[Erratic Gravy]

I believe that in Schiit‘s True Multibit DACs the interpolation is done by Mike’s unique MegaComboBurrito Filter which creates the interpolated samples using its dual time- and frequency-domain optimisation, rather than a simple geometric interpolation algorithm. This filter is a key asset for Schiit, based on many man-years of work, and they are obviously not willing to provide details.

Regards, Nigel

Its the 'secret sauce' on the multibit smorgasbord of Digital and for me anything else sounds nice but wrong

 

[Paladin79]

Due to circumstances that aren't really relevant to you all, I got about 10 days behind in this thread. And just got caught up.

My advice is:

DO
NOT
EVER
DO
THAT.

(You're welcome.)

Loved the reminder of the Transmorgifier. Back in the day, I recall writing a few memos describing things (or processes) are being transmorgified. I guess everyone I worked with understood the term.

Good to see you back Doug.

This mornings listen will pertain to brothers.

The Avett Brothers ........... Head Full of Doubt/Road Full of Promise
The Brother Brothers ............ Tugboats
The Neville Brothers ............ A Change is Gonna Come

 

[OldRoadToad]

https://emeapp.org/

"Welcome back my friends, to the show that never ends"...

Indeed. Thank you for that link!

ORT

 

[yonson]

Going way off the sine wave combining two of my favorite hobbies, auto racing and audio. This is the Turner GTD Pro BMW racing at the Rolex 24hr race this weekend. Apologies to Schiit for posting a big-ass meter.

I do a decent amount of sim racing and I have audio stuff on my virtual suit and helmet hehe.

 

[Erratic Gravy]

Still smaller than the BMW grille.

Sine wave amplitude increased - Le Mans 1989 taken by me with access to the old pits and paddock - blisstory

 

[Erratic Gravy]

Luna

Letter from the UK's premier cat Larry - I think it relates to his intention to become the latest Prime minister !!

 

[theeclone]

Something I’ve long wondered is how the upsampling is done. Is it a linear interpolation between any two points or some sort of spline that goes through multiple points? I would think the latter could help reduce the filtering requirements but might have a difficult time with transients.

Ray

I've wondered the same thing. But then I also wonder, what kind of shape would you draw? If it's effectively 2x, do you use exactly halfway between the amplitude of the surrounding samples? And if so are you creating some kind of stretched triangle wave that contains ugly harmonics? If you try to use sine or cosine type shapes does that become computationally complex? And then with 8x it probably becomes even tougher to choose a shape.

 

[bevanc]

Me ? ...

Me too

 

[ImagesbyMurray]

I've wondered the same thing. But then I also wonder, what kind of shape would you draw? If it's effectively 2x, do you use exactly halfway between the amplitude of the surrounding samples? And if so are you creating some kind of stretched triangle wave that contains ugly harmonics? If you try to use sine or cosine type shapes does that become computationally complex? And then with 8x it probably becomes even tougher to choose a shape.

 

[Ableza]

Without getting into technicalities that my memory might screw up, with conventional "up-sampling" schemes the PCM signal has its apparent sample rate increased by inserting zero value samples between the original values. This can cause undesirable noise, and conversely it can enable more sophisticated filter circuits because of more data to work with, which is why some up sampled signals may sound "better" than the original. If a signal is actually modified by some sort of interpolation between original samples and insertion of non-zero values occurs, this is a synthesizer function and not typical up sampling. Resampling (both up or down) is often done in multiple steps to get from one rate to another depending on the designer's approach. The reason to do this is a theory that multiple resamples results in less digital noise due to zero insertions happening in smaller steps.

 

[Pondoro]

I've wondered the same thing. But then I also wonder, what kind of shape would you draw? If it's effectively 2x, do you use exactly halfway between the amplitude of the surrounding samples? And if so are you creating some kind of stretched triangle wave that contains ugly harmonics? If you try to use sine or cosine type shapes does that become computationally complex? And then with 8x it probably becomes even tougher to choose a shape.

Non-uniform Rational B-Splines are my favorite because of the acronym - "NURBS." If an acronym sounds good it will sound good. (Just like - if an airplane looks good it will fly.)

 

[Birch]

Something I’ve long wondered is how the upsampling is done. Is it a linear interpolation between any two points or some sort of spline that goes through multiple points? I would think the latter could help reduce the filtering requirements but might have a difficult time with transients.

Ray

I believe that in Schiit‘s True Multibit DACs the interpolation is done by Mike’s unique MegaComboBurrito Filter which creates the interpolated samples using its dual time- and frequency-domain optimisation, rather than a simple geometric interpolation algorithm. This filter is a key asset for Schiit, based on many man-years of work, and they are obviously not willing to provide details.

Regards, Nigel

Neophyte question (with rough paraphrasing) - The Bifrost 2/64 uses four Texas Instruments DAC8812 D/A converters. This allows Schiit to dedicate one converter to each individual phase, a "hardware balanced" configuration to do the conversion, while the digital filter is implemented on an Analog Devices SHARC DSP.

Can someone name (and expand on) what these four phases are?
.

 

[ImagesbyMurray]

Neophyte question (with rough paraphrasing) - The Bifrost 2/64 uses four Texas Instruments DAC8812 D/A converters. This allows Schiit to dedicate one converter to each individual phase, a "hardware balanced" configuration to do the conversion, while the digital filter is implemented on an Analog Devices SHARC DSP.

Can someone name (and expand on) what these four phases are?
.

Sure, 2-channels, each channel has a positive going phase and a negative going phase (hardware balanced)

 

[theeclone]

Neophyte question (with rough paraphrasing) - The Bifrost 2/64 uses four Texas Instruments DAC8812 D/A converters. This allows Schiit to dedicate one converter to each individual phase, a "hardware balanced" configuration to do the conversion, while the digital filter is implemented on an Analog Devices SHARC DSP.

Can someone name (and expand on) what these four phases are?
.

My guess: Left positive, left negative, right positive, right negative. Technically I'm giving each individual polarity but sometimes phase and polarity get used interchangeably.

 

[Bonehead7171]

Neophyte question (with rough paraphrasing) - The Bifrost 2/64 uses four Texas Instruments DAC8812 D/A converters. This allows Schiit to dedicate one converter to each individual phase, a "hardware balanced" configuration to do the conversion, while the digital filter is implemented on an Analog Devices SHARC DSP.

Can someone name (and expand on) what these four phases are?
.

I KNOW THIS ONE! Right Positive, Right Negative, Left Positive and Left Negative.... I think?

EDIT - Dang! Left in the dust by @theeclone and @ImagesbyMurray