Revel F328Be Speaker Review

[amirm]

@amirm I am not trying to cast doubt on your subjective impressions or your measurements, but rather to highlight two signals that appear to contradict, in a way that many of us are struggling to reconcile. Do you have any thoughts on why the F328Be had such a deeper audible bass response vs e.g. the F208, despite the NFS measurements showing the exact opposite?

I don't like the way the bass response in F328Be flattens for a bit. I think that is a sign of measurement error in that region. With both ports being in the back, the low frequency soundfield is more complex than F208, resulting in more measurement error. I think this error is also visible in Harman measurements with wavy response in that area.

On listening results, this is the first speaker I have tested in our living room so that is a nasty variable as well.

For a bunch of analytical types, this situation is the worst we could have been handed! Maybe Harman knew this and hence the reason they did not want to cooperate with me to measure this speaker.

I think at this point we need to decide how to move forward, rather than debating a fuzzy picture of data we have. I have a ton of speakers to measure with the backlog now substantially worse than electronics. Do I spend the time messing with bess measurements of these complex and difficult to test speakers or move on to others?

 

[richard12511]

Kal Rubinson couldn't reliably tell whether he preferred the Salon 2 or the F228Be. That would imply, if there are any differences, they are at best subtle.
https://www.stereophile.com/content/blind-listening-harman-international

Yeah, I would think the differences are quite small. Amir's quote of Kevin Voeks saying the Salon2 smokes it in blind tests seems really hard to believe. I'm sure it beats it, but the measurements aren't all that far apart. Kevin may have been exaggerating a bit

 

[amirm]

And oh, I have a logistical problem here. The F328Be comes in massive road boxes (hard sided cases). It is taking up a ton of space. So ideally, I like to sell this speaker to someone at a good price and free up the space. We literally can't walk through our garage anymore.

 

[amirm]

Yeah, I would think the differences are quite small. Amir's quote of Kevin Voeks saying the Salon2 smokes it in blind tests seems really hard to believe. I'm sure it beats it, but the measurements aren't all that far apart. Kevin may have been exaggerating a bit

That was relative to what became F226Be, not F328Be.

 

[phoenixdogfan]

Kal Rubinson couldn't reliably tell whether he preferred the Salon 2 or the F228Be. That would imply, if there are any differences, they are at best subtle.
https://www.stereophile.com/content/blind-listening-harman-international

Kal Rubinson couldn't reliably tell whether he preferred the Salon 2 or the F228Be. That would imply, if there are any differences, they are at best subtle.
https://www.stereophile.com/content/blind-listening-harman-international

Kind of think the difference is going for the most part going to be bass response and loudness capability across the entre Revel line. If you don't have a room for Salon 2s (to say nothing of a budget), then 226s may be just the thing.

 

[richard12511]

I don't like the way the bass response in F328Be flattens for a bit. I think that is a sign of measurement error in that region. With both ports being in the back, the low frequency soundfield is more complex than F208, resulting in more measurement error. I think this error is also visible in Harman measurements with wavy response in that area.

On listening results, this is the first speaker I have tested in our living room so that is a nasty variable as well.

For a bunch of analytical types, this situation is the worst we could have been handed! Maybe Harman knew this and hence the reason they did not want to cooperate with me to measure this speaker.

I think at this point we need to decide how to move forward, rather than debating a fuzzy picture of data we have. I have a ton of speakers to measure with the backlog now substantially worse than electronics. Do I spend the time messing with bess measurements of these complex and difficult to test speakers or move on to others?

How many big tower speakers with ports do you have coming up? If you have a bunch of tower speakers coming up in the queue, my vote would be to first get to the bottom of this bass issue, as it's likely to come up again in those reviews.

 

[amirm]

How many big tower speakers with ports do you have coming up? If you have a bunch of tower speakers coming up in the queue, my vote would be to first get to the bottom of this bass issue, as it's likely to come up again in those reviews.

They are infrequent. I say the ratio of bookshelf to tower is about 10:1. I currently have a couple of very low end ones to measure and an offer of another Revel in the line from local members.

 

[flipflop]

I think at this point we need to decide how to move forward, rather than debating a fuzzy picture of data we have. I have a ton of speakers to measure with the backlog now substantially worse than electronics. Do I spend the time messing with bess measurements of these complex and difficult to test speakers or move on to others?

If the error is approximately 1 dB, as you mentioned on page 13, I'd say it's time to move on.
Even if this model had 1 dB more bass, it wouldn't change the fact that it's a bass-light speaker.

 

[minnend]

What's the argument for these Revels at $16k/pr over something like the Linkwitz LX521.4, Kii Three, or Dutch & Dutch 8c? Well, other than aesthetics?

My home system isn't in this price range, but if I upped my budget, aren't the benefits of the non-traditional speakers (primarily in the bass response due to constant directivity as a design goal) huge compared to the F328Be or Salon2? I see the Revels as a very expensive way to get minor improvements over lesser speakers (i.e. Revel is doing an excellent job designing speakers, but we're deep into the diminishing returns region). In contrast, the Linkwitz, Kii, and Dutch are using different designs to directly address known shortcomings. This seems like a huge win at a lower price point (though still not *low* at around $12k/pr).

What am I missing?

 

[andreasmaaan]

I think at this point we need to decide how to move forward, rather than debating a fuzzy picture of data we have. I have a ton of speakers to measure with the backlog now substantially worse than electronics. Do I spend the time messing with bess measurements of these complex and difficult to test speakers or move on to others?

I don't think there's any need to remeasure these speakers specifically, but I think that, if there are doubts about the accuracy of the measurement system when it comes to floorstanders and/or rear-ported speakers, those should be put to rest before more speakers of this kind are measured.

(FWIW, personally I don't have any doubts that the measurements are sufficiently accurate.)

 

[DDF]

With a sample size of only 70 and a model fitting this many parameters, any cross-validation you do is not going to be able to make highly confident generalizations. It's a great start, but it's not even remotely "well validated" in a general or pure statistical sense -- this simply isn't possible with the sample sizes involved given the dimensionality of the space being modeled, so I'm confused by how you can say this.

I agree that as an over-all generalization, its not statistically robust. However, within the constraints of the speakers studied, it significantly advanced understanding except for how speakers interact with the room which is a glaring weakness. I hadn't seen much if any discussion regarding that flaw and I was attempting to point it out to add something new to this circular discussion.

I agree "well validated" was too strongly worded but my motivation was in offering a fig leaf to the inevitable attackers that would derail the conversation by reading too much into the criticism and pointing out "its the best we have". You can't please everyone.

In fact, there are strong arguments that an accurate speaker preference score actually need more feature parameters to avoid being entirely blind to characteristics that are unquestionably important to speaker preference -- for example, independent measures of horizontal and vertical beam width! Unfortunately, adding such parameters would mean even larger sample sizes are likely required for robust results. Regularization methods can help, but such tricks only go so far when you are fundamentally bottlenecked by insufficient data.

I would agree with that but I don't know if Harman would. For example, Bech's DBT JAES papers point out that vertical reflections clearly have a more audible effect on tonality than do horizontal reflections. I know this will be controversial, but the outcomes of these Bech studies get confusingly short shrift in Dr. Toole's book. Other designers such as PSB's Paul Barton give them much higher priority.

P.S. Taking this a step further, we might even question why we should trust any preference score which uses human-designed summaries/features boiled down from the higher resolution datasets. Why not fit a model to the full raw measurements, with no hand-tuned dimensionality reductions at all? In 2020 we now have many tools now to do this (though fitting to such a high-dimensionality problem would yield a model which is not very human-interpretable, even if the predictive power is exceptionally high).

There's always that balance: does one assign expert knowledge to a model (ie Olive's) or do we let the model learn independently? The second option as adopted in modern AI/Ml techniques requires a significant number of data samples to train on, that are just not available here. So, adding expert knowledge is sensible.

 

[echopraxia]

What's the argument for these Revels at $16k/pr over something like the Linkwitz LX521.4, Kii Three, or Dutch & Dutch 8c? Well, other than aesthetics?

Well, it should be capable of much greater undistorted SPL than the D&D and Kii simply due to the much greater woofer surface area and cabinet volume. I don’t know about the LX521.4 but one big advantage of the Revel is that you can actually buy it.

 

[lovemusic]

Interesting -- I think you're the first to express a preference of the F328Be over the Salon2's. So far, the majority of direct comparisons report that the Salon2 still beat the F328Be (including several people on this thread), in both blind and sighted tests. I'd be interested to hear more of your thoughts on this. I'm also interested about how you find the bass response, since the measurements here would imply they have quite a lot less bass extension than the Salon2's.

I used primarily classical music to test as first that's what I love to listen and with a full orchestra it captures the entire frequency spectrum. I'm just sharing my audition experience from recollection.

The wave-guides are years ahead of Salon2, it was particularly visible when I played Pavarotti's "Ah! mes amis". The high Cs were as clear as air! And in La Traviata Anna Netrebko Rolando Villazon brindisi, the integration between the cello piece and the soprano aria was flawless.
The true reveler was Erich Kleiber's Le Nozze de Figaro Overture recorded in 1955. The recording has distortion and noise especially in the base frequency [Drum, Timpani, Cello portions]. The Salon2 added to this distortion especially at higher volumes, while in F328s it was reduced. I have listened to this classic live (Vienna and Berlin Philharmonic) and in recordings over hundred times, and for the first time I heard the subtle piccolo piece towards the end. It brought new life to a classic piece which I enjoyed for decades. However, it was not as clear in the Salon2.

Spins for both 328s and Salon2s are available online, and its clearer that 328s has much smoother performance.

Few things to consider. Owner/listener bias: if I owned/ used to listen to Salon2 for years, I would have been used to it and would have a preference for its signature. FWIW, I went in with a clean slate when listening/buying [upgrade on a 18 yr old active speaker, yikes!], I would have picked Salon2 had it been better. If I factor in aesthetics, Salon2 would be my pick and all in I can see why it costs more. It is a statement piece with incredible performance, the 328s are too understated looks wise.

I'm not a harman employee or a dealer or a reviewer or a tech guy, so don't play weight to my opinion. I'm just a regular guy who likes listening to good music that's reproduced right in the comfort of my humble home.

 

[Sprint]

Will do. At some, point I'll probably do a blind(similar to how I did with the M105, 308p, and 8030c). Did you listen to the 8351A, or the 8351B?

It was older model 8351A. Finally I got 3 * 8340 as LCR which I absolutely love with 8330 as surrounds. The Queens Gambit , Kung Fu Panda absolutely shine and is a treat.

 

[pierre]

I don't like the way the bass response in F328Be flattens for a bit. I think that is a sign of measurement error in that region. With both ports being in the back, the low frequency soundfield is more complex than F208, resulting in more measurement error. I think this error is also visible in Harman measurements with wavy response in that area.

On listening results, this is the first speaker I have tested in our living room so that is a nasty variable as well.

For a bunch of analytical types, this situation is the worst we could have been handed! Maybe Harman knew this and hence the reason they did not want to cooperate with me to measure this speaker.

I think at this point we need to decide how to move forward, rather than debating a fuzzy picture of data we have. I have a ton of speakers to measure with the backlog now substantially worse than electronics. Do I spend the time messing with bess measurements of these complex and difficult to test speakers or move on to others?

Hi Amir,

can you measure once in room? If you get flat bass in your room, combine with the low distorsion, that may explain why you like them.

i don’t think it is worth the time to remeasure except if you get more large speakers to measure. An explanation by Klippel would be nice.

pierre

p.s.: congrats on this 100th reviews. You have double the amount of spin that are publicly available and they are higher precision than most. that’s outstanding.

 

[valerianf]

Congrats Amir for this 100th review. I am impressed with your review output rate.
Regarding the low frequencies measurement of the F328be I have some thoughts related to my tower speakers that have 2 x 8" woofers and a front BR port.
I looked at my own measurement and concluded that what is missing from the F328be graphic is the added energy from the 2x back BR ports.
I made 3 different guesses:
● the measurement room configuration does not allow the reflected energy to come back to the Mic.
● the BR ports are not working properly (they are firing backward and may be too small for 3x 8" drivers).
● 3x 8" drivers are placed in a too small enclosure related to the driver VAS.

My empirical way of verifying the BR port is to place the Mic directly at the output of the port and look at the curve.
If buying a new tower speaker I may buy a F208 (2x 8" drivers and front BR port) but not a F328be.

 

[amirm]

Hi Amir,

can you measure once in room? If you get flat bass in your room, combine with the low distorsion, that may explain why you like them.

Ah, it is work. I tore up the audio system which I need to put back in, plus a bunch of other bits to capture using my good mic. Not looking forward to it....

 

[Kal Rubinson]

Kal Rubinson couldn't reliably tell whether he preferred the Salon 2 or the F228Be.

If you think that 2 out of 3 is statistically valid to indicate reliability or not.

That would imply, if there are any differences, they are at best subtle.

Implication is applicable.

Amir's quote of Kevin Voeks saying the Salon2 smokes it in blind tests seems really hard to believe.

Agreed.

 

[dtaylo1066]

The Revel speakers I listened to at the last Rocky Mountain Audio Fest were among the best at the show. While this model is out of the price range of many on this forum, it is nice to see a company's flagship offering excel in performance and measurements. The design and technology no doubt flows down to lesser priced models. I also like to see it when brands and models under a large conglomerate's umbrella perform well and demonstrate attention to design, componentry, finish and performance.

 

[RC231]

If I can pull off moving back to the Seattle area next year and the pandemic makes it safe I will help Amir move some heavy speakers around if he lets me listen to them