Selah Audio RC3R 3-way Speaker Review

[amirm]

This is a review and detailed measurements of the Selah Audio vintage RC3R 3-way speaker. It is on kind loan from a local member. From what I can tell, the Rc3R came out around 2003 and it cost full built for around US $1,300. Kits were available for half the price.

Please excuse the poor mobile phone picture as the unit is still on the measurement stand:

The cabinet seems quite solid and finish nice.

As you can tell, it is a 3-way design. Here are the specs for the drivers which I had a hell of a time to find:

Drivers
(1) Scan-Speak 18W8545
(1) Morel MDM-55
(1) ATD LeRibbon

I am only going to present the measurements as I have to rush to our audiophile society meeting where I can return this and other speakers to their owners.

Measurements of the Selah Audio vintage RC3R were performed using the Klippel Near-field Scanner (NFS). This is a robotic measurement system that analyzes the speaker all around and is able (using advanced mathematics) to subtract room reflections. It also measures the speaker at close distance ("near-field") which sharply reduces the impact of room noise. Both of these factors enable testing in ordinary rooms yet results that can be more accurate than anechoic chamber. In a nutshell, the measurements show the actual sound coming out of the speaker independent of the room.

Spinorama Audio Measurements
Acoustic measurements can be grouped in a way that can be perceptually analyzed to determine how good a speaker can be used. This so called spinorama shows us just about everything we need to know about the speaker with respect to tonality and some flaws:

We see good integration between the drivers as the on-axis curve (black) is mostly smooth. However, it is rising up in level as frequencies climb to 20 kHz. As a result this speaker should sound rather bright if you full absorb the reflections.

Digging into the directivity indices gives us this:

Two observations are apparent as noted on the graph. The early window directivity index (dashed blue) at a glance tells us how similar the on-axis and important reflections are. If the two were identical, the graph would be a flat line at 0. If it is rising, it means that the highs are rolled off relative to on-axis/direct sound. This is fine and a good thing. What is not a good thing is if you see variations along the line. And that is what we see. This means that the sound that bounces from surfaces around your room will have a different frequency response than what you hear. Speakers like this tend to not do well in controlled, preference testing.

The above also means that the sound that you hear will be room dependent with respect to what absorption there is at various angles to the speaker. Block some of them and the overall response/tonality changes even though the direct response is not changed. Such kinks also cannot be fixed with EQ because they occur in the room, not upstream of it.

Looking at an estimate frequency response you would get in a typical room we get:

So depending on what you play, the tonality of the music is changed a lot. An ideal response would look like the arrow I have drawn.

High level analysis is done at this point. All else being equal, this is not a speaker you want to buy.

Speaker Impedance Measurement
Speaker is spec'ed at 5.5 ohm impedance with lowest at 4.5 Ohm and we have that more or less:

Advanced Measurements
I have to run to the meeting!!! So here are the rest of the graphs for you all to chew on:

"Spin" data is enclosed.

Conclusions
We have been "unlucky" in starting our measurements with more recent designs which are more or less following what is right in speaker design. Good to go back in time and look at a speaker that while not terrible, doesn't follow the rule of having good "off-axis" sound. Hopefully this helps calibrate you all better in what to look for in these measurements.

------------
As always, questions, comments, recommendations, etc. are welcome.

Have to drive 150 miles roundtrip to the meeting and to return speakers to owners. Hopefully that makes you feel sorry enough to donate using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/

 

[mi-fu]

No pink panther rating for this?

 

[LTig]

This is a review and detailed measurements of the Selah Audio vintage RC3R 3-way speaker. [..]
Measurements of the Emotiva 6s was

Wrong name ...

 

[MZKM]

Hmm, that vertical performance ain't pretty, and odd about how the SPL is noticably different vertically off-axis in the bass, as seen in my graphs and yours.
Preference Rating
SCORE: 4.2
SCORE w/ subwoofer: 6.7
https://docs.google.com/spreadsheet...sfUGYtfkXnbpS7tDsqOSSKgVv34f9gpO1nMGi/pubhtml

 

[renaudrenaud]

No pink panther rating for this?

Pink Panther was gone to the meeting before Amir.

 

[thewas]

Digging into the directivity indices gives us this:

Thank you again for the detailed measurements, one small correction though, a similar hump in angle measurements is usually a sign of a resonance, but not a hump in the directivity indices, there it means a discontinuity in the directivity. A similar hump in the angle measurements like a resonance almost disappears in the directivity indices as those are something like the difference of the angle measurements.

 

[mhardy6647]

Looks like a fairly challenging load in the midbass.

 

[flipflop]

Preference Rating
SCORE: 4.2

After looking at the PIR, I expected the the preference rating to be awful. Since it's not, I had another look at the PIR and noticed it's now shown in 2 dB increments instead of 5 dB like in the previous reviews. It's really not that bad at all.

 

[MZKM]

After looking at the PIR, I expected the the preference rating to be awful. Since it's not, I had another look at the PIR and noticed it's now shown in 2 dB increments instead of 5 dB like in the previous reviews. It's really not that bad at all.

Yeah, I always try to keep the same scale for my graphs (this one had low SPL so had to shift the y-axis while keeping the scale in mind, 50dB for SPIN, 45dB for Directivity).

 

[Severian]

It seems like the y-axis scale on the Directivity Index and Estimated In-Room Response charts is exaggerating the directivity issues? Other reviews have had the in-room response in 5dB increments. The spin looks good aside from the rise in the high frequencies, which would be easily corrected with a bit of EQ.

 

[MZKM]

rise in the high frequencies, which would be easily corrected with a bit of EQ.

Or less toe-in.

 

[hardisj]

View attachment 48261

View attachment 48262

Hey, fellas. First, thanks to Amir for posting the raw data. Since I am writing some scripts to help automate some of my own testing analysis (a thread soon about this) I thought I'd use this data as a way to help me iron out the details of how I want to do it. With the data, I re-drew the horizontal polar map because the one provided by Klippel is a bit hard to read in some areas. Particularly, if you look at the 2-3khz region of the "SPL Horizontal" graph you can see some of the off-axis response bits are a bit "hotter" in SPL vs the on-axis response. While the Klippel spectrogram shows this it's really hard to see due the way it is presented.

Spectrograms are either shown "normalized" with the 0 degree axis being the reference point or they are shown in pure SPL figures. I tend to prefer the former view. Just personal choice. Within the normalization method I've seen two different methods:
1) Data normalized not by frequency but rather by maximum SPL. Using the maximum 0-degree SPL - no matter the frequency - and reference everything else against that. Basically SPL @ n degrees - Max(SPL @ 0 degrees).
2) Using each axis' response at each frequency and comparing that to the 0-degree axis frequency/SPL. SPL @ n degrees - SPL @ 0 degrees.

I prefer #2 because #1 doesn't highlight the cases where off-axis response may be higher in SPL than the on-axis.

So, here's my plots of the above for illustration.
#1: is the "bad" way of normalizing (where max SPL, no matter the frequency, of 0-degrees is the sole reference). You'll notice the colorbar goes from 0 to -30.
#2: the "good" way of normalizing (where each axis is referenced to the 0-degrees axis, frequency by frequency. Notice the colorbar starts at +3 and goes to -30.
#3: the same as #2 above but with a finer color gradient.

#1

#2

#3

Also sorry about the tag. I'm not trying to imply the data came from me. I'm just building tags in my scripts for when I start testing and posting my data and didn't think to remove it before I posted this. But surely since I'm posting in this thread as a reply everyone understands this is from Amir's supplied data and not mine.

 

[Karu]

It’s nice the main graphs have the comments on them again

 

[MZKM]

Hey, fellas. First, thanks to Amir for posting the raw data. Since I am writing some scripts to help automate some of my own testing analysis (a thread soon about this) I thought I'd use this data as a way to help me iron out the details of how I want to do it. With the data, I re-drew the horizontal polar map because the one provided by Klippel is a bit hard to read in some areas. Particularly, if you look at the 2-3khz region of the "SPL Horizontal" graph you can see some of the off-axis response bits are a bit "hotter" in SPL vs the on-axis response. While the Klippel spectrogram shows this it's really hard to see due the way it is presented.

Spectrograms are either shown "normalized" with the 0 degree axis being the reference point or they are shown in pure SPL figures. I tend to prefer the former view. Just personal choice. Within the normalization method I've seen two different methods:
1) Data normalized not by frequency but rather by maximum SPL. Using the maximum 0-degree SPL - no matter the frequency - and reference everything else against that. Basically SPL @ n degrees - Max(SPL @ 0 degrees).
2) Using each axis' response at each frequency and comparing that to the 0-degree axis frequency/SPL. SPL @ n degrees - SPL @ 0 degrees.

I prefer #2 because #1 doesn't highlight the cases where off-axis response may be higher in SPL than the on-axis.

So, here's my plots of the above for illustration.
#1: is the "bad" way of normalizing (where max SPL, no matter the frequency, of 0-degrees is the sole reference). You'll notice the colorbar goes from 0 to -30.
#2: the "good" way of normalizing (where each axis is referenced to the 0-degrees axis, frequency by frequency. Notice the colorbar starts at +3 and goes to -30.
#3: the same as #2 above but with a finer color gradient.

#1

View attachment 48279


#2
View attachment 48280

#3
View attachment 48281


Also sorry about the tag. I'm not trying to imply the data came from me. I'm just building tags in my scripts and didn't think to remove it before I posted this. But surely since I'm posting in this thread as a reply everyone understands this is from Amir's supplied data and not mine.

It depends what purpose you want the polar plot to show; how it actually performs, or how good the directivity control is. This is why for my horizontal/vertical plots I show both, but if you only want to show one, it matters what I do you want to obtain from it.

 

[hardisj]

It depends what purpose you want the polar plot to show; how it actually performs, or how good the directivity control is. This is why for my horizontal/vertical plots I show both, but if you only want to show one, it matters what I do you want to obtain from it.

I understand that. Which is why I gave my rationale for the one I prefer.

Any data I provide from my own testing will also include the individual axes' response in a standard frequency vs spl format. But when I look at a spectrogram I want to see the normalized response across the axes. So seeing the normalized response against each frequency step is important to me to quickly spot issues in directivity control.

 

[aarons915]

I disagree about these not being that great of a speaker, they seem very well behaved to me. The only real issue is the rising response, which can be remedied somewhat by pointing them straight ahead.

 

[MZKM]

I disagree about these not being that great of a speaker, they seem very well behaved to me. The only real issue is the rising response, which can be remedied somewhat by pointing them straight ahead.

They are about as good as the LS50 for far-field listening.

 

[RayDunzl]

We have been "unlucky" in starting our measurements with more recent designs which are more or less following what is right in speaker design. Good to go back in time and look at a speaker that while not terrible, doesn't follow the rule of having good "off-axis" sound.

I calmly await the entertaining comments to follow the first speaker you measure that doesn't rely on drivers mounted in a box.

 

[napilopez]

After looking at the PIR, I expected the the preference rating to be awful. Since it's not, I had another look at the PIR and noticed it's now shown in 2 dB increments instead of 5 dB like in the previous reviews. It's really not that bad at all.

Yep, similar for the directivity curves, or at least the early reflections one. It's better than most Revels. Please pay attention to scaling everyone.

I'm actually a bit surprised it didn't score higher. From the spins alone this looks like a great speaker other than a rising top couple of octaves.

@MZKM can you help me understand the difference between the NBD and smoothness scores a bit? I'm having trouble grasping how the NBD_PIR score is so high but the SM_PIR score is so low.

 

[MZKM]

Yep, similar for the directivity curves, or at least the early reflections one. It's better than most Revels. Please pay attention to scaling everyone.

I'm actually a bit surprised it didn't score higher. From the spins alone this looks like a great speaker other than a rising top couple of octaves.

@MZKM can you help me understand the difference between the NBD and smoothness scores a bit? I'm having trouble grasping how the NBD_PIR score is so high but the SM_PIR score is so low.

NBD just cares about FR deviation in small, separate segments.
SM cares about FR deviation and tonal balance, by looking at the segments together.

If you look at the smoothness chart, it does look pretty dang smooth, not a lot of jaggedness, so NBD is high; however, <500Hz the points are above the regression line, 500-1000Hz is below the line, etc., this leads to a lower SM score, as it shows slight de/emphasis in certain ranges, even if those ranges are consistent.

Amir’s graph of the PIR shows this to a magnified degree.