[EliasGwinn]

Hello Bizkid...

Quote:

Originally Posted by bizkid /img/forum/go_quote.gif I've owned the DAC1 USB for some months now and basically i'm happy with it. However there are 2 things where i see the need for improvement. First the easy one, the volume pot. Atleast on my DAC1 the L/R Balance varies more than i'd like throughout the range. That means the headphone out is unusable if you don't attentuate the volume in the digital domain to reach the (very small!) sweetspot of the poti. That's just not good enough for 1300 Euro, sorry.

Any potentiometer that has a 'full-off' position will have a short range at the beginning where the channels don't track well. This is because the wipers are completely off the elements at 'off', and the first part of the range is the steep part of the logarithmic curve.

The potentiometers we use track very tightly after the first 10-15% of their range. If your's isn't tracking well beyond the first 10-15%, your unit may have a defective pot.

Quote:

Originally Posted by bizkid /img/forum/go_quote.gif The second one goes a little deeper, i'm not happy with the resampling algo that is used, especially since EVERYTHING get's resampled to the DACs optimal rate (which makes sence) and there's no way turning that off thus limiting the DACs capability. The problem is that the resampling algo in the DAC1 doesn't appear to be a good one, atleast sonically. You can see measured comparisons of various resampling soft/hardware here: SRC Comparisons Yesterday i ABX'd a propper resampled 96khz file and the original 44khz file (thus the DAC1 has more resampling to do). Out of 10tries i scored 100%.

I'm not sure I understand your testing procedure, but it sounds like you were ABX'ing the SRC performance of the software. The DAC1 will resample everything to 110 kHz, and the performance of that SRC won't vary based on the incoming sample rate.

Can you describe in more detail how you determined your opinion of the DAC1's SRC?

Thanks,
Elias

 

[HeadLover]

BTW, why did it upsample from the USB only to 110MH and not the 192KHZ it can ?
And if it so, isn't it better try and use it with COAX or what?

 

[EliasGwinn]

Quote:

Originally Posted by Lil' Knight /img/forum/go_quote.gif After reading about 50 pages of this thread, I decide to pull the trigger on the DAC1 USB. Hopefully, it'll be a big improvement from my DacMagic. Just some small questions: + Since it already has a native USB driver, I don't have to use ASIO4ALL with foobar2k, right? + How do the headphones out on the DAC1 drive low impedance phones, like the ESW10JPN and PK1? I'll sell my portable amp if I buy the DAC1. + Can I use both the headphones out and XLR outputs at the same time? + And lastly, anyone tried it with some tube amps? How do they sound together?

Hello Lil' Knight,

- No, you don't have to use ASIO4ALL. Btw, are you using XP or Vista? ...shouldn't make a difference, just curious.

- The HPA2 (the DAC1's headphone amp) maintains its performance even with 30 ohm headphone loads, mainly because of the 0-ohm output impedance.

- Yes, you can use both headphones and XLR at the same time. However, the left headphone jack has a feature that automatically mutes the XLR and RCA outputs whenever headphones are plugged in. This feature makes it easy to go from loudspeakers to headphones without having to turn off your amplifier, etc. This feature is defeatable if you don't want it. Also, the right-hand headphone jack doesn't have this feature.

- I don't have much experience with the DAC1 driving a tube amp, but a lot of DAC1 users do that.

Thanks,
Elias

 

[EliasGwinn]

Quote:

Originally Posted by dmashta /img/forum/go_quote.gif i have always been curious about this but never quite sure what it means. on the dac1 pre, i've adjusted the volume pot through its range but never noticed a 'sweet spot'. sure it sounds best within a certain range but that's just my loudness preference, no? so the inner workings of the pot notwithstanding, what exactly is this sweet spot we talking about?

Dmashta,

The sound of the pot doesn't vary, but the L/R track balance is optimized after the first 10-15% of the full rotation. In other words, it is ideal to operate the volume pot beyond the 9-10 o'clock position.

As I mentioned earlier, this is simply due to the nature of 2-channel, full-off, logarithmic pots. The first 10% above 'full-off' is a huge jump in value...too much to expect an inter-channel accuracy less then 0.1 dB, which is required for good L/R balance.

Thanks,
Elias

 

[dmashta]

Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif Dmashta, The sound of the pot doesn't vary, but the L/R track balance is optimized after the first 10-15% of the full rotation. In other words, it is ideal to operate the volume pot beyond the 9-10 o'clock position. As I mentioned earlier, this is simply due to the nature of 2-channel, full-off, logarithmic pots. The first 10% above 'full-off' is a huge jump in value...too much to expect an inter-channel accuracy less then 0.1 dB, which is required for good L/R balance. Thanks, Elias

Okay, thanks. I've had no issue with channel imbalance (not that i listen to music that low) so i went looking for a sweet spot of another kind.

 

[bizkid]

Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif Dmashta, The sound of the pot doesn't vary, but the L/R track balance is optimized after the first 10-15% of the full rotation. In other words, it is ideal to operate the volume pot beyond the 9-10 o'clock position. As I mentioned earlier, this is simply due to the nature of 2-channel, full-off, logarithmic pots. The first 10% above 'full-off' is a huge jump in value...too much to expect an inter-channel accuracy less then 0.1 dB, which is required for good L/R balance. Thanks, Elias

Hi Elias, then i just have bad luck with my pot, the L/R balance varies until i reach the sweetspot (around 1clock) and then again from 2oclock on.

I'll make some measurements and upload them here to make my point a little clearer concering the SRC. Anyway i'm curious if the AD1896 can be used without SRC but still make use of it's re-clocking capabilities? The chip has a "bypass" pin but it's not really clear to me if it just bypasses the SRC or the whole chip alltogether.

 

[EliasGwinn]

Quote:

Originally Posted by HeadLover /img/forum/go_quote.gif BTW, why did it upsample from the USB only to 110MH and not the 192KHZ it can ? And if it so, isn't it better try and use it with COAX or what?

I can explain it to you in full detail if you like, but I've explained it earlier in this thread (here, for example).

The short answer is this: all digital inputs (coax, USB, optical, XLR) are re-sampled to 110 kHz because this frequency optimizes the performance of the digital filters, even better then 192 kHz.

Thanks,
Elias

 

[HeadLover]

Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif I can explain it to you in full detail if you like, but I've explained it earlier in this thread (here, for example). The short answer is this: all digital inputs (coax, USB, optical, XLR) are re-sampled to 110 kHz because this frequency optimizes the performance of the digital filters, even better then 192 kHz. Thanks, Elias

Ok thanks!

 

[EliasGwinn]

Quote:

Originally Posted by bizkid /img/forum/go_quote.gif Hi Elias, then i just have bad luck with my pot, the L/R balance varies until i reach the sweetspot (around 1clock) and then again from 2oclock on.

Who did you buy the unit from?

Quote:

Originally Posted by bizkid /img/forum/go_quote.gif I'll make some measurements and upload them here to make my point a little clearer concering the SRC. Anyway i'm curious if the AD1896 can be used without SRC but still make use of it's re-clocking capabilities? The chip has a "bypass" pin but it's not really clear to me if it just bypasses the SRC or the whole chip alltogether.

Well, re-clocking is actually SRC. Even if we are talking about re-clocking to the 'same' sample rate (i.e., 96k -> 96k), the algorithm used will be the same. No two clocks are exactly the same (96.0001 kHz -> 95.9999 kHz). So, re-clocking will always convert the sample rate, and the choice of final sample rate won't affect the performance, even if its the 'same' sample rate.

The thing about your ABX test is that both 'A' and 'B' are going through the exact same SRC/algorithm, and so any differences between the two are inherent in the audio data.

Thanks,
Elias

 

[poo]

Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif I can explain it to you in full detail if you like, but I've explained it earlier in this thread (here, for example).

To be honest, I would find it really helpful if you could explain it in full detail; but also in lay terms if possible. The post that you linked to was helpful in regards to a general understanding, but for someone like myself with no technical understanding of what it actually means, your explanation just resulted in me having faith that you know what your talking about. I'm not suggesting you don't - just that I'm really non the wiser

So when you say: Quote:

Originally Posted by EliasGwinn /img/forum/go_quote.gif all digital inputs... are re-sampled to 110 kHz because this frequency optimizes the performance of the digital filters, even better then 192 kHz.

it sounds like (for example) there could have been filters to better handle the 192kHz signal, but we didn't use them for "other reasons".

 

[EliasGwinn]

Quote:

Originally Posted by poo /img/forum/go_quote.gif To be honest, I would find it really helpful if you could explain it in full detail; but also in lay terms if possible. The post that you linked to was helpful in regards to a general understanding, but for someone like myself with no technical understanding of what it actually means, your explanation just resulted in me having faith that you know what your talking about. I'm not suggesting you don't - just that I'm really non the wiser So when you say: "all digital inputs... are re-sampled to 110 kHz because this frequency optimizes the performance of the digital filters, even better then 192 kHz." ...it sounds like (for example) there could have been filters to better handle the 192kHz signal, but we didn't use them for "other reasons".

I'll try to explain it in layman's terms...feel free to ask me to clarify anything.

The filtering is done by performing math operations on the digital data. However, only so many operations can be performed per second (thats why your computer slows down whenever there are too many things happening). With such an increase in data (192 kHz is 2x the amount of data as 96k and 4x as much as 48 kHz), the filter must reduce the amount of math operations to keep up with the data. With fewer math operations, the precision of the filter is compromised.

The highest amount of data the filter can process before reducing the math is about 117 kHz. We choose 110 kHz to be absolutely certain that the filter is always in the full-performance mode.

Thanks,
Elias

 

[HeadLover]

amm, so why not just put a better and faster processor?
How does some DAC's out there goes even up to 768MHZ ?
It sure be nice if all data that goes into the DAC can be upsampled to 192KHZ

 

[EliasGwinn]

Quote:

Originally Posted by HeadLover /img/forum/go_quote.gif amm, so why not just put a better and faster processor? How does some DAC's out there goes even up to 768MHZ ? It sure be nice if all data that goes into the DAC can be upsampled to 192KHZ

Keep in mind, just because a DAC goes up to a certain sample rate doesn't mean that the filters perform well. In fact, it probably means the opposite.

When a circuit is designed, one must consider all the tradeoffs...the strengths and weaknesses of the available technology. Then, the design engineer must decide which tradeoffs to take based on what factors are most important.

The tradeoff in this case is system bandwidth versus sonic performance.

System bandwidth is, more or less, the highest analog frequency (aka Nyquist Frequency) of a digital converter, which is always one-half of the sample rate. The system bandwidth of the DAC1's digital filter is 55 kHz, whereas the system bandwidth of a 192 kHz conversion is 96 kHz.

Sonic performance, in this case, is defined by the pass-band ripple and stop-band attenuation. Pass-band ripple is another way of saying frequency-response distortion. Stop-band attenuation is necessary to prevent aliasing, which is a type of digital distortion that is very unflattering. Stop-band is the audio at frequencies above the Nyquist Frequency. If these frequencies are not properly filtered out, they will be 'folded' back into the audio band as a tone with a frequency equal to f_audio - f_Nyq. If you're unfamiliar with what aliasing sounds like, listen to a 96 kbps MP3.

We decided that we would rather have a very high performing circuit with an analog bandwidth that is limited to 55 kHz then a less accurate circuit whose bandwidth goes to 96 kHz.

Thanks,
Elias

 

[HeadLover]

I see
Well I agree with you here, I mean I (and I guess others) will prefer a much better Sound Quality, than just numbers

 

[bizkid]

Grrr now i found the culprit of my resampling troubles.

The problem with my ABX test was... windows. When i set it up some time ago I set the audio options of the DAC1 connected via USB to 24bit 96khz in windows, and using direct sound in foobar.

That means when i compared the 44khz file to the 96khz one i actually ABXd windows resampling algo which resampled the 44khz stream to 96khz. OUCH! Sorry for the missunderstanding to Elias and the other people i were in disagreement.

One note for windows users: If you listen to 44khz most of the time, it's the best to use that value in the windows audio setup. You DON'T want to resample twice, once in software (be it windows itself or foobar), one time in the DAC 1. I measured the results of 2x software resampling and it's a big NO NO. If anyone is curious i can post the graphs.

Atleast we brought this thread back to live a bit