June 1, 2009

A Bit More About Blind Listening Tests

Last month, when I wrote about the benefits to audio reviewing of blind listening tests, I ended by saying that although I believe strongly in such tests, they’re not always the most easy and practical thing to do -- which is why, in the past, we haven’t conducted these tests in the way I’d like. This article details some of the challenges involved in conducting blind tests, and explains what we’re currently considering and working on to overcome them.

Environment and setup

Two of the biggest obstacles to conducting blind listening tests are creating the proper listening environment and having a proper setup. The room must be of sufficient size and quality that the component under test remains the focus and isn’t overshadowed by the anomalies of the room. As well, the component being evaluated must be integrated into the system without adding any more variables to the setup, or modification to the component, that could obscure the results. Basically, you want to be able to assess only the component under evaluation.

The testing of source components, cables, and other electronics doesn’t present that many obstacles -- it’s not difficult to set up a test that’s fair to the component under test and produces results that hold up under scrutiny. I’ve done blind tests of CD players, D/A converters, and cables in my own room, using the products exactly as their designers intended them to be used, and with little modification of my system. The results have been quite telling. For example, we recently connected two DACs to one CD transport and one preamplifier, matched the volume levels, and were able to switch between the DACs while playing a single CD in the same CD player, without knowing which DAC was which. We’ve tested some cables with a similar setup.

Not all components are easy to test this way; certainly, speakers aren’t. Provided the room is appropriate, the biggest obstacle in testing speakers is setting them up -- the sound a speaker makes is highly dependent on where it’s placed in the room. Furthermore, no two speakers can occupy the same space at the same time; even if you were evaluating identical speakers, there’s a good chance the two samples would sound different solely by virtue of their placement in the room.

This is a difficult issue to overcome, and few solutions have been ideal. Canada’s National Research Council (NRC) deals with it by using multiple trials in which the speakers’ positions are changed; they end up being listened to from different points in the room. Listeners rate the speakers’ sound in each position, and the results are then usually averaged. It’s not perfect, but it works fairly well, giving a broader representation of how the test speakers perform. But there are better solutions.

The folks at Harman International, who are big believers in blind tests, recognized this limitation of speaker testing years ago and took it into consideration in the design of their current listening room. Harman built what they refer to as a "speaker shuffler": a fully automated system that, with the flick of a switch, moves one speaker out of a certain spot and the next speaker into that same spot. Therefore, all speakers are listened to in the exact same position. Smart -- but also expensive, and difficult to build. To my knowledge, Harman is the only company that has such a thing, although as far as I can tell, it’s the best solution so far.


In Harman International’s blind-listening room, test speakers are listened to from behind a visually opaque but acoustically transparent curtain, unseen by listeners; an automated "speaker shuffler" moves test speakers into and out of position, so that each occupies precisely the same position in the room.

When it comes to setting up the SoundStage! Network testing environment, we’re carefully considering all of this to ensure that, when we do blind tests, we have the best setup possible, that products are assessed fairly, and that the results are valid. I plan to visit Harman’s facility to evaluate their speaker-testing system before we finalize the room here, and some other component categories will present their own challenges. Issues such as these are why, up till now, we at the SoundStage! Network haven’t been able to implement blind testing as we would have liked to -- it’s not something you just set up and run with.

While assessing these challenges, we also recognize that the typical testing done today by the "sighted reviewing community" doesn’t hold itself to the same sort of standards or rigor. I’ve often read reviews of speakers that were assessed in the wrong size room (a small speaker in very big room, or a big speaker in a small one); or of components whose reviewers hooked them up in such a way as to introduce more variables than did the component itself, making the results from their testing even more suspect. I find it ironic that so many are happy to poke holes in blind tests, no matter how carefully set up, but so few criticize sighted tests, which can have so many flaws that they’re laughable.

Proximity and convenience

Another problem has to do with the number of blind-testing rooms we can afford to set up and the number of people who will do the listening, particularly when our writers are so widely scattered across North America. We will be fortunate if we can get one room set up well, but that will greatly limit the number of people who can listen there.

The only thing I can hope for is that the idea of blind testing will appeal to enough reviewers and readers that they’ll be willing to take the time to come and listen, and we’ll be able to afford to bring them here. I also hope that it will bring more credibility to our publications, and that, therefore, growth will occur, and that we then might be able to set up more rooms. For now, the focus is on creating one listening/testing room.

Resistance

The last obstacle to point out has to do with the surprising level of resistance still met by even well-thought-out blind tests. After my first article, in which I merely discussed the blind-testing process, went live on May 1, there was a flurry of activity about it on the various audio forums. Some posters were appreciative of the article, but many others weren’t so kind. In fact, some of the posts were quite hostile, and interactions among some participants got nasty. What most interested me about this was how emotional some people get over this subject. You’d swear I was writing about politics or religion, not the evaluation of audio components.

Therefore, if we do more of these -- which I believe will be a great thing for the reviewing community -- I suspect we’ll see the same sorts of responses: some will applaud our efforts, others will dismiss it, and others will react with anger. But while the latter two responses will be obstacles, there’s nothing we can do about them. Besides, our goal should not be to try to convert to our way of thinking those who are steeped in their own. Rather, our goal is to present the best information we can to our readers, let them do with it as they please, and let the cards fall where they may. Blind testing will help us do that.

To be continued . . .

In these two articles, I’ve talked as openly as I can about the subject of blind testing -- the first detailed why I believe in the methodology, and this one has focused on some of the obstacles we have to overcome in order to do it in a meaningful way. But at least for now, I think, I’ve said enough on the subject. My next step is to work behind the scenes to make blind testing happen here at GoodSound!, and perhaps elsewhere in the SoundStage! Network, at some time in the future. The time frame for implementation will depend largely on how difficult some of these obstacles are to overcome. Look for more updates on this topic, likely in the fall.

. . . Doug Schneider
editor@goodsound.com