This is one of the most debated topics on internet audio forums, probably a close second behind cable arguments. The topic comes up again and again and usually goes something like this. The objectivist “spec guy” claims flat response is the end-all-be-all for judging a speaker. The subjectivist “Golden Ear Guy” claims that all you should trust is your ear and there are flat speakers that sound terrible. What is the right approach? In my opinion, both are wrong and I’ll explain why in the following article.
First, let me start by saying you can train your ear pretty effectively, not only to hear correctly, but to hear incorrectly. If you have a speaker system you have been living with for the last year that has a large peak in the 2-4 kHz range, if you listen to something that measures flat, it might sound dull to you. However, if you reverse this and start on the flat system, the system with the peak will sound shrill and forward.
So getting back to the original question, let’s start with what we know. In blind listening tests, axial first response (which translates to on-axis frequency response) is the predominant indicator of listener preference. This is usually what “Spec Guy” cites as the reason why this is what you should worry about most. In general, this is the easiest thing for our brain to discern, especially when doing short listening tests that these studies normal involve. “Spec Guy” is right here.
However, “Golden Ear Guy” is going to argue that there are speakers that measure flat that sound horrible. He might also be right. There are a number of reasons why a flat measuring speaker might not sound good.
Now, he might be used to listening to non-flat speakers, which is why he doesn’t like the sound of the flat speakers. This is a problem with his ear not being properly adjusted to what flat sounds like. But in the end, I chalk this up to personal preference. It doesn’t really do any good to argue with him. In the end, it only matters that he likes what he hears and not some internet warrior.
On the other hand, he could have a highly reflective room and a flat speaker just sounds too bright because of the early reflections he is picking up from his room. Room treatment could help solve this issue, and with it in place, he might change his opinion.
It also could be an issue with the crossover. There are number of factors that should go into any good crossover design, but a lot of people (and companies) either don’t understand or overlook them. Often, the most blatant offender is often a distortion issue. Something like a tweeter being pushed too low or a metal cone woofer with an improperly suppressed breakup will not show up in a single on-axis frequency response sweep but can be very audible.
I was working on a pair of moderately expensive Canton speakers for a client who complained about some harshness in the tweeter. I could hear what he was talking about when I listened to them, but when I measured, they were very flat on-axis. When I did a distortion sweep, there was a big distortion spike centered at around 3 kHz due to the breakup on the metal cone midrange unit. There are two ways to fix this. The distortion can make that area seem like the level is higher than it really is, so one way is just to lower the on-axis response in this region. Using an EQ, the speaker sounded better with a 3-4 dB drop in the 2-4 kHz region. The second is to get rid of the distortion by using a better crossover. This is the optimal solution because you still end up with a flat on-axis frequency response.
Directivity and by extension, power response, are two other issues that can have significant effects on the sound of a speaker that can measure flat but sound bad. Without going into too much detail, basically you want a smooth curve not only on-axis but as you transition off-axis you shouldn’t see any drastic changes in your response. It should be a smooth transition from 0 degrees to 90 degrees. This can be accomplished and displayed in a number of ways and I won’t say one is better than the other. Some people like narrow directivity and others prefer wide, but having a smooth change from 0 to 90 degrees is always preferred over drastic changes.
Drastic changes are caused by improperly mating a very large diameter driver to a tweeter at a high frequency. On something like an 8” woofer, the frequencies about 2 kHz will start to fall off in intensity just a few degrees off axis. At 30 degrees off axis, you might start to see this as low as 1200 Hz. Crossing to a tweeter at 3 kHz would result in a large depression developing between 1-3 kHz off axis. Because of the reduced off-axis energy, the tonal balance of the speaker is affected.
Taking all this in to consideration, here is my design philosophy. Flat on-axis response is still king, but that doesn’t mean that flat will sound good. All these things have to be taken into consideration. I always shoot for flat response first. This means you have to pick drivers that will work together so you can get all the other issues sorted out. Deviations from flat are a conscious choice to address other issues.
For instance, in my last design, the Standards, I chose to put a downward tilt on the frequency response. My new apartment has much lower ceilings (8’ in the new versus 12’ in the old) and this has made things seem overly bright to me when flat. I can’t convince my wife to let me put acoustic treatments on the ceiling, so I made the design decision to go with a downward tilted response instead.
Your ears tend to be easily fooled, especially when it comes to short listening tests. What might seem like more detail at first because of a slightly higher level in one region, can lead to listening fatigure or irritation over time when compared to a flat repsonse. How many times have you heard a speaker that sounded great on acoustic music, but harsh or shouty on other music?
So the next time you see and argument on the internet about this topic, try and remember there are a lot of variables that could be at play that can make flat on-axis sound bad. However, you should also remember that in blind tests, it is still almost always the preferred response and generally over longer periods of time, most people will prefer it as well.