[Jawed]

Convolution filter is another matter because it can correct even more room acoustics problems for me so I always hear an even bigger improvement in frequency response. But the more I listen, the more I notice how the transient response is less accurate when you engage any convolution filter. I suspect the problem is that some convolution filter introduce a minimal phase filtering which alters transient accuracy. But I suspect all convolutions are done via FTT convolution not direct convolution so I wonder if the FTT conversions (even at 64-bit) introduces slight low level errors that are audible. My theory is further supported by a software program that is supposed to mimic M-Scaler where originally it upscales at 64-bit using FFT but recently released the ability to do it at 128 or 192 bits and users report improvements, likely because they get more accurate convolution.

"Direct convolution" still has an accuracy problem. With computers, the order in which you do multiplies and adds affects the accuracy of the overall result.

With direct convolution this can be seen when computing from the taps in the same order as when the data arrives. Half of the tap coefficients are in ascending order of size, e.g. the first few coefficients might be (I'm ignoring positive/negative coefficients, by the way):

0.0000000000011234
0.0000000000012345
0.0000000000013456

then the coefficients rise to their highest values:

0.4567
0.5678

then after that, the coefficients fall:

0.4567
0.3456

down to the final coefficients::

0.0000000000013456
0.0000000000012345
0.0000000000011234

So by the time you get to use these coefficients their results are so small that they do not correctly affect the final result of the convolution. This all depends on the precision of the accumulator.

There are two solutions:

1. massively increase the precision used by every multiply and add (e.g. 128 bits)
2. re-order the convolution to work with the smallest (absolute value) coefficients first, then proceed to the largest

The second solution is tricky because it requires constantly shuffling the incoming data so that its order aligns with the "smallest first" coefficients. This means the newest sample has to be placed next to the oldest sample. Both samples will be multiplied by the smallest coefficient, in this case 0.0000000000011234. Then repeat with "next newest" and "next oldest" samples, to go with the next coefficient, 0.0000000000012345. etc.

 

[ecwl]

"Direct convolution" still has an accuracy problem. With computers, the order in which you do multiplies and adds affects the accuracy of the overall result.

Great points.

But my understanding of FFT convolution is that you’re converting your signal with FFT into the frequency domain first and then doing a direct convolution in the frequency domain and then doing an inverse FFT back to the resulting signal. So the potential errors in direct convolution in the time domain would still apply for FFT convolution.

Hence, I’m curious more about equivalency, for example does FFT convolution in 64-bit have the same accuracy as direct convolution at 60-bit or 50-bit or 40-bit? I don’t know but someone skilled in math can probably figure it out.

 

[FunkyBassMan]

When running the MScaler on battery, what is the right voltage to pick?

 

[Reactcore]

When running the MScaler on battery, what is the right voltage to pick?

It takes any voltage between 9v to 15v
Go for a unregulated battery.
12v is most common to find.

 

[FunkyBassMan]

It takes any voltage between 9v to 15v
Go for a unregulated battery.
12v is most common to find.

Thanks!

 

[pichler]

When running the MScaler on battery, what is the right voltage to pick?

I use a 12v camper battery, works fantastic!

 

[pkny]

I am using unregulated direct-battery output (9v-12.3v) from a Bixpower BP-90 power bank. It is better than Krisdonia (9v regulated) and Maxoak K2 power banks (12v regulated).

 

[FunkyBassMan]

I am using unregulated direct-battery output (9v-12.3v) from a Bixpower BP-90 power bank. It is better than Krisdonia (9v regulated) and Maxoak K2 power banks (12v regulated).

Thank you for that!

 

[FunkyBassMan]

I am using unregulated direct-battery output (9v-12.3v) from a Bixpower BP-90 power bank. It is better than Krisdonia (9v regulated) and Maxoak K2 power banks (12v regulated).

I'm using a Krisdonia 25000 now and it sounds incredible. But you say regulated is not a good idea?

 

[pkny]

Switching regulator always generates noise. Unregulated direct battery output eliminates that noise.

 

[organicmac]

@RobWatts There have been a few posts now on head-fi referring to a future ultima range DAC and many relating to the expected choral x-scaler.
I would like to know whether there will be an additional scaler in the ultima range to accompany the ultima DAC please.

 

[mikecheck95]

Why do snare hits and other transients sound soft on my Hugo 2 compared with my Macbook pro’s onboard audio?

 

[ecwl]

Why do snare hits and other transients sound soft on my Hugo 2 compared with my Macbook pro’s onboard audio?

To me, two possibilities (because they don’t to me)
1) you just like noise floor modulation which makes everything sound brighter and more exciting. Which is fine because people like what they like. But you’ll find even smooth slow songs will have vocals that sound extra exciting with noise floor modulation that you won’t with Hugo 2.
2) the more likely possibility is that you’re not feeding Hugo 2 bit perfect signal and the over sampling from the MacBook is causing a loss of transient accuracy
For maximum transient accuracy, the Hugo 2 should be using the white or green filter instead of the red or orange one.

 

[mikecheck95]

To me, two possibilities (because they don’t to me)
1) you just like noise floor modulation which makes everything sound brighter and more exciting. Which is fine because people like what they like. But you’ll find even smooth slow songs will have vocals that sound extra exciting with noise floor modulation that you won’t with Hugo 2.
2) the more likely possibility is that you’re not feeding Hugo 2 bit perfect signal and the over sampling from the MacBook is causing a loss of transient accuracy
For maximum transient accuracy, the Hugo 2 should be using the white or green filter instead of the red or orange one.

Thanks!

I’m just running it in Roon’s exclusive mode? I thought that should be enough to bypass the macbook’s oversampling.

I also don’t know if I like it per se… but it is one of the bigger differences I notice between the two.

 

[ecwl]

Thanks!

I’m just running it in Roon’s exclusive mode? I thought that should be enough to bypass the macbook’s oversampling.

I also don’t know if I like it per se… but it is one of the bigger differences I notice between the two.

Yes. Roon exclusive mode should bypass MacBook oversampling. I also presume you’re not using Roon’s DSP functions which can also lead to loss of transient accuracy. This would even include the crossfeed function in Roon.

My favourite vocals to test is the musical Hamilton track Its Quiet Uptown. On DAC chip DACs, the song is dynamic and exciting. On Chord DACs, it’s sad and soft as it should be.

Ultimately, it’s what you prefer. I personally think snare drums sound more accurate through Hugo 2 than my MacBook headphone Jack.

The other issue might be that the LCD-5 has a treble roll off. So if you add noise floor modulation back, things would sound more balanced and brighter. Whereas Hugo 2 would just let you hear the treble roll off without noise floor modulation so it might sound less dynamic because of less high frequency levels.