Apple Homepod in-room measurement

[RayDunzl]

Referring to https://www.audiosciencereview.com/...to-reinvent-home-music.1700/page-5#post-65105


There are three files:

Known test conditions:

The microphone was a UMM-6, with a calibration file named narrow_band_response_0_degree.frd

I'll assume the absolute SPL levels shown are reasonable (in the first file).

These traces were taken at 1m and the only difference between them was the level of the sweep tone, and the HomePod did what it does on its own.

I don't know the relationship of the speaker to the walls around it

---

Update:

HomePod placement was in a corner, 2 feet from rear/side wall and tweeter level about 38" off the ground, mic was 1 meter from the HomePod out in the room with a 30º angle in line to listening position. The audio was directly streamed from REW on a Mac with REW's sweep level held constant. I have GIK triangle corner traps, 6" bass traps on first reflection points with a BAD diffuser panel and 1D styrofoam diffusers directly to left/right of LP. Despite the diffusers, I've been told the room is exceedingly dead which gives me warm and fuzzies.

If streaming from iTunes, the volume slider in the software directly controls HomePod's volume level. For instance, if you are playing a track and ask Siri "Hey Siri, what's the volume?" Siri will respond with "the volume level is 20%." If you adjust the slider in iTunes to the right and ask Siri again, she will respond with "the volume level is 40%." In these measurements, I asked Siri to change the volume level to 10%, 20%....100% but REW's level was not adjusted.

Ok, so, the traces show Homepod's Volume Control action, and not necessarily its response to differing input levels.

---

I don't have a Homepod, or any other modern Apple stuff, though I do have an antique Mac in the garage someplace. Anyone need a 200meg hard disk?

---

The Homepod Sweep file contains 10 sweeps, named 10% to 100%:

This is a set of sweeps with increasing volume, stepped about 3dB (except the lowest). These sweeps were used to create the data in the other two files below.

---

The Homepad Normalized file contains 10 sweeps, named 10% to 100%:

These are the same set of measurements, but the "level" is adjusted on the display so they all overlap at about 1kHz. Relative levels are maintained, absolute levels are meaningless. (example: the softest sweep shows over 110dB in the bass, but the actual was 70dB)

This shows the range of difference in the bass and treble at different playback SPL. In this case, the loudest trace is the lowest line (most attenuation of the extremes).

---

The Homepod Loudness Compensation file contains the 10 Normalized sweeps, and does another manipulation of the data, to compare the 10% through 90% files to the baseline, the 100% measurement (shown as the 0dB flatline).

This shows bass and treble boost (or maybe just lack of attenuation) at lower SPL. The lowest SPL trace is at the top, louder as you go down (on the bottom display)

(more to come)

 

[RayDunzl]

Balance at SPL:

At lower levels, it looks pretty flat.

To maintain its composure (4" woofer) it progessively attenuates the bass, beginning at higher and higher frequencies, as the overall SPL is raised.

30Hz bass is locked down at 65db. I see the level reduction (compression) up to about 500 hz. Maybe it goes higher, but the sweeps didn't go higher, so, don't know.

Similar compression occurs above 5k.

For example, 40Hz bass is locked down at 78dB in this display, Higher bass frequencies get locked at higher SPL as the overall SPL rises.

15khz locks at 70dB

The speaker is pretty flat at 60-70dB. It becomes progressively bass-shy (and high treble-shy) at higher levels.

 

[RayDunzl]

Distortion:

Physics intrudes, and distortion levels are high at lower frequencies. The progressive SPL locking at the lows can be seen locking the low frequency distortion levels.

Ambient noise dominates the HF distortion levels until the speaker SPL rises far enough. The lowest HF distortion occurs with trace 8.

Showing trace 8, 9, and 10, rising distortion levels with each (normal).

 

[RayDunzl]

Impulse and Step:

If there's a surprise, it could be here.

Maybe it inverts the signal, so I'll invert the traces, for a more normal display. Also, I'll let REW adjust the timing based on the peak (Estimate IR Delay).

HomePod placement was in a corner, 2 feet from rear/side wall and tweeter level about 38" off the ground, mic was 1 meter from the HomePod out in the room with a 30º angle in line to listening position. ... I have GIK triangle corner traps, 6" bass traps on first reflection points with a BAD diffuser panel and 1D styrofoam diffusers directly to left/right of LP. Despite the diffusers, I've been told the room is exceedingly dead which gives me warm and fuzzies.

100ms Impulse:

The ~85dB SPL trace is shown (#8)

It doesn't take long for the room to dissipate the sound, about 220ms.

The waves in the display before the peak are indicative of distortion products, but I don't know how to interpret it beyond mentioning it.

The levels immediately after the Impulse are rather high, so the ratio of Direct to Reflected is low. I don't know if that is due to the nearby wall reflections, or to action of the DSP "enhancing" the signal. The highest reading after the peak is -4.5dB (seems very high).

Zooming in, showing 20ms, and changing the Impulse display type:

I'm unable to say if the high levels after the peak are the walls (some should be) or the speaker's DSP playing around with the presentation.

And Step Response:

Hmmm...

 

[RayDunzl]

reserved

 

[RayDunzl]

reserved

 

[Blumlein 88]

Maybe a little off topic. The Tact units I have used have this capability. To vary the target curve based upon volume setting using F-M curves. You also could tweak those F-M curves as well as specifying less than full compensation. I found with most music it was a plus. The other thing is you were allowed to specify a 0 reference level of unaltered response. Setting this level was the key to this feature being a benefit and not some coloration. I typically paired it with 80 db in room levels. I also typically used about half the compensation at lower volume levels say 50 db and below.

I wonder if the Homepod doing this automatically will get it right in one size room and be wrong in another size room where its 0 db reference level results in too much compensation for the SPL level in the different sized rooms.

 

[Soniclife]

I'm wondering how it handles room correction vs where you are listening in the room. If it fixes the sound for the other side of the room how does it sound nearfield? Seeing it measured at 1M in different rooms might be required.

 

[Soniclife]

Maybe a little off topic. The Tact units I have used have this capability. To vary the target curve based upon volume setting using F-M curves. You also could tweak those F-M curves as well as specifying less than full compensation. I found with most music it was a plus. The other thing is you were allowed to specify a 0 reference level of unaltered response. Setting this level was the key to this feature being a benefit and not some coloration. I typically paired it with 80 db in room levels. I also typically used about half the compensation at lower volume levels say 50 db and below.

I wonder if the Homepod doing this automatically will get it right in one size room and be wrong in another size room where its 0 db reference level results in too much compensation for the SPL level in the different sized rooms.

RME have something similar in their DACs. This seems like one of the least tricky things for Apple to get right compared to some of the other ideas in this product.

 

[Blumlein 88]

I'm wondering how it handles room correction vs where you are listening in the room. If it fixes the sound for the other side of the room how does it sound nearfield? Seeing it measured at 1M in different rooms might be required.

If you talk to it, the directional miking will know which direction, and maybe distance you are. Does it alter things for this as well?

 

[RayDunzl]

If two people talk to it, does it swap back and forth?

(personally, I think too much is being read into the ad copy)

 

[12B4A]

HomePod placement was in a corner, 2 feet from rear/side wall and tweeter level about 38" off the ground, mic was 1 meter from the HomePod out in the room with a 30º angle in line to listening position. The audio was directly streamed from REW on a Mac with REW's sweep level held constant. I have GIK triangle corner traps, 6" bass traps on first reflection points with a BAD diffuser panel and 1D styrofoam diffusers directly to left/right of LP. Despite the diffusers, I've been told the room is exceedingly dead which gives me warm and fuzzies.

If streaming from iTunes, the volume slider in the software directly controls HomePod's volume level. For instance, if you are playing a track and ask Siri "Hey Siri, what's the volume?" Siri will respond with "the volume level is 20%." If you adjust the slider in iTunes to the right and ask Siri again, she will respond with "the volume level is 40%." In these measurements, I asked Siri to change the volume level to 10%, 20%....100% but REW's level was not adjusted.

 

[Thomas savage]

If you talk to it, the directional miking will know which direction, and maybe distance you are. Does it alter things for this as well?

She’s very annoying ...

I definitely would not share my chips with Siri.

 

[Phelonious Ponk]

She’s very annoying ...

I definitely would not share my chips with Siri.

Currently, she’s the only woman I care to share living space with...

 

[Soniclife]

Great stuff so far Ray.

Do I read that right, 50% distortion at 90db (midrange)

I wonder if the lack of hf gain at higher levels is deliberate to balance the bass not going up.

 

[RayDunzl]

Do I read that right, 50% distortion at 90db (midrange)

It wouldn't appear to be a low distortion device. Where is a 4" cone most comfortable?

Measurements (at least these) look a little sketchy as far as the "Reinventing Home Music" meme is concerned.

I wonder if the lack of hf gain at higher levels is deliberate to balance the bass not going up.

That's as good a guess as any.

 

[Blumlein 88]

Wow on that step response. Trying to think was does that.

???

 

[Soniclife]

It wouldn't appear to be a low distortion device. Where is a 4" cone most comfortable?

Compared to something like a conventional (DSP active crossover) that costs less (but no clever stuff) it seems much worse, e.g.
http://www.eve-audio.com/index.php?page=SC204
If you trust manufacturers measurements.

 

[Phelonious Ponk]

I’ll need to see this repeated to trust it. They sound clearer in the mids than my LSR 305s. Hell of an illusion if these distortion measurements are accurate.

 

[12B4A]

How the Homepod DSP separates the signal and sends it out to the multiple tweeters is kinda a black box situation for how it attempts to reproduce a REW sweep signal. I felt confident that I could at least get a decent loudness compensation measurement. Does anyone think that LF loudness compensation is too aggressive at lower volumes?