[haloxt]

You could put some silicone pad on the desk, and the nfb-12 on top of it. Or sticky tack or for less mess silicone putty.

 

[holland]

Hi Kingwa,
 
The rolloff is not caused by the caps, but by the filter selected for 44.1KHz/48KHz data.  The filter selected by selecting the 192KHz oversampling setting sets -6dB @ 22.05KHz.  This is fine, but the filter is a slow roll-off, filter, which then sets the corner frequency to around 15KHz.  With 96KHz data, the -6dB point is at 48KHz instead of 22.05KHz.
 
For 44.1/48KHz data, I would suggest the apodising filter, which is a sharper filter, but was set up to copy the Meridian CDP.  This is set by floating the filter pin and setting the OSR to 44.1KHz.  There are whitepapers on the Meridian apodising filter, presented at AES.  There is a slight roll-off on the top end, but it is not a very wide filter, like the slow roll-off chosen by default.  This is the filter I use on my DIY dual WM8741, FWIW.  I think it sounds good, for what it is.
 
For 96KHz/192KHz, the filter chosen is theoretically better, this is the type of filter Ayre uses in their CDP.
 
The problem really is that the standard 44.1KHz stream has the Nyquist frequency at 22.05KHz.  Since audio is 20-20, that doesn't give much room to implement a good roll-off filter.  Unfortunately, the chosen filter starts to roll off at 2KHz, which is audible.
 
Quote:

What cause the roll off at NFB12.
The WM8741 setting is no fault at NFB12.
This  setting is  for accept 192KHz input.
The NFB12 can had better specs on page.The two caps in the red loops cause the roll off.
The NFB11 is same with the NFB12 except the digital parts and TWO caps.
In NFB11 we applied 100PF, but in NFB12 we applied 470PF.
An electronical university man can easy design a excellent specs gear. But the sound maybe not excellent.
We also can build a excellent specs gear, but we want more listen not only meter test.
In NFB11 we plan to design its sound very revealing, so we choice 100PF caps for it.
The NFB12 is finished design with NFB11 at same time, we want to design its sound a bit smooth , and forgive so we spend more days than NFB11 just for adjust parameter and listen, final choice the 470P.

For people who want better specs, you just simply replace these two caps by 100PF (or lower).
Sorry my poor English, I want to study English again, but I like listen musices and design PCB more than study English.

 

[madwolf]

Finally I get to mod the 8741 with Jumpers so that I could RMAA test this 
The following 3 graph are obtain with me changing only the voltage of Pin 22 on both 8741 
 
Input is from USB at 44khz 
 

 

This RMAA test is done NFB12 to Creative X-fi, The noise, dynamic range, THD, IMD .... are at the limit of the X-fi 
So should not see these numbers. 
 
I did not change any capacitors. 
 
This is a picture of my temporary mod. 

I added jumpers to both Pin4 and Pin22, so that I could play around with the different filter and with the sampling frequency. 

 

[Audiophil]

Hi Holland,
very interesting. I am not sure wether we are talking about the same filter, you mean the filter before ADC, right? If I've got Kingwa right, he was talking about the lowpass after DAC. But anyway, you are imo on the right track. As mentioned few times the filter after DAC would influence each signal the same.
But your approach would mean to redesign?
 
Hi madwolf,
also interesting, but I guess a bit, if your setup is correct. All results are worther than the other ones before: All graphs are not even touching 20 kHz down to the -4.5 dB, none of them

 

[madwolf]

Quote:

 
 
Hi madwolf,
also interesting, but I guess a bit, if your setup is correct. All results are worther than the other ones before: All graphs are not even touching 20 kHz down to the -4.5 dB, none of them

The x-fi on loop back would only go to 18Khz at best , So it is a limit with my test setup/test equipment. 

 

[holland]

Quote:

Hi Holland,
very interesting. I am not sure wether we are talking about the same filter, you mean the filter before ADC, right? If I've got Kingwa right, he was talking about the lowpass after DAC. But anyway, you are imo on the right track. As mentioned few times the filter after DAC would influence each signal the same.
But your approach would mean to redesign?
 

Hi Audiophil, I am saying it is not the low pass filter Kingwa has implemented.  There are filters in the WM8741, which have been shown to be the limiter.  The filter chosen, with a 44.1KHz stream comes in before the filter implemented by Kingwa.
 
The external analog filter is not an issue.

 

[LiqTenExp]

 
It so happens that the spec calls this out too as everyone probably already knows who has looked into it.
 
Ideally I am going to look and see how the input switch is wired.  Ideally I am going to get it to ground pin 0 when i switch to usb and float pin 22 when its in any other mode.  I don't have a 192kHz source or source material so this will work for me.

 

[LiqTenExp]

update, the center contact on the switch is ground.  technically one could lift both pin 22s, then connect to them together and run a single 30awg wire over to the bottom contact on the input select switch.  this would give you the correct filter for 44.1 usb and 96k for opt/coax.
 
i may have some ceramic pad isolators around here somewhere to properly isolate and locate pin 22 safely.
 
just general note... this thing is pretty well put together for it's price range.  I am J-Standard Class III Cert. and it is definitely above Class I work, it is Class II+ looking at first glance (no scope and no extra lighting used at the moment).  It has good part placement, little light on the solder in some places, some left over flux in some spots, and 1 solder splat was found stuck to a capacitor.  
 
Now here's my comment.  I want to see this run on a device that has the BW to actually RMAA this thing.  Good efforts so far to show the filter options work, but we don't want people taking those graphs to heart.

 

[supercurio]

Thanks guys for your investigations !

To me everything indicates that it's defective by design.
No matter how good or burn-in can be the analog part, nothings good with a misconfigured DAC.
(The analog part is okay but IMO pretty boring at least with HD650)
 
I appreciate the creator is here to discuss with us, but at the same time I must admit I'm really disappointed that he seems to discover all this.
Looks like no serious test has been done during the development or manufacturing of this product.
 
Defects (rolloff + resampling artifacts + lack of good oversampling) It took me 1 minute to hear, and a few hours to produce the graphs & measurements in order to share with you.
 
I bought this DAC/amp to use it mainly connected to a set-top-box that outputs 48KHz signals, upsampling is not an option.
 
Defective product = refund.
Kingwa, I send you a message to ask how to proceed.
 
 
PS: I hear till 19kHz and it's not everybody's case, impressions on what I consider a defect may vary.

 

[LiqTenExp]

had asked a question, then a comment and need to do some more research.... place holder for now.

 

[Kingwa]

I don't gree it is a defective product.

At first we design the NFB12 we want to make it different flavor to the NFB11 and I want to make it with smooth sound , and we have mark on our web it is a bit smooth and forgive for source .
It is work on hardware mode, so it is setting is follow the performance request and this is the best setting at 192KHz accept ability .
The table of #577 post, it is so the max oversample setting. The NFB12 want to accept 192KHz oversample so the Pin22 must setting at 1 . Otherwise the NFB12 only can up to 48KHz or 96KHz.
Does anyone had better setting in hardware mode?
LiqTenExp post at #578 just customize setting, in face the USB can accept 96KHz and coaxial/optical can accept 192KHz but he limit USB at 44K and coaxial at 96KHz.
About the high frequency roll off not mean it is a defective product, just read the diagram of #556 I was post, the Simaudio Moon Evolution Andromeda Reference CD player which is a colstly more than USD20K gear, also so the high frequency roll off around -3DB ,does anyone think it is a defective products?
 
The high frequency roll off define sound flavor, it bring a smooth flavor.
But the roll off not mean its sound unneutral, the neutral is depend on the circuits components less coloration, like some tube gears even without roll off at 20KHz but sound tube like (coloration).
For supercurio: If you define the NFB12 is a defective product, please offer the effective evidence.
For LiqTenExp post at #580, addition a MCU can check the oversample rate and setting the Pin22.
But I think a MCU join in not a good idea for this price rang gear. In this price rang gear , the power supply is limit and the size is quite small, the MCU maybe bring more disturb to the sound make the sound quality degrade .

 

[LiqTenExp]

Kingwa,
 
I agree with you that this thing is designed right for its price tag and size.  I think it sounds better than my LD DAC_I
 
I don't hear any rolled off highs, and plus there is nothing good above 15k anyways, just trash and harmonics.  I personally dont ever plan on using 192khz so I may do what I was suggesting above anyways.

 

[treaves]

Quote:

Defective product = refund. Kingwa, I send you a message to ask how to proceed.
 
 
PS: I hear till 19kHz and it's not everybody's case, impressions on what I consider a defect may vary.

That it's not designed as you want it to be in no way makes it defective.  Stop being a petulant brat, and sell it if you are not happy.

 

[borrego]

From the RMAA test above, there is just 1db of rolloff at 15KHz.....It is nothing!
 
Not meant to being rude but I think the slight placement difference of putting headphone over the ears and the amount of ear wax in one's ear canal would cause more difference....
 
I think the 100pf and 470pf capacitors would make make more difference. But as Kingwa said, it was intentionally designed to be that way.
 
And for the different digital filter used, I wonder if Wolfsen actually did it for a purpuse as it is more complicated for pin 22 to control more than just the max incoming frequency but also to change the filter.

 

[supercurio]

Quote:

I don't gree it is a defective product.

At first we design the NFB12 we want to make it different flavor to the NFB11 and I want to make it with smooth sound , and we have mark on our web it is a bit smooth and forgive for source .
It is work on hardware mode, so it is setting is follow the performance request and this is the best setting at 192KHz accept ability .
The table of #577 post, it is so the max oversample setting. The NFB12 want to accept 192KHz oversample so the Pin22 must setting at 1 . Otherwise the NFB12 only can up to 48KHz or 96KHz.
Does anyone had better setting in hardware mode?
LiqTenExp post at #578 just customize setting, in face the USB can accept 96KHz and coaxial/optical can accept 192KHz but he limit USB at 44K and coaxial at 96KHz.
About the high frequency roll off not mean it is a defective product, just read the diagram of #556 I was post, the Simaudio Moon Evolution Andromeda Reference CD player which is a colstly more than USD20K gear, also so the high frequency roll off around -3DB ,does anyone think it is a defective products?
 
The high frequency roll off define sound flavor, it bring a smooth flavor.
But the roll off not mean its sound unneutral, the neutral is depend on the circuits components less coloration, like some tube gears even without roll off at 20KHz but sound tube like (coloration).
For supercurio: If you define the NFB12 is a defective product, please offer the effective evidence.
For LiqTenExp post at #580, addition a MCU can check the oversample rate and setting the Pin22.
But I think a MCU join in not a good idea for this price rang gear. In this price rang gear , the power supply is limit and the size is quite small, the MCU maybe bring more disturb to the sound make the sound quality degrade .

Hi
 
I accept the roll-off is design a choice to smooth the sound by killing the highest frequencies.
What I consider as a defect is something else, it's the sampling artifacts due to a misconfiguration of the DACs:
 
Here is how it appears on previous graphs:
 
On udial:
 

 
 
On the RMAA test file:
 

 
On this graphs, remember that everything that's higher than the center (around 20kHz) is artifacts.

For reference, here is how the first udial should look like: (nothing above 20kHz)
 

 
I would suggest you to try by yourself: http://dl.project-voodoo.org/killer-samples/udial.wav
When playing this file on headphones at moderate/high volume clocked at 44K, you'll hear high frequencies artifacts in the second part of the file.
 
Compare with the NFB-12 clocked at 96 or 192k with a high quality upsampling (or another non-buggy DAC), you'll see the difference.
 
This is the part I consider as a defect. Simply: bug in the implementation. It's hard to hear because of the additionnal roll-off, but it's there.
 
But again, if people's are happy with the sound, I'm perfectly happy for them.
My goal is not to trash a product but show technical deficiencies if any in order to see them fixed in future products or revision of the same product.