REW measurement results from an ESL

[kmartinka]

I am newer to the forum and attempting my first calibration with my home theater system. Below is my 5.1 setup.

Sound Card - EVGA NU Audio Pro 7.1 (running on a windows 10 HTPC)
Amp - Emotiva XPA-5 Gen2
Front - Martin Logan EM ESLs
Center - ML EM C2
Surround - ML EM FX2
sub - (2x) ML Dynamo 1000x
Room size 24'x28"x18'(sloped ceiling)

• Front mains are calibrated(using flashlight technique), 9' apart and 15' from listening position
• All speakers are calibrated to 75dBC(at an offset of 33dBFS from 0dB)
• Using EQ APO in Windows 10 (64-bit) (no filters applied)
• miniDSP UMIK-1 for measurement
• REW 5.20 RC5

I took 9 measurements of the left front speaker in a parallel pipe type array from the listening position(the couch). I was very surprised of the REW results with a roll-off starting at 2kHz. I am wondering what I can do to improve this. I have read and understand how difficult it can be to use dipole ESLs. The only "remedy" I can do was use multiple HS 6dB filters to boost the upper frequencies to match Harman house curve. This obviously made a major improvement but seems extreme to have use these cascaded filters like this. Below is the uncorrected graph.

I have attached the .mdat file for further review. There was no rhyme to the reason for anything that I did, as this is all experimental and I'm still learning the tools, software, and techniques. Any advice would be much appreciated. Thanks in advance.

 

[John Mulcahy]

Try pointing the mic directly at the speaker and using the zero degree cal file to see how the measurement looks then.

 

[jtalden]

Oops wrong thread.

 

[ddude003]

A couple of thoughts... Can you tell us more about your room... Any acoustic treatments? How/where are the ESLs positioned in the room... Furniture locations... Floor carpeted? Photos or drawings might help... What does a measurement look like at 9 feet? Try a little more toe-in or toe-out from the flashlight position... Maybe try a higher mic placement to see if the highs are beaming over your head... That is an issue I had... Lower the front feet and raise the back feet...

Then there is the amp... Do you have a 4 ohm setting or terminals you can try? Do you have any idea how the amp behaves when the speaker dives to 1 ohm at those high frequencies?

Lastly, I would not boost highs... I would cut lows...

 

[kmartinka]

Try pointing the mic directly at the speaker and using the zero degree cal file to see how the measurement looks then.

John, thanks for the reply. Below is the measurement result from the mic(with standard cal file) pointed directly at the left speaker. The result is compared against the average of the 9 original measurements.

 

[kmartinka]

A couple of thoughts... Can you tell us more about your room... Any acoustic treatments? How/where are the ESLs positioned in the room... Furniture locations... Floor carpeted? Photos or drawings might help... What does a measurement look like at 9 feet? Try a little more toe-in or toe-out from the flashlight position... Maybe try a higher mic placement to see if the highs are beaming over your head... That is an issue I had... Lower the front feet and raise the back feet...

Then there is the amp... Do you have a 4 ohm setting or terminals you can try? Do you have any idea how the amp behaves when the speaker dives to 1 ohm at those high frequencies?

Lastly, I would not boost highs... I would cut lows...

Thank you for the reply.

The room size is roughly around 9000 ft^3. The only acoustic "treatments" are medium density drapes(9'L and 6'W) on the side wall to front left and rear left of the listening position.

There is a full size fire place that is 13' wide that is between the 2 side windows with drapes. To the right(pseudo right wall) of the listening position is a mostly open floor plan that leads to a dining room and bar area

The ESLs sit ~4.5' offset from the center line of the room and are 2" offset from the front wall. The ESLs are between a 7.5' wide wood TV stand. There 2 full length 8' couches in a "L" shape formation in the center of the room with a coffee table. The floor is concrete with thick carpet installed over it.

The measurement taken at 9' away was very similar but only couple dBs higher across the frequency range.

I have calibrated the toe-in/out to be(flashlight beam) about 1/3 from the inside corner on each ESL. This seemed to be the best position as every other position resulted in a slightly worse measurement.

I also took a measurement at 3'(1m) away (and 3' fro. the floor) from the speaker with the mic pointing directly at the center of the ESL. This also resulted in a similar wave at around 6 dBs higher across the frequency range.

The amp is 4-ohm capable and there is no setting or terminals. I am using unbalanced RCA inputs. The RCA cables are high quality double shielded cables at 10' length. If I push the amp beyond 80 - 90% at very high frequencies (15kHz - 25kHz), it will go into protect mode. Otherwise anything less than that is fine.

 

[Sonnie Parker]

I am a huge ML fan... have 13 ML's in my room.

What is your movies to music ratio... and are you the only one in the room most of the time, or do you have others in there that you are really trying to impress, or that have the capability to know the difference in what they are hearing?

 

[kmartinka]

Hi Sonnie, nice to speak with another ML fan. I've had my ML setup for 8 years now and love it. The wife loves it too which is always a plus. I have not had the opportunity to listen to 13a model which I'm sure sounds amazing.

I've built my own home server which has all my media stored on it and the server is also used as a home automation PC.

My movies to music ratio... Do you mean owning or preferring? I weigh them equally.

It is just my wife and I (my daughter as well) that enjoy the HT setup. And we enjoy the unique dipole wide dispersion sound. But of course we get plenty on compliments on the setup and sound.

 

[Sonnie Parker]

Something you might consider if you listen to music is simply optimize for the center most listening position. That way your imaging, sound stage and depth acuity is going to be much more improved over trying to optimize for more than one listening position... and you'll likely find that the other two positions beside you will sound as good as they did when you optimized for all three positions, yet you'll get the benefit of the primary listening position being improved.

You might also try pulling your two front main speakers out from the front wall more, closer to you... maintaining the 9' distance between the speakers. I have found that less of an equilateral triangle sounds significantly better in my room. Say keep the 9' distance between the two speakers... maybe even try 10'... and keep your distance from the center position at 1' less... 8' if 9' and 9' if 10'.

At many of the shows we covered we would take a chair and move it forward while listening and nearly every time it sounded better. Of course as always, to each their own and YMMV.

 

[ddude003]

@kmartinka Thank you for the additional information... Have you measured just your center speaker to see if it has a similar roll off or the other ESL... Could it be something in using EQ APO in Windows and/or miniDSP?? Where are you connected with the mic? Computer USB or the EVGA NU Audio Pro 7.1? or? Are you sure that the EVGA software might be adding some "Advanced EQ"? Is there a filter jumper in the board itself? Many AKM based DACs do...

 

[kmartinka]

@Sonnie

Yes, I have spent many hours trying different combinations of speaker angle and depth to achieve the best sound for the room. I have to say I have no complaints there at the moment. I do agree, even a small change can have a large impact. But I may revisit this again.

 

[kmartinka]

@ddude003

I have done some more testing and I am starting to question the miniDSP UMIK-1 (connected to the USB port on the motherboard). I have taken a measurement with the center and surround speakers and both show a very similar result in the graph. I have also tested the front speakers with the onboard Realtek sound card and I get a similar result as well. Either Windows 10 is providing some "advanced EQ" or the miniDSP mic has a problem. EQ APO was disabled when the tests were taken.

 

[kmartinka]

I decided to try the UMIK-1 on my Dell Laptop that has JBL 2.1 speaker setup. I tested the UMIK-1 against the integrated mic. Below are my results.

The UMIK-1 measurement result of the laptop speakers seems similar to the ESL response. The integrated mic also had an interesting response. But since my lack of knowledge and experience with the UMIK-1, I am not sure what to make of this. I will send an email to miniDSP support asking for assistance

 

[xrazor77]

 

[BenToronto]

Measure the signal going into your speakers.

(Without going into details of using REW, alligator clips, and a DAC, you can simply use a $10 voltmeter at a few tell-tale frequencies.)

B.

 

[BenToronto]

I decided to try the UMIK-1 on my Dell Laptop that has JBL 2.1 speaker setup. I tested the UMIK-1 against the integrated mic. Below are my results.

The UMIK-1 measurement result of the laptop speakers seems similar to the ESL response. The integrated mic also had an interesting response. But since my lack of knowledge and experience with the UMIK-1, I am not sure what to make of this.

Absolutely true. In the interest of stimulating more REW by fearful audiophiles, I've posted results for my laptop mic showing how EASY it is to do REW at zero cost using the laptop. There are obvious limitations, of course, but the general benefit is wonderful.

BTW, the calibration correction of that family of condenser mics is so tiny, makes little difference if you plug it in or not. Ummm, I think I recall you shouldn't point the mic at the speakers..... but at 90-degrees

 

[kmartinka]

I took a few measurements of the ML EM ESL with the UMIK-1 connected to my Dell laptop. I took 3 measurements with UMIK-1 pointed at the speaker(with normal cal file), then without the cal file, and finally at 90 degrees with cal file for 90. Then I took 1 measurement with the Dell integrated mic(not calibrated). See results below.

The Dell integrated mic is measuring much higher in dBs in the higher frequency range. I am starting to not trust the results from the miniDSP UMIK-1.

 

[BenToronto]

Relative to the gross absence of treble, those curves look pretty much the same to me and as far as I can make out anything in the tiny legend, make sense in their differences and add cred to your "lab skill".

Interesting you have separate calibration curves for straight and 90-degrees. You can open and read the cal files (at least on a Mac). How do they differ, just generally speaking?

Any chance your computer is using its internal mic for some or all the curves when you think you are using your mic? My MacBook sometimes does that if I'm not careful.

Please run a curve or do some freq spot checks on the electric signal going INTO the speakers. Confirming the input is an essential step.

B.

 

[kmartinka]

I agree the trace for each measurement is similar but the diff in amplitude is what concerns me. I captured two images that contain each mic cal data trace. See attached examples.

I am certain the laptop's internal mic was only used when selected. I tested the feedback each time before I took a measurement.

I will get the freq measurement from the amp with my DVOM as requested. Will post later on this.

 

[BenToronto]

The height of the curves are of no importance since they reflect the connections and mixing and various amplifications and conversions etc along distinctive long paths.... unless you scrupulously matched them all to absolute levels using an SPL meter.

Thanks for posting the calibration curves. As you see, hardly worth losing sleep over using them or not up to maybe the limits of adult hearing about 14kHz - which is far above what many recording engineers can unknowingly hear. Being able to pinpoint sound levels over 8 kHz of little importance since it has little value in predicting your choice of house curve.

How about some data on inputs to speakers? Voltmeters aren't ideal for this measurement but certainly truthful enough to see if the input looks pretty flatish north of 2kHz.

 

[kmartinka]

Thanks for all your replys and assistance with this. I am learning more about this as I go along.

I performed a few measurements on the amp using the signal generator in REW. I used the default FS sine wave at 1 Vrms. I measured the amplifier speaker output with my Fluke 87 DVOM. I took measurements at 100, 1k, 2k, 5k, 10k, 12.5k, and 16k Hz. I measured AC voltage and frequency with windows volume at 100% and -33dBFS pk(and 0 dBFS pk) in the generator settings.

I correctly measured each discrete frequency input on the speaker output. I also measured ~1Vrms at -33dBFS and ~50Vrms at 0dBFS for each frequency point. Does this seem reasonably normal? This test was done with eq apo disabled.

 

[BenToronto]

First some clarifications, please.

Was your speaker connected during the tests (it must be for electrics to work right and for you to hear if the speakers are playing in a more or less normal range)? The exact level doesn't matter as long as you are consistent.

Why did you disable the the EQ (or anything else) that ordinarily would be in the circuit?

What were the voltages you measured at say, 100 Hz, 1kHz, and 5 kHz?

In your last paragraph, can you explain better the meaning of "I correctly measured each discrete frequency input on the speaker output"? Aren't you measuring the the speaker input?

B.

 

[kmartinka]

Please forgive my COVID brain. I've been battling this horrible virus for the past 3 weeks and it's been causing some brain fog.

I did not follow your instructions correctly and measure the signal going into the speaker. I was attempting to test the hardware without any signal conditioning from the EQ, which is why I disabled EQ APO in my first test.

So I restarted the test with the signal generator and measured the frequency and voltage at the speaker inputs (with the speaker connected to the amplifier). I took measurements at 100 Hz, 1kHz, and 5kHz. I measured the following voltages at each specified frequency.

1.) 0.52Vrms at 100 Hz
2.) 0.50Vrms at 1kHz
3.) 0.52Vrms at 5kHz

Let me know if this what you were looking for.

 

[BenToronto]

That appears to be a legitimate test of the signal going into your speaker and it seems to be flat. And fortunately it took very little time.

Subjective judgments on an audio system you live for any length time are not too trustworthy. But have you played something with acoustic bells, triangle, "classical" percussion music, or - wait for it - cowbells? Sound OK on earphones but not speaker under test?

Have you tried unplugging the MLs to see if that makes a difference? (I don't know enough about the ML electronics - might turn off entirely or just the ESL part.)

That leaves at least two other general sources of the substantial drop-off you've posted in the treble: (1) the speaker or (2) the measurement system.

Is there any way you can borrow another speaker or even just an umounted driver and see how it measures? Using REW to sweep 800 Hz to 3kHz would be informative.