[MartinWT]

I don't know about the rest of you but post mod power-up excitement is always filled with mixed emotion. Fear first sets in.....did I botch this up permanently??!!

Yes! When I used to rework my Beresford Caiman SEG DAC, it was an easy board to work on, double-sided only and I even replaced SMDs without too much bother. Perhaps the relatively low cost of the DAC made me unconcerned at working on it, as well as having Stan around for advice and tips. Some of us really made that DAC sing.

The LKS is another matter: the board is harder to work on, especially when sucking solder from tight spaces. It's Chinese, so effectively no support and a month at least to purchase a replacement board should the worst happen. It makes me much more nervous of bricking it.

Thank goodness for this forum, an oracle of information.

 

[b0bb]

Now 2 questions, my voltage rails from the analog source were at 16.5v and -16.5v, do you have B0bb and the rest of GEN1 ?, What do you think about using "styroflex" capacitors in parallel with the I / V resistors instead made of polypropylene?

Mine is ±15.5V, If you replace any of the components like the filter caps just re-adjust the grey trimmer.
I did this when I replaced the shunt resistors in the analog power supply with Caddock MP915s.

Styroflex Polystyrene (PS) caps will round out the midrange making it richer, polypropylene is a little on the cold side.
The problem with PS is getting in tight enough tolerance to match the resistor tolerance.
You have to be extremely careful not to melt them when soldering.
Temp control during soldering is tricky on the LKS board.

 

[b0bb]

The LKS is another matter: the board is harder to work on, especially when sucking solder from tight spaces. It's Chinese, so effectively no support and a month at least to purchase a replacement board should the worst happen. It makes me much more nervous of bricking it.

Thank goodness for this forum, an oracle of information.

The board has to be pre-heated to bring the surrounding area up to temperature otherwise the goundplanes will suck the heat out of the iron quicker that it can heat it.

 

[fmzip]

Yes! When I used to rework my Beresford Caiman SEG DAC, it was an easy board to work on, double-sided only and I even replaced SMDs without too much bother. Perhaps the relatively low cost of the DAC made me unconcerned at working on it, as well as having Stan around for advice and tips. Some of us really made that DAC sing.

The LKS is another matter: the board is harder to work on, especially when sucking solder from tight spaces. It's Chinese, so effectively no support and a month at least to purchase a replacement board should the worst happen. It makes me much more nervous of bricking it.

Thank goodness for this forum, an oracle of information.

So funny you mention Stan and the Caiman. That was the first DAC I had. I modded it as well. Needless to say, I destroyed it from lifting a few pads with my poor soldering. Stan actual was kind enough to sell me just a board to replace mine. I destroyed the second one as well after swapping parts one too many times!

Maybe this is why I am getting to the point of leaving this board inside the DAC once and for all. The investment in this one is far steeper, the sound is 1000x better as well. I think I'd cry if I were to scrap this DAC at this point in time. At least I am smart enough now to let the pros at my shop who are skilled in soldering do the mods when I am in over my head.

The thing they teach in our IPC soldering classes is you really don't ever want to touch a solder joint unless you have to. The stress from heating up that pad is not a good thing. On a multilayer board like this one, a lifted solder pad could be costly. Not as simple as a doubled sided board where you can run a jumper wire if you screw something up and lift a pad.

The board has to be pre-heated to bring the surrounding area up to temperature otherwise the goundplanes will suck the heat out of the iron quicker that it can heat it.

exactly what my tech said

 

[oldearwax]

By the way, why did they put this 10uf / 500v EPCOS capacitor, why this model with so much voltage?

ESL57,
I *think* EPCOS cap in the Vcom position is for "sound" and not for voltage.
Capacitor in this Vcom position has a profound effect on the sound of this DAC.
The EPCOS cap has a lot of details and ambience/air , but not as much bass as a tantalum cap.
The voltage @ Vcom is only 1.21 volt !



You have replaced your EPCOS with a yellow capacitor and blue cap. What are those ?
Can you describe the new sound ? Thank you in advance.

 

[oldearwax]

Not advisible, the values have been increased almost 100x from stock to compensate for the 100x lower cap value, this increases the impedance 100x, making the Vcom reference more vulnerable to RF noise pickup, you will have to shield the resistors to avoid RF noise pickup.

The biggest source is the adjacent XO @100MHz followed by any strong FM radio stations nearby.

The input capacitance of the jfet input of the opamp plus the high resistance (470k) forms the input of a primitive RC radio tuner.
Stock DAC avoids this by using a much lower 5k resistor value to dampen any pickup.


I am not sure if the 0.1uF Mundorfs is worth this kind of trouble, you have seen how sensitive the Vcom is to even a few millivolts offset.
(The stock value is already marginal in suppressing RF pickup as @Xoverman and others have shown thru experiments with shielding, you should try and avoid making the problem worse)

You have the ultra low leakage 150uF cap, you could scale down the value of the resistive divider chain by 5x to 1kohm, this is still large enough compared to the 3ohm ESR of the cap, there are 2 main benefits

1) Vcom offset errors reduced by 5x from cap leakage current
2) 5x lowering of the impedance of the reference chain resulting in improvement in RF pickup resistance

RF noise pickup vulnerability is one of the main weaknesses of the 004.

RC remains the same.

Stock
A) R=5K C=10uF RC= 5x1000x10x10**-6 = 5x0.01

0.1uF mod (LKS005 ??)
B) R=500K C=0.1uF RC= 500x1000x0.1x10**-6 = 5x0.01

Bass is strong with the wet tantalum Vcom mod, background is darker too. Perhaps too strong for my taste.
The delicate top end and very fine details are lost.
Most unfortunate is that solo piano music does sound right to my ears.
The lower registers sound impactful.
But the higher register notes sound more like hammering, the resonace decay is lost.

 

[b0bb]

RC remains the same.

Stock
A) R=5K C=10uF RC= 5x1000x10x10**-6 = 5x0.01

0.1uF mod (LKS005 ??)
B) R=500K C=0.1uF RC= 500x1000x0.1x10**-6 = 5x0.01

That is the trap you fell into, did you actually trace the 005 board or merely speculated on what might be the case based on the photos?
Do you have detailed photos showing what LKS did with their new design on the 005?

The following from the online calculator: https://keisan.casio.com/exec/system/1258032649

Your mod increased the input impedance of one of the differential inputs by 100x but did nothing to address the impedance seen by the other input.
This results in unbalancing the differential performance of the I/V.

The impedance seen by the opamp inverting input is about 25ohms, but the non inverting input goes from 16ohms to 1600ohms, that will change the sound for sure.
The I/V input impedance is completely unbalanced by your mod in addition to creating a radio tuner frontend as mentioned before

The mod you proposed messes with the frequency compensation, the input capacitance + input impedance causes a peak in gain at the resonant frequency.
Now that the impedance is altered, the scheme LKS put in no longer applies.
Frequency compensation affects the stability, it can bite you later if your mods eat away too much of the stability margin LKS built into their opamp.

You got lucky the discrete opamp did not become unstable with your mods, others like @ESL57 and @ktm15 were not so fortunate.

 

[b0bb]

Gen1 004s have a 22ohm resistor between Vcom and the non-inverting input.
Impedance seen by the inverting input is about 22ohm at high frequencies.
This helps to keep the impedances approximately equal between the inverting and non-inverting inputs when this point is reached.

With 5k//150uF Tantalum cap the impedance of the RC network at 350Hz is about 3ohm which is the min ESR of the cap
This means Vcom sees a flat impedance of 25ohms (22 from resistor + 3 ohm Vcom network ) from 350Hz -22KHz
This is about equal the impedance of the inverting input

Gen2s seem to be missing this resistor, if it is indeed missing this might be something worth adding back to the Gen2 boards.
Without the resistors the impedance falls with frequency and the differential balance varies with frequency.

The Gen1 board could also do with a mod to move the flat part of the impedance lower say 35Hz but the cost of the caps becomes prohibitively expensive need about 680uF

 

[b0bb]

This shows the topology of the 9038Pro strapped for mono operation.
VCM is Vcom (non-inverting input) should be seeing the same impedance as the DAC output impedance at the inverting input.
DAC output impedance is 25ohm.

 

[oldearwax]

That is the trap you fell into, did you actually trace the 005 board or merely speculated on what might be the case based on the photos?
Do you have detailed photos showing what LKS did with their new design on the 005?

The following from the online calculator: https://keisan.casio.com/exec/system/1258032649

Your mod increased the input impedance of one of the differential inputs by 100x but did nothing to address the other input there by unbalancing the differential performance of the I/V.

The impedance seen by the opamp inverting input is about 25ohms, but the non inverting input goes from 16ohms to 1600ohms, that will change the sound for sure.
The I/V input impedance is completely unbalanced by your mod in addition to creating a radio tuner frontend as mentioned before

The mod you proposed messes with the frequency compensation, the input capacitance + input impedance causes a peak in gain, now that the impedance is altered, the scheme LKS put in no longer applies.
Frequency compensation affects the stability, it can bite you later if your mods eat away too much of the stability margin LKS built into their opamp.

You got lucky the discrete opamp did not become unstable with your mods, others like @ESL57 and @ktm15 were not so fortunate.

It is not "my mod", I posted the mod in #3332 for Tiago.
@tiago please join the discussion and share your impression of this mod.

The RC that I was referring to is the Time Constant to replenish the voltage due to leakage.
See diagram.
Vcom ideally should stay constant. Practically it is performed by the (equivalent) resistor R and capacitor C.
I don't know why these values were picked.

I know nothing about op-amp frequency compensation.
I agree that the input impedance of both inverting and non-inverting inputs should not be varied too much.
But this is not possible due to frequency dependency of the impedance of Vcom capacitor C.

The 150uF wet tantalum cap mod also modified the input impedance by 15x too.

What is the spectrum of frequency of this Vcom capacitor C subjected to ??? Do you know ?
The output of the DAC is a sequence of sigma-delta spikes (I could be totally wrong here) which
may have all kinds of frequencies. This is my guess.

 

[b0bb]

It is not "my mod", I posted the mod in #3332 for Tiago.
@tiago please join the discussion and share your impression of this mod.

The RC that I was referring to is the Time Constant to replenish the voltage due to leakage.
See diagram.
Vcom ideally should stay constant. Practically it is performed by the (equivalent) resistor R and capacitor C.
I don't know why these values were picked.

RC values are a scale factor, the actual impedance is determined by the frequency
Merely looking at RC values means missing a big chunk of what is actually going on.

If you want to implement some of the changes from the 005, at least get the detailed pictures to understand what LKS did.
The blurry marketing photos should not be used as any form of reference
All we are able to see is the 0.1uF.
Do you even know it is the Vcom cap without evidence of the physical connection from the trace on the board?

I know nothing about op-amp frequency compensation.
I agree that the input impedance of both inverting and non-inverting inputs should not be varied too much.
But this is not possible due to frequency dependency of the impedance of Vcom capacitor C.

That is why Gen1s have the 22ohm resistor, impedance never drops below 22ohms.
The way you have drawn things is definately not the case in the Gen1 board.

If you want to mod why not do this mod instead?

The 150uF wet tantalum cap mod also modified the input impedance by 15x too.

That is what gives you the extra bass extension.

Capacitor reactance increases with falling frequency, it becomes less effective with lowering frequency
With 10uF//5k, the reactance is 3 ohms at 5000Hz (most sensitive frequency band of human hearing)
With 150uF//5k, the reactance is 3 ohms at 350Hz
3ohms is the min ESR of the cap so its reactance never falls below 3ohms.
Below the frequency where the reactance hits 3 ohms the bypass becomes less and less effective as rectance rises, the lower you go in frequency range.

This means the bypass operates up to a frequency 15x lower without the penalty of crazylow reactance at the higher frequency upsetting the I/V balance too much.
(It works better if you have the 22ohm in series, it then becomes 25 ohm instead of 3ohms).

This is why the tantalum caps work as well they do, the 150uF extends bypass down to 350Hz compared to stock which starts to become in-effective at 5kHz
This is the reason the low midrange and upper bass performance on the 004 is lifted.

What is the spectrum of frequency of this Vcom capacitor C subjected to ??? Do you know ?
The output of the DAC is a sequence of sigma-delta spikes (I could be totally wrong here) which
may have all kinds of frequencies. This is my guess.

Depends on the sampling frequency typically 22MHz, the main impact is on the audio signal not the sampling noise as the impedence rises as the frequency falls.
At 22MHz the cap has effectively removed the resistor from the signal path, very undesirable as it unbalanced the I/V.
Gen1 used a 22ohm to isolate the opamp input.

 

[b0bb]

@ESL57 remove the 2 yellow caps and see if you like the result.
@Whitigir I think you have a Gen1, you might also see some benefit for your 004

This removes the frequency bypass of the I/V input impedance compensation resistor, the I/V is now compensated across it entire operating frequency range.

 

[b0bb]

deleted duplicate response

 

[piaseczek]

@ESL57 remove the 2 yellow caps and see if you like the result.
@Whitigir I think you have a Gen1, you might also see some benefit for your 004

This removes the frequency bypass of the I/V input impedance compensation resistor, the I/V is now compensated across it entire operating frequency range.

If I understand correctly the resistor near the blue capactor is 22ohm? In the Gen2 there are no such resistors, could it be added?

 

[piaseczek]

The Gen1 board could also do with a mod to move the flat part of the impedance lower say 35Hz but the cost of the caps becomes prohibitively expensive need about 680uF

So the bigger the tantalum cap the better and the optimum is 680uF? For the Gen2 it isn't true?