[shamu144]

24/96 upsampling through Pure Music sounded much better to my ears than native 16/44, but I have to admit something was still a miss (slight hardening of sound or not completely natural flow yet). I have been playing a little further with upsampling, and I have decided to install Fidelia on my Mac. Fidelia takes advantage of the iZotope SRC and dithering engines, so kind of world class as I understand it when it comes to digital manipulation.
 
I am getting absolutely amazing results through upsampling at 24/96 with 24 bits dither enabled (perfered standard setting in Fidelia SRC is the minimal phase filter). The Ref 7.1 is now a completely different beast to my ears and in my system. I can't believe how much difference those SRC operations and digital filters settings can make. I would be very curious to know what others experience is wrt software upsampling into the Ref 7.1.
 
Now I can say for sure the Ref 7.1 is definitely a keeper for me.
 
 
Quote:

My prefered settings so far:
- Pure Music with target upsampling at 96khz.
- "NOS Classic" upsampling (there are 2 minimum phase upsampling method to choose from)
- PLL off
- Stopband attenuation -90dB
Sounds sublime to my ears now even with CD rips. It was certainly not the case prior to these changes.

 

 

[IPodPJ]

Quote:

The RE7.1 not quite hot.
If it have good heat dissipate, like have not anythings cover on its top board, far to the heat source, it can power on 24 X5 hours. Then want to had a restart.
 

Then why do you say in your online manual (that you formerly had available for download from your Chinese site) to only keep it on 8 hours per day?  Now you are providing conflicting information with what you previously recommended.
 
I agree that the Ref 7.1 does not get hot like the Ref 1 did, but I never use it for more than a few hours at a time anyway.
 
Shamu,
You want to make sure it has warm up time if you turn it off.  I don't recommend leaving any gear on all the time.  It shortens the life of your component and excess heat generated by the device will accelerate the drift of the DSP-1's slave oscillator by decreasing frequency stability.  Oscillators function best at a precise temperature which is why the better ones are thermally controlled (TCXO).  Never underestimate how vital clocking is in digital audio components.

 

[IPodPJ]

Quote:

 
Seeing as how most audio is presented with 44.1 KHz sampling rate, it's a bit troubling to hear that the DAC performs better when you resample to 96 KHz before feeding in the signal.  From your posts it seems you've been burning the unit in for less than a week.  It would be interesting to hear the results of an A/B comparison between feeding 44.1 and 96 once the unit has been fully burned-in.  I look forward to more of your impressions.
 
Happy listening!

That only occurs if the filter isn't doing a good job, or if he lacks a precision clock in his source.  Since I know the Ref 7.1 has a good filter and is capable of sounding just as good if not better at 16/44.1 than 24/96 with the same material upsampled or not, then it's likely his source.  Clean power is also important, and a regenerator isn't the only thing that takes care of that.  It starts with the trafos and regulators inside your source.
 

 

[shamu144]

That's interesting because I had the opposite experience in my system. 24/96 files sound much better (better flow, prat, liquidity, natural) to me than native RBCD ripped at 16/44. I would have thought indeed the SRC algortythm or first digital filter in the DSP1 was the weak link. At least, this is my impression after comparing native RBCD playback vs the upsampled version through Fidelia or PM.
 
I reckon some users had similar negative experience when playing back RBCD formats. See here
 
Quote:

That only occurs if the filter isn't doing a good job, or if he lacks a precision clock in his source.  Since I know the Ref 7.1 has a good filter and is capable of sounding just as good if not better at 16/44.1 than 24/96 with the same material upsampled or not, then it's likely his source.
 

 

 

[IPodPJ]

Quote:

That's interesting because I had the opposite experience in my system. 24/96 files sound much better (better flow, prat, liquidity, natural) to me than native RBCD ripped at 16/44. I would have thought indeed the SRC algortythm or first digital filter in the DSP1 was the weak link. At least, this is my impression after comparing native RBCD playback vs the upsampled version through Fidelia or PM.
 
I reckon some users had similar negative experience when playing back RBCD formats. See here

See my post in the Reference 7 thread.  There is a reason the DSP-1 is on its 5th iteration.  I love the Ref 7.1 but if there was one weak spot it would be what you said.  Again refer to my post above about needing precision clocking to hear how good 16/44.1 can and will sound.  I'm not saying a 24/96 file won't sound better if it was mastered that way, but if you are working from the same file and merely upsampled it then 24/96 should not sound any better with precision clocking.
 
I know exactly what you are hearing because I once felt the same way as you.... until I started focusing on the clock circuit in my source and upgrading it.  There are other contributing factors too, but I can't discuss those here.
 

 

[shamu144]

I think we hear the same symptoms, but use different medicines

 
Btw, merely upsampling a 16/44 file to 24/96 is a lot more meaningfull than what you suggest, since it will condition the use of the first digital filter. This first digital filter is said to impact most the sound, and I would definitely agree based on my limited experience.
 
According to the DSP1 documentation, the filters included in that hardware when oversampling are steep linear phase filters (especially with the stock -130dB stopband attenuation setting)... Don't quote me on this, but trying to make sense of what I read, this means such filters can exhibit significant pre and post ringing. This will be heard as the sound becoming harder and unnatural, since pre-ringing can smear transients ealier in time. This is why lot's of people report improvements with the DSP1 when the stopband attenuation is set to -90dB, as the filter slope is not as steep, and should exhibit less ringing, making for a more natural listen (on the other side, there might be some aliasing occuring, but at -90dB, I don't really mind).
 
Look at what Ayre has to say regarding digital filters here: Ayre MP White Paper
 
Now, upsample all 16/44 material through a very high quality software SRC (like iZotope) with a different filter setting, and you will hear a significant difference (or not, because not everyone is sensitive to the response in the time domain). I have tried to replicate a minimum phase apodising filter (as Meridian and Ayre implement in their hardware) through the advanced settings of the SRC in iZotope (following those recommendations here), and the listening results are trully outstanding to my ears. Of course, YMMV.

 

[glenda]

Its just sad to me that the ref1/ref7 have such an archaic oversampler.  Once I tried the DSP rev 5 and didn't like it I knew it was time to move one.  Kingwa had been taken advantage off by the dsp programmer and dropped support, then he moved away from the superior PCM1704, rest is history.  Those of us lucky enough to get the full sized PCM1704 Audio-gd's when he was using a professional silicon-in-wire filter are holding on to them, to my ears they blow away the ref 1 & 7,  the Assemblage 2.7, etc,  but I am afraid the legacy will be a bit tarnished by the DSP-1. 
 
I hope you are able to  come up with a computer upsampling/filter that makes up for the deficiencies.  Ultimately ripping out the DSP1 and making the ref run like a Phasure DAC would probably be a huge winner,  you may want to look at the Phasure for ideas.   Probably a year or two away from 8x oversmapling on the computer,  but the potential is there for a ref to be modded and be future proof,   then folks will be paying $5k for them

 

[shamu144]

That is an interesting idea I am considering myself... It is possible to run the Ref 7.1 in NOS mode bypassing the DSP1 (hence no internal oversampling), while upsampling 16/44 through Fidelia/iZotope... I am just not that convinced that 24/96 would be enough for the PCM1704 to operate optimally for that matter... Would than even be possible at all ? It would probably need higher rates. Is it possible to send a 24/192 upsampled signal into the Ref 7.1 when bypassing the DSP1 ?
 
The phasure takes 24/768 if I recall correctly.
 
 
Quote:

 
I hope you are able to  come up with a computer upsampling/filter that makes up for the deficiencies.  Ultimately ripping out the DSP1 and making the ref run like a Phasure DAC would probably be a huge winner,  you may want to look at the Phasure for ideas.   Probably a year or two away from 8x oversmapling on the computer,  but the potential is there for a ref to be modded and be future proof,   then folks will be paying $5k for them

 

 

[glenda]

the dsp1 would have to go
 
Quote:

That is an interesting idea I am considering myself... It is possible to run the Ref 7.1 in NOS mode bypassing the DSP1 (hence no internal oversampling), while upsampling 16/44 through Fidelia/iZotope... I am just not that convinced that 24/96 would be enough for the PCM1704 to operate optimally for that matter... Would than even be possible at all ? It would probably need higher rates. Is it possible to send a 24/192 upsampled signal into the Ref 7.1 when bypassing the DSP1 ?
 
The phasure takes 24/768 if I recall correctly.
 
 
Quote:


 

 

 

[Nada]

Quote:

Apparently, bypassing the DSP1 also defeats oversampling, which is kind of a problem

 
However, I might have nailed down my issue with the Ref 7.1. Some high res material at 24/96 sounded absolutely stellar (talking only from a timing perspective), while regular CDRB format file simply did not. But if I am not mistaken, I realize that the master clock of the DSP1 only supports 32khz and 48khz (256Fs, 384Fs, 512Fs and 768Fs). So I decided to use Pure Music excellent SRC build in engine to upsample everything to 24/96. Improvements were obvious in my system when playing back native CDRB files. I am very curious to hear what other Ref 7.1 owners think about this.

I have a REF7. Ive played around with the DSP-1 jumper setting and have heard marked changes. I was surprised at what I heard as other guys said it made little or no difference and that was a consistent view so I wasnt expecting it to do anything. I wonder why others didn't hear a difference? Perhaps their gear imposed so much of a flavour on the music the changes were obscured? Or maybe the jumpers were put in the wrong pins?
 
 
 
There are many reason the 24/96 can sound better but the minimal preringing upsampling Izotope filter could be a huge one.
 


Quote:

If the music was released as 16/44.1, I don't understand how resampling it would introduce any improvements unless the equipment you are feeding it through somehow works better when being fed a signal with a certain sample rate. 
I would love to learn how REsampling with a higher sample rate would increase quality.
Anybody else care to comment on this?

Sample rates are essential for DA conversion to push the noise beyound the audio spectrum without creating artiacts throughout the music with filter noise.
This is the big issue with achieving analogue quality with digital storage. Basically the higher the sample rate the better for music quality or an upsampler with filter that minimises the horrible pre and post-ringing artefacts.
 
Izoptope and espiecally XXHighend-Phasure are doing that.
Quote:

My prefered settings so far:
- Pure Music with target upsampling at 96khz.
- "NOS Classic" upsampling (there are 2 minimum phase upsampling method to choose from)
- PLL off
- Stopband attenuation -90dB
Sounds sublime to my ears now even with CD rips. It was certainly not the case prior to these changes.

Nice work.
That all makes sense so far.
 
Quote:

 I do leave it constantly on though, which might not be recommended.
I also quite do not like the proximity of the AC socket and power cord with the digital circuit, and especially the AP2 gets very close to the IEC connector.
Other than that, I certainly do find that the sound is improving with burn-in. I still want to let it burn-in for another 200 hours before serious evaluation.

If you want the best sound leave it on in my experience.
It might be worth buying or rigging some mains plug shielding - let us know if you hear a difference please.
I think my Ref7 took about four weeks to mature and went through some interesting phases on the way - not always reassuring for me at the time!
 
Quote:

The RE7.1 runs.
If it have good heat dissipate, like have not anythings cover on its top board, far to the heat source, it can power on 24 X5 hours. Then want to had a restart.
 

Ive taken temperature reading of the REF7 using an  laser thermometer and most of the components from memory ran slightly warm - perhaps just 10C above ambient room temperature. I find it takes at least an hour of being on to start sounding good and have noticed repeatedly the sound keeps getting smoother and smoother up to three days. I use very transparent speakers to notice that and I don't think I would notice with a tube amp on my HD800's even. I wonder why a restart is needed after five days? Wound the DSP-1 need restarting I wonder?
 
 
Quote:

That is an interesting idea I am considering myself... It is possible to run the Ref 7.1 in NOS mode bypassing the DSP1 (hence no internal oversampling), while upsampling 16/44 through Fidelia/iZotope... I am just not that convinced that 24/96 would be enough for the PCM1704 to operate optimally for that matter... Would than even be possible at all ? It would probably need higher rates. Is it possible to send a 24/192 upsampled signal into the Ref 7.1 when bypassing the DSP1 ?
 
The phasure takes 24/768 if I recall correctly.

 

 
Definitely put the DSP-1 in NOS mode and check it out. It sounds to me, well different. I used SOX upsampling to push the 44.1 up to 88.2 and prefered that in NOS DSP-1 mode - lots clearer and more details while still very smooth. High res 24/96 tracks sounded better again but maybe thats the recording quality?

I think Kingwa could say if the REF7.x    SPDIF receiver and DSP-1 could accept any higher sample rates above 24/96.
Somehow I doubt it.
The PCM1704 can accept 24/96 X 8 oversampling so most receivers and filters used in PCM1704 machines take a max of 24/96.
Check it out as in NOS mode 192 probably will perhaps sound even better then 96 according to theory and the trend.
 
Watch out for any noise experimenting like this. Use a preamp or integrated amp volume control  turned all the way down and a cheap pair of headphones to start with!
 
I like your questions and hope you try it for yourself and report back for us?
 
 Its doubtful anyone else has done this with iZotope filtering so believe your own ears and look after them with that volume knob.
 
The Phasure USB is stated to go to 24/768. That means ALL the up-sampling and filtering is in the software domain under the control of the users fingertips. The Phasure designer must be exceptional to achieve that. I think its a first. Hes also got a minimal ringing filter in the software. This is like the holy grail of this design ethos.
 
As good as the PMD100/200 filters are and how amazing the DSP-1 is, the XXHighend filter (if it actually performs as stated) is going to blow them all away.
 
As well if a better filter comes along all that is needed is a quick download, update the software and then listen for the difference. No lid to take off, no expensive filter to take out, send away for reprogramming or in the case of the early filters replace. This is clearly the way of the future. I mean now. If you have a spare 4.5 grand dropping out of you pocket that is.

 

[vrln]

My initial impression was also that regal is back - which would be a good thing, I enjoyed his posts a lot I think the musical series drop is maybe more about business, since with the ACSS line he can lock people into buying his amps as well. That said, I´ve never heard a musical series Audio-gd DAC. It would be very interesting though.

 

[IPodPJ]

Quote:

My initial impression was also that regal is back - which would be a good thing, I enjoyed his posts a lot I think the musical series drop is maybe more about business, since with the ACSS line he can lock people into buying his amps as well. That said, I´ve never heard a musical series Audio-gd DAC. It would be very interesting though.

I disagree.  When I owned the Phoenix, I felt it sounded better using the XLR inputs from the XLR outputs on the Ref 1 and Ref 7 than it did from ACSS, so there is good IV conversion in the DACs.  I don't own the Phoenix anymore so I can't test the same with the 7.1, but I assume I would still prefer the XLR connection.  It sounded more totally balanced to my ears.
 
I also built an input with passive IV conversion into my amp using the best Texas Components resistors you can buy, with jumpers for configuring the voltage output I want (16 choices) based on the Ref 7.1's current output mA, and I still prefer using the voltage-based XLR inputs fed by the Ref 7.1.  I wound up wasting a lot of money as all those resistors weren't cheap, but I never knew until I tried.  Also, my attempt was passive while I believe Audio-gd uses active conversion.
 
This is a very good thing, since it gives customers their choice of amplification and they are not locked in, as you say, to buying an Audio-gd amp if they purchase a DAC with ACSS.

 

[vrln]

Good points. Actually I also liked XLR more when I owned the Phoenix. But I think my argument is mostly a psychological one: people wonder about the ACSS outputs/inputs and on the site they see it has higher SNR. Not that it really means anything though... But it´s an extremely effective incentive for creating the urge to get the "optimal" all-ACSS solution. Yes, while in reality XLR vs ACSS is all down to subjective taste, I still think that the ACSS lineup has a very clear commercial advantage to the musical line. Unfortunately so few (I can´t think of anyone here actually) have heard the REF 7 vs REF 8 side by side both using XLR it´s all speculation in the end. What I do know however, is that the neutral lineup sounds so good via XLR that I don´t really wonder about the lost musical line.

 

[IPodPJ]

That SNR threshold is so far beyond the levels we would ever listen to music at that it's irrelevant.....  Unless you listen to sustained levels of 126dB.

 

[vrln]

Exactly, that´s what I meant with the "not that it really means anything" line. But when it comes to marketing, it works.