Best Schiit Bifrost 1 Multibit modification practice

https://www.mouser.com/ProductDetail/Analog-Devices/ADA4620-2ARZ?qs=B6kkDfuK7/DMPAbFAY8pBA==

 

Sorry there was a mistake yesterday, the convert "dsp" is what I made earlier, it was MultiResampler 2x -> mda dither -> MultiResampler 2x

the MultiResamplers configured as Upsampler 2x with Linear Interpolation
the mda Dither configured with Second-order noise-shaped -> Word len 24 bit -> Dither amp 4.0 lsb -> DC trim -2.0 lsb -> Zoom 0dB

 

72 software Sample Rate Converters put to the test | Audio Science Review (ASR) Forum

Some dithers you can select

I tried saracon dither and converted to 192khz and it is really really good. and also, you need to turn on High-definition audio output on Foobar2000 for iPad

recall: The combination of ad826 + ada4620 is really really amazing, as precise as a scalpel, all the details and very small changes of all the sources can be clearly reflected in the sound, and any ultrasonic part can be perfectly filtered out by ADA4260, which is a perfect DC servo

The reason why I select AD826 is because it is a pure Class A folded-cascode architecture, and it actually gets a little hot when it's working. Don't worry about the 10-bit described in the datasheet, because it's for ultrasound standards. Over time, the op amp will gradually stabilize to the appropriate accuracy, which is very similar to Delta-Sigma. If you look at page 6 of the ADA4620 datasheet, 12-Bit takes 485ns and 16-Bit takes 3.3us.

Page 42 of the ada4620 has an extraordinary feature that makes it a perfect filter: Mux-Compatible Glitch Representing, I am not sure if its previous version ada4625 has the same function, if ada4620 is sold out

I would say that Schiit has done a very, very good job in making NOS DACs. If they can work on the DSP, especially the dither, it will be a very good product. As they mentioned, aliasing is the elephant in the room. My personal opinion is that "listening comfortably" is far more important than "hearing details"

 

Actually I think powr1 is the most natural sound, best 3D but little bit dull, means dither only and without noise shaping, and powr2 is good for 99% music
http://audio.rightmark.org/lukin/dither/dither.htm
I tried different sound clip on the website.
powr2 and powr3 have varying degrees exaggerated descriptions of the sound, UV22 is little bit light, in between powr1 and powr2. powr1's sound is closest to the original, but powr2 imporved dynamic range, The positioning of instruments is not as accurate as powr1, but more dynamics will sound more hearty. powr3 is over clean at high frequency range, and lost some texture.
http://audio.rightmark.org/lukin/dither/

I actually found that Logic Pro on Macbook have Powr1/2/3 and UV22HR that is so cool. but I can not select uv22hr for 16bit 192k

 

All process recommendation:

IV opamp: AD826
DC-Servo opamp: ADA4620
Source: iPad with Foobar2000 with High-definition audio output turned on
USB: Unison
Audio file process: Saracon or Logic Pro upsampling to 16 bit 192Khz, with Powr2 noise shaping

 

After carefully comparing the processing methods of foobar, Powr1, Powr2, and Powr3, I am sure that I prefer the processing method of Foobar2000. It is more natural, does not compress dynamics, has more air, and has a more grand aura. Which is the #22 method. On this basis, if we add Apple's magical audio processing method, that is, turn OFF the High-definition audio output & measurement mode, this is the best sound I've ever heard.

 

After a long period of torn, about making a trade-off between sound realism, cleanliness, and dynamic range
The final setting is
Saracon:
Upsampling audio files with POWR3 dither to 16 bit 192K
Foobar2000 on iPad:
Turn ON High-definition audio output and Turn ON Measurement mode

 

Regarding the digital Processing part, I don't think the filter is the most important. There are two factors that play a big role in audio, one is quantization noise, and the other is dynamic range. In order to remove noise and increase the dynamic range within a limited bit, It is impractical to increase the dynamic range in all frequency ranges. It is only possible to increase the dynamic range in the range of 2K~5K where the ear is most sensitive. It is the frequency range of scratching a blackboard with your fingernails.

However, human ears become less sensitive above 16k. Some possible methods involve exchanging sample frequencies for bits, or even sacrificing signal quality above 16k in exchange for a dynamic range between 2K and 5K. First, add a dither to drown out the quantization noise, then use aggressive shaping technology (actually a negative feedback) to offset this part of the mixed noise, and finally make the limited signal content audible.

During this process, a lot of the original dynamic range is sacrificed, and you can clearly feel the spatial sense of the sound and the stage being pulled forward. Because some of the audio information representing the sense of space is indeed mixed with the quantization noise, that is, the better the sense of space, the rougher the sound (within a limited bit)

ADI and TI's existing technology can only produce 20-bit DAC chips at most. Others can only rely on Delta-Sigma technology, which is essentially a DSD technology. Its advantage is that it is really possible to increase the SNR ratio to 24 bits in the 2K~5K range, but it will sacrifice a lot of audio information in other frequency bands. You can clearly feel that the sound in other frequency ranges is fuzzy. I don't know what Schiit's next generation Delta-Sigma product will be like, but the first thing that needs to be solved should be Delta-sigma modulation technology. This should be the unique skill of each professional audio manufacturer, and it is the Mount Everest of audio technology.

In fact, I think that if the Dither and Noise shaping can be done well, 20 bits should be good enough. The key point is that the standard of audio technology is 16, 24, and 32 bits. You can't use a computer to convert the audio signal to 24 bits and then feed it to the DAC, and then cut off the last 4 bits. The sound will be strange and hollow, just like your 3.5mm headphone jack is halfway inserted and not grounded.

 

If you are gonna wax poetic about digital audio theory, you need to do it another thread. You've already muddied up this thread about computer audiophile what not.

 

Okay, tried Saracon POWR3 converted to 24 bit 192K feed to the DAC, acturally not too bad, 24 - 16 = 8 bit discarded.
Not hollow as I thought, the sound will actually become rough and less bass, not hollow. but the dynamic improvement is huge. It is worth to do with this.
If you buy a 20 bit NOS DAC from schiit and then convert with POWR3 and then actually should pretty good.

 

In fact, after using Saracon to convert to 24bit by POWR3, it sounds quite enjoyable. Although there is less bass and the focalization of the sound is a little worse because 8bit discarded, but the sound dynamics and sense of space are full, and the sound is clean and smooth.

 

After tolerated a period of time of the sound loses focus, that is, 8 bits are discarded.
Actually, NOS is pretty acceptable and actually very good in "Hi-Fidelity", that is using foobar2000 with zero order hold upsample 4x to 176.4K
I prefer to use Zero order hold rather than linear interpolation because Zero order hold can reduce bit flipping inside the DAC chip.
Foobar2000 on iPad:
Turn ON High-definition audio output and Turn ON Measurement mode
Foobar2000 on computer:
Convert music file to WAV with: MultiResampler -> Target 4x -> Zero Order Hold

In addition:
I added a breadboard behind the DAC, a simple passive RC low-pass filter, R is 1K, C is 2700P, the sound will not be too bright, NOS now sounds really good. As I guess Schiit will not add LPF in the circuit. The capacitor can be polypropylene or any film capacitor. I used Nichicon QYX2A272JTP

 

Well, QYX1H272JTP is alternative selection. I personally like Nichicon's QYX series. I don't know why such excellent capacitors were discontinued.
Actually I am using 2x QYX1H222JTP parallel which is 4400p, it seems 2700p is not quite enough in my audio system

 

Apparently you are unaware of the multicomboburrito filter and what really makes Schiit multibit DACs so special.

 

I think the multicomboburrito filter actually makes the music seem more distant and has more false dynamics, which makes the sound more tense, but I think the NOS sounds better.
And you get ear fatigue pretty quickly when you turn on the filter, don't you? And you need to endure the sibilance from S's and T's. And I believe more than one person has mentioned this problem
If you've listened to LampizatOr's decoder, this is actually the more correct sound.
The sound of NOS is rough but correct, you should improve your hearing
I also recommend you to copy my steps and compare the performance of AD826 + ADA4620 to your OPA1656
If you think Schiit is always right, you won't be in this sub forum.
I'm just sharing my modification experience. You can dislike it as you like. I don't care, as long as you think your theory can win over your ears.

If you compare it with the filtering of iPhone, you can find that Schiit's filtering technology is quite bad. Of course, this is not the key point, because 20K to 24K cannot be heard by human ears, but it is certain that there must be aliasing, You may not be able to hear 25kHz, but it may intermodulate with 16kHz to produce 9kHz noise, which you can hear.
Review: Battle of Schiit Audio DACs | Page 4 | Audio Science Review (ASR) Forum

iPhone X White Line


Also, If you can test Apple's square wave, and it does not have any Pre-echo, which is the most un-natural part caused by Gibbs phenomenon. The reason is simple. In nature world, you cannot produce an echo unless you hit an object, so pre-echo is unnatural. Pre-echo can make your ear fatigue, which is another side effect of digital filters.

Archimago's Musings: MEASUREMENTS: Apple's Lightning to 3.5mm Headphone Adaptor [MMX62] (CES 2018 Creative Super X-Fi audio, and A/C cable "proof"?)

 

So I often ask myself, should I buy a multibit decoder for the clarity or detailed information of part of the signal or sound, at the cost of auditory fatigue, or an Apogee once and for all?