iFi's "Gibbs Transient Optimized" filter whitepaper

[derp1n]

iFi have introduced a new "Gibbs Transient Optimized" digital filter in the latest firmware for their flagship "Pro iDSD" DAC and made it available in previously released DACs via a firmware update. The whitepaper is an entertaining read full of pseudoscience and misunderstanding of basic digital signal theory.

 

[amirm]

Indeed.

Clever though in using the horsepower of the Xmos USB interface to implement DSP functionality. This is how they are able to retrofit older DACs to support what is basically low number of taps, FIR filter.

 

[Blumlein 88]

I see nothing, but misleading BS. I especially dislike the totally wrong-headed, intentionally deceptive crap about temporal resolution of 44.1 khz sample rates being 22.7 microseconds. It is not true. They know it is not true (if they don't they should close up shop from embarrassment). And they are using it to bamboozles customers. The bit about the length of transient filtering is also BS. They basically are making excuses for short tap length and allowing aliasing to occur. And the cherry on top of the pile of steaming bull doo is the MQA at the end. A pox on both of their houses.

MQA Malarkey Quotient Assured.

 

[solderdude]

Would Chord finally see the error of their ways ?
Will I finally be rid of that infernal 'echo' and 'reverb' out of all my DAC's ?
MQA was right with their time-smear BS afterall but could not make it work nearly as well as IFI ?
Do I finally get to own a GTO for pennies ?
Am I going to reproduce 'illegal' signals such as needle pulse and squarewaves near perfectly that do not exist in music nor are they recorded ?
Have been wanting to hear 'real music' instead of a reproduced recording.

I really can't wait ... take my money ... please

GTO Gibberish Text Optimized ?

 

[FrantzM]

What is "temporal resolution" ? Serious question.

 

[Wombat]

What is "temporal resolution" ? Serious question.

Google: https://en.wikipedia.org/wiki/Temporal_resolution

 

[FrantzM]

Google: https://en.wikipedia.org/wiki/Temporal_resolution

I was trying to post about the laziness of my question ...Under the weather and all that

Googling it , I found this interesting site, I plan to visit later ..
https://science-of-sound.net/2016/02/time-resolution-in-digital-audio/

iFi makes good products but cannot help themselves to try profiting from audiophiles tendencies. If one wants want to have an idea of how crazy things can take a strong hold constrictions then visit Computer Audiophile and be edified ... IFi is catering to the market thus these strange components they come up with from time to time .. The bulk of their (other) products, however is seriously good.

 

[SIY]

Monte Montgomery from xiph has a delightful YouTube video where he debunks the time resolution crap in real time (the entire video is worth watching). That part starts at about 20:54. He also anticipates the Gibbs stuff.

 

[solderdude]

I suspect they mean pre- and post-ringing which would 'mask' the actual speed of the transients. Hearing is close to logarithmic so a signal has to become very small in signal level when it isn't perceived any more. All the nice squarewave and impulse plots are all linear. So even when a signal appears to not be there any more in such a plot it is still there but say -30dB.
The longer the pre-and post ringing the 'worse' the temporal resolution is said to be.

The ear canal rings at 3kHz and attenuates around 20kHz already. The 0.7ms is the time it takes for the ear canal to stop ringing at around 3kHz. Most likely a certain attenuation 'border' is reached after that time. No idea what amplitude it is at that point 10% ? (-10dB) or more... no idea.

The filter rings at around 20kHz (for 44kHz files) which is inaudible and only rings when the recording has those steep transients in it or close to it.
It doesn't start to magically 'ring' or pre-echo or whatever at 20kHz when lower frequencies are present.
It never is a 0dB 'spike' as per the test signal. The 20kHz level is low in amplitude compared to the mids anyway. The 0.7ms thus only applies to the 3kHz signal in the ear canal which is NOT 'activated' by a 20kHz frequency.

Another thing I encourage anyone to do. You know what..... I am not spilling the beans and want those that want to know let them search for it themselves.

Here is what you should do:
Search for a piece of music that has transients, the fast sharp ones....
Then open the file with something like audacity.
Zoom in on those transients till you arrive at sample frequency levels.
Closely inspect that transient and see if you can spot an instant 'rise' in amplitude in the file .... within 1 or 2 samples to see if they are there during a perceived transient.
Then report back if you find anything in the recording that is that fast related to that perceived transient.
Highly educational....

have fun.

 

[Liberal Democracy]

I see nothing, but misleading BS. .

I think you can get a cream for this now..

 

[derp1n]

I think you can get a cream for this now..

Yeah, but it's pretty gritty.

 

[danadam]

Googling it , I found this interesting site, I plan to visit later ..
https://science-of-sound.net/2016/02/time-resolution-in-digital-audio/

Also this: https://troll-audio.com/articles/time-resolution-of-digital-audio/

 

[Blumlein 88]

I suspect they mean pre- and post-ringing which would 'mask' the actual speed of the transients. Hearing is close to logarithmic so a signal has to become very small in signal level when it isn't perceived any more. All the nice squarewave and impulse plots are all linear. So even when a signal appears to not be there any more in such a plot it is still there but say -30dB.
The longer the pre-and post ringing the 'worse' the temporal resolution is said to be.

The ear canal rings at 3kHz and attenuates around 20kHz already. The 0.7ms is the time it takes for the ear canal to stop ringing at around 3kHz. Most likely a certain attenuation 'border' is reached after that time. No idea what amplitude it is at that point 10% ? (-10dB) or more... no idea.

The filter rings at around 20kHz (for 44kHz files) which is inaudible and only rings when the recording has those steep transients in it or close to it.
It doesn't start to magically 'ring' or pre-echo or whatever at 20kHz when lower frequencies are present.
It never is a 0dB 'spike' as per the test signal. The 20kHz level is low in amplitude compared to the mids anyway. The 0.7ms thus only applies to the 3kHz signal in the ear canal which is NOT 'activated' by a 20kHz frequency.

Another thing I encourage anyone to do. You know what..... I am not spilling the beans and want those that want to know let them search for it themselves.

Here is what you should do:
Search for a piece of music that has transients, the fast sharp ones....
Then open the file with something like audacity.
Zoom in on those transients till you arrive at sample frequency levels.
Closely inspect that transient and see if you can spot an instant 'rise' in amplitude in the file .... within 1 or 2 samples to see if they are there during a perceived transient.
Then report back if you find anything in the recording that is that fast related to that perceived transient.
Highly educational....

have fun.

And isn't it interesting that time based stereo perception only works up to 1500 hz above which stereo is perceived mainly by sound level differences. And further that in artificial signals .7 mseconds is about the point of maximum left or right perception for frequencies below 1500 hz.

 

[solderdude]

ITD and IID you mean for sound localization (660ms) ?
The threshold for ITD seems to be determined at 10 microseconds (at 1kHz).
That ITD= 10us (corresponds to 100kHz).
I guess that is where the 'we need at least 100kHz bandwidth comes from' even when we are just able to hear to 16kHz (for medium aged folks, older farts like me are lower) yet that 10us limit does not seem to be age dependent.

Of course temporal resolution to 10us can still be reached with bandwidths < 100kHz for this we don't need 192kHz sample frequencies.
we can still localize sounds very well with just a RBCD bandwidth it seems.

> 1500Hz timing differences could not be 'determined' which would be important for localization.
However, the 10us limit (22us at 1.5kHz) still suggests that the hearing can detect those timing differences.
Why would the brain only use this available 'info' for localization ?

Have no idea how that relates to perception and enjoyment of music and how the ears/neurons/brain handles that.
All the research is usually done with test tones and (sometimes suspect) transducers.

I think it is more complex than the numbers suggest when music is involved, instead of localization with test tones only.

Interesting indeed nonetheless.

 

[andreasmaaan]

Of course temporal resolution to 10us can still be reached with bandwidths < 100kHz for this we don't need 192kHz sample frequencies.
we can still localize sounds very well with just a RBCD bandwidth it seems.

Exactly. As applies to all bogus claims along these lines citing the 10us figure for ITD, there is no relationship between sampling rate and ITD. The former determines the bandwidth of the signal, whereas the latter relates to arrival time differences between the L and R channels, which (if present) are neither caused by "low" bandwidth nor mitigated by higher bandwidth.

 

[Veri]

Just upgraded my micro BL to see if iFi's hype is worth glossing over,

volume is reduced on this new filter. Possible a bug, but I can't reliably compare lol.

 

[Werner]

They know it is not true (if they don't they should close up shop from embarrassment). And they are using it to bamboozles customers.

Thorsten L certainly knows this. But like most others, iFi/Abbingdon have found that selling BS is more profitable.

 

[FrantzM]

Thorsten L certainly knows this. But like most others, iFi/Abbingdon have found that selling BS is more profitable.

Hi

I am not trying to defend iFi.. This product is likely a pile of manure ..but .... but ...

The audiophile market is not rational. People come with wants and whims and are even willing to reject a preferred manufacturer/provider/vendor if he/she/it/they do not align with their perception. As an example audiophiles rejected John Dunleavy and his wares once he started rejecting their views on Audio cable , e.g Read this, Many want the B.S. IOW. So manufacturers oblige and sell them the B.S. and they buy .. vide the commercial success of B. Schiit ...

 

[mansr]

iFi foe fum

 

[Westsounds]

In the interest of science, what I want to know is. Can anyone subjectively hear this filter and what if any are the benefits?