"early" anti-alias rolloff on the iPhone 4S ADC
"early" anti-alias rolloff on the iPhone 4S ADC
- Subject: "early" anti-alias rolloff on the iPhone 4S ADC
- From: Keith Lazuka <email@hidden>
- Date: Wed, 21 Dec 2011 14:12:06 -0800
Hello,
Our application relies on a relatively flat frequency response around
20kHz. This is no problem on the iPhone 4 and the 3GS, but on the
iPhone 4S we are seeing severe attenuation in this band. I did some
measurements, and it appears that the transition band for the iPhone
4S ADC anti-alias filter is more gradual than the iPhone 4 but it starts
significantly lower (about 2700Hz lower).
This image shows the magnitude frequency response curves for both the
iPhone 4 and the 4S given an input sine sweep from 0kHz to 24kHz over
the course of 10 seconds:
http://dl.dropbox.com/u/441716/iPhone_4s_early_rolloff/iphone_4_vs_4s_sine_sweep.png
Test Setup:
I generated a sine sweep from 0-24kHz using Baudline running on an
iMac. Baudline was attached to the iPhone-under-test via an Edirol
UA-25 USB audio capture box sampling at 48khz. I connected the output
of the Edirol to the iPhone's linein using an iRig harness and then
connected the iRig's/iPhone's lineout to the Edirol's linein, completing the
loopback. The iPhone-under-test was sampling at 44.1kHz and was
running a custom pass-through app using the RemoteIO audio unit with
the PlayAndRecord audio session active and Measurement mode enabled.
(While in this test I did a full duplex test, I have also done a
separate test of
just the iPhone 4S's ADC and saw similar results.)
Now I understand that there are no guarantees regarding the iPhone's
frequency response above 20kHz, but I was wondering if there was a way
to configure CoreAudio such that the anti-alias filter transition band
starts higher? I'm already setting the kAudioSessionMode_Measurement
flag, which--according to the headers--should minimize the effect of
system-supplied DSP. But are there any other flags that need to be set
to get closer to the raw input from the ADC? Or am I out-of-luck?
I also tried running the iPhone 4S at 48khz sample rate, but despite
the increased distance from Nyquist, the anti-alias filter transition
band starts at the same frequency as when I ran the iPhone 4S at 44.1khz.
The only difference I saw with 48kHz is that the filter's transition
band is about 2x wider, resulting in gentler rolloff, but still not
gentle enough
for our application.
Any help would be much appreciated.
Thank you,
Keith Lazuka
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden