Review and Measurements of Topping DX3Pro DAC and Headphone Amp

[Killingbeans]

Ahhh okay, I begin to see the grand scheme here. Just keep doubling the bitrate in each new chip generation until we hit the bandwith limit of USB 2.0 And then we can start to push $100K USB 3.0 cables on the market as replacement for those pathetic old $10K USB 2.0 cables

...Sorry for going off topic

 

[bravomail]

1. Remove the main power supply cap, the one right after the power supply plug socket. Use a 330uf/25V or higher good solid polymer cap..

It is the stuff like this you say, funnybear, will make you call out as "BS maker". The material from which capacitor made will make absolutely no difference. Repeat it until u remember. It is the capacity and rated voltage that will make any difference. And you call yourself knowledgeable. Your made up facts show yourself in your full glory.

 

[777]

It is the stuff like this you say, funnybear, will make you call out as "BS maker". The material from which capacitor made will make absolutely no difference. Repeat it until u remember. It is the capacity and rated voltage that will make any difference. And you call yourself knowledgeable. Your made up facts show yourself in your full glory.

Are you sure ?

 

[splkn]

It is the stuff like this you say, funnybear, will make you call out as "BS maker". The material from which capacitor made will make absolutely no difference. Repeat it until u remember. It is the capacity and rated voltage that will make any difference. And you call yourself knowledgeable. Your made up facts show yourself in your full glory.

Finneybear wrote it right. The type of capacitor is important because our power supply produces high-frequency noise, a capacitor with a low ESR is needed.

 

[freaksamael]

It is the stuff like this you say, funnybear, will make you call out as "BS maker". The material from which capacitor made will make absolutely no difference. Repeat it until u remember. It is the capacity and rated voltage that will make any difference. And you call yourself knowledgeable. Your made up facts show yourself in your full glory.

Well, good to know! Why should I pay almost 3 EUR per one Simlic instead make myself a limited series from chocholate foil with cornflakes as dielectric? Also I could choose a proper bright colour for them...Very informative, indeed...

 

[simonchretien]

Go to Page 150, post #2,996 for instance. Early post also complained about unbearable HF shrills, lack of mid bass, etc.

Well, you call this DIY crap when it solves the DX3 stability problem?

Thats what I said, this is only related to line out, which most people dont care about, while others pointed out this is BS. I cant say, I havent tried line out. But stop saying generic comments like "the dx3 sounds so bad people use it as paperweight", really...

 

[simonchretien]

Well, good to know! Why should I pay almost 3 EUR per one Simlic instead make myself a limited series from chocholate foil with cornflakes as dielectric? Also I could choose a proper bright colour for them...Very informative, indeed...

Well, it is said that corn flakes make full bodied bass and remove the harshness of middle-aged lesbian female singers.
Or you could try rice krispies if you prefer warmer mids.

 

[finneybear]

I can not find a 100uf 16v capacitor with a size of 3.2 mm.

You meant the 4493 output coupling cap, size 3216? You only need voltage rating 6.3v there. 47uf nobium oxide is the largest I can find. Dont use tantalum there. If you want strong full bass, no coupling will be the way to go.

 

[finneybear]

If it was economically feasible for bridge engineers to do multiple full scale prototyping before final implementation, I'm pretty sure it would be considered a felony if they didn't.

I never said that the theoretical design phase is unimportant. I'm just saying that testing and measurements are equally important. A prototype is riddled with uncertainties, no matter how many calculations have been done on its design. Measurements are a great tool for lowering those uncertainties. "Am I really hearing a difference or am I most likely imagining things?" You can't make a bridge stronger via confirmation bias, but you can easily get that illusion going in an audio product.

Actually it's not prototype. It's more like scale model built for experiment and validation. The tests done are far from being called real measurements. Nowadays the models are built in computer. Software is used to do simulation and finite element analysis, etc. Circuit and chip designs are done through simulations, too. You can get fairly accurate estimate on final measurement numbers through simulation.

As I have said, I can see that the measurement on the fixed DX3 will have a bit higher THD with much less noise spikes. If you only look at the new 1K FFT chart, the result will be in-conclusive. No way you can figure out the drastic change on sound quality. If you have access to 400Hz and 100Hz FFTs and know how to interpret the different distribution of harmonics then probably you will have some clues. Again, the basic measurements we have here are solid indicators on performance yet far from telling you the whole story.

Japan built a lot of super low distortion amps in 80s. Did majority of them sound any good? Another typical example is Halcro. They built machines with the lowest distortion numbers through RF amp and feed forward design technique. Had they ever taken the whole audio industry by storm? The reality is they almost went under and got acquired by another company, and now becomes relatively irrelevant. I never like their amps.

 

[finneybear]

Thanks for the detail description.
There are smd to dip socket. You can replace the smd opamp with the socket and play with swaping dip opamps.
You can also get DC-DC isolated converter similar to below to have the input 15V converted into +/-15V instead of the existing +/-10V using the tps54331. The higher voltage should improve the drive for the opamp output.

https://www.mornsun-power.com/uploads/pdf/A_S-1WR2.pdf

Thanks for the tips. I have the tweezer to remove SOIC OPs easily so directly swapping SOIC OPs is not an issue. Less issues with inductance, power decoupling, etc. The decoupling caps need to stay as close to the OP as possible.

The +-10V is also used for the headphone amp. Moving it up to 15V needs to redo calculation on the buffer section. This board is so small. There is just this much you can do. This is why I am looking for a decent 4497 DAC now. 4493 sounds so warm and natural when ESS DAC tends to be too artificial.

The next major thing is to use better XOs. XMOS chip has a fairly inconsistent delay time on data outputs. The I2S re-clocking circuit is crucial.

 

[finneybear]

Just keep in mind that it will make the tarces to the decoupling capacitors substantially longer. It could cause high bandwith chips to become unstable. At least in theory

This is correct.

 

[finneybear]

Given that you personally can hear up to 384kHz and that ther's so many music material originally recorded at 768kHz that is a real shame.

I am aligned with what most recording engineers believe. 48KHz/24bit is all what you need. For classical music, probably you can go up to 96KHz/24bit.

The catch here is that most DACs now upsample the data to 192KHz internally. A lot of them will do PCM to DSD conversion internally.
Feeding DAC with 768K or DSD512 data will bypass most of those upsampling functions inside the chip.
Usually sophisticated computer software can do a better job here. Sure, if you are willing to pay big $$$, machines like Chord or dCS will have powerful on board DSP to do a job equally as good.

How about PCM 1536K or DSD1024? The problem is the clock speed now is very high. A good low phase XO will be very expensive to get.
It's also a problem with ESS DACs which use 100MHz for the DPLL function. DSD1024 will pretty much render the 100MHz DPLL ineffective.

4493's on board DSP is fairly good. It does very well with CD 44.1K data.

 

[finneybear]

It is the stuff like this you say, funnybear, will make you call out as "BS maker". The material from which capacitor made will make absolutely no difference. Repeat it until u remember. It is the capacity and rated voltage that will make any difference. And you call yourself knowledgeable. Your made up facts show yourself in your full glory.

Mmm.... since we all talk about measurements here. How about some measurement numbers for you?
Just wondering whether you can read those curves properly?

 

[finneybear]

Well, good to know! Why should I pay almost 3 EUR per one Simlic instead make myself a limited series from chocholate foil with cornflakes as dielectric? Also I could choose a proper bright colour for them...Very informative, indeed...

47uf/25v Silmic II is only USD 68 cents here.

I have measured the stock caps at 120Hz, 1KHz, etc, the numbers are terrible. If you are fine with those caps then go ahead use them and be happy. Just kidding.

 

[finneybear]

Thats what I said, this is only related to line out, which most people dont care about, while others pointed out this is BS. I cant say, I havent tried line out. But stop saying generic comments like "the dx3 sounds so bad people use it as paperweight", really...

Didnt you just make a generic statement that it's only a problem for line out? Surely a lot of people care about the line output. I am using DX3 to feed a separate headphone amp.

Inside DX3. the line out pretty much feeds the headphone amp section directly. There is no such things as the line out is bad yet the headphone amp outputs are good. It's all from the same line out.

Another interesting thing is the coupling caps in the headphone amp section are very bad. Whether you can hear the problem or measure the problem is entirely another topic to discuss.

 

[amirm]

Mmm.... since we all talk about measurements here. How about some measurement numbers for you?
Just wondering whether you can read those curves properly?

View attachment 23022

I do. How about you? Have you read the paper this came from? And the reason those two graphs are higher than the rest???

 

[MusicNBeer]

This is ridiculous. So all these wonderful DIY mods are being done without even measuring (quantifying) the changes in performance? And it's also assumed if any change is measurable, it's automatically better? I'm seeing the ugly audiophile snake oil tendencies rear their ugly heads.

By the way, I think the stock line out and headphone outs both sound fantastic. No weakness whatsoever.

 

[ChuckT]

Thanks for the tips. I have the tweezer to remove SOIC OPs easily so directly swapping SOIC OPs is not an issue. Less issues with inductance, power decoupling, etc. The decoupling caps need to stay as close to the OP as possible.

The +-10V is also used for the headphone amp. Moving it up to 15V needs to redo calculation on the buffer section. This board is so small. There is just this much you can do. This is why I am looking for a decent 4497 DAC now. 4493 sounds so warm and natural when ESS DAC tends to be too artificial.

The next major thing is to use better XOs. XMOS chip has a fairly inconsistent delay time on data outputs. The I2S re-clocking circuit is crucial.

The dc-dc module is actually smaller than the space for the 2 tps dc switcher. This is a higher quality one from Murata in mouser.

https://www.mouser.com/ProductDetai...=sGAEpiMZZMvGsmoEFRKS8Kn8hpQIdAem/1LMX04mJrk=

I have seem a dip to smd socket with cog cap solder accross the +/- pin. This is one of the opamp tweaks that improve the sound.
If I do this, I would use a audio ecap and solder one of the leg directly on the socket and the other to the board.

Topping had an initial offer of the D70 but ended without further news.
https://www.aliexpress.com/item/Top...-Decoder-December-20-Arrival/32959667114.html

The ak4497 would be more difficult to implement correctly and more demanding on the power supply.

You can play with battery power directly to the digital circuit. Trust me, it will be a revelation. The lifepo4 battery output 3.2-3v and you can easily tap it directly. Just using the battery on the clock will bring improvement.

 

[eziitis]

so the first approach to the dx3pro:
- switching regs gone, linear PS instead.
- 2200uF on top for each Vref

let see how it sounds when fed from audirvana

 

[eziitis]

it sounds pretty good indeed.... not matching yet my heavy es9028pro box, but... velvet sound you know
since there is no way how to objectively forward listening impressions (whatever you think about that), i took the liberty to measure 192/24 1khz sinewave, which should show the things if done right.

next SOX upsampled to PCM768

and now upsampled to DSD128

interesting, right? you almost can hear these spectrograms when back in the box it will be a little better. next thing to do would be to put a decent output stage in a piggyback fashion instead that poor opa1612 ... and more advanced power rearrangement perhaps