June 25th, Clifton and myself were winding down a discussion and Ethan Winer entered in and compared graph "A" (20-2khz measurements of room response) to the frequency response of a Bryston amplifier. I immediately saw a couple of problems.
1) Comparing the two graphs is completely inappropriate as I shall discuss near the bottom of the post.
2) Graph "A", which was suposedely a measurement of the room response in a normally furnished but untreated room at low frequencies, immediately raised red flags as to its accuracy.
(Other subjects and information will be presented as well.)
I continued to discuss the situation with Ethan and was able to obtain more information including response graph "B" (200 to 20khz measurements), his microphone being accurate from 20-800hz, and Mackie HR824 amp powered speakers that have a response of +/- 1.5db from 39-20khz.
I have obtained enough information to evaluate his measurements/graph. Some of the information I will present is fairly widely known, so some will already understand the material. Below is graph "A".
The red line is a representation of the frequency response with associated comb filtering/reflection problems.
For those who don't know what comb filtering is, as defined by Wikipedia:
"a comb filter adds a delayed version of a signal to itself, causing constructive and destructive interference.
(peaks and nulls/valleys for those new) The frequency response of a comb filter consists of a series of
regularly-spaced spikes, giving the appearance of a comb."
It is correct that the effects of comb filtering occur at regular intervals. Notice the spikes at approx 47, 94, and 141hz (blue line). However the maximum total deviation of 35 db produced red flags.
After Ethan posed graph B, more red flags appeared. Graph B is a representation of the measurements Ethan gathered
between 200hz to 20khz.
Let's start with graph B. The blue line shows the room response and the effects of comb filtering. But the 1/3 octave response (red line) is what we are primarily interested in. The red line basically shows the woofer's average response dropping 11db or so from 300hz to 2khz, and then the response suddenly rises at 2khz, where the speaker crossover "transfers" the signal from the woofer to the dome tweeter. Notice after peaking at approx 3.7khz, the red line falls until the spl is 5db down at 10khz, 7db at 12khz, 10db at 14khz, and a whopping -23db at 20khz.
The spl reduction over such a wide frequency clearly indicates comb filtering/reflections are not the cause of the spl reduction over a broad 3.7khz to 20khz range, 16.3 khz width. What this indicates is that the measurement is severely "off axis" (speakers are not facing the microphone).
This evidence is supported by the 8.75 inch woofer's average output (red line) also dropping approximately
11db from 300hz to 2khz. There is less spl drop from the woofer because of the much lower frequencies involved.
If the speakers were facing directly from the speaker to the mic, the average spl -11db drop would not be there and the tweeter's average spl drop would also not be there, as mentioned near the bottom of my post.
One can mimick this response (if one has a meter and signal generator) by simply rotating the front of the speakers
away from the microphone. One can also check out the tweeter's on and off axis response curves on the internet/review magazines and compare them to graph "B".
Another suspected problem is the huge 35db total deviation from 30 to 200hz in Ethan's measurements (graph A). However, by simply reversing the polarity of one speaker (reverse the leads to one speaker), one can mimick Ethan's measurements. By the way, one can minimize total comb filter/reflection deviation by placement of the speakers.
In fact, one can raise the deviation up to nearly double the max/min db change (measured response) then when both speakers are of the same polarity. We have all heard the low frequency response dramatically change by reversing the polarity of one speaker. That is because the total deviation of peak to null has dramatically increased.
Now consider this. If Ethan's totally deviation measurement of 35db were correct, then reversing the polarity of one speaker would result in a nearly 60db or more maximum/peak to minimum/null spl deviation. Of course this is absurd.
When one hears the speakers with one speaker reversed polarity, one hears a total deviation of 20s-30s db, not near 60 or more db. Anyone with an spl meter can measure this change as an experiment.
With a 60db+ deviation, and the reference bass level set to 80db spl, some frequencies/bass notes would peak at 110db while the weakest/null bass notes would measure only 50db.
According to the "Threshold of Audibility Curves" based on Fletcher the "very critical" listeners ("top 5%") would barely hear 50hz at 50db spl, as 50db is the threshold of hearing at 50hz ("in the absence of noise").
The average listener's threshold is 63db spl so they would not hear 50hz at all, let alone if the null were lower than 50hz.
If 90db is the reference, a 50hz valley/null spl would be 60db, so average listeners would still be just at the hearing threshold at 50hz.
If we change the frequency to 70hz, the threshold for the "top 5%" is 40db spl while the average listener's threshold is 53db spl.
So if 90db were the reference, a null at 70 hz would barely be heard, while nothing would be heard with a null below 50hz. (Fletcher curves and quotes courtesy of RCA Radiotron Designers Handbook, written by 26 engineers, page 620.)
It is obvious that a 35db deviation for correct polarity is incorrect, except possibly in some very rare circumstances.
Another red flag that I do not see on Ethan's websites, but mentioned in a post in "Entry Level":
"Even if it was done with an SPL meter, mine is flat within 1 dB from 20 Hz to 800 Hz. The graph below shows side-by-side tests using my SPL meter and the AKG microphone."
No account is made concerning the accuracy of his microphones above 800hz, so we have no idea how accurate the mic actually is, nor could I find any information on Ethan's websites or online.
I would have to assume the microphone(s) are not accurate. Otherwise he would have stated the mic would have been accurate up to 10khz or 20khz, not 800hz.
If the mic response is down (-Xdb) from accurate at the higher frequencies, then leaving out this information leaves the impression of a poorer room response than it actually is. If the mic response is plus (+Xdb) from accurate, then he still did not reveal the rooms true response. Most mics are -db accurate at high frequencies.
Since Ethan left this important information out of the post and website, visitors have to assume the measured response/graphs are actually accurate.
Ethan has posted on Entry Level Forum that 35db deviation is typical for a room. Here are his comments found at http://forum.stereophile.com/forum/showf...part=4&vc=1
"The first graph below shows the low frequency response you'll have in a typical bedroom size listening room." (Upper post on page.)
In another post at http://forum.stereophile.com/forum/showf...part=4&vc=1
"The LF graph I posted earlier is at an even higher resolution. And that is even closer to the true response arriving at your ears. By the way, these graphs are absolutely typical for home-sized listening room."
(Lower post on page. If typical then the maximum deviation can easily exceed 35db.)
Ethan states on his website at http://www.realtraps.com/facts.htm
"What's the point in buying gear that's ruler flat from DC to microwaves when the acoustics in your room create peaks and dips as large as 35 dB throughout the entire bass range?"
So I conclude 35db is not typical and above 35db would be rare.
It is also interesting that the perceived sound of a speaker/room with less than 1/3 octave deviation spikes (comb filtering/reflection) is much much different than with wideband tonal deviations that electronic components can produce.
The speaker/rooms can measure high spike deviations and the ear has a hard time detecting them while any wide band tonal imbalance/measurement in components is easily detectable.
So electronic components need to be incredibly close to flat response to sound correct. Trying to compare the Bryston amplifier to speaker/room response is absurd and amateurish.
At http://www.realtraps.com/rfz.htm Ethan states:
"Some people use thin panels made of fiberglass or foam, or blankets, and believe that's sufficient. But those materials absorb reflections at higher frequencies only."
This is correct, thin panels of fiberglass are poor for treating low frequencies. However, an entire roll of R-19 fiberglass insulation in the corners is very effective and offers an inexpensive alternative to more expensive options.
Keep the plastic cover intact though, and yes it does look bad, but it also works. At a house in Madison, Wisconsin a few years ago, the owner had one roll in each corner behind the speakers and it really helped. When a 2nd roll was added to those corners, the results were even better. Boomy bass became nice.
Notice figure 2 on Ethan's personal site. It uses the same speaker/amp in the same room as graph "B", yet figure 2 appears quite different than graph "B". Notice the differences in average spl response from 4khz to 18khz.
In figure 2 between 4khz and at least 16khz, the average response does not drop. At 4khz, 8khz, and 16khz the spl averages approximately 78db (not counting comb filter). At 18khz approx 75db. (Close to the speakers 3db window, +/- 1.5db to 20khz.)
However, the red line in graph "B" is on average dropping from 4khz to 18khz. At 4khz we see 93db, at 8khz approx 89db, at 12khz approx 85db, at 16khz approx 83db, and at 18khz approx 80.5db,
so approx 13db drop for graph "B" vs 3db drop in figure 2. Same speaker, amplifier, and untreated room.
In conclusion, anyone can produce similar results to the graphs presented by simply placing the microphone in the listening position, changing the location and rotating the speakers so the drivers are not facing the microphone, and reversing the polarity of one speaker by reversing the cable leads.
If possible please re-read each sentence carefully and consider what is actually said because there are many subjects and lots of technical information covered.