The Mike Moffat (#2 at Schiit) Blog

You can tune so that a certain subset of chords are more in tune. For example you could tune it so that C major, F major and G major and C minor, F minor and G minor are perfectly in tune. But then you try and play other chords involving the same notes and they would sound horrible. All this this is independent of choosing the base pitch (A=440 vs A=432 etc).

Yes, exactly. It's a global problem, to do with the relationship of chords with each other in time, and not a local one. Doing this in real time sounds like a hard computational problem quite aside from the DSP that does the retuning. One really needs to analyse the music and figure out the harmonic relationships to decide how to tune things, and some of those relationships are only apparent after the fact, so you'd need some machine-learning style understanding of chord progressions. A short-cut would be to guess the key at the start, retune things to a scale which makes all the common chords in that key sound good, and hope there aren't any wild modulations later on.

I hope the speculations here are on the money. The more I think about this the more mind-bending it is.

 

Yes and no. Equal temperament distributes the errorequally among all intervals. Non-equal temperaments strive to make the intervals of some keys more in tune than others; the keys that are not as in tune sound "bad".

The concept behind JS Bach's Well-Tempered Clavier was a temperament system that supposedly sounded good in all keys. Unfortunately, nobody knows exactly what his system was because it was never documented.

 

I agree about the mind bending part and the computations needed to determine what and how to make 'adjustments' could very well be in the form of an iterative loop, all of which adds to a fair amount of latency.
Which could help to explain why MP won't work for home theater.

Just more wild a$$ed rampant speculation on my part.

JJ

 

Well then their ML model previously could be pretty relevant although an interesting application.

As I understood the paper it was a "new time variant decaying history model".

Analyse the note pitch at the current time stamp, and use the time variant decaying model (so past notes have an effect on current note but not too far back) to apply varying degrees of adjustment to the note.

 

Thanks @Garns and @SSL for all of the explanations! Previously I had a tough time wrapping my head around things like temperament, but now I'm starting to see the light.

For some reason this, by @baldr, seems relevant here:

I imagine that the "totally autoseeking" versions would require oodles of computational power, rely on machine-learning (which would explain the need to employ the charms and skills of a certain Dr. Ivana), and would effectively be a self-adjustable real-time DSP...

And revolutionary. Isn't this what the schiitsters have promised us?

 

Yet more tea-leaves on Manhattan, by @baldr. The unveiling seems to be impending... (If I'm reading things right.)

 

My post got lost over the weekend, but I was revisiting @baldr's breadcrumbs and the first chunk of bolded text made me think that the rumoured DDC and MP could come in the same box.

But reading it again just now, the underlined/bolded text implies that they'll be separate devices, as I think @landroni said he inferred from previous hints.

 

Sounds to me like it's not impending, as Mike still hasn't figured out whether to release it with any configurable parameters or to go full bore into auto adjustments, which would require more computing power and memory, plus I'd imagine a lot of testing to ensure that the MP actually adjusts the way it's supposed to. Reading through Jason's chapters, especially the one on the Rag, that software testing can take a while before it's in a state ready to be released.

One non-MP thing I gleaned out of that is that Mike said up until now he spent most of his time working to release multibit DACs at a variety of price points--and that most of those have been released. Most. Not all. What other multibit tricks does Mike have up his sleeve I wonder?

 

Fulla MB will be astonishing. Or a Yggdrasil topping Odin?

 

Releasing vs Unveiling, but I guess it's mostly semantics. Definitely agree with your points there!

There were strong hints on a Bifrost upgrade, possibly -- and I speculate -- making it balanced, though I'm wondering if that's physically possible and if it would require sending in existing Bifrosts for casing adjustments...

 

I would hope it's more than just a Bifrost update, but of course I will be happy for those with Bifrosts!

As for releasing versus unveiling, that's a good point, although we already know it's a form of DSP that affects frequency and time domain, so the veil has already been lifted an inch.

 

From what I understand from @baldr, Dr. Ivana will be presenting the Gadget at Canjam LA... That's less than a month away.

 

That could also be read as she will simply be at Canjam LA along with the rest of the Schiit team.

Anyone got any tea leaves that can shed some definitive light on this subject?

 

I feel like MP could end up being somewhat divisive in the audio world.

 

To be fair it's not often an audio manufacturer offers a DSP box, and for the "original codes" diehards that's tantamount to Ragnarok.

 

There might be one other thing: An R2R ADC. Mike designed in the past the Mobile Fidelity Gain system (the original, as Gain 2 was D/S), and last year has been dropping some hints on a possible modern replacement for that.

 

Yes, and then donate that box to shaizada so he can record us all of his different setups and we can wish we were there.

 

Yup, she will be there with the rest of us. I already told you to go easy on her.

Is the gadget revolutionary? Well, such a thing has never been built for quality audio use. There have been built feature stripped products for entirely different applications, but they are not suitable for an audio system which is not a toy. What is for sure is that the first will be the most simple, the one I listen to as I write. The target price is the low (very) hundreds of $$$. We shall see how lucky I get.

 

Apropos of this:

@baldr can I suggest having it fixed or auto-adjusting by default, but then fully adjustable via a USB connection for control only. There would be software involved (maybe VST plugin?) but purely to send parameter adjustments to the hardware DSP. Something like a Nord Modular.

(Some of us would consider the ability to adjust it very badly to be a feature, not a bug...)

 

I can see it all now.
Exchanging 'recipies' for the settings for different styles of music, perhaps even specific albums.
Especially if there are more than just 3 variables involved.

JJ