Review and Measurements of Topping DX3Pro DAC and Headphone Amp

[eziitis]

The major issue I can see is those DIP OPs. SOIC is much better.

just wonder, what is the SOIC superiority audio wise?

 

[eziitis]

What kind of piece is this ? Perhaps voltage regulator, from 5,3v to 5.0v ?

Huston, we have a problem! to me that 5V testpoint rings to VREF and VDD and AVDD pins. would that mean there is just one reg for all these +5V? it is a must to have VREF innocently clean for these AK beauties. if that is the case, then shame on you Topping!

basically this means we are facing VREF pin lifting and putting a buffer something like below to make VREF happy:

 

[amirm]

Those curves are valid measurement data.

Of what? They show a THD of -60 dB. What DAC do you know that has that performance? DX3 Pro clocks at -105 dB. No way what you are reading there is a typical measurement or circuit.

 

[finneybear]

No, I meant capacitors for power OPA1612.

In this case, you need 16V or 25V cap. 10uf is fine here. If you want something better, you can use solid polymers, as what I had done.

 

[finneybear]

just wonder, what is the SOIC superiority audio wise?

signal integrity wise, it is much better. shorter feedback path as there are a lot of HF signals going through.
Go ahead check out the layout guidelines in those high performance OPs datasheets

 

[finneybear]

Huston, we have a problem! to me that 5V testpoint rings to VREF and VDD and AVDD pins. would that mean there is just one reg for all these +5V? it is a must to have VREF innocently clean for these AK beauties. if that is the case, then shame on you Topping!

Yep, the 100uf value used in DX3 is bad, too. 4493 is suffering with all the ripples.

 

[finneybear]

Of what? They show a THD of -60 dB. What DAC do you know that has that performance? DX3 Pro clocks at -105 dB. No way what you are reading there is a typical measurement or circuit.

The paper stated that there's a 80.5db total gain. You have to deduct the values by 80.5db. The general shape of the curves look right to me. Have seen those in too many reports.

 

[amirm]

The paper stated that there's a 80.5db total gain. You have to deduct the values by 80.5db. The general shape of the curves look right to me. Have seen those in too many reports.

In which case the differences are -140 dB down. In other words, it is beyond the realm of audibility. On that front, I suggest reading this article I wrote: https://www.audiosciencereview.com/...-digest-audio-capacitors-myth-or-reality.176/

 

[finneybear]

In which case the differences are -140 dB down. In other words, it is beyond the realm of audibility. On that front, I suggest reading this article I wrote: https://www.audiosciencereview.com/...-digest-audio-capacitors-myth-or-reality.176/

It is not the amount of THD, it is the shape of the curve. After signal going through the cap, the distortion amplitude changes among different frequencies. The impedance curve is another important factor, too. Then after the cap, if the signal gets amplification, the THD will go up again. This THD is just a component of the signal chain, not the final THD number of the whole circuit. From the curve we can see coupling cap would degrade the low frequency signal quality more. How serious the problem is, all depends on what's at the next stage. If the cap is for the final output, things may be OK. But as the output coupling cap for DX3's 4493 output, it can be a problem.

 

[ChuckT]

They are not valid data when you cut off even the label under the graph let alone the link to the paper. The piezoelectric effects are known and nothing new. The focus of the paper is also on acoustic interference from what I can gather. It certainly has no bearing on headphone listening.

If all the caps are tested in exactly the same way and shows a difference. Don't you think there is really a difference between the caps?

 

[amirm]

But as the output coupling cap for DX3's 4493 output, it can be a problem.

What problem? THD+N is -106 dB. 2nd harmonic distortion is at -110 dB.

 

[amirm]

If all the caps are tested in exactly the same way and shows a difference. Don't you think there is really a difference between the caps?

The only ones that showed a "problem" were those affected by piezoelectric effect. There is no such factor here.

 

[777]

Huston, we have a problem! to me that 5V testpoint rings to VREF and VDD and AVDD pins. would that mean there is just one reg for all these +5V? it is a must to have VREF innocently clean for these AK beauties. if that is the case, then shame on you Topping!

View attachment 23082
basically this means we are facing VREF pin lifting and putting a buffer something like below to make VREF happy:

View attachment 23083

The same LDO for all 5v of the one AK4493. To the other AK 4493 5v, there is another small LDO. The ground is common for all the voltages, digital and analog, without any isolator.
For 3.3v there is only one regulator for both AK4493's. The same 3,3v is split in two rails trough the ferrites, one for the analog 3.3v and one for the digital 3.3v.

Upstream, there is a 5.3v regulator(tps54331) which split the output for AK4493's 5v and 3.3v LDO's and LDO's for signal inputs and digital processing.

For the price of the unit everything it's more than ok but as you can see, we did not talking about highend implementation.

 

[777]

The only ones that showed a "problem" were those affected by piezoelectric effect. There is no such factor here.

I was remove the capacitors from the signal path and use only one AK4493. For me, the sound is the same, probably the measurements are the same too or almost there. In my opinion, like Amirm one, those caps doesn't represent a problem.

 

[noobie]

Ummm... at the risk of putting my foot in it, can I, rather timidly, once again as I did a few pages back ask @finneybear and others modding their units to send them to @amirm for measurement and comparison to stock units? Brighter minds than mine can then debate the findings, but can there first be findings to debate?

Also, two points if I can dip my little toe into the debate...

1. This is a forum that is founded on measurement. There actually has been a very tactful avoidance of the discussion as to whether the measurable differences are actually audible. I don’t believe this forum was established to deal with that issue. It is about identifying how well a product was engineered - and no product is perfect. There are other forums such as SBAF and Head-fi for more subjective evaluations. Folks shouldn’t be surprised when “measurement” keeps coming up. When you are in Church, God comes up, when you are in a bar, whiskey comes up. Comes with the territory.

2. I think that there is no doubt that better components and cleaner engineering would result in a better product than the DX3PRO... it is a product with a budget trying to hit a particular set of marketing and engineering constraints, and even then mistakes and sub-optimizations occur... Many, many posts back @finneybear said his changes to date would result in a $500 product that would compete well with $1000-$2000 products. I would think everyone would grant that makes sense in concept, add $300 underlying value to a $200 product, and it probably does get better! But, there is a desire to actually measure the results of those modifications.

So, going back to my original suggestion... can some modified units be provided to @amirm for measurement so we can have the discussion about the resulting measurements?

As I tell my wife all the time, sometimes people fall into the trap of talking about the talking rather than talking about the issues...

 

[eziitis]

signal integrity wise, it is much better. shorter feedback path as there are a lot of HF signals going through.
Go ahead check out the layout guidelines in those high performance OPs datasheets

sorry, what kind of HF signals we are talking about in an audio application? I would kindly ask to stop confusing people knowing even less that use of DIP8 opamp package in an audio application is a major design fault. just opposite, it may have even some benefits as it allows more solid power traces.

 

[ChuckT]

Thanks for the link to Murata part. Have you ever tried it? It can only do 1w though.

Some SOIC/DIP adapters would put silver mica across the +- lines but this may not work all the time.
Another concern is the feedback loop. The adapter will pull the line too long. I like SMD parts much better. Easier to work with, too.

This weekend I will replace the XOs. Probably add a 1.5A 15V linear power supply later then that will be it. I will be done with DX3. This thing already sounds much better than most $1k DACs. The machine has been running non-stop for over 60 hours now. Solid stable. The sound keeps getting better and better. Warm and smooth. Lovely vocals. The biggest steal in 2019!

I have read some early feedback on Topping's D70, pretty much in line with what I heard from DX3.
After looking into DX7, most likely I will skip D70 - I see the same bad 100uf SMD caps again.

This one seems to be a good candidate. The implementation looks good, at least it follows whatever 4497 datasheet asks for, the four big caps around 4497 are good. The output stage looks correct. The major issue I can see is those DIP OPs. SOIC is much better.

Those dc-dc parts also has 2w and 3w version. I have a similar Chinese parts A1212s inside a Little Bear B4-X tube-opamp headphone convert battery 12v to +/- 12v. It uses 2 of those powering a 5899 tube and opamp for left right channel and it sound pretty good (the heater was power by 12v battery direct). So I will use those in the future whenever I need to have single rail to +/- rail. BTW, those are isolated converter meaning the ground from input and output are not connected/isolated.

The linear supply will certainly improve the DX3. I used a 19v linear on the DX3 and the overall sound improved compare to the original wall wart.

Funny I was looking at the JF Digital ak4497 dac too and it was my top choice if I were to get another akm dac. The overall power supply seems quit decent compared to other dac internal that I have seem. Since I now have an easy way to get +/- rail from single supply, I am thinking of battery power the whole DX3 dac with 12.8v lifepo4 (4 of those)

For me, in general, digital stuff I prefer small and analog stuff I prefer larger.

 

[ChuckT]

The only ones that showed a "problem" were those affected by piezoelectric effect. There is no such factor here.

Was the test done while knocking on the ceramic caps? If not, then I think the test is valid as piezoelectric effect clearly exist.

 

[Krunok]

I wonder what is in this thread that attracts so much BS..

 

[eziitis]

Ummm... at the risk of putting my foot in it, can I, rather timidly, once again as I did a few pages back ask @finneybear and others modding their units to send them to @amirm for measurement and comparison to stock units? Brighter minds than mine can then debate the findings, but can there first be findings to debate?

Also, two points if I can dip my little toe into the debate...

1. This is a forum that is founded on measurement. There actually has been a very tactful avoidance of the discussion as to whether the measurable differences are actually audible. I don’t believe this forum was established to deal with that issue. It is about identifying how well a product was engineered - and no product is perfect. There are other forums such as SBAF and Head-fi for more subjective evaluations. Folks shouldn’t be surprised when “measurement” keeps coming up. When you are in Church, God comes up, when you are in a bar, whiskey comes up. Comes with the territory.

2. I think that there is no doubt that better components and cleaner engineering would result in a better product than the DX3PRO... it is a product with a budget trying to hit a particular set of marketing and engineering constraints, and even then mistakes and sub-optimizations occur... Many, many posts back @finneybear said his changes to date would result in a $500 product that would compete well with $1000-$2000 products. I would think everyone would grant that makes sense in concept, add $300 underlying value to a $200 product, and it probably does get better! But, there is a desire to actually measure the results of those modifications.

So, going back to my original suggestion... can some modified units be provided to @amirm for measurement so we can have the discussion about the resulting measurements?

As I tell my wife all the time, sometimes people fall into the trap of talking about the talking rather than talking about the issues...

good points indeed unfortunately sending mod unit elsewhere for measurements is absolutely impractical for me.
so if the thread is about measurements, and the DAC performance as it is in original state has been measured in very beginning, what more we can measure?
in order to come up with sensible mods one should identify critical points, and at the moment it appears VREF to me. so I am putting back those
TPS54331 switchers, what I prematurely removed in favour to linear PS and will try to measure what noise we have on +5VL/R test points. I would say first one should look for presence of switching frequency and signal frequency. upon the extent of contamination one can decide whether and how things could be changed. anyone with a decent scope willing to join this measurement?

kind of OT, but it would be very helpful if Topping would provide schematics. such a request is not so silly as it may sound. for instance Khadas tone board has all schematics public, and I can say from the point of EE they have done very good job. yes there is a switcher for opamp power as well, but it is not so critical, because opamps have high PSRR. another matter is impedance of course .... nevertheless the critical AVCC comes from a dedicated reg and is pretty clean in fact.