I received today a new ROM upgrade to one of my SDTrans 352.8 SD card players
The SDTrans 352.8 are upgraded with JFET (LU1014D) regulated powersupplies with a virtual capacity of 20F placed after yet another JFET regulated powersupply that are supplied from batteries.
The ES9018 are also equipped with JFET regulated powersupplies and shares the first JFET regulator level with the SDTrans 352.8. There are minimum of three regulator levels on all powersupply voltages.
Well - back to the ROM...
I had ordered a different programming of the SDTrans 352.8 due to the ES9018 DAC receives 32bit of data even if the source data are 16bit (I could have programmed the ES9018 to use only 16bit of the data, but the ES9018 must then have been reprogrammed to receive 24bit or 32bit when I play 24bit or 32bit sources, and reprogrammed to 16bit again etc..).
Standard programming of the SDTrans 352.8 clocks out "0" for the 16 LSB bits (of the 32bits) for both positive and negative samples with a 16bit source, and the same happens with the 8 LSB bits with 24bit sources...
The new ROM causes the SDTrans 352.8 to clock out "1" for the 16 LSB for negative samples, or for the 8 LSB with 24bit sources..
This upgrade are the most positive audible upgrade I have ever done
Absolutely everything sounds better, but the easiest way to describe the difference are that ALL "digitalis" are gone and replaced with details and natural sounding instruments and voices..
The depth and width of the soundstage are much better and a lot of "new" information in the records are giving new meaning.
I will "salute" Bunpei and Chiaki for the efforts they have used to make this fantastic product and for their willingness to have done several changes after my request...
The SDTrans 352.8 are upgraded with JFET (LU1014D) regulated powersupplies with a virtual capacity of 20F placed after yet another JFET regulated powersupply that are supplied from batteries.
The ES9018 are also equipped with JFET regulated powersupplies and shares the first JFET regulator level with the SDTrans 352.8. There are minimum of three regulator levels on all powersupply voltages.
Well - back to the ROM...
I had ordered a different programming of the SDTrans 352.8 due to the ES9018 DAC receives 32bit of data even if the source data are 16bit (I could have programmed the ES9018 to use only 16bit of the data, but the ES9018 must then have been reprogrammed to receive 24bit or 32bit when I play 24bit or 32bit sources, and reprogrammed to 16bit again etc..).
Standard programming of the SDTrans 352.8 clocks out "0" for the 16 LSB bits (of the 32bits) for both positive and negative samples with a 16bit source, and the same happens with the 8 LSB bits with 24bit sources...
The new ROM causes the SDTrans 352.8 to clock out "1" for the 16 LSB for negative samples, or for the 8 LSB with 24bit sources..
This upgrade are the most positive audible upgrade I have ever done
Absolutely everything sounds better, but the easiest way to describe the difference are that ALL "digitalis" are gone and replaced with details and natural sounding instruments and voices..
The depth and width of the soundstage are much better and a lot of "new" information in the records are giving new meaning.
I will "salute" Bunpei and Chiaki for the efforts they have used to make this fantastic product and for their willingness to have done several changes after my request...
Last edited:
I have listened more to the upgraded SDTrans 352.8 and absolutely all of my music are now like "new" and I am now enjoying playing music I thought I knew - only to discover what I never have heard in any setup before.
Not only have everything become much more natural, but the biggest difference are that both the placement in the soundstage and level of natural details are extremely much better.
On many records I now are able to hear non-music noises from the record studio and hiss and hum from recording equipment etc..
This new level of details are NOT disturbing, but a great relief as all the records I previously did not like due to I heard something was wrong - but not exactly what - are now playable.
Music that I previously believed had been "destroyed" during recording and mixing are now playable..
I do not want to put to much OT in Bunpei and Chiakis thread,
but here are some info regarding the amplifiers:
http://www.diyaudio.com/forums/digital-line-level/117238-ess-sabre-reference-dac-8-channel-56.html
#559 and #560
Some info regarding the AMT:
http://www.diyaudio.com/forums/full-range/156466-one-only-full-range-amt.html
Not only have everything become much more natural, but the biggest difference are that both the placement in the soundstage and level of natural details are extremely much better.
On many records I now are able to hear non-music noises from the record studio and hiss and hum from recording equipment etc..
This new level of details are NOT disturbing, but a great relief as all the records I previously did not like due to I heard something was wrong - but not exactly what - are now playable.
Music that I previously believed had been "destroyed" during recording and mixing are now playable..
I do not want to put to much OT in Bunpei and Chiakis thread,
but here are some info regarding the amplifiers:
http://www.diyaudio.com/forums/digital-line-level/117238-ess-sabre-reference-dac-8-channel-56.html
#559 and #560
Some info regarding the AMT:
http://www.diyaudio.com/forums/full-range/156466-one-only-full-range-amt.html
Last edited:
Interesting Ray,
I wonder would you have achieved the same improvement if you programmed the ES9018 to only use 16 bits for 16bit sources, etc? Again, in other words, could this be made more universal & less specific to SDtrans (not wishing to go OT here)
What do you reckon lies at the heart of the sonic improvement?
I wonder would you have achieved the same improvement if you programmed the ES9018 to only use 16 bits for 16bit sources, etc? Again, in other words, could this be made more universal & less specific to SDtrans (not wishing to go OT here)
What do you reckon lies at the heart of the sonic improvement?
Hi John,
As I did not know how Dustin have programmed the ESS chips with regard to how to increase the bit length from 16 to 32, and due to the SDTrans 352.8 already before the upgrade by a large margin was the best player I ever have heard I felt it as safest to have the SDTrans to deliver the data perfectly.
Then ALL DACs receiving data from the SDTrans 352.8 either via SPDIF (24bit) or I2S (32bit) will benefit from the upgrade.
The sonic improvement are simply a removal of a digital "crossover" distortion due to the faulty values in the LSB bits below the 16 bit.
Some calculations I did earlier indicated a distortion of 1/2 to 1 LSB at clipping that increased up to close to 100% distortion at the lowest level where only the LSB bit are changing - this distortion are present in the whole level range - but worst around 0 volt.
And to my understanding - this problem exists possibly in MOST hardware and software that sends 16bit data at 20, 24 or 32bit length...
Thanks Ray, if your analysis is correct & this is a widespread phenomena in most hardware & software that outputs 16bit data at 20,24 or 32, this is fundamental information. I don't know too much about digital audio processing but how has this not been noted before? So from what you say I gather that the 16 LSB of a 32 bit word are padded out incorrectly in negative samples with 0s instead of 1s. Is this not a basic flaw in digital? Please excuse my ignorance in this.
PS. Have you tried the ES9018 in synchronous mode i.e operating from the clock in the SDtrans? Well worth a listen although you seem to be happy with the sound you are now getting. I would be interested in your impressions of the sound of synchronous clock operation in the Sabre DAC.
PS. Have you tried the ES9018 in synchronous mode i.e operating from the clock in the SDtrans? Well worth a listen although you seem to be happy with the sound you are now getting. I would be interested in your impressions of the sound of synchronous clock operation in the Sabre DAC.
Last edited:
Hi Patrick,
I don't understand your biasing comment - is there a good primer on digital audio processing that I can read?
But what about positive samples where the empty LSBs are still all 0s - as far as I understand it? I realise I don't really understand much though
For negative samples the 8 (24bit data) or 16 (16bit data) LSB to fill out to 32bit are clocked out as "1".
For positive samples the 8 (24bit data) or 16 (16bit data) LSB to fill out to 32bit are clocked out as "0".
Ray,
Can you explain the effect of this change in the digital processing sense. I was thinking that it works like quantization -adding noise to the signal which helps the distortion particularly at low signal levels but this does not hold up for the positive going samples which are not padded with 1s.
No not the ES9018 - it is clocked at 100MHz.
But the ES9022 and ES9022 balanced version was tested with synchronous master clock from the SDTrans and 40 MHz, 50 MHz and 80 MHz asynchronous clocks..
50 MHz asynchronous clock was the best alternative.
80 MHz asynchronous clock appeared to play correct, but there was a "hiss" - background noise so it could not be used.
I have not tried to over-clock the ES9018, but I will test this on the new PCB where I can change clocks under software control from the remote..
A "0" in the LSB bits for a positive sample are a "0"...
A "1" in the LSB bits for a negative sample are a "0"...
So when negative samples wrongly are padded with "0" values in the LSB bits it represents a value - just like if you padded the LSB bits of a positive sample with "1"..
OK, I have to try to get my head around this - need to do some reading, methinks
BUT again, isn't this such a fundamental error in digital processing? How could it be overlooked in most hardware/software implementations?