-
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!
Let's develop an ASR inter-sample test procedure for DACs!
- Thread starter splitting_ears
- Start date
staticV3
Grand Contributor
- Joined
- Aug 29, 2019
- Messages
- 10,143
- Likes
- 16,391
The Mojo2 can handle +1dBFS, but not +3dBFS: https://goldensound.audio/wp-content/uploads/2022/05/USB-Isolated.pdf
The M-Scaler reduces all incoming signals by about 2.8dB, so improves intersample over performance of all DACs plugged into it, at the cost of less volume:
https://goldensound.audio/2022/03/17/chord-hugo-m-scaler-measurements-and-technical-evaluation/
That's all I could find.
Some some improvements people might feel they can hear with the M-scaler could just be the inter-sample overs being cleaned up
staticV3
Grand Contributor
- Joined
- Aug 29, 2019
- Messages
- 10,143
- Likes
- 16,391
I have yet to see proof that intersample overs are an actual, audible issue.
Hmm. Digital clipping can certainly be heard. But I suppose clipping of such short duration perhaps not?
Experiment using Audacity, 16 bit integer samples, then upsampled to 384kHz float (to imitate reconstruction):
4.211dB "over"
4.211dB "over"
Last edited:
spigot
Active Member
- Joined
- Jun 3, 2018
- Messages
- 183
- Likes
- 227
John Siau at Benchmark can hear a brightness resulting from distortion, with some songs having hundreds of examples of this phenomena. I suspect that those that came on here claiming that their DAC sounded bright, only to be told that noise/distortion was below the threshold of hearing and they're imagining it, were probably right. It depends on the recording. Kinda looks like all these 'transparent' DACs might sound a bit rough sometimes.
Roland68
Major Contributor
Is there a list of these songs, or examples?
JSmith
Master Contributor
KSTR
Major Contributor
You can always actively create you own proof, as this can be analytically tested with little effort, see my post earlier in this thread.
My aging ears can hardly hear a difference between IS-over reproduced faithfully and properly clipped IS-overs but that doesn't mean anything what is audible to others.
What I still do hear is IS-overs causing wrap-arounds, though.
Just two pages earlier there are audio samples for download, and two participants submitted ABX results.
Read the post below and the subsequent replies.
www.audiosciencereview.com
Read the post below and the subsequent replies.
Let's develop an ASR inter-sample test procedure for DACs!
Source: https://www.sfu.ca/sonic-studio-webdav/cmns/Handbook Tutorial/Audiology.html Thanks for reminding this book! Assuming that music has similar effect like noise, after 15 minutes of listening at 85dB SPL average, we will temporarily lose ability to hear noise at 8dB SPL. Does nature abhor...
www.audiosciencereview.com
It couldnt be that John Siau is imagining it?
What are "wrap-arounds"?
KSTR
Major Contributor
When the output reaches a certain level where it would normally clip, it goes to the opposite polarity, full scale, rather than simple saturating.
The result is a rather loud "click" because it goes from full positive output voltage to full negative (or vice versa).
DAC chips normally don't do this, but it can happen in digital pre-processing like sample rate converters or other stuff. Some older YAMAHA CD-players had a dithering chip (for use with digital volume control) that extended the 16-bit data to 20-bit which caused these nasty wraparounds when reaching full scale values.
This shows "wrap-arounds" of computer integer arithmetic when we have overflows (or underflows when dealing with negative numbers). The upper panel is the Python code, the lower shows the output.
x is specified as a signed 16 bit integer, and assigned with a value of 32766. We then print the values of x, x + 1, x + 2, and x + 3.
The output should be 32766, 32767, 32768, 32769. However, the maximum value of a signed 16 bit integer is 32767. Adding 2 to 32766 causes an overflow and the value "wrapped-around" to -32768. Add 3 to 32766 wrapped-around to -32767 after overflow.
If DSP processing is done with integer math and not implemented carefully to deal with over- and underflows, we can get signal "wrap-arounds".
Here's a modern (albeit cheap) consumer DAC that will hit the opposite side when presented with intersample peaks. I don't think we should assume DACs with this defect are rare.
KSTR
Major Contributor
But you came to the conclusion that it's an Android software issue, with DAC chip proper innocent, didn't you?
I've tested quite a few of those phone dongles including the Apple and Google V2 and not a single one showed misbehavior with IS-overs. So far, those YAMAHA CDPs (CDX-496 and CDX-596, IIRC) were the only case where I found wraparound.
danadam
Major Contributor
- Joined
- Jan 20, 2017
- Messages
- 1,121
- Likes
- 1,735
So according to him intersample-overs can only happen during recording (or downsampling too, I guess)? That's some new definition I've never seen before.
Last edited:
Read it again, that's not what he is saying.
He is saying intersample overs hardly ever occur. His reasoning is, that for it to happen, the signal before downsampling must have clipped. This clipping gets picked up at recording/mixing, and any engineer worth his salt, would have taken care of it, before downsampling.
At least, that is my understanding.
Similar threads
- Poll
