[Setmagic]

Is there an English language manual available for the Z4?

Basic one:

 

[Mr.HiAudio]

today replase:
1) opa2156 -> lpf;
2) lp5907 -> ldo x2;
3) 3 capactior 22uF muRata -> 2 DACs (3pc x2).

Result:
deep soundstage, more details, more highs, cleaner sound. Just wow sound. Will replace today/weeks other 22uF capactior(around ldo and DACs)

 

[Inspector Gadget]

Greate stuff. You can measure for 300 ohm as well?

Main problem with based on opa1622 circuits is design for low impedance usage. In datasheet it is even 600 Ohm but it's a number. To work best with heavy loads need to be set up in parallel. Single 1622 isn't capable to drive it in proper way.

How about 2 opa1622?
I know good DIY amp with 4x 1622 in parallel at the end of headphone stage is capable to drive 300 and 600Ohm in proper way.

@edit
For comparison this is stock previous flagship from Zishan DSDs AK4499.
Mesurment by @slavalun translated from Russian.

1) stock


2) With very complex and expensive slavalun MOD [source from Player.ru]

I read few articles about adding too many capacitors, that just creating more problems than benefits.
Data sheet of 9038, 1656 and 1622 amount of capacity is just on right spot on stock Z4.

Yes, they don't do any harm I add as well 56nF Panasonic best performance film capacitors with ultra low ESL ESR.
So it's now 15nF + 56nF + 10uF (stock one)
I all data sheets for 1622 are only 10uF I know it's a newest generation opamp and TI also recommend 10uF bit I'm just wonder if it helps to add more.
Same with LDO in Zishan we have 10uF, TI recommend 1uF and I was solder 15nF + 47uF but seems it's changes nothing.

Previous Zishan had insufficient amount of capacity but Z4 seems to have right amount. Someone can confirm?

OK, so dont take me the wrong way, you are clearly putting in hard work here for the community, but I think there are some things you could do to help yourself and others.

1. Spend some time drawing out a schematic for the analogue stages (this is a very simple application and not multilayer PCB, so should not be too difficult) then spend some time understanding the circuit. I dont expect you to reverse engineer the whole circuit, but if you want to improve a circuit, it is best to know what the circuit does and why values were chosen in that circuit.
2. Understand where the limitations are and focus mods on those areas.
3. One mod at a time (or at least one section). this allows easier troubleshooting if something goes wrong and also easier to determine if you are making an improvement and if so, what made the improvement.
4. power supply decoupling/bypass caps. more capacity does not always = better smoothing, especially if placing many caps in parallel.


also npo/c0g, while good (one of the best) if you know exactly what target frequency you are trying to attenuate and have the means to measure the result, but it is almost guaranteed to cause issues you can't measure if you don't. they have very high q and thus can easily cause resonances when they interact with other capacitors (and their parasitic secondary effects (mostly the inductance)). Films are even worse, so when you place—for example—15nF c0g || 56nf film (very close, but not the same value) and then do that on a rail of each opamp, supplied with the same regulator/ldo and with tolerances, meaning you have for example LDO -> 11uf || 9uf || 10.5uf || 8uf || 49nf || 52nf || 59nf || 60nf || 16nf || 14nf || 17nF || 15nf combined with ESR for each respective cap and cap type being slightly different too, you can see how the 'DC' nolonger is DC (it never is, there is always some AC/ripple). when placing caps of different types in parallel, best to keep them at least a decade apart in value (as a rule of thumb) 15nf and 56nf is about the worst, not close enough to be the same, but not far enough apart to avoid issues, particularly with cap types with high q, like film and c0g. it is not so problematic with lossy cap types (low q), such as x7r, but still best avoided. parallel same value, or far apart.

One last thing, 600Ω headphone loads require less current (but more voltage) than 32Ω, not more. Adding a parallel 1622 helps with current capability, not voltage.

 

[Mr.HiAudio]

Does anyone know what capactior can be use for (5) ? Can i use just ceramic(not tantal) capactior 220uF 6.3 v?

 

[Inspector Gadget]

Those are for filtering the 3.3VDC AVCC left and right voltage reference pins on the 9038. they are the most important caps on the board for objective performance (SINAD) and they are on the output of a regulator. I would stick with tant polymer, but doesnt have to be AVX. I would definitely avoid ceramic. not only would it be difficult to find a 220uf ceramic in that size, if you did, it would have to be something that has terrible temp coefficient spec, like x5 or y5. these are effectively in the signal path, as they supply the voltage reference the dac uses to turn code into output.

 

[Mr.HiAudio]

Those are for filtering the 3.3VDC AVCC left and right voltage reference pins on the 9038. they are the most important caps on the board for objective performance (SINAD) and they are on the output of a regulator. I would stick with tant polymer, but doesnt have to be AVX. I would definitely avoid ceramic. not only would it be difficult to find a 220uf ceramic in that size, if you did, it would have to be something that has terrible temp coefficient spec, like x5 or y5. these are effectively in the signal path, as they supply the voltage reference the dac uses to turn code into output.

thank you so much

 

[Inspector Gadget]

No problems. Give me a few, i'll find a couple options on mouser for you.

 

[Inspector Gadget]

Do you have any calipers by chance? they look like 1411 size, but I cant be sure.

https://au.mouser.com/ProductDetail...GAEpiMZZMsh%2B1woXyUXj1Xm5b4oEbJgfHyIq1246v0=

or

https://au.mouser.com/ProductDetail...GAEpiMZZMsh%2B1woXyUXj9v/LaJorgg0zs02G26XeYA=

will work just fine IMO

 

[Mr.HiAudio]

Do you have any calipers by chance? they look like 1411 size, but I cant be sure.

https://au.mouser.com/ProductDetail/ROHM-Semiconductor/TCOB0J227M8R-EN1?qs=sGAEpiMZZMsh%2B1woXyUXj1Xm5b4oEbJgfHyIq1246v0=

or

https://au.mouser.com/ProductDetail/Panasonic/6TPE220MVB?qs=sGAEpiMZZMsh%2B1woXyUXj9v/LaJorgg0zs02G26XeYA=

will work just fine IMO

Im already find tantal in my country but size will be B

 

[Inspector Gadget]

which tantal? tantal polymer? there are fairly large performance differences between tantalum caps. I generally stick with panasonic SP, rubicon, or ROHM. ok yes, B or 1411 (imperial size code) as I thought.

 

[Mr.HiAudio]

which tantal? tantal polymer? there are fairly large performance differences between tantalum caps. I generally stick with panasonic SP, rubicon, or ROHM. ok yes, B or 1411 (imperial size code) as I thought.

this: https://ru.mouser.com/ProductDetail/AVX/TCJB227M006R0070?qs=BZqFI1yMojzD%2BJPU1LyuWw==

 

[Setmagic]

OK, so dont take me the wrong way, you are clearly putting in hard work here for the community, but I think there are some things you could do to help yourself and others.

1. Spend some time drawing out a schematic for the analogue stages (this is a very simple application and not multilayer PCB, so should not be too difficult) then spend some time understanding the circuit. I dont expect you to reverse engineer the whole circuit, but if you want to improve a circuit, it is best to know what the circuit does and why values were chosen in that circuit.
2. Understand where the limitations are and focus mods on those areas.
3. One mod at a time (or at least one section). this allows easier troubleshooting if something goes wrong and also easier to determine if you are making an improvement and if so, what made the improvement.
4. power supply decoupling/bypass caps. more capacity does not always = better smoothing, especially if placing many caps in parallel.

Hello,
Thank You for this post. I'm learning on those Zishans players, because I find it more difficult and fascinating than full sized DIY "hi-end like" projects.
I have done more than 20 separate mini-mods attempts for research on Z4 I hope will survive next few ones
This is the lasts summary.
Answers:

1. Yes, this will be helpful and quite challenging to make it.
2. So far I based mods on work of experienced ones from this forum and Russian Player.ru who made mods for DSDs. I read evaluation boards and recommended values from developers. I try to find as many applications and projects from DIY community as i can. I'm rolling these applications and try to find the sweet spot.
3. I follow that way. So far only 2 times bricked Z4 who was repaired in easy way
4. I find to follow implementation scheme and use 1uF MLCC X7R (stable) or more in that case i put 2,2uF + 100nF is best so far. 48uF X5R works as well, but I can't tell it was better, way more unnecessary inductance for sure, so I decide to follow instructions from Texas Instruments.

also npo/c0g, while good (one of the best) if you know exactly what target frequency you are trying to attenuate and have the means to measure the result, but it is almost guaranteed to cause issues you can't measure if you don't. they have very high q and thus can easily cause resonances when they interact with other capacitors (and their parasitic secondary effects (mostly the inductance)). Films are even worse, so when you place—for example—15nF c0g || 56nf film (very close, but not the same value) and then do that on a rail of each opamp, supplied with the same regulator/ldo and with tolerances, meaning you have for example LDO -> 11uf || 9uf || 10.5uf || 8uf || 49nf || 52nf || 59nf || 60nf || 16nf || 14nf || 17nF || 15nf combined with ESR for each respective cap and cap type being slightly different too, you can see how the 'DC' nolonger is DC (it never is, there is always some AC/ripple). when placing caps of different types in parallel, best to keep them at least a decade apart in value (as a rule of thumb) 15nf and 56nf is about the worst, not close enough to be the same, but not far enough apart to avoid issues, particularly with cap types with high q, like film and c0g. it is not so problematic with lossy cap types (low q), such as x7r, but still best avoided. parallel same value, or far apart.

That was my first attempt to set up values of capacitance, and now I'm not using 15nF in that way- gives nothing visible but can cause problems.
At this moment, it looks like that:
- LPF filter decoupling 10uF+1uF X7R +100nF C0G
- AMP 10uF electrolytic low ESL/ESR +10uF X7R + 100nF C0G and testing this film 56nF, but I think I will pass due to no changes.
- added 1000uF electrolytic to this original 1900uF pool of X5Y one's by manufacture, no visible improvements sadly, but it will for now.

One last thing, 600Ω headphone loads require less current (but more voltage) than 32Ω, not more. Adding a parallel 1622 helps with current capability, not voltage.

In that case, it's balanced amp so i its add more voltage (I hope) as well 4Vrms in SE and 8Vrms in BAL, official specification from Zishan factory.

 

[Inspector Gadget]

No. balanced or single ended, if you parallel a chip, it allows higher current output, not higher voltage. If you put 2 amps in series (bridged, or balanced) that allows for double the voltage swing, but the headphone load appears halved to each amp. a balanced, or bridged amp, is effectively 2 amps placed in series with the load (one for + and one for -). 1 x 1622, or 2 x 1622 (in parallel) produces the same voltage. +/-9vdc power rails on the 1622 isnt exactly a lot of swing, but it should be enough for 600Ω. yes, the balanced swing here is double the SE output, because the SE output only uses half the amp plus GND, rather than adding a subsequent summing stage.

so for example, if a balanced amp is able to swing from -7V -> +7 (allowing for some losses on +/-9V rails) that is 14V, but +7v -> GND is only 7V

 

[Setmagic]

No. balanced or single ended, if you parallel a chip, it allows higher current output, not higher voltage. If you put 2 amps in series (bridged, or balanced) that allows for double the voltage swing, but the headphone load appears halved to each amp. a balanced, or bridged amp, is effectively 2 amps placed in series with the load (one for + and one for -). 1 x 1622, or 2 x 1622 (in parallel) produces the same voltage. +/-9vdc power rails on the 1622 isnt exactly a lot of swing, but it should be enough for 600Ω. yes, the balanced swing here is double the SE output, because the SE output only uses half the amp plus GND, rather than adding a subsequent summing stage.

so for example, if a balanced amp is able to swing from -7V -> +7 (allowing for some losses on +/-9V rails) that is 14V, but +7v -> GND is only 7V

How about design of balanced amplifier?
I saw in Z4 1 LPF opamp have R+/L+ and second one have R-/L- in balanced mode (4.4/2.5 and 3.5 is working as well we can listen music from 3 pairs of headphones (amplified) technically even from 4 (+1 line-out). In single ended, half of rail is working 1xDAC-LPF-AMP balanced outputs are off.
It should be different - like 2 audio rails (DAC-LPF-AMP) first +/- Right and second +/ Left?

 

[Setmagic]

I made some updates in Z4 mod
An electrolytic 1000uF capacitor in analog 9V can't change sound, so it was removed.
Additional 56nF in parallel of 100nF change nothing so was removed as well.
Testing Elna 10uF in 1622.

1. Best improvements was changing capacitors in audio rail near op amps
2. LPF OP275 replaced AD8620 or OPA1656. OPA1656 is on the lid right now.
3. LDO for TI5907 plus recommended ~2,2uF+100nF - clean sound.
4. Resistors in audio rail for Panasonic thin-film 0.1% instead of thic-film probably >2%
5. Changing capacitors for X7R and additional 100nF G0C in analog supply part improve smoothness.
6. DAC section additional 4x220uF tantal-polimer and replaced 8x 10uF X7R +100nF G0C cleaner sound as well.