Soekris Dam Dac

[jkeny]

nige2000, many thanks for brave steps, i waiting for clock changes long time. for sure i will try the same. I like this thread. Here is lots of interesting researches. Thanks for sharing your experience.

5 different Soren DACs were researched and heavily moded(almost all mods described here and even more) by me, and I clearly noticed, that all dacs has different measurements more than normal spreading in comparison with another dacs. All has a different level of harmonics. Sometimes 20-30db of difference(The measurements were made using a Lynx Aurora ADC predominantly at a frequency of 96K 24-bit)

My idea, that the answer is a difference(spreading) of resistors and their mutual ratio. And this is the one of significant factor that influence on measurements, harmonics and sound quality in the final. Even more, I have changed resistors and received a clear change in harmonics level.
Look at this.



i see that the r2r soekris schema is


There is 1021 .01 version dac from the first delivery. Having re-measured of each resistor there. And have received the following results:

Yellow marked real accuracy percentage that are outside of the specified accuracy - 0.01.

so my conclusions:
- It was stated that the precision resistor is 0.01, but in reality there are resistors with an accuracy of less than stated
- By itself, the accuracy is not so important, more important it is 2R and R mutual ratio (should be 2), and the same resistors in series (bits).
- Since the resistors in the matrix has a significant variation (over 0.01), which means that the result of these relations and also provide spread.

I want to ask an advice here on r2r matrix construction. For me it remains full unclear until the end the sense resistor in bit from 0 till 2 rows. Soren calls it like “thermometer”. What is the practical sense of 0,1,2 rows? Any thoughts?

Thanks Alecm - great work
Are you sure that is the correct schematic for the resistor network - it can't work if bits 0,1 & 2 aren't connected to a a 574 chip output? You traced this schematic from the Soekris board, I presume? Or are you saying that 0, 1 & 2 are connected to the register just through the 4.99K resistor & not a R2R ladder?

Shame about the resistors that are out of spec - it isn't very reassuring
The accuracy of the resistors does determine the linearity of the DAC's output & the location of that out-of-spec resistor will determine which bit's voltage level it is affecting - obviously resistor on the most significant bits (MSB) will output the highest voltages so any linearity issues here will result in larger voltage errors than least significant bit errors

[jkeny]

I want to ask an advice here on r2r matrix construction. For me it remains full unclear until the end the sense resistor in bit from 0 till 2 rows. Soren calls it like “thermometer”. What is the practical sense of 0,1,2 rows? Any thoughts?

wonder if therses any point asking soren?
At worse case there will be no reply or a political type non disclosing reply

Listened for an hour with fixed 22 clk feeding fpga and the pre fpga reclocker
Im positive the adjustable clock shoud be removed and replaced with fixed clk and i2s prealigned
Resolution is immense with darker background and a level of coherance ive not experienced before
The biggest improvement since vref ;)

Wow, that's quite a jump in SQ, Nige - well done! Must get a listen to it - put it in a box & lock it :)

[alecm]

if theres any mods we haven't done let us know please

thanks, i will do)

MOD - arms adjustment
Sometimes this mod is very effective and significant. But very depends from each sample of Soekris DAC. The idea is voltage adjustment in the arms of soekris dac. This modification is very simple to implement and good for harmonics reduction. I have received a very good sound results that increase clarity and sound clearness. This is very good to hear when you twist trimmer in the opposite direction.

Here is difference btw 2mV and 0mV (-1db, 1khz)


You have to solder trimmer only. The opamp (which makes from -4 to +4) has two resistors - standard connection of inverted schema. You need to solder trimmer legs on resistor to aligning their balance. Without adjustment the difference btw -4 and 4 is 2-3mV. So you have to set as less as possible.

[alecm]

Thanks Alecm - great work
Are you sure that is the correct schematic for the resistor network - it can't work if bits 0,1 & 2 aren't connected to a a 574 chip output? You traced this schematic from the Soekris board, I presume?

Yes, each resistor connected to corresponding 574 output. And yes, schematic was traced from DAM pcb.

[alecm]

The accuracy of the resistors does determine the linearity of the DAC's output & the location of that out-of-spec resistor will determine which bit's voltage level it is affecting - obviously resistor on the most significant bits (MSB) will output the highest voltages so any linearity issues here will result in larger voltage errors than least significant bit errors

Yes for sure. This is the idea of my question. Recently I started to fix the resistors and found lots of unintelligible things.
Look at here (-4db is -1db in reality)



Nevertheless, on fixing resistor I didn’t received any significant results. The tricky thing, with the falling of signal, harmonics are hardly reduced. But what the most uninterpretable is their random behavior. Something is making additional instability to the result.
There are two assumptions:
-somehow affect the final result bits 0 1 2
- filters

[nige2000]

I want to ask an advice here on r2r matrix construction. For me it remains full unclear until the end the sense resistor in bit from 0 till 2 rows. Soren calls it like “thermometer”. What is the practical sense of 0,1,2 rows? Any thoughts?

wonder if theres any point asking soren?
At worse case there will be no reply or a political type non disclosing reply

Listened for an hour with fixed 22 clk feeding fpga and the pre fpga reclocker
Im positive the adjustable clock shoud be removed and replaced with fixed clk and i2s prealigned
Resolution is immense with darker background and a level of coherance ive not experienced before
The biggest improvement since vref ;)

Wow, that's quite a jump in SQ, Nige - well done! Must get a listen to it - put it in a box & lock it :)

its a big jump to my mind as this type of benefit is hard to obtain
it might not be very in your face type of difference
it seems to blend mid level detail with low level without much drowning effect, decay seems to fade naturally into a dark background
just seems more right

theres still an async between sd player and reclocker that needs to be made sync i think theres a little more to get
every async clocking of i2s seems to hurt coherency and low level detail

[jkeny]

The accuracy of the resistors does determine the linearity of the DAC's output & the location of that out-of-spec resistor will determine which bit's voltage level it is affecting - obviously resistor on the most significant bits (MSB) will output the highest voltages so any linearity issues here will result in larger voltage errors than least significant bit errors

Yes for sure. This is the idea of my question. Recently I started to fix the resistors and found lots of unintelligible things.
Look at here (-4db is -1db in reality)



Nevertheless, on fixing resistor I didn’t received any significant results. The tricky thing, with the falling of signal, harmonics are hardly reduced. But what the most uninterpretable is their random behavior. Something is making additional instability to the result.
There are two assumptions:
-somehow affect the final result bits 0 1 2
- filters

OK, thanks but I'm not sure what this is a graph of (or many graphs overlaid)? I presume you changed all out-of-spec resistors & you are getting no reduction in harmonics but rather random increases in different harmonics?

[rickmcinnis]

Nigel,

So you can use either "22" or "24" mHz clocks? I am assuming these are abbreviations for the typical cd player values?

Hoping I can use the 24 in the DAM and the 22 in the SDTrans while I wait.

[nige2000]

Nigel,

So you can use either "22" or "24" mHz clocks? I am assuming these are abbreviations for the typical cd player values?

Hoping I can use the 24 in the DAM and the 22 in the SDTrans while I wait.

yea audio frequency values
you need to match them so two 22mhz for 44.1khz, 88, 176 etc
so pairing a 22 with a 24 will result in no lock on dam

[nige2000]

The accuracy of the resistors does determine the linearity of the DAC's output & the location of that out-of-spec resistor will determine which bit's voltage level it is affecting - obviously resistor on the most significant bits (MSB) will output the highest voltages so any linearity issues here will result in larger voltage errors than least significant bit errors

Yes for sure. This is the idea of my question. Recently I started to fix the resistors and found lots of unintelligible things.
Look at here (-4db is -1db in reality)



Nevertheless, on fixing resistor I didn’t received any significant results. The tricky thing, with the falling of signal, harmonics are hardly reduced. But what the most uninterpretable is their random behavior. Something is making additional instability to the result.
There are two assumptions:
-somehow affect the final result bits 0 1 2
- filters

OK, thanks but I'm not sure what this is a graph of (or many graphs overlaid)? I presume you changed all out-of-spec resistors & you are getting no reduction in harmonics but rather random increases in different harmonics?

so matching the resistors doesnt resolve or stablise?