• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Revel F328Be Speaker Review

HooStat

Senior Member
Joined
May 11, 2020
Messages
349
Likes
258
Location
Calabasas, CA
The weighting of parameters is arbitrary and assessing the outcome by one of the variables is a complication. Is there a single detailed explanation of the entire process?
The weighting isn't arbitrary. The weights are based on a statistical model describing the systematic (not random) relationships in the data. The selection of parameters to include in the model was based on the experience of the researchers. This has an element of "arbitrary" to it, but it is no more arbitrary than a reviewer. The statistical model could be formulated differently -- it seems to under predict at higher predicted scores. However, there is a limit to the number of inputs that can be used with only 77 observations.

Statistical models are definitely a simplification of the real world, but they do work (e.g., weather forecasting).
 

beagleman

Active Member
Joined
Sep 24, 2020
Messages
104
Likes
103
This is still far from flat down to 30hz or even 40hz, which IMO is a pretty fair thing to ask of a big expensive set of towers. In contrast for example, my much smaller Genelec 8351B’s are +-6db down to 32hz anechoic. Unless I’m missing something, this means the much smaller Genelec 8351B’s should sound much deeper than the giant Revel F328Be towers.

Of course, I know the laws of physics say the Genelec 8351B will not be capable of as much maximum bass SPL as the F328Be independent of the frequency response tuning, but I don’t see any real reason why the Revel F328Be can‘t be tuned to be flat anechoic down to 32hz also. With 3x the woofer area of the 8351B, it seems the F328Be should be able to achieve this even at reference SPL.

I think the manufacturer specs show the -3db point far lower than the test on here did.
 

CDJ123

New Member
Joined
Mar 25, 2020
Messages
3
Likes
2
The Revel is a very tempting purchase for me. I am also curious how a Vivid Giya or Magico A5/S5 Mk II would stack up against the Revel given their exotic cabinets. Also, somewhere on this thread, someone mentioned DIY speakers. There is a top end Troels Gravesen ATS4-HE for sale on USAudiomart. It would be great if anyone who owns these speakers would volunteer them for testing.
 

echopraxia

Addicted to Fun and Learning
Forum Donor
Joined
Oct 25, 2019
Messages
818
Likes
1,351
Location
Seattle Area
I think the manufacturer specs show the -3db point far lower than the test on here did.
Yes, which just adds to the confusion I guess. Manufacturer specs and Amir’s subjective review indicate it has solid bass response, but the NFS measurements show it should sound less bass heavy than many tiny bookshelf speakers or active monitors reviewed by ASR. I’m not doubting any specific signal here, just expressing that these are mixed and conflicting signals in many ways and it’s not clear why.
 

Kal Rubinson

Major Contributor
Industry Insider
Joined
Mar 23, 2016
Messages
2,598
Likes
3,652
Location
NYC/CT
The weighting isn't arbitrary. The weights are based on a statistical model describing the systematic (not random) relationships in the data. The selection of parameters to include in the model was based on the experience of the researchers. This has an element of "arbitrary" to it, but it is no more arbitrary than a reviewer. The statistical model could be formulated differently -- it seems to under predict at higher predicted scores. However, there is a limit to the number of inputs that can be used with only 77 observations.
OK. Have you a link to a full description of the process? I admit to a general bias against such procedures and a preference for the underlying data.
 

richard12511

Major Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
1,939
Likes
2,617
The Revel is a very tempting purchase for me. I am also curious how a Vivid Giya or Magico A5/S5 Mk II would stack up against the Revel given their exotic cabinets. Also, somewhere on this thread, someone mentioned DIY speakers. There is a top end Troels Gravesen ATS4-HE for sale on USAudiomart. It would be great if anyone who owns these speakers would volunteer them for testing.
Man, I would really love to see Vivid or Magico speakers on the Klippel. A lot of the uber high end boutique speakers would probably be “exposed” by the Klippel, but those two manufacturers seem to be guided by really solid engineering.

The odds of owners sending in such speakers for testing are quite low I imagine, unfortunately :(.
 

echopraxia

Addicted to Fun and Learning
Forum Donor
Joined
Oct 25, 2019
Messages
818
Likes
1,351
Location
Seattle Area
Quick search shows that “sink” means reverse polarity ( out of phase). No clue on benefit though.

3 woofers is +4.8dB over a single woofer. It is around 91.5dB from 200Hz-1kHz, so that would put each woofer at 86.7dB. So, only if their individual sensitivity is higher than that could they keep the high sensitivity as well as extend the bass. Also, do keep in mind Amir said the bass may be a tiny bit higher in level than his measurements show.
The roll off makes sense if it was a compromise made for higher sensitivity, but again this is where I’m seeing mixed signals on this thread: Many people are saying this design is intentional to tune them to expected boundary reinforcement. But that also conflicts with the idea that the F208 is a well tuned speaker regarding bass output.

You can't have it both ways -- the F208 and F328Be cannot both be simultaneously tuned to perfection regarding bass:
  • If the F328Be's bass is not too weak, then it mathematically follows that the F208's bass must be too strong!
  • If the F208's bass is not too strong, then it mathematically follows that the F328Be's bass must be too weak!
So yeah I think your explanation makes more sense than anything else on the thread so far: The bass roll off is an engineering compromise made by Revel to achieve higher sensitivity. But that means it definitely is a compromise, and a step backward in some sense vs e.g. the tonal balance of the F208 (and of course the Salon2’s). If not, then someone must claim either that the F208’s are excessively bass heavy, or that there’s something wrong with the measurements.
 
Last edited:

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
2,909
Likes
6,746
Location
Land O’ Lakes, Florida
OK. Have you a link to a full description of the process? I admit to a general bias against such procedures and a preference for the underlying data.
https://www.researchgate.net/public...e_Ratings_of_Around-Ear_and_On-Ear_Headphones
It says headphones, but scroll down for the speaker paper.

I don’t think anyone thinks it’s super accurate, especially as how it’s been identified that it favors narrow directivity. But, in regards to tonal balance, it should give you a good ballpark.
 

MediumRare

Major Contributor
Forum Donor
Joined
Sep 17, 2019
Messages
1,196
Likes
1,294
Location
Chicago
Imagine you are a big loudspeaker manufacturer seriously committed to science and you want to formalize a scientific method that would guide your engineering activities for several years.
What you would do is invest millions by getting the best audio scientist in your r&d doing large scale experiments. You want them to come out with a formula that almost perfectly predicts, from objective measurements, what most consumers will subjectively perceive as a quality sound.
Now that you have the secret sauce you have three possibilities:
1- Publish your results entirely so that other manufacturers can join the campaign for a better audio and a better planet, no matter if one day they will leverage your investments to grab your market share
2- Hide the secret sauce, but then the whole story of your epic r&d effort cannot be used to strengthen your brand reputation, your company will be one of the many in the high end industry that claim to have found the best snake-oil ever, with no evidence. You might become like the enemy you are fighting.
3- Do something in the middle, keep the special ingredients of the sauce secret, now publish a recipe that is good enough to celebrate your r&d superiority, your competitors are still far away to get even closer to it, gain some time, Keynes said "in the long term we are all dead".

(some years later... destiny is sometimes ironical...)

One of the former snake-oil merchants from Britain has become theoretically superior to you according to the published formula, thanks to a proper design of coincident drivers.
According to the published formula, there is no objective reason to buy your top-of-the line speaker costing several times more than your mid-line.
A real threat is taking place: manufacturers of "pro" market are launching everyday high performance active monitors for few bucks, just pray none of them wants to make a nicely looking floorstander...

What an exciting story! Any clue on the sequel?
Your scenario makes complete sense and isn't nefarious at all. Patents don't necessarily give away the secret sauce either, even though they purport to divulge the invention. They include what is necessary to protect the invention, not to replicate the results (trade secrets). Revel has said that the measures are only a part of the development process and that listening tests are still required.
 

echopraxia

Addicted to Fun and Learning
Forum Donor
Joined
Oct 25, 2019
Messages
818
Likes
1,351
Location
Seattle Area
OK. Have you a link to a full description of the process? I admit to a general bias against such procedures and a preference for the underlying data.
https://www.aes.org/e-lib/browse.cfm?elib=12847
https://www.aes.org/e-lib/browse.cfm?elib=12794

You’re not wrong to be skeptical of a model’s reliability without knowing the robustness of the underlying data. In this case, we’re talking about a hand-designed model whose parameters/weights were statistically derived to fit to a sample size of just 70 speakers (which are now ~20+ year old designs). Before half the people here attack me for criticizing one of their beloved authority figures (which is absolutely not what I’m doing), I’ll point out that even Olive himself is very open about the statistical weaknesses of the model, just like any good scientist.

The model’s weights are not arbitrary, but just because a statistical method decided them doesn’t implicitly make the model perfectly reliable beyond question. I can fit statistical models to small datasets and yield models all day long that are better than random chance but still deeply wrong in many important ways. I’m not saying the Olive score is deeply wrong by any means, but am merely making the point that it’s not nearly as strong enough to justify the level of trust that some people here put in it. I see a lot of people citing it as a source of settled scientific fact while simultaneously demonstrating many fundamental misunderstandings of how statistical modeling works (and the spectrum of reliability/unreliability yielded in various circumstances).

That said, this same debate of the Olive score comes up in every thread, so this info is probably not new to many reading this. When a review correlates well to the Olive score, we have all the fans of the score professing the Olive score’s indisputable reliability and perfection. And when it doesn’t (like here), we have those fans questioning Amir’s taste in speakers and of course a lot of people come out to attack the Olive score as utterly useless. It’s not useless, it’s just far from settled science.
 
Last edited:

Sprint

Active Member
Forum Donor
Joined
Sep 19, 2019
Messages
210
Likes
107
That’s very similar to my views with the 8030c and M105. It’s not a strong preference either way, but I think I prefer classical music on the Revels, and modern music on the Genelecs.

I’ve got a pair of 8351b on the way. Really curious how they will compare to the 8030c. I’m hoping there will be more to the difference than just bass and headroom, but the measurements do look pretty similar.
Please share with us your experience on 8030 Vs 8351. A couple of years ago, I did a AB comparison of 8050 and 8351. I was surprised that 8050 was on par with 8351 with respect to details. the 8351 had a bit more of separation.
 

HooStat

Senior Member
Joined
May 11, 2020
Messages
349
Likes
258
Location
Calabasas, CA
It’s not useless, it’s just far from settled science.
I agree completely. And science often makes is best progress when its understanding (i.e, a model) isn't sufficient. This often leads to new investigations to improve the models. Which is why weather forecasting is so much better today than in the past.
 
Joined
Oct 29, 2020
Messages
21
Likes
46
Amir, congrats on the 100th speaker review.

The measurements just proves my listening experience. I went in looking for speakers under $50k (auditioned everything from Wilson, B&W, Sonus Faber, Paradigm, KEFs) and landed on these. To me, Revel has produced a gem at around $15k price range and has set a benchmark for others to follow both in terms of performance and price.

To my ears they were slightly better than Salon2 (listened on the same day, equipment, place and music) on the tweeter and base response; and being $6k cheaper the 328s were an easy pick.

The only downside is the cabinet, but hey I got these to listen not to look.
 

franspambot

Addicted to Fun and Learning
Forum Donor
Joined
Mar 30, 2019
Messages
533
Likes
1,105
Can anyone explain what the blue line (total sound power) is doing? Does it perhaps speak to the confusion over the bass tuning?

Of interest to some may be the fact that this official measurement was taken the day before the one for the F226Be. The results are similar in the bass range. I put together a slew of measurements of the latter from Harman, Erin, and myself that may be helpful in comparison back in page 13.

Spin - Revel Performa3Be F328Be raw.png
 

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
2,909
Likes
6,746
Location
Land O’ Lakes, Florida
Can anyone explain what the blue line (total sound power) is doing? Does it perhaps speak to the confusion over the bass tuning?

Of interest to some may be the fact that this official measurement was taken the day before the one for the F226Be. The results are similar in the bass range. I put together a slew of measurements of the latter from Harman, Erin, and myself that may be helpful in comparison back in page 13.

View attachment 93031
It measures sound in all directions. It is showing that at the rear of the speaker you are getting more output due to it being rear-ported. To my knowledge though, it should still be omnidirectional or close to it, but it of course is moving air and hence why people generally want front-ported speakers if using close to the front wall.

So, I am assuming that it shows that Harman’s chamber/measurement is indeed flawed in the bass.

EDIT: It’s not a simple average though, it is weighted by measurement angle:
https://www.audiosciencereview.com/...gs-for-loudspeakers.11091/page-26#post-474432
 
Last edited:

echopraxia

Addicted to Fun and Learning
Forum Donor
Joined
Oct 25, 2019
Messages
818
Likes
1,351
Location
Seattle Area
To my ears they were slightly better than Salon2 (listened on the same day, equipment, place and music) on the tweeter and base response; and being $6k cheaper the 328s were an easy pick.
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.
 

DDF

Addicted to Fun and Learning
Joined
Dec 31, 2018
Messages
534
Likes
850
I’m not saying the Olive score is deeply wrong by any means, but am merely making the point that it’s not nearly as strong enough to justify the level of trust that some people here put in it
I think the model was well validated across different loudspeaker types, and very well validated across different listener types.

Its main weakness is that it wasn't validated against very many different room types. A main point of the model is to study how preference is affected by off axis behaviour, which is strongly dependent on room construction and placement within the room.

They did find that relative (not absolute) ranking seemed to hold up reasonably well across a very limited number of different rooms, but the number of rooms studied was far from enough for statistical relevance. I think if a large number of room and placement types were considered, the model may end up splitting into 2 or 3 unique sets of coefficients, optimized for 2 or 3 broadly different room environments. Unfortunately, executing that testing would be prohibitively expensive.

It's good to learn how to read the spins and extrapolate those to how a speaker would sound in a particular room.
 
Joined
Mar 9, 2017
Messages
54
Likes
32
Location
My kind of town
Amir, congratulations on the 100th review! What an amazing milestone!

I would be very interested in seeing the measurements for the cheapest model in the line, the F206. This is not only the cheapest speaker, but with its smaller size it may appeal to a much broader audience then even the F208.
 

phoenixdogfan

Major Contributor
Joined
Nov 6, 2018
Messages
1,027
Likes
895
I'd like to have those, but they are too big for my home. Until Revels gets some of similar quality that fits into an apartment or I buy a new house this is for me like looking at lamborghinis.
Same boat here. There are any number of smaller speakers which are excellent. From the Kef Reference1s to the Gradient 1.4s, to the D & D 8Cs, Revel 226s, GGNTKT M1s, Genelec 8341s, or even now the Mesanovic RTM 10s. No shortage and they should all excel in smaller rooms.
 

Similar threads

Top Bottom