Do iPhones sound better, and if so, why? What about Huawei mobile phones, how do they sound? Different models of mobile phones will sound different, and although I’ve been working on mobile games for a couple of years now, I realized that I had never experimented with comparing the sounds of mobile phones. I decided to look into this as I was curious about what we can we do to improve audio in mobile games. I set aside some time to test:
- Loudness on popular smart phones
- Frequency response on popular smart phones
Of course “popular” mobiles are different in different places. The models I collected for testing represent the most popular devices in China (during the time of testing). While I was not expecting to have concrete conclusions, I was hoping to identify common phenomenons, similarities or differences, and perhaps come up with some tricks towards better sounding mobile games. The mobile phones I used for testing:
- iPhone 7P
- Huawei Mate 7
- Vivo Y51A
- Oppo R9 Plus
When choosing the games for testing, I decided to select games that would represent different types of game sounds, with a range of loudness and game play modes. I selected seven different mobile games that were very popular in the Chinese mobile game market and that represented different game types.
- Fight the land lord
- Happy cancellation
- Arena of Valor
- Clash Royale
- Onmyoji
- Homescape
- PUBG (mobile version by Tencent - and I actually tested 2 different versions of PUBG mobile by Tencent just for fun. )
The Testing Method
The Facilities:
- Microphones: Sanken CO-100(master) and Earthworks TC-20
- AD: Apogee DUET FW
- Mac Book Pro 15"
- Senheisser HD25 AMP headphone
- Signal A: Stereo white noise, PCM WAV, 60s long, generated in ProTools. -11.9 LKFS, -8.1dBTP. This is the mastering signal.
- Signal B: Stereo white noise, PCM WAV, 60s long, generated in ProTools. -6 LKFS, -2.1dBTP (The target level is desined to 12LKFS)
The Steps:
- Play the Signal A on the smart phones, with the original and default APP of each phone, at the max level of each.
- Modify the microphone gain for each phone. Make sure the microphone input LKFS level is -11.9. So the -11.9LKFS is the target reference to each phone.
- Play the games one by one and record directly into Pro Tools 12HD.
Specifications:
- Reference Level: -11.9LKFS (-12LKFS)
- Sliding Term: 30s
- Recording time: over 5 minutes for each game
About the Speaker Box in mobile devices
- A smart DSP handles the amp of DA
- Temperature and Amplify detector feed the DA info back to DSP.
- And then the DSP will excude the Auto Gain Control(AGC) to limit the output power and then protecting the hardware. The AGC works like a compressor with a limiter.
- Also, every manufacturer can adjust and define the EQ and other in the DSP for getting different sound taste and response.
This research is suppose to:
- Help us understand Loudness on popular mobiles, today.
- Loudness in popular mobile games.
The common Loudness reference charts
Film |
-24LKFS |
|
CD Mastering |
-15LKFS |
(Also in Soundcloud) |
iTunes |
-16LKFS |
|
YouTube |
-13LKFS |
|
Spotify |
-14LKFS |
|
K-lab recommend to mobile |
-16LKFS |
|
All the max peak target is -1.5 ~ 4dBFS |
My personal Loudness target
Music mastering |
-16LKFS |
30s term |
In-game SFX |
-16dB RMS |
-45dB threshold. Peak < = -3dBFS |
Game master output |
-16LKFS |
PCOL/Mobile |
The Loudness of speakers output, with White Noise |
Reference: 30cm. -12LKFS white noise on iPhone 7P with -12LKFS target recording level. |
Some findings:
- Content louder than -12LKFS doesn’t sound louder. The inside auto-gain-control works in the devices just like a compressor.
- The roll-off level between 30cm and 40cm is about 3dB. So loudness loss is actually fast.
- iPhone 7Plus doesn’t sounds quietest, with white noise. And iPhones don't mean better dynamics. iPhone games however are the quietest!
- Vivo Y51A is the loudest one, with white noise. But the games on Oppo R9 are the loudest. That's odd.
- The loudest game amongst the ones I tested with was the AOV on Oppo R9, -14.8LKFS.
- Louder content over -12LKFS doesn't create louder sound. -12LKFS might be a physically critical parameter for most of phones. Actually -16LKFS is almost the max level to all the phones.
- The same source, the loudness roll-off is quite obvious between 30 and 45cm
So …
Any source louder than -16LKFS is meaningless, and only sharper sounds can be perceived as louder at times. However sharper sounds on mobile phones are aggressive.
More charts for reference ( click on images to enlarge )
The frequency response on popular smart phones
Some conclusions
- On all the devices, the most sensitive frequencies are: 220,500, 1k, 2.5k, 3.5k, 10k, 11k, 14k
- All the devices, the speakers roll off from 1kHz. All the strong and rich games by hearing, the roll-off between 220-1000Hz are slower.
- All the Android games start raising from 220Hz. But the iPhone 7P starts from 400Hz slowly. That’s probably the reason of why iPhone 7P sounds lighter and softer, but more lower frequencies can be felt.
- All the Android game stop high frequencies or roll off from 10kHz. iPhone 7P is quite same. But, it goes upper again from 16kHz to 22kHz.
- Reduction around 11kHz, and the mean time slower roll-off above 11kHz, shall be good for getting smooth brightness. Please check Clash Royale and Homescape. PUBG can be another reference to prove it.
- 4-5kHz shall be very careful band to be controlled and designed on all the devices.
Frequency response on smart phones is really hard to explain in one article. So, I've included a 30 page PDF for you to dive into in your spare time. In the 30 pages PDF, there are lot of details you can observe from the spectrums, and you can probably come up with more audio tricks for mobile games. I hope you'll find this information helpful!
I'd like to thank Next Studio for their support. Many thanks to Mr.ZhangXin and Mr.Zhanglei from the Next audio team.
Comments
Juan Baez
January 30, 2019 at 12:46 pm
Great post and resource! Especially helpful as I've done a few VR/360 videos where there seems to be no current spec, at least knowing that they're still primarily viewed on mobile, and having this data, one can at least mix towards the platform. I was mixing to -17LKFS (+/-2)[-2dBFS peak] based on a stereo youtube spec established by one of my larger clients. But I'm happy to have this data for any future projects. Thanks for doing all of this!
Jian Yu
February 18, 2019 at 02:37 am
This has always been an interesting topic that I often discuss with my friends, and there are such meticulous results. THANKS for sharing
Nemo Steven
January 25, 2024 at 05:52 am
Um actually according to my calculations I've been listening to phone speakers and I guarantee Motorola phone speaker go down 30 hz but 30 hz is were I start hear it closely then Samsung speakers go down 60 hz no lower iPhone speakers go down 40 hz I been using fequency generators for 4 years