neutralguy
Member
- Joined
- Apr 16, 2019
- Messages
- 68
- Likes
- 157
Does bi-amping make a difference? My Revel Salon2 speakers come with two pairs of binding posts, and the manual says:
Revel does not endorse one particular connection method over another.... The design of this loudspeaker is such that optimal performance can be attained using the standard connection method.
By connection method, do they mean bi-wiring, or is bi-amping included? When I switched to vertical bi-amping (one stereo amp per speaker), I sensed a subtle but definite improvement in clarity of high frequencies.
To start, let's compare frequency responses. In the vertical bi-amp configuration, each Salon2 is powered by one Benchmark AHB2 in its low gain mode, with the high and low frequency binding post each driven by one of the amp's two channels. In the normal configuration, the manufacturer supplied shorting bridge is used and the speaker is powered via one pair of binding posts. Note that in the normal configuration, each Salon2 is still powered by a separate AHB2, i.e. monoblocking using a single AHB2 channel per amp. This means any differences would be attributable solely to separation of high and low frequency amplification, and not to elimination of stereo crosstalk between the amp's two channels.
As we can see, frequency response is unchanged. Variations below 50hz are within usual variation on each measurement, so I wouldn't put any credence on differences there. Still, I was not convinced that the difference I was hearing was all placebo. It's impractical for me to do a blind test, given the time and effort required to change from bi-amp configuration to normal. An ABX with enough repetitions to prove audibility would take dozens of changes, each of which takes longer than ideal for our short audio memories.
Recently in my previous experiment, it was shown that recorded speaker outputs can be used in an ABX test to demonstrate an audible difference for a change of speaker feet, so I wanted to see if this ABX method would also work here. I recorded the output of my Revel Salon2 speakers playing music in stereo using a pair of high quality cardioid microphones (Shure SM81), each mic 2' in front of the tweeter's logo.
I've attached sample clips of the recordings. The original is available free on 2L.co. I think you'll agree upon listening that these recordings are a good enough representation of the original to plausibly reveal differences. I did the ABX test using a piece I'm more familiar with, another recording of a classical string quartet.
The difference was minute, and ABX is hard. There's the time, effort, and mental fatigue. The default of 16 trials requires at least 32 intense listenings of the same passage, likely more. It takes training to tease out perceived differences that are real and can be repeatedly ascertained, versus those that turn out to be due to memory blur or too weak to be detected repeatedly with sufficient accuracy. Repeated listening causes a kind of burn in of the senses and memory. Eat a bite sized piece of single-origin chocolate and you can taste the nuances down to the trees that were planted next to the cocoa. Now imagine eating 32 pieces in a row and then asked to distinguish the nuances between the last two pieces.
I only take a test when I'm able to get good accuracy in training mode, and still my first two tests were only 10/16 and 9/16 correct, which led me to devise techniques to limit fatigue and its effects. It helped to identify an extremely short passage of 2 seconds and to limit listenings by reducing trials from 16 to 10, which is a tradeoff that requires a higher accuracy for statistical significance. I did the third test after training and improved to 9/10 trials correct (p = 0.0107).
It's no wonder to me now that ABX tests have come up null for all manners of changes. If I were a random subject brought in off the street, with no practice and no incentive to get the right answer, I would have performed even worse than my initial trials and we would have concluded there was no difference.
I do think the real difference is greater than is captured by the recordings, but that cannot be shown by this ABX setup.
TLDR
Vertical bi-amping of Revel Salon2 using 2x Benchmark AHB2 has the same frequency response but makes an audible difference.
Revel does not endorse one particular connection method over another.... The design of this loudspeaker is such that optimal performance can be attained using the standard connection method.
By connection method, do they mean bi-wiring, or is bi-amping included? When I switched to vertical bi-amping (one stereo amp per speaker), I sensed a subtle but definite improvement in clarity of high frequencies.
To start, let's compare frequency responses. In the vertical bi-amp configuration, each Salon2 is powered by one Benchmark AHB2 in its low gain mode, with the high and low frequency binding post each driven by one of the amp's two channels. In the normal configuration, the manufacturer supplied shorting bridge is used and the speaker is powered via one pair of binding posts. Note that in the normal configuration, each Salon2 is still powered by a separate AHB2, i.e. monoblocking using a single AHB2 channel per amp. This means any differences would be attributable solely to separation of high and low frequency amplification, and not to elimination of stereo crosstalk between the amp's two channels.
As we can see, frequency response is unchanged. Variations below 50hz are within usual variation on each measurement, so I wouldn't put any credence on differences there. Still, I was not convinced that the difference I was hearing was all placebo. It's impractical for me to do a blind test, given the time and effort required to change from bi-amp configuration to normal. An ABX with enough repetitions to prove audibility would take dozens of changes, each of which takes longer than ideal for our short audio memories.
Recently in my previous experiment, it was shown that recorded speaker outputs can be used in an ABX test to demonstrate an audible difference for a change of speaker feet, so I wanted to see if this ABX method would also work here. I recorded the output of my Revel Salon2 speakers playing music in stereo using a pair of high quality cardioid microphones (Shure SM81), each mic 2' in front of the tweeter's logo.
I've attached sample clips of the recordings. The original is available free on 2L.co. I think you'll agree upon listening that these recordings are a good enough representation of the original to plausibly reveal differences. I did the ABX test using a piece I'm more familiar with, another recording of a classical string quartet.
The difference was minute, and ABX is hard. There's the time, effort, and mental fatigue. The default of 16 trials requires at least 32 intense listenings of the same passage, likely more. It takes training to tease out perceived differences that are real and can be repeatedly ascertained, versus those that turn out to be due to memory blur or too weak to be detected repeatedly with sufficient accuracy. Repeated listening causes a kind of burn in of the senses and memory. Eat a bite sized piece of single-origin chocolate and you can taste the nuances down to the trees that were planted next to the cocoa. Now imagine eating 32 pieces in a row and then asked to distinguish the nuances between the last two pieces.
I only take a test when I'm able to get good accuracy in training mode, and still my first two tests were only 10/16 and 9/16 correct, which led me to devise techniques to limit fatigue and its effects. It helped to identify an extremely short passage of 2 seconds and to limit listenings by reducing trials from 16 to 10, which is a tradeoff that requires a higher accuracy for statistical significance. I did the third test after training and improved to 9/10 trials correct (p = 0.0107).
It's no wonder to me now that ABX tests have come up null for all manners of changes. If I were a random subject brought in off the street, with no practice and no incentive to get the right answer, I would have performed even worse than my initial trials and we would have concluded there was no difference.
I do think the real difference is greater than is captured by the recordings, but that cannot be shown by this ABX setup.
TLDR
Vertical bi-amping of Revel Salon2 using 2x Benchmark AHB2 has the same frequency response but makes an audible difference.
