Do you need linear power supply for DACs?

[Miska]

The tests were done with the same SoX DSD signal with the same modulator settings, so all differences were originated from the DAC.


All RMAA tests were done with the ADI-2 ADC, so with this factor being constant, all differences were originated from the DACs under tests.

You just don't seem to understand how the DAC + ADC combination affects the results. But I don't have motivation to continue arguing about this.

I was making measurement systems and passive sonar systems already in mid 90's, I do have some sort of idea how these things work.

 

[Miska]

One thing to pay attention to when measuring ESS DACs is whether they run in synchronous mode (two separate clocks) or asynchronous mode (from single 100 MHz clock). There are both kinds of implementations out there.

 

[bennetng]

You just don't seem to understand how the DAC + ADC combination affects the results. But I don't have motivation to continue arguing about this.

I was making measurement systems and passive sonar systems already in mid 90's, I do have some sort of idea how these things work.

So simply ran out of arguments to continue. You mentioned aliasing first but it turned out that higher DSD rates for ES9038 actually introduced more noise, instead of higher DSD rates pushed the noise further and get filtered at analog output, and if it is related to different clocking modes then it is a hardware issue and cannot be solved by software upsampling.

 

[Miska]

So simply ran out of arguments to continue. You mentioned aliasing first but it turned out that higher DSD rates for ES9038 actually introduced more noise, instead of higher DSD rates pushed the noise further and get filtered at analog output

No, I just don't want to keep repeating same thing over and over. I have shown that noise floor is 0.1 dB lower with DSD source on the Pro-Ject which is synchronously clocked ES9038Q2M. But of course this is with my modulators, there could be some differences with other modulators.

And the "higher noise" in <20 kHz is just aliasing from the ADC depending on how the noise slopes hit the weak spots in anti-alias filtering. Especially evident in the results from ADI-2 Pro ADC.

and if it is related to different clocking modes then it is a hardware issue and cannot be solved by software upsampling.

Sure, could be some fault in Topping DAC. Such as IMD from it's analog stages. I don't own any of those, nor have much interest to.

OTOH, small difference in noise floor is less of an issue compared to other bigger issues with PCM inputs.

 

[bennetng]

No, I just don't want to keep repeating same thing over and over. I have shown that noise floor is 0.1 dB lower with DSD source on the Pro-Ject which is synchronously clocked ES9038Q2M. But of course this is with my modulators, there could be some differences with other modulators.

And the "higher noise" in <20 kHz is just aliasing from the ADC depending on how the noise slopes hit the weak spots in anti-alias filtering. Especially evident in the results from ADI-2 Pro ADC.

When measuring other AK4493 DACs with the ADI-2 Pro ADC, Topping DX3 Pro showed nearly no differences in different DSD rates, and in PCM.

Earlier tests with JRiver DSD resampling:

MEASUREMENTS: Topping DX3 PRO V2 (LDAC version) - plus frequency/amplitude stepped sine measurements, a quick look at the 8kHz USB PHY packet noise, and AirPods Pro "Spatial Audio" coming...

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

Recent tests with SoX DSD resampling:

MEASUREMENTS: A look at DSD and using SoX-DSD as a standard for test signal conversion. The 1/10-decade Multitone 32 test. And retro-pop remixes... ;-)

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

So if there is a "weak point", the weak point is not on the AK557x ADC in the ADI-2 Pro ADC.

RME DAC measured with another RME ADC, DSD Direct, 150kHz filter:

...50kHz filter. So, with filter settings remain the same, there is no difference in different DSD rates.

Sure, could be some fault in Topping DAC. Such as IMD from it's analog stages. I don't own any of those, nor have much interest to.

The same happens on his Oppo UDP-205 as well, so not a Topping specific issue. It's simply the Pro-Ject is unable to adequately utilize the PCM potential of the ESS chips, which clearly lags behind the Topping D10B in Amir's reviews.

 

[Miska]

When measuring other AK4493 DACs with the ADI-2 Pro ADC, Topping DX3 Pro showed nearly no differences in different DSD rates, and in PCM.

So if there is a "weak point", the weak point is not on the AK557x ADC in the ADI-2 Pro ADC.

That is because of the differences in DSD D/A conversion between AKM and ESS chips and how that matches the weak point (evident from the chip data sheets).

The same happens on his Oppo UDP-205 as well, so not a Topping specific issue. It's simply the Pro-Ject is unable to adequately utilize the PCM potential of the ESS chips, which clearly lags behind the Topping D10B in Amir's reviews.

It happens with the Oppo too because it has the same D/A conversion. The fallacy is that you think you are looking at the DAC performance while in fact you are more testing the ADC performance with two different input signals.

 

[bennetng]

I edited my previous post to show the RME DAC with adjustable DSD Direct filters which is not available on Topping DX3 Pro. Clearly, the only difference is the 150kHz filter resulted in more noise, but the noise remain the same with different DSD rates. So "weak point" makes no sense.

 

[Miska]

I edited my previous post to show the RME DAC with adjustable DSD Direct filters which is not available on Topping DX3 Pro. Clearly, the only difference is the 150kHz filter resulted in more noise, but the noise remain the same with different DSD rates. So "weak point" makes no sense.

What does that have to do with ESS DAC combined with ADI-2 Pro's ADC?

Do you have any data of your own, or just repeat what other people do?

 

[bennetng]

What does that have to do with ESS DAC combined with ADI-2 Pro's ADC?

Read my post carefully. Both Topping DX3 Pro and RME ADI-2 series are using AK4493 DACs.

 

[Miska]

Read my post carefully. Both Topping DX3 Pro and RME ADI-2 series are using AK4493 DACs.

So!? Irrelevant about using ESS DAC with ADI-2 Pro ADC or some other ADC for measurements.

I know ADI-2 pretty well, I have two ADI-2 Pro's. And my software supports also recording DSD256 with it.

 

[bennetng]

So!?

Topping D10B (ES9038Q2M), Oppo UDP-205 and Topping D90SE (ES9038Pro) showed more noise in higher DSD rates in the same tests.

 

[Miska]

Topping D10B (ES9038Q2M), Oppo UDP-205 and Topping D90SE (ES9038Pro) showed more noise in higher DSD rates in the same tests.

So? I said that is because differences in D/A conversion makes the ADC used for measurements perform differently. Not because there would be actual noise level differences in the DAC. But because there are noise level differences in the ADC due to different kind of input stimulus.

 

[bennetng]

So? I said that is because differences in D/A conversion makes the ADC used for measurements perform differently. Not because there would be actual noise level differences in the DAC. But because there are noise level differences in the ADC due to different kind of input stimulus.

Did you see the RME DAC DSD Direct tests with 50kHz and 150kHz filter? Even allowing much more noise to induce more aliasing, the differences among 50kHz and 150kHz and PCM is only about 4dB.

Topping D90SE for example showed more than 10dB of differences. Therefore the claim about ADC filter weak point makes no sense.

 

[Miska]

Did you see the RME DAC DSD Direct tests with 50kHz and 150kHz filter? Even allowing much more noise to induce more aliasing, the differences among 50kHz and 150kHz and PCM is only about 4dB.

That is still irrelevant as it is different kind of noise profile. First warning sign should be already if you get lower noise results for DSD64 compared to higher rates.

The AKM ADC chip digital decimation filter converting SDM from the converter to PCM output has only 85 dB attenuation. In addition, typical analog anti-alias filter is 2nd order. That combined to how skirting in the sampling and decimation works causes such differences.

For similar reasons, even devices like AP have a separate external analog filter for measuring things like class-D amplifiers. Because otherwise you'd get incorrect results. (since I work also on direct digital conversion amps, I have such filter too)

Here's the Pro-Ject narrow band noise floor with DSD512:

And here for 768k PCM input:

You can see how it's internal modulator noise floor behaves. Here too, the 20 kHz noise level is slightly lower with DSD512 input. With less spurious tones.

 

[bennetng]

That is still irrelevant as it is different kind of noise profile. First warning sign should be already if you get lower noise results for DSD64 compared to higher rates.

The AKM ADC chip digital decimation filter converting SDM from the converter to PCM output has only 85 dB attenuation. In addition, typical analog anti-alias filter is 2nd order. That combined to how skirting in the sampling and decimation works causes such differences.

For similar reasons, even devices like AP have a separate external analog filter for measuring things like class-D amplifiers. Because otherwise you'd get incorrect results. (since I work also on direct digital conversion amps, I have such filter too)

The multitone tests I posted on the previous page done with the E1DA Cosmos ADC also showed exactly the same relative differences in PCM vs different DSD rates. Therefore the results are deterministic among different recording hardware.

E1DA Cosmos ADC links:

E1DA Cosmos ADC

@IVX kindly sent me a sample of the new Cosmos ADC to try out. Here are my impressions of it so far. The Cosmos is housed in a compact metal case with rubber feet, USB Type C connector on one and dual XLR inputs on the other. Power is over USB only. Beside the USB connector is small LED that...

www.audiosciencereview.com

...and a much longer thread:

Best spec ADC Chip currently.... ??

I never use sine gen, only cosine ;) alright - why not tan ? actually - since THD is basically non linearity I could also imagine to do a static calibration of the transfer function. but guess it not really effective.

www.audiosciencereview.com

 

[MalinYamato]

Probably the most asked about question about DACs on audio forums is whether an external, usually linear, power supply improves performance. It just makes sense that it would do that. Doesn't it? Well, let's see if we can figure this out.

I have been kindly loaned a Wyrd4Sound PS-1 power supply for testing. It has replaceable modules that let you configure it for different voltages. The base PS-1 price is USD $499 and includes one regulated output. Additional modules cost $100. The unit I have has three modules in it. Two are 9 volt and one is 15 volt. Sadly, I see no regulatory and safety certification on the PS-1.

I was also kindly loaned an Sbooster BOTW ECO years ago so I thought I test it too. It costs $399 and comes in many variations. Sbooster provides OEM products to many companies and does proper engineering as far as safety standards.

For switching power supply I compared the one that comes with Topping DX3 Pro. Topping also has a strong commitment to safety and certification so the supply comes with such. Here is the whole contraption:

View attachment 23441​

Note that this is a long term project and I hope to improve and refine the tests. So consider this "draft 1."

Let's get into some measurements, shall we?

DAC Measurements
As a rule, I am not a fan of showing raw power supply measurements although I will be showing that toward the end. What is most important is whether the signal coming out of a DAC is improved with an aftermarket power supply. After all, those are the waveforms we hear. We don't listen to power supply wires. The impact of power supply noise is very indirect as the DAC usually performs secondary filter and re-regulation. And its circuits may be designed to have high rejection of power supply noise and ripple (called PSRR).

I thought I start my testing with Topping D50. It is a very good DAC and as it turns out, it has a 5 volt input jack that I could mate with either USB power from my computer, or that of the external linear power supplies. I started the test by powering the D50 with USB and the compared the performance to Sbooster BOTW:

View attachment 23442

The graph is kind of hard to red. But essentially I am feeding the DAC a 10 kHz tone at 192 kHz sampling. The harmonics of 10 kHz tone are visible at multiples of it so ignore that. Everything else is noise and spurious distortion. We see fair bit of them because I have zoomed way in, setting the top of the graph to -80 dB. Blue is the USB port on my computer which I use for DAC testing. We see that when we switch to Sbooster BOTW power supply, a lot of higher frequency spikes go away. Mind you, the worst case one is at -125 dB and all of these are at frequencies we can't hear. But objectively there is some improvement in "noise."

Next I went to test the PS-1 with the D50 and I realized it didn't have a 5 volt output. Thinking there is a pre-regulator in the D50, I plugged it in anyway. Sadly, the D50 didn't power on, nor did it do so with the other supplies anymore. She has died and gone to DAC heaven. May it rest in peace....

For plan B, I decided to test with Topping DX3 Pro since it takes 15 volt supply which lets me test the PS-1. Unfortunately the Sbooster is only 5 to 6 volts so it is out of the running here. I switched the test conditions around to 12 kHz tone at 96 kHz sampling (told you this is a draft work in progress):

View attachment 23443

Nothing is changed really if you compare the two graphs but one thing has gotten worse: there are some noise spikes between 1 and 1.2 Megahertz. Note: that is Megahertz, not Kilohertz. We are way, way outside of audible range but folks always think there are bad things up here so I have decided to show it with this 1 Megahertz bandwidth test. We will come back to these extra tones shortly.

I broke my usual rule of only testing audio signals and connected the output of the power supplies directly to the analyzer. I used a differential input so to keep external noise interference at bay (some probably still getting through). Let's compare the spectrum of noise between Topping's switching power supply and Wyrd4sound PS-1:
View attachment 23444

Looking to the left, the PS-1 is clearly much cleaner. What is disappointing is to the right: yes, we see those noise spikes around 1 Mhz just as we did in the output of the DAC! I have no idea what the cause of this would be other than I see a display on the PS-1. Maybe there is a microprocessor running in there at this frequency. It is sad that they did not measure the unit the way I have and found a way to eliminate that.

Let's now overlay the noise spectrum of PS-1 against the Sbooster BOTW:

View attachment 23446

Sbooster (in green) has somewhat higher noise but none of those spikes at 1 to 1.1 Mhz.

Let's focus down in the audio band and see how the three compare:

View attachment 23447

The general noise level of the PS-1 is the best. We do see a spray of spikes though which I suspect are mains frequency leakage. Sbooster has a smooth spectrum that falls in between. Topping power supply is the worst but again, look at the vertical scale: the highest noise spike is at -85 dB. No wonder none of this matters when we look at the output of the DAC.

Conclusions
It is clear that whether you use the USB power, or the supplied switching power supply, there is absolutely no audible improvement in the output of the DAC with linear power supplies. One can help himself believe otherwise by looking at the noise spectrum alone as I have shown in the last graph. But again, we don't listen to power supply wires. Those waveforms go through filtering stages even in cheapest DACs.

If you do want to get a linear power supply, my strong recommendation is to get it from a company that understands safety and has regulatory certification. In that regard, my recommendation would be for the Sbooster BOTW over Wyrd4Sound PS-1. It is also $100 cheaper.

------------
As always, questions, comments, recommendations, etc. are welcome.

I used a ton of electricity for this testing. So please donate some money so that I can pay my bill at the end of the month:
Patreon: https://www.patreon.com/audiosciencereview), or
upgrading your membership here though Paypal (https://audiosciencereview.com/foru...eview-and-measurements.2164/page-3#post-59054).

okay, so I will totally forget about linear power supplies and go on with life.

 

[Miska]

The multitone tests I posted on the previous page done with the E1DA Cosmos ADC also showed exactly the same relative differences in PCM vs different DSD rates. Therefore the results are deterministic among different recording hardware.

What ADC does it have and how much anti-alias attenuation does the converter have in total?

Results are not deterministic across any of my measurement gear. But yeah, I don't have that cheap gear...

 

[bennetng]

What ADC does it have and how much anti-alias attenuation does the converter have in total?

The links I posted showed measurements by different users and reviewers, including AP users.

Results are not deterministic across any of my measurement gear. But yeah, I don't have that cheap gear...

Dismissing a product's ability based on price, appearance and personal impression is also a cheap way of respond.

[EDIT]
Anyway, let's see how the Pro-Ject's noise floor looks like with "expensive" measurement gear.

Project Pre Box S2 DAC & Headphone Amp Review

This is a review and detailed measurements of the Project Pre Box S2 Digital USB DAC and Headphone Amplifier. It is on kind loan from a member. The S2 costs US $399 including Prime shipping from Amazon. I reviewed an early unit before I have my new audio analyzer so request was made to...

www.audiosciencereview.com

All filters are bottlenecked by the Pro-Ject's inherent noise floor.

Topping D10B

Topping D10 Balanced Review (USB DAC)

This is a review and detailed measurements of the new Topping D10 Balanced. It was sent to me by the company and costs US $139. The D10 Balanced as the name implies, comes in my favorite USB DAC package with the same cute orange display: Back panel is where the difference is: Balanced...

www.audiosciencereview.com

No selectable filter and only "Fast Minimum" is used, but the noise floor is 10dB lower.

Basically what I said:

It's simply the Pro-Ject is unable to adequately utilize the PCM potential of the ESS chips, which clearly lags behind the Topping D10B in Amir's reviews.

 

[Miska]

The links I posted showed measurements by different users and reviewers, including AP users.

OK, where are the AP measurements showing this PCM vs DSD difference.

Anyway, let's see how the Pro-Ject's noise floor looks like with "expensive" measurement gear.

All filters are bottlenecked by the Pro-Ject's inherent noise floor.

Not they are not, they are bottlenecked by the AP's settings.

I have just shown that I can even analyze behavior of the ESS Sabre modulator from Pro-Ject's output. So it is certainly not limited by the device's noise. You just need to know how to use the measurement gear.

Topping D10B

Topping D10 Balanced Review (USB DAC)

This is a review and detailed measurements of the new Topping D10 Balanced. It was sent to me by the company and costs US $139. The D10 Balanced as the name implies, comes in my favorite USB DAC package with the same cute orange display: Back panel is where the difference is: Balanced...

www.audiosciencereview.com

No selectable filter and only "Fast Minimum" is used, but the noise floor is 10dB lower.

This filter is also bottlenecked here by the AP settings.

It's simply the Pro-Ject is unable to adequately utilize the PCM potential of the ESS chips, which clearly lags behind the Topping D10B in Amir's reviews.

I have also Mytek Stereo192-DSD DAC which is ESS Sabre based (first generation) which doesn't have such behavior. I also had Mytek Brooklyn DAC+ which didn't have such behavior either. In addition I have other devices like Merging Hapi, etc.

We can also look at Jtest24 figures, at DSD512:

And the same at 44.1k PCM input:

Where we can nicely see extra -145 dB sidebands at 44.1k PCM input that are absent in the DSD512 input case. Noise floor being the same in both cases. Which is expected, knowing how the ESS Sabre converts PCM to analog compared to how it converts DSD to analog.

IIRC, Pro-Ject uses ESS' special analog/regulator chip, designed for the particular purpose.

 

[bennetng]

I have also Mytek Stereo192-DSD DAC which is ESS Sabre based (first generation) which doesn't have such behavior. I also had Mytek Brooklyn DAC+ which didn't have such behavior either. In addition I have other devices like Merging Hapi, etc.

Probably true if they use different chips.

We can also look at Jtest24 figures, at DSD512:

And the same at 44.1k PCM input:

The definition from Julian Dunn is that the main tone is a high amplitude, phase-shifted 1/4 fs tone, which means 705.6kHz for DSD64. In this sense it is simply impossible to perform any DSD J-Test. I am not surprised about your results, given the fact that noise floor is limited by other factors, and not from the DAC chip itself.

[EDIT]

OK, where are the AP measurements showing this PCM vs DSD difference.

See above. You cannot measure the Pro-Ject's differences in PCM vs DSD simply because the Pro-Ject has limited noise performance, and has nothing to do with measurement gear and methodology.

Not they are not, they are bottlenecked by the AP's settings.

I have just shown that I can even analyze behavior of the ESS Sabre modulator from Pro-Ject's output. So it is certainly not limited by the device's noise. You just need to know how to use the measurement gear.

So other than "cheap" gear, now another excuse of "don't know how to use the gears". Basically you denied everything. No fun at all if you keep replying in this way.