versatile. The woofer tester is just a basic usb audio codec with sense resistor built in after all.
If VituixCAD is on your radar with full spatial information, XLR mic for dual channel measurement is a must. Absolute SPL is truly not as important as you might think, relative SPL is good enough for
Low signal level is not the issue, but rather probe impedance vs load impedance, and how much output you can reliably measure before clipping the sound card input.
I personally don’t use the zener diodes, IMO they are only required if you are prone to doing silly things. I find the clipping indicators on my USB audio interface to be more than adequate. If you have built the jig properly it should be no more complicated than the woofer tester, but much more versatile. The woofer tester is just a basic usb audio codec with sense resistor built in after all.
If VituixCAD is on your radar with full spatial information, XLR mic for dual channel measurement is a must. Absolute SPL is truly not as important as you might think, relative SPL is good enough for