[jj69]

I made the mistake of trying the new firmware - which is a disaster. 
 
Now I can't go back to the original 2.0.0.  Every time I try it, I get the following error message:
 
"Cannot found firmware image or invalid image.  Installation aborted." 
 
What method are you guys using to go back to 2.0.0? 

 

[borrego]

By design the Sharp/Slow roll off filters will affect ultrasonic frequencies. In short if the source file is well mastered with little ultrasonic noise, then using the Slow filter or No filter will be best. If the source file contain much ultrasonic noise, then using Fast filter can give a "cleaner" sound.

I suspect the 2.00 firmware does not have the filter enabled.

The Fast/Slow rolloff filters will have no or very little effect on mp3 or other compressed files playback, as ultrasonic frequencies were discarded durung the encoding process already.

 

[goodvibes]

  I like the new firmware, and also the ability to choose filters. I missed this feature as I usually prefer the sharp filter with my Sabre implementations. I did so with the Concero and also now with the Ibasso.
 
Pre-ringing is the artifact I would rather eliminate instead of phase incoherence. For those of you new to this, you guys should go educate yourselves about inherent artifacts created by modern sigma-delta conversions. I believe Resonessence had a nice long article explaining it somewhere on their site.
 
And I found it for you guys:
 
http://resonessencelabs.com/digital-filters/

Fast roll exhibits both more pre-ringing and more group delay than slow roll. Neither filter is minimum phase. With the amount of over/upsampling all these DACs use, there's little need for fast roll brickwall filters. I think you'll find that the slow filter shows significantly less pre-ringing than the fast one. http://hifiduino.wordpress.com/sabre32/
 
Note the lack of pre-ring but the large amount of general ringing in the AK dac with minimum phase filter. There are tradeoffs here as well and the slow ESS filter has very little pre-ring. Actually the least I've seen outside of a minimum phase.

 

[Pudu]

It is so interesting how this FW has generated such opposite opinions, in general. If each FW can satisfy each person's desire for what they hear, perfect! Very interesting. 

Well only in part. If each firmware is offered in all the various sound options (which will quickly become a management nightmare, I would guess), fine. Otherwise having to choose between sound and interface seems a bit daft to be honest. Never understood that with the DX100.

Sound is always going to be subjective. In the end you have to either make it user adjustable or pick your house sound and stick with it. Constantly making changes to it makes no sense whatsoever in my opinion.

 

[goodvibes]

By design the Sharp/Slow roll off filters will affect ultrasonic frequencies. In short if the source file is well mastered with little ultrasonic noise, then using the Slow filter or No filter will be best. If the source file contain much ultrasonic noise, then using Fast filter can give a "cleaner" sound.

I suspect the 2.00 firmware does not have the filter enabled.

The Fast/Slow rolloff filters will have no or very little effect on mp3 or other compressed files playback, as ultrasonic frequencies were discarded durung the encoding process already.

There definitely needs to be a filter for proper function and I think fast vs slow is audible on higher bit rate compressed files as well. (tried it on AAC 400 CBR) I'm not going to discuss this with you but please look it up before giving more seriously bad technical info.

 

[goodvibes]

Well only in part. If each firmware is offered in all the various sound options (which will quickly become a management nightmare, I would guess), fine. Otherwise having to choose between sound and interface seems a bit daft to be honest. Never understood that with the DX100.

Sound is always going to be subjective. In the end you have to either make it user adjustable or pick your house sound and stick with it. Constantly making changes to it makes no sense whatsoever in my opinion.

I doubt that they are changing the codecs. It's likely a function of buffers and options etc. Every clip/fuze update had a sound change for me as well and some were huge compared to this.

 

[M-13]

  Fast roll exhibits both more pre-ringing and more group delay than slow roll. Neither filter is minimum phase. With the amount of over/upsampling all these DACs use, there's little need for fast roll brickwall filters. I think you'll find that the slow filter shows significantly less pre-ringing than the fast one. http://hifiduino.wordpress.com/sabre32/
 
Note the lack of pre-ring but the large amount of general ringing in the AK dac with minimum phase filter. There are tradeoffs here as well and the slow ESS filter has very little pre-ring. Actually the least I've seen outside of a minimum phase.

You're absolutely right. I guess I was spoiled by the IIR filter from Resonessence Labs. I assumed the sharp filter was in place to reduce pre-ringing, but No! I guess the two default filters from Sabre kind of suck. No wonder Resonessence designed their own filters. I miss my IIR and Apodizing filters.
 
I guess I will go an reevalute the fast/slow filters to see which I like better. My initial thought was that the fast filter sounded sharper/crispier, while the slow filter was more blurry and enveloping with better staging/imaging cues. Which was kind of pleasant, but I wanted my details razor sharp to give me that sense of extra resolution.

 

[M-13]

By design the Sharp/Slow roll off filters will affect ultrasonic frequencies. In short if the source file is well mastered with little ultrasonic noise, then using the Slow filter or No filter will be best. If the source file contain much ultrasonic noise, then using Fast filter can give a "cleaner" sound.

I suspect the 2.00 firmware does not have the filter enabled.

The Fast/Slow rolloff filters will have no or very little effect on mp3 or other compressed files playback, as ultrasonic frequencies were discarded durung the encoding process already.

There is no such thing as "no filter". All DACs need to have a filter. This is how they work, otherwise there would be no sound.
 
Ibasso was using either the fast or slow filter before. Unless they had a custom filter they made but then got rid of suddenly.

 

[Pudu]

I doubt that they are changing the codecs. It's likely a function of buffers and options etc. Every clip/fuze update had a sound change for me as well and some were huge compared to this.

I don't care about the methods or the precedent. It's the end result that matters.

 

[Pottsy2]

  I made the mistake of trying the new firmware - which is a disaster. 
 
Now I can't go back to the original 2.0.0.  Every time I try it, I get the following error message:
 
"Cannot found firmware image or invalid image.  Installation aborted." 
 
What method are you guys using to go back to 2.0.0? 

This method: http://www.head-fi.org/t/697035/ibasso-dx90-dual-sabre-1st-page-to-reformat-and-fw-download-interest-in-balanced-mod/5040#post_10552968
 
Link really needs to be on Page 1!

 

[jj69]

Finally got 2.0.0 back using the Rockchip batch tool (what a hassle). 
 
Rockbox for DX90 cannot get here fast enough. 

 

[musicheaven]

Talk about complicating your life

Over sampling filter is a must for removing digitalization effects, you just can't go by without doing it. Having said that, the vendor usually offers the standard band pass filter response and the slow-roll off one. It's up to the designer to program it to be turned on or off (just a one bit toggle on the control line).

Most of what everyone has said partially explains the effect of either not using it or switching between the two modes. I wish ESS tech would publish their roll-off curves the way Wolfson does but unfortunately they don't. Assuming they use a similar technique than Wolfson, the band pass attack in both roll-offs are about the same, it's how wide each allow frequencies to be filtered and at what dB level signal they do. The big difference between the two is about 1/10 of the bottom and top band pass frequencies which comes to about 2 hz at the bottom and roughly 2 kHz at the top. As most instruments and female voices have a top range that's way below the top cut-off frequency, the harmonics are what you'll capture and hear, not the top band which is too high for most listeners. What most listeners described could either be pleasing to the ears of clinical to others but what's nice is you got a choice. Wolfson as such uses the standard filter action which is the sharp one as the default. ESS does not mention which one but I would suspect it's the sharp too.

 

[dafos58]

  I made the mistake of trying the new firmware - which is a disaster. 
 
Now I can't go back to the original 2.0.0.  Every time I try it, I get the following error message:
 
"Cannot found firmware image or invalid image.  Installation aborted." 
 
What method are you guys using to go back to 2.0.0? 

I got the same message as you mentioned, so I used the RKBatchtool, you can find it in this forum. This worked perfectly. 
 
Oeps, you already found that solution 

 

[goodvibes]

 
You're absolutely right. I guess I was spoiled by the IIR filter from Resonessence Labs. I assumed the sharp filter was in place to reduce pre-ringing, but No! I guess the two default filters from Sabre kind of suck. No wonder Resonessence designed their own filters. I miss my IIR and Apodizing filters.
 
I guess I will go an reevalute the fast/slow filters to see which I like better. My initial thought was that the fast filter sounded sharper/crispier, while the slow filter was more blurry and enveloping with better staging/imaging cues. Which was kind of pleasant, but I wanted my details razor sharp to give me that sense of extra resolution.

They far from suck and pre-ring effects are debatable (I agree with you

). The slow filter exhibits very good ring and group delay characteristics in all areas. Doesn't mean that you can't find one that you prefer.

 

 

[goodvibes]

Talk about complicating your life

Over sampling filter is a must for removing digitalization effects, you just can't go by without doing it. Having said that, the vendor usually offers the standard band pass filter response and the slow-roll off one. It's up to the designer to program it to be turned on or off (just a one bit toggle on the control line).

Most of what everyone has said partially explains the effect of either not using it or switching between the two modes. I wish ESS tech would publish their roll-off curves the way Wolfson does but unfortunately they don't. Assuming they use a similar technique than Wolfson, the band pass attack in both roll-offs are about the same, it's how wide each allow frequencies to be filtered and at what dB level signal they do. The big difference between the two is about 1/10 of the bottom and top band pass frequencies which comes to about 2 hz at the bottom and roughly 2 kHz at the top. As most instruments and female voices have a top range that's way below the top cut-off frequency, the harmonics are what you'll capture and hear, not the top band which is too high for most listeners. What most listeners described could either be pleasing to the ears of clinical to others but what's nice is you got a choice. Wolfson as such uses the standard filter action which is the sharp one as the default. ESS does not mention which one but I would suspect it's the sharp too.

I think steep is always the default but before the option was given, the DX90 could have been using either, Just clarifying for others. You can see the filter slopes ESS uses in the link I gave on the previous page.