• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

AirPlay1 and AirPlay2 vs direct HDMI input measurements

gags11

Senior Member
Forum Donor
Joined
Feb 8, 2019
Messages
356
Likes
528
A bit of background. I recently changed my sublime (to me) 2-channel active setup. Bought new speakers and amp. Now, I had my heavily moded Klipsch with ScanSpeak 12mu midrange, instead of the horn, and TAD / Pioneer beryllium ribbon tweeters, left to upgrade my theater mains. Long story short, I put a passive crossover into these and hooked them up into my 7.2.2 Atmos setup in my Theater.

I decided to do some measurements without EQ in pure direct mode with no Sub.

I’m using MacBook Pro HDMI into Yamaha CX-A5100 processor to pure direct mode. Mic is about 6 meters away, so these are relatively loud measurements.

after doing some tests, I decided to test Airplay. Yamaha has a built in Airplay, it is Airplay 1. I also have an Apple TV 4K hooked up to my processor. So I decided to test both.

I was shocked at what I saw, what a shame. Measurements are done with 44.1khz, 48khz and 96khz sweeps. Very little difference with HDMI input. But with AirPlay, oh my.... Airplay 2 is a bit better, and 44.1khz signal performs better too, but still ridiculous frequency response. Also, distortion skyrockets for some reason.

now this cannot be a Yamaha issue as I have used Apple TV to stream the signal, which connects via hdmi. Is this what one would expect.?

Figure 1. frequency response with 44.1, 48, 96 kHz input signals. HDMI direct to processor vs Airplay 2 with Apple TV to HDMI input. Note worsening frequency response with 48khz and 96khz signals. HDMI is not really effected

50CD55CF-E805-43FB-B229-F2C5B4E9F689.png


Figure 2. Same graph with Airplay 1, streaming directly into Yamaha.

BC4999DF-4A29-4CA5-AD3A-19C0F10B263E.png


Figure 3. To clarify the picture, here is the HDMI input with 44.1, 48, 96 kHz signals on their own

DDB03547-9422-4AD9-96C2-D48DABCCA7C9.png


I knew AirPlay is so so. I could not imagine it cuts out frequencies. Is this reality true or am I doing something wrong here? I really cannot explain this. It seems when playing the sweep, I can here the higher frequencies, but they do not register. ...or I am be I’m a bit deaf possibly.

BTW: These ScanPeak 12mu midrange woofers and Pioneer tweeters have some nicely low distortion. Klipsch woofers are no slouch either, producing relatively low distortion at extreme SPLs.

0AF1DB67-87A1-4B99-85F2-DA5B631C8D35.png
 
Last edited:
  • Like
Reactions: trl

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,478
Likes
3,316
Location
Detroit, MI
As you mentioned if you can hear high frequencies it is highly unlikely those results are real.

What mic are you using?

It looks related to a sample rate mismatch. What do your MIDI settings look like?

Michael
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,425
Likes
7,941
Location
Brussels, Belgium
I’m gonna assume you’re using an UMIK-1

UMIK-1 uses 48 KHz sample rate. So there is no way you can test Airplay (44.1 KHz) with it.

You will have f’ed up results, that’s my experience.

Same thing with Roon or any lossless streaming. You have to stream 48 KHz or your results are invalid.
 

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,478
Likes
3,316
Location
Detroit, MI
I actually just played around with this, I think it is related to the Airplay delay. I get similar results if I do a loopback without using any sort of loopback or acoustic timing reference. If I use a timing reference it is fine for all sample rates.

Michael
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,425
Likes
7,941
Location
Brussels, Belgium
When you use Airplay everything is getting downsampled to 44.1

No point in using different sample rates.
 

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,478
Likes
3,316
Location
Detroit, MI
I actually just played around with this, I think it is related to the Airplay delay. I get similar results if I do a loopback without using any sort of loopback or acoustic timing reference. If I use a timing reference it is fine for all sample rates.

Michael

See attached for the results. This is using Airplay from a MacBook pro to an Apple TV to a cheap DAC to a MOTU M4 audio interface running at 48 kHz (to simulate a UMIK running at 48 kHz). As you can see when you do not have a timing reference the results look similar to what you show. However, when I use a timing reference the response is what you expect.

Michael
Airplay Test.png
 

Music1969

Major Contributor
Joined
Feb 19, 2018
Messages
4,641
Likes
2,810
Also, distortion skyrockets for some reason.

Does it stop the frequency response sweep once distortion skyrockets? I think that's an option in REW.

So need to find the cause of the distortion spike with Airplay? Higher gain?
 

DeLub

Active Member
Joined
Jun 16, 2020
Messages
135
Likes
178
Location
The Netherlands
UMIK-1 uses 48 KHz sample rate. So there is no way you can test Airplay (44.1 KHz) with it.

I'm no expert, but I don't think that's correct. A digital signal captures all frequencies below the nyquist frequency perfectly. This means that certainly below 20 kHz it is impossible to distinguish between a recording with a 44.1kHz and a 48kHz sampling rate. Assuming that before the AD conversion a filter has cut off all frequencies above Nyquist. In the analogue domain, your speakers, the sound wave are identical. Therefore you can measure perfectly fine with the UMIK-1, even if it was originally a 44.1kHz sampling rate.

... right?
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,425
Likes
7,941
Location
Brussels, Belgium
I'm no expert, but I don't think that's correct. A digital signal captures all frequencies below the nyquist frequency perfectly. This means that certainly below 20 kHz it is impossible to distinguish between a recording with a 44.1kHz and a 48kHz sampling rate. Assuming that before the AD conversion a filter has cut off all frequencies above Nyquist. In the analogue domain, your speakers, the sound wave are identical. Therefore you can measure perfectly fine with the UMIK-1, even if it was originally a 44.1kHz sampling rate.

... right?

I had really really weird FR curves when i used 44.1 KHz files in roon (with acoustic timing) with an UMIK-1.

When i used 48 KHz files in roon the FR curves became comprehensible. Since MiniDSP insists multiple times that you should only use 48 KHz stuff with it, I’m pretty sure it’s important somehow.

Test tones =\= music btw.
 
OP
G

gags11

Senior Member
Forum Donor
Joined
Feb 8, 2019
Messages
356
Likes
528
See attached for the results. This is using Airplay from a MacBook pro to an Apple TV to a cheap DAC to a MOTU M4 audio interface running at 48 kHz (to simulate a UMIK running at 48 kHz). As you can see when you do not have a timing reference the results look similar to what you show. However, when I use a timing reference the response is what you expect.

Michael
View attachment 113960

thank you, now it makes sense, it’s a timing delay .
 

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,478
Likes
3,316
Location
Detroit, MI
thank you, now it makes sense, it’s a timing delay .

You are welcome.

It is definitely a bit of a weird issue. I noticed that the signal seemed delayed when running loopbacks and it is almost like the measurement ends early because the sweep starts late. It seems the acoustic reference fixes this because the sweep does not start until the acoustic reference is measured by REW. As a result a loopback reference does not solve the issue because both the sweep and the reference are delayed.

I also found that if you increase the length of the sweep (even without an acoustic reference) the frequency response seems OK, presumably because the sweep just takes a lot longer so when it ends early it only cuts off the very end.

Michael
Screen Shot 2021-02-21 at 2.39.14 PM.png
 
Top Bottom