Speakers sound exceptionally harsh, ran measurements, where to go from here?

[materialbasis]

Hello all!

I recently picked up a pair of speakers I've wanted for quite some time, a pair of Dunlavy SC-III speakers. The upper mid to lower high range frequencies sound unbearably harsh. They didn't sound this way when I picked them up, given I did not listen to them for a long time. I'm trying to rule out if there is something wrong with the speakers themselves, or maybe my environment is just seriously hard for them to work with? I was running Goldenear Triton 5 speakers before in the same spot, which had the opposite issue, the upper mid/highs were pretty recessed.

I tried changing amplification from a Peachtree Nova150 to an Arcam Radia A25+, did not change much on the front of harshness. I decided to run some pretty amateur measurements, and they seem to confirm what I am hearing.

I did a few sweeps from listening position. I'm using a Dayton Audio IMM-6C, via HouseCurve, using a the calibration file for my specific mic. I also tested using my iPhone's internal mic, similar pattern just at different db levels. So I'm pretty confident in this measurement, unless both the Dayton and iPhone have terribly fucked up mics.

Would I be able to detect something wrong with the drivers from a measurement from listening position? I ran the same tests using one speaker at a time and they look to be relatively the same, so it wouldn't be an issue with one speaker. Are there other tests I should run?

Speakers are around 6 feet apart, listening position is 10 feet away (I know, not ideal.) Speakers are toed in to where I can hardly see the inner panel. I tried changing the position of the speakers, away from the windows and without sidewalls being an issue. The harshness persisted. The speakers were about 8 feet apart, and I was listening from about 8 feet away. I also tried putting some blankets behind the speakers when they were by the windows, it seemed to take down the harshness by 1 or 2 DB but not nearly enough. At its worst, around 2-4k, it's 9db too high.

The details in music sound great, staging is good given the position, I don't necessarily hear obvious distortion. I've checked each speaker individually (except for tweeters). No coil rub, no separation at the spider, tinsels look good. Each speaker seems to be producing sound, none of them appear to be completely dead.

I'm kind of at a loss with these, I've never had to try so hard to get speakers to sound right. Should I have someone come in and test the drivers in a more serious fashion? Could it be a crossover issue? I'm thinking about moving on from these but I'd hate to sell them not knowing if something is wrong with them.

sidewall_stereo.png - both speakers measured at same time

no_sidewalls_stereo.png - measured from new position, without sidewalls being nearby

sidewall_separate - more prominent green is left side, faded green is right. Right side is closer to corner, explaining the different bass response

If folks would like, I can attach a zip with the WAV impulse responses.

 

[Salt]

Crossover issue would look something more weird.
If You have an option to/for EQ, this could be solved easily.

 

[materialbasis]

Crossover issue would look something more weird.
If You have an option to/for EQ, this could be solved easily.

EQ did a pretty good job at resolving a lot of the harshness. I'm pretty new to room correction, I've had somewhat of a luddite/old school approach to it despite being relatively young for an audiophile.

I'm not sure if there is an option to use it for analog listening but I mostly listen to records and would really like for the harshness to not exist for my vinyl listening as well. I would also like to eventually use my Dunlavy's with my TV setup. If I can do that, I'll probably just rest easy and use the RC. It just bothers the hell out of me that these measure so different than they do for so many others who own these.

Attached is a graph for measurements with RC. Used Roon to apply RC.

 

[Daverz]

It looks like your SC-III is a later iteration than mine. I know for mine that the grill is part of the baffle because it has thick felt on each side of the grill for reducing diffraction. It's not clear if you are running them with the grill on or off.

 

[materialbasis]

It looks like your SC-III is a later iteration than mine. I know for mine that the grill is part of the baffle because it has thick felt on each side of the grill for reducing diffraction. It's not clear if you are running them with the grill on or off.

I've mostly listened to them while the grills are off. The harshness existed when I had the covers on as well. Mine is serial 136a and b.

 

[AnalogSteph]

To verify basic response, measure individual speakers at relatively close distance, with appropriate windowing. In-room, again, each speaker separately.

This room looks like it would be much too lively for its own good. That wouldn't be helping matters.

For comparison, here are @Daverz's in-room measurements of his:

Dunlavy Speakers

I was quite interested in the Dunlavy speakers, back in the day. They used cheapish VIFA drivers that I used in a couple of designs I was working on at the time. They were using the D"appolito configuration (Dome tweeter in the middle, a midrange both above and below as well as woofers above and...

www.audiosciencereview.com

 

[Crook]

"i'm not sure if there is an option to use it for analog listening'.

I run my phono pre out through a minidsp pocket adc and digitally into my dac with EQ. Works great, being able to EQ the turntable same as the streamer is much more of a benefit than any loses through extra conversion (inaudible to me).

 

[Daverz]

I've mostly listened to them while the grills are off. The harshness existed when I had the covers on as well. Mine is serial 136a and b.

Mine are SC-III, not SC-IIIA, and have serial numbers 806A/B.

Do your grills not have felt on the sides? The felt tucks into a lip on the side of each grill. The grills on these are definitely meant to stay on when listening.

 

[Keith_W]

It would be better to attach a ZIP file with REW .MDAT measurements.

Harshness is not necessarily seen in the frequency response alone. You have to know where to look - I would start by looking at the freq response, reflections (ETC) and distortion. Note that distortion measurements are SPL dependent - the louder you play, the more distortion you get. Loudspeakers are nonlinear devices - measure at a different volume, and you may get a different frequency response. For this reason, a measurement that "looks OK" may be inaccurate if it was taken at the wrong SPL. I would suggest these measurements:

1. Place your mic at the listening position. Sweep left and right speakers independently at normal listening volume using a 512k or 1M sweep.
2. Increase the volume by 10dB in REW and repeat LR sweeps. Then increase the volume again by 10dB and repeat the sweeps.

If you have an SPL meter or a UMIK-1 or UMIK-2, you could do this measurement formally.
1. Sweep the speakers at normal listening volume at the listening position (this measurement is still needed to look at the ETC),
2. Move the mic 1m on-axis to the speaker and play pink noise. Adjust the output volume until it reads 76dB.
3. Sweep at 76dB, 86dB, and 96dB, adjusting the volume in REW by 10dB steps (and NOT by acoustically measuring the SPL!).

You should also listen carefully when you are doing the sweep. At some point you may hear the harshness. It would be very useful if you could indicate at what frequency you think the harshness is. This is why I recommend a slower sweep (512k/1M). Also, slower sweeps have better SNR and it helps separate the distortion components better.

Note that proper documentation of loudspeaker distortion requires a different procedure, a stepped sine wave. Don't bother doing that for now, just do this and we'll take a quick look.

 

[Alan O]

That room photo screams "square, bare, with hard surfaces" to me.

 

[materialbasis]

Mine are SC-III, not SC-IIIA, and have serial numbers 806A/B.

Do your grills not have felt on the sides? The felt tucks into a lip on the side of each grill. The grills on these are definitely meant to stay on when listening.View attachment 509351

Mine do have felt on the sides. Yours is actually a later version than mine by far, as my felt is lighter and the serial number being lower. Mine is not an A version, I was just relaying the serial number. 136A and 136B.

EDIT: tried measuring with cover on/off, there is a .7 decibel difference in frequency response around 1.5k. It is slightly flatter, but I'd imagine not a huge part of the problem being the covers (at least in terms of FR, I'm sure there are other measurements that would make it different)

 

[materialbasis]

It would be better to attach a ZIP file with REW .MDAT measurements.

Harshness is not necessarily seen in the frequency response alone. You have to know where to look - I would start by looking at the freq response, reflections (ETC) and distortion. Note that distortion measurements are SPL dependent - the louder you play, the more distortion you get. Loudspeakers are nonlinear devices - measure at a different volume, and you may get a different frequency response. For this reason, a measurement that "looks OK" may be inaccurate if it was taken at the wrong SPL. I would suggest these measurements:

1. Place your mic at the listening position. Sweep left and right speakers independently at normal listening volume using a 512k or 1M sweep.
2. Increase the volume by 10dB in REW and repeat LR sweeps. Then increase the volume again by 10dB and repeat the sweeps.

If you have an SPL meter or a UMIK-1 or UMIK-2, you could do this measurement formally.
1. Sweep the speakers at normal listening volume at the listening position (this measurement is still needed to look at the ETC),
2. Move the mic 1m on-axis to the speaker and play pink noise. Adjust the output volume until it reads 76dB.
3. Sweep at 76dB, 86dB, and 96dB, adjusting the volume in REW by 10dB steps (and NOT by acoustically measuring the SPL!).

You should also listen carefully when you are doing the sweep. At some point you may hear the harshness. It would be very useful if you could indicate at what frequency you think the harshness is. This is why I recommend a slower sweep (512k/1M). Also, slower sweeps have better SNR and it helps separate the distortion components better.

Note that proper documentation of loudspeaker distortion requires a different procedure, a stepped sine wave. Don't bother doing that for now, just do this and we'll take a quick look.

Thanks for this. I haven't been using REW because my mic is primarily made for phone use. The only thing I could think to do is either get a tripod that is strong enough to hold up my laptop, while connecting into its USB-C port. Or finding some kind of tripod that can hold the microphone alone, and use a USB-C female to male extender so I can take measurements from a distance. I'm using a Dayton IMM-6c.

Thanks again!

 

[DSJR]

As above, my first thought was "Where's the rug on the floor, or a carpet?"

I know it's been a fashion in more recent times to have living rooms that are seriously lively in acoustic, but I have to say it's not a thing that many loudspeakers like much... The right hand speaker jammed into a corner may not help so much either. Can you try shifting the whole system to the left a bit?

 

[Daverz]

Mine do have felt on the sides. Yours is actually a later version than mine by far, as my felt is lighter and the serial number being lower. Mine is not an A version, I was just relaying the serial number. 136A and 136B.

EDIT: tried measuring with cover on/off, there is a .7 decibel difference in frequency response around 1.5k. It is slightly flatter, but I'd imagine not a huge part of the problem being the covers (at least in terms of FR, I'm sure there are other measurements that would make it different)

The felt is for controlling diffraction, so the effect doesn't necessarily show up in a direct mic measurement.

 

[materialbasis]

Update: I feel much less like this is an issue with the speakers themselves and the issue is definitely with my room. Response is looking and sounding a whole lot more flat. Putting them closer to the rear wall/putting them on the long wall actually helped improve the response quite a bit. Stage sounds much better too. The covers helped a little, the faux rug helped too (u-haul padding and an old comforter). The upper mids and lower highs sound so pleasant now. There is that bit of weirdness in the upper highs, but I guess I'll just have to RC that or try a few more room treatments.

Side note, one side of one cover has a part of the felt missing. I removed the wool/felt and attached it as the upper diffuser on one of the speakers, as the upper diffuser was apparently missing. I'll have to go on the hunt for replacement wool/felt, if anyone has recommendations for these speakers. It'd be nice to use identical material but I'm sure it's near impossible to find now.

Big shoutout to all of the advice and patience from everyone!

The rest below is basically just tangental, as it relates to trying to keep the speakers in this spot and have it work with my living situation:

For now, the position is not ideal for my living situation. I'd like to be able to wire my turntable from across the door, and really don't think it'll look good to have it running across the door to our bedroom. If I want to add a sub, I don't think there'd be a good spot to put it with the TV console in the way. I'm thinking about moving the TV to where the speakers were (where you can see the record cabinet currently), and then getting an audio rack so all my components can sit between the speakers instead of the TV. If anyone else has some genius ideas on how to accommodate this new speaker position I'd be happy to hear them. Really trying to get creative here. Very complicated room, added some pictures of it to see if anyone has some extra ideas.

 

[Daverz]

Update: I feel much less like this is an issue with the speakers themselves and the issue is definitely with my room.

Looks like a tough room for any speakers, especially the SC-III with their distance requirements.

 

[Keith_W]

Thanks for this. I haven't been using REW because my mic is primarily made for phone use. The only thing I could think to do is either get a tripod that is strong enough to hold up my laptop, while connecting into its USB-C port. Or finding some kind of tripod that can hold the microphone alone, and use a USB-C female to male extender so I can take measurements from a distance. I'm using a Dayton IMM-6c.

Thanks again!

I think that all Dayton IMM-6's should be collected and incinerated. The stupidest idea for a measurement microphone ever! Microphones need to be absolutely still when taking a measurement, and they should be clear of any reflective surfaces in close proximity, such as a laptop or a phone. The smaller the distance between the mic and the reflective surface, the higher the frequency of the measurement artefact. The larger the reflective surface as a proportion of the microphone capsule, the larger the amplitude of the error. I can't imagine what kind of moron would design a microphone like that. I don't blame consumers, you don't know any better.

If you MUST use an IMM-6, get an extension cable for it and duct tape it to a broom handle or a tripod.

 

[tineared]

Update: I feel much less like this is an issue with the speakers themselves and the issue is definitely with my room... If anyone else has some genius ideas on how to accommodate this new speaker position I'd be happy to hear them. Really trying to get creative here. Very complicated room, added some pictures of it to see if anyone has some extra ideas.

I don't have any genius ideas, but I think if you go to a carpet store and buy a remnant that will fill the space from your record cabinet to your dining table and from your speakers to your couch you will significantly tame the high frequencies for a much more relaxed sound. That floor is not your friend.
A remnant on top of your coffee table wouldn't hurt either, but since I don't do that I really can't advise you to do that...