Nick,
I'm responding to your PM here, since I think it is more appropriate.
This is a fairly complex subject. Stereophile's nearfield measurements are inaccurate for a couple of reasons. I'll list them in order of highest to lowest error:
1) Measuring radiators covering the same frequency range or closely overlapped frequency ranges and very different radiating areas. An example of this is measuring a 10" woofer and a 3" port. When you measure either the port or the woofer, there is far too much crosstalk in the measurement from the other radiator. To minimize the crosstalk to a low enough level to get acceptable results, you need to have several radiator diameters between them. When you have the port and woofer nearly touching each other, and then splice these measurements together, the results are very, very inaccurate. I consider them worthless.
To determine how bad they are, build a subwoofer with two ports. One in front and the other in back. Plug one port while measuring the other port and the woofer. Do this twice, once for the front port and another time for the rear port. Compare the results.
2) When the wavelength of the sound produced by the radiator is large relative to the cabinets baffle dimension, the output from the radiator radiates into a complete sphere. When the wavelength is short, the sound only radiates into half of a sphere. If you have a flat woofer alignment, it will measure flat in the nearfield, since the sound can't diffract around the cabinet into a full sphere at low frequencies. When you measure this system in the far field, the response will be very rolled off in the bass, since only the low frequencies will diffract around the cabinet. To make the system flat in the far field, it must have a response that rises as frequency decreases. How much rise needed is a function of a lot of things (speaker placement in the room, baffle dimensions of the cabinet, listener location, etc.)
This makes Stereophile's nearfield measurements inaccurate due to the effects of diffraction. It has nothing to do with the fact that they are splicing the nearfield bass measurement to the high frequency anechoic measurement.
3) The third error is of a much smaller magnitude. I lump a number of things into this category. The driver in use may have some breakup modes starting at the highest frequency of the splicing. There may be standing waves in the nearfield due to driver recessing into the cabinet. It is difficult to get an accurate phase response curve from both the nearfield and far field measurements. Without both of these being accurate, the resulting summation will not be accurate.
In Stereophile's defense, JA is aware of all of these shortcomings in the measurement technique. He has written about them a couple of times in the magazine over the years. JA has done long articles on his speaker measurement techniques at least twice, I believe. They are actually quite good articles that do a good job at dispelling some myths.
Has you have found out, the response rise in Stereophile's measurement of the NHT Three's bass response is due to diffraction behavior. You can see that most of it goes away when measured in a semianechoic chamber. The Soundstage measurements have been corrected with a compensation curve to correct for the low frequency standing waves in the chamber they use. If you measure the Three in a completely anechoic environment, you will find no rise in the bass at all.
I'm responding to your PM here, since I think it is more appropriate.
This is a fairly complex subject. Stereophile's nearfield measurements are inaccurate for a couple of reasons. I'll list them in order of highest to lowest error:
1) Measuring radiators covering the same frequency range or closely overlapped frequency ranges and very different radiating areas. An example of this is measuring a 10" woofer and a 3" port. When you measure either the port or the woofer, there is far too much crosstalk in the measurement from the other radiator. To minimize the crosstalk to a low enough level to get acceptable results, you need to have several radiator diameters between them. When you have the port and woofer nearly touching each other, and then splice these measurements together, the results are very, very inaccurate. I consider them worthless.
To determine how bad they are, build a subwoofer with two ports. One in front and the other in back. Plug one port while measuring the other port and the woofer. Do this twice, once for the front port and another time for the rear port. Compare the results.
2) When the wavelength of the sound produced by the radiator is large relative to the cabinets baffle dimension, the output from the radiator radiates into a complete sphere. When the wavelength is short, the sound only radiates into half of a sphere. If you have a flat woofer alignment, it will measure flat in the nearfield, since the sound can't diffract around the cabinet into a full sphere at low frequencies. When you measure this system in the far field, the response will be very rolled off in the bass, since only the low frequencies will diffract around the cabinet. To make the system flat in the far field, it must have a response that rises as frequency decreases. How much rise needed is a function of a lot of things (speaker placement in the room, baffle dimensions of the cabinet, listener location, etc.)
This makes Stereophile's nearfield measurements inaccurate due to the effects of diffraction. It has nothing to do with the fact that they are splicing the nearfield bass measurement to the high frequency anechoic measurement.
3) The third error is of a much smaller magnitude. I lump a number of things into this category. The driver in use may have some breakup modes starting at the highest frequency of the splicing. There may be standing waves in the nearfield due to driver recessing into the cabinet. It is difficult to get an accurate phase response curve from both the nearfield and far field measurements. Without both of these being accurate, the resulting summation will not be accurate.
In Stereophile's defense, JA is aware of all of these shortcomings in the measurement technique. He has written about them a couple of times in the magazine over the years. JA has done long articles on his speaker measurement techniques at least twice, I believe. They are actually quite good articles that do a good job at dispelling some myths.
Has you have found out, the response rise in Stereophile's measurement of the NHT Three's bass response is due to diffraction behavior. You can see that most of it goes away when measured in a semianechoic chamber. The Soundstage measurements have been corrected with a compensation curve to correct for the low frequency standing waves in the chamber they use. If you measure the Three in a completely anechoic environment, you will find no rise in the bass at all.
Wow, so this means you could have a speaker that measures perfectly flat with a gated/nearfield measurement and models flat but in reality has very rolled off bass in an anechoic environment. Here I am building speakers with a calibrated mic thinking they are totally flat and feeling smug, and in reality my bass could be a totally deficient mess.
Thank you for the detailed response, it was enlightening. It makes me want to get a giant telescoping pole or a crane and measure my speakers in free air.
I'll try the 2 port trick next time I build a test box for a vented enclosure, but it sounds like that won't go too far in eliminating the problems associated with this type of measurement.
Thank you for the detailed response, it was enlightening. It makes me want to get a giant telescoping pole or a crane and measure my speakers in free air.
I'll try the 2 port trick next time I build a test box for a vented enclosure, but it sounds like that won't go too far in eliminating the problems associated with this type of measurement.
I very much respect his work and dedication (guru speakers), but he was banned from this forum/website and he was called a liar by some posters, in what started as a joke, because people had to wait for 2 years before getting his speakers.
What matters is what it sounds and measures like in your room.
Because you are not selling this speaker to anyone else. If its flat in room then that is an accurate design, You can then tweak all you want with your own curves.
He is still a registered member in good standing.
http://www.diyaudio.com/forums/members/ingoehman.html
dave
Is it Dave, his IP was not blocked(?), maybe I was wrong (that was not the impression I got at the time, or what was said... maybe that he didn't like)?!He is still a registered member in good standing.
http://www.diyaudio.com/forums/members/ingoehman.html
dave
Anyway, this was one of his last threads in this forum.
http://www.diyaudio.com/forums/multi-way/106389-vented-alignments.html
Last edited:
The comparisons show how quasi they really are...
dave
Actually, it looks to me like the measurements are pretty similar, the issue is only in how you match the nearfield measurement to the farfield. This would simply be a calibration issue.
For instance, if you measured one speaker in a proper anechoic chamber, and then performed both a near/farfield measurement on that speaker, then, upon splicing the two measurements, you adjust the gain on the nearfield so that it matches the anechoic measurement.
Then, for further measurements, as long *** you remembered that gain offset, and finely controlled the amplifier settings, microphone gain, and distance from the microphone to the driver, keeping all those attributes the same, you SHOULD be able to produce a fairly repeatable and accurate frequency response measurement, no?
i mean, if you look at these, they're very very close, the only notable differences, aside from the extra gain in the nearfield, could easily be chalked up to production variance
marantz vari-Q comes to mind
adjustable bump -
from a store window in downtown L.A.
```````````````````````````````````````````````````````````````````````````````````````````````````
something else -
http://sports.yahoo.com/nba/blog/ba...f-Bad-Idea-Jeans-Pranking-Keny?urn=nba,231919
Last edited:
I was just alerted to this thread. Yes, my published measurements have the 2-Pi boost in the upper bass. I mention this fact in every review, so it is not like I am trying to hide anything. I decided years back that I would publish the uncorrected measurement rather than try, somewhat arbitrarily, to apply a correction.
It could be argued, BTW, that the true anechoic LF response is also misleading, as this will always have less LF than a speaker would deliver in-room. For example, a speaker tuned to be maximally flat in the bass under anechoic conditions will always sound bass-heavy in all but very large rooms.
John Atkinson
Editor, Stereophile
A warm welcome to the forum John.
Indeed you are very careful to be very clear about your measurement conditions. I could probably skip the 1st couple paragraph in your measure sidebar, but still read thru them.
Your measures are consistant, but here, with many different people measuring in their homes there is considerablr room for inconsistency. Anechoic measures usually imply at least some professional experience with measurement technique and therefore less concern with inconsistancies.
In the end it is really a matter of the end-reader being able to put an appropriate interpretation on the results. Your averaged room measure using Fuzzmeasure is very helpful in that respect and also captures (in a manner less subject to interpretation) the gestalt of the on, and off axis responses and the way the speaker actually performs in room.
That measure in a diverse set of rooms would be very useful, but would suck up at least as many resources as a big anechoic chamber.
dave
PS: you should get Dudley to review (and maybe build) some diy designs... he seems to have that bent.
Indeed you are very careful to be very clear about your measurement conditions. I could probably skip the 1st couple paragraph in your measure sidebar, but still read thru them.
Your measures are consistant, but here, with many different people measuring in their homes there is considerablr room for inconsistency. Anechoic measures usually imply at least some professional experience with measurement technique and therefore less concern with inconsistancies.
In the end it is really a matter of the end-reader being able to put an appropriate interpretation on the results. Your averaged room measure using Fuzzmeasure is very helpful in that respect and also captures (in a manner less subject to interpretation) the gestalt of the on, and off axis responses and the way the speaker actually performs in room.
That measure in a diverse set of rooms would be very useful, but would suck up at least as many resources as a big anechoic chamber.
dave
PS: you should get Dudley to review (and maybe build) some diy designs... he seems to have that bent.
Hello and welcome to our forum.
Wouldn't it be handy to show the rough room gain trend of your usual measuring space so we get an idea of its contribution against averaged RTA matrix window at listener's position in same space? Or you use different rooms?
Welcome to the forum John, and I hope you post on somewhat of a regular basis. First my room rant: I learned a set of what I call, "Golden Rules" from your magazine involving speaker design theory which I will touch on in a sec. I've designed a few hundred or so, "virtual" speakers using Linearx's LEAP. I've built a few out of those designs, and the one thing that sticks out at me is my room, "gain." Measured with LMS, it is +- 10-15 decibels. It has HUGE peaks and valleys, and it is a normal cigar box shaped room. The room is 22.5 ft long, 12.5 ft wide, with a 7 ft ceiling. On each side is an adjoining bedroom, and kitchen of average size. If somebody thinks that they can measure a room with a 20 band RTA or something, forget about it. I've seen dozens of posts on the net by people who think they know what the room is doing, but they don't have a clue. With the old fashion LMS, it shows you exactly what you are up against. Another thing I don't believe in are warble tones for room measurement, because Stanley Clarkes bass line does not warble when we listen to him! LOL
I once pulled out all the stops to get the room, "flat." I used the Audio Control Richter scale, Linearx's LMS for measuring, two diferent high and low pass crossover frequencies, etc. I got it to about +- 3-4DB. I didn't like the sound. It didn't seem like enough bass!! I think by accident many speaker companies wind up with the right amount of bass, if they design theirs reasonably flat, because even though the response in your room is rough, it still adds a LOT to the final amount of bass.
I prefer the tight bass of a sealed box. If one is venting, and the design indicates a tuning frequency around 40 HZ, CHEAT a little and tune it to 42.5 - The low string on a Fender precision bass. (Got that one from Stereophile review of a speaker) I did that with small speakers with 2-4.25" drivers in each and I was SHOCKED at the amount of bass that they would put out. LOL Design your vents with as much cross-sectional area as possible, and make sure you subtract the internal capacity of the vent from your box volume.
Other things I learned from Stereophile: In one mag, they talked of the NRC, in Canada. They said that the NRC research, with BLIND listening tests, indicated that audiophiles preferred the following: Speakers with vertical arrays, speakers with flat frequency response meaning the flatter the better. They showed charts of percentage of people who preferred flatter speakers - the flatter the speaker, the higher the percentage went. Speakers with low distortion. - There was a lot more info than that......
Maybe John can help us out with more input?
I once pulled out all the stops to get the room, "flat." I used the Audio Control Richter scale, Linearx's LMS for measuring, two diferent high and low pass crossover frequencies, etc. I got it to about +- 3-4DB. I didn't like the sound. It didn't seem like enough bass!! I think by accident many speaker companies wind up with the right amount of bass, if they design theirs reasonably flat, because even though the response in your room is rough, it still adds a LOT to the final amount of bass.
I prefer the tight bass of a sealed box. If one is venting, and the design indicates a tuning frequency around 40 HZ, CHEAT a little and tune it to 42.5 - The low string on a Fender precision bass. (Got that one from Stereophile review of a speaker) I did that with small speakers with 2-4.25" drivers in each and I was SHOCKED at the amount of bass that they would put out. LOL Design your vents with as much cross-sectional area as possible, and make sure you subtract the internal capacity of the vent from your box volume.
Other things I learned from Stereophile: In one mag, they talked of the NRC, in Canada. They said that the NRC research, with BLIND listening tests, indicated that audiophiles preferred the following: Speakers with vertical arrays, speakers with flat frequency response meaning the flatter the better. They showed charts of percentage of people who preferred flatter speakers - the flatter the speaker, the higher the percentage went. Speakers with low distortion. - There was a lot more info than that......
Maybe John can help us out with more input?Thank you.
With respect to the quasi-anechoic responses Stereophile publishes, the room reflections are windowed out of the impulse response, so the characteristics of the measuring room don't matter. Downside, of course, is that frequency resolution is limited to the reciprocal of the time window, so you need to arrange the geometry of the measurement set-up to push reflections back in time as far as possible. I measure out-of-doors for this reason, weather permitting.
The measuring room is also not a factor in nearfield measurements. Downside, however, is that, as have been discussed in this thread, the measurement assumes that the radiators are embedded in an infinite plane.
I do publish in-room measurements when logistically practical. These are performed in the reviewers' listening rooms (including my own). The amount of room gain at low frequencies will depend on the size of the room, so there is no universal correction factor. However, if I remember correctly, Martin Colloms estimated that room gain typically tends to follow a 4dB/octave slope below an arbitrary corner frequency.
John Atkinson
Editor, Stereophile
Last edited:
Thanks, Dave. And thanks for your comments.
While we did occasionally cover some DIY speaker designs in the 1980s through the mid-'90s, it is not currently feasible to do so, I am afraid.
John Atkinson
Editor, Stereophile
The most detailed coverage of that is probably Floyd Toole's book Sound Reproduction. Required reading IMHO.
dave
Thanks. Those are the most indicative of the final perceived balance IMO.
Since there is no fixed reference auditioning room, comparing historical review averaged RTA data is moot.
Also different loudspeakers and support, would show different low frequency gain curve depending to the proximity of their LF systems to the floor even if their far field non gated averages were compared in the same room.
Back in the 1990s, I used to measure 1/3-octave in-room spectra using pink noise. In recent years, however, I publish a 1/6 octave-smoothed spectrum. However, in discussion with Sean Olive of Harman's research center, he told me that he feels that the 1/10 octave-smoothed correlates best with a speaker's perceived balance.
John Atkinson
Editor, Stereophile
Hi John,
A warm welcome to the site here. I hope you find it interesting.
I'm a tech kind of guy who used the raw T/S parameters I measured myself when designing and measuring loudspeakers. Something I haven't really done for 10 years now, but I will always remember how far off a manufacturer's specifications can be from the actual product.
Do you remember the "standard European living room that Philips used, along with the famous DIN 45500 limits for "HiFi"? What might you consider to be an average listening space these days?
Now, you mentioned 1/6th octave measurements for a speaker systems response. I have to admit that todays methods are a practical mystery for me. Do you use an impulse test, or near-field mic for this? I have an Audio Control SA-3050 I should dust off and play with I guess. I never did use it to measure response in a room as it always showed giant peaks and dips. Then there was that mic ground issue that I still haven't corrected (lost interest).
Hi Dave,
The book by Floyd Toole is one I don't have, and didn't even know existed to be honest with you. Thank you for mentioning it.
-Chris
A warm welcome to the site here. I hope you find it interesting.
I'm a tech kind of guy who used the raw T/S parameters I measured myself when designing and measuring loudspeakers. Something I haven't really done for 10 years now, but I will always remember how far off a manufacturer's specifications can be from the actual product.
Do you remember the "standard European living room that Philips used, along with the famous DIN 45500 limits for "HiFi"? What might you consider to be an average listening space these days?
Now, you mentioned 1/6th octave measurements for a speaker systems response. I have to admit that todays methods are a practical mystery for me. Do you use an impulse test, or near-field mic for this? I have an Audio Control SA-3050 I should dust off and play with I guess. I never did use it to measure response in a room as it always showed giant peaks and dips. Then there was that mic ground issue that I still haven't corrected (lost interest).
Hi Dave,
The book by Floyd Toole is one I don't have, and didn't even know existed to be honest with you. Thank you for mentioning it.
-Chris
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.