Topping Centaurus R2R DAC Review

[AndyZ]

Differences in perceived sound stage could be related to phase modulation when using non linear filters.

 

[VientoB]

I would still love to do a blind test between this R2R and your favorite Delta Sigma.

Blind tests like that always sound tedious to me!

 

[Mnyb]

Blind tests like that always sound tedious to me!

+1 Yes there is somekind of hope that after decades of trying that a DAC will come and show all naysayers that they sound different and glorious

We already established limits of human hearing, therefore it’s unecessary to blind test DAC’s that are better than humans it’s basically already done for all off us for all time .

So if you do this you have some likely outcomes.

• you find no differences after many many tries , it’s expected as they test objects works as expected .

• you find differences but after many retrials and consultations with others you find a small experimental error . This is expected if an amateur like myself would do this as it very hard to do properly with extremely small differences involved.

• you find a true difference, congratulations one of the test objects is a broken design.

 

[Parakmiakos]

Following up on the review and GoldenSound's write-up (Cyan 2/Centaurus comparison on Headphones.com) with something neither one tested directly. Equipment is modest - MacBook, MOTU M2 loopback, REW - so this isn't meant to compete with a proper audio analyzer on absolute SINAD. It's well suited to the question below, which is about a large, relative effect rather than fine noise-floor numbers.

For context, this started from a practical concern: I listen to a lot of CDs (mostly metal, heavily weighted toward loudness-war-era masters) through a Denon DCD-900 (line out) into my DACs, and wanted to tame the signal so it doesn't run into intersample-overshoot clipping in the DAC's reconstruction filter during playback. The usual advice is to pull the level down a few dB to leave headroom - but that only works if the attenuation happens before the reconstruction filter. If it happens after, you're just scaling a signal that has already clipped, and you've gained nothing.

So the question is: on any given DAC, which control actually attenuates before the filter?

Method: feed the DAC a full-scale test signal engineered to overshoot, then attenuate using the control under test and see whether the overshoot artifact actually goes away or just scales down along with everything else. If it disappears, that control is pre-filter and genuinely protects. If it scales proportionally, it's post-filter and doesn't.

Results across two DACs, and the controls turn out to be different on each:

- D70 Pro Sabre (ESS): the main digital volume knob is the right control. Attenuating with the knob makes the overshoot artifact disappear - it's pre-filter. So on the D70 you just leave the volume a few dB down and you're protected.

- Centaurus (R2R): the main volume knob does NOT work for this - the artifact only scales down proportionally, meaning it's post-filter and you're not actually preventing the clip. What does work is PEQ pre-gain (the pre-amplification cut in the PEQ section): with that, the artifact disappears entirely. So on the Centaurus you have to leave the main volume alone and pull headroom via PEQ pre-gain instead, setting listening level downstream on the amp.

I tested the Centaurus main volume in both OS-4x/Preamp and Best-OS/DAC-Fixed modes to be sure - same result both times.

Takeaway: don't assume the volume knob protects you against ISP clipping. On the Centaurus specifically it doesn't; PEQ pre-gain is the control that does. Worth checking on any DAC rather than assuming, since the answer isn't consistent even within one brand.

Would be curious whether others measuring these DACs see the same.

 

[Roland68]

Following up on the review and GoldenSound's write-up (Cyan 2/Centaurus comparison on Headphones.com) with something neither one tested directly. Equipment is modest - MacBook, MOTU M2 loopback, REW - so this isn't meant to compete with a proper audio analyzer on absolute SINAD. It's well suited to the question below, which is about a large, relative effect rather than fine noise-floor numbers.

For context, this started from a practical concern: I listen to a lot of CDs (mostly metal, heavily weighted toward loudness-war-era masters) through a Denon DCD-900 (line out) into my DACs, and wanted to tame the signal so it doesn't run into intersample-overshoot clipping in the DAC's reconstruction filter during playback. The usual advice is to pull the level down a few dB to leave headroom - but that only works if the attenuation happens before the reconstruction filter. If it happens after, you're just scaling a signal that has already clipped, and you've gained nothing.

So the question is: on any given DAC, which control actually attenuates before the filter?

Method: feed the DAC a full-scale test signal engineered to overshoot, then attenuate using the control under test and see whether the overshoot artifact actually goes away or just scales down along with everything else. If it disappears, that control is pre-filter and genuinely protects. If it scales proportionally, it's post-filter and doesn't.

Results across two DACs, and the controls turn out to be different on each:

- D70 Pro Sabre (ESS): the main digital volume knob is the right control. Attenuating with the knob makes the overshoot artifact disappear - it's pre-filter. So on the D70 you just leave the volume a few dB down and you're protected.

- Centaurus (R2R): the main volume knob does NOT work for this - the artifact only scales down proportionally, meaning it's post-filter and you're not actually preventing the clip. What does work is PEQ pre-gain (the pre-amplification cut in the PEQ section): with that, the artifact disappears entirely. So on the Centaurus you have to leave the main volume alone and pull headroom via PEQ pre-gain instead, setting listening level downstream on the amp.

I tested the Centaurus main volume in both OS-4x/Preamp and Best-OS/DAC-Fixed modes to be sure - same result both times.

Takeaway: don't assume the volume knob protects you against ISP clipping. On the Centaurus specifically it doesn't; PEQ pre-gain is the control that does. Worth checking on any DAC rather than assuming, since the answer isn't consistent even within one brand.

Would be curious whether others measuring these DACs see the same.

You don't need to perform any measurements for this, however.
You simply need to check where the volume control takes place in the specific DAC.
It could be digital—before the DAC chip (as with RME ADI units or those with PEQ)—or digital within the DAC chip's DSP section (the most common method, known as "preamp mode"), or analog—occurring only after the DAC chip and filtering stages. Any method that controls volume in the analog domain is ruled out.

 

[Parakmiakos]

You don't need to perform any measurements for this, however.
You simply need to check where the volume control takes place in the specific DAC.
It could be digital—before the DAC chip (as with RME ADI units or those with PEQ)—or digital within the DAC chip's DSP section (the most common method, known as "preamp mode"), or analog—occurring only after the DAC chip and filtering stages. Any method that controls volume in the analog domain is ruled out.

Agree on RME and PEQ, and on ruling out analog-domain controls. But "preamp mode"/DSP-internal isn't one consistent answer either - both my D70 and Centaurus main volumes are that same category, and they came out opposite (D70 pre-filter, Centaurus post-filter). So knowing which category a control falls into doesn't actually tell you pre- or post-filter for a given unit - it still has to be checked per DAC, even within that category.

 

[MaxwellsEq]

Following up on the review and GoldenSound's write-up (Cyan 2/Centaurus comparison on Headphones.com) with something neither one tested directly. Equipment is modest - MacBook, MOTU M2 loopback, REW - so this isn't meant to compete with a proper audio analyzer on absolute SINAD. It's well suited to the question below, which is about a large, relative effect rather than fine noise-floor numbers.

For context, this started from a practical concern: I listen to a lot of CDs (mostly metal, heavily weighted toward loudness-war-era masters) through a Denon DCD-900 (line out) into my DACs, and wanted to tame the signal so it doesn't run into intersample-overshoot clipping in the DAC's reconstruction filter during playback. The usual advice is to pull the level down a few dB to leave headroom - but that only works if the attenuation happens before the reconstruction filter. If it happens after, you're just scaling a signal that has already clipped, and you've gained nothing.

So the question is: on any given DAC, which control actually attenuates before the filter?

Method: feed the DAC a full-scale test signal engineered to overshoot, then attenuate using the control under test and see whether the overshoot artifact actually goes away or just scales down along with everything else. If it disappears, that control is pre-filter and genuinely protects. If it scales proportionally, it's post-filter and doesn't.

Results across two DACs, and the controls turn out to be different on each:

- D70 Pro Sabre (ESS): the main digital volume knob is the right control. Attenuating with the knob makes the overshoot artifact disappear - it's pre-filter. So on the D70 you just leave the volume a few dB down and you're protected.

- Centaurus (R2R): the main volume knob does NOT work for this - the artifact only scales down proportionally, meaning it's post-filter and you're not actually preventing the clip. What does work is PEQ pre-gain (the pre-amplification cut in the PEQ section): with that, the artifact disappears entirely. So on the Centaurus you have to leave the main volume alone and pull headroom via PEQ pre-gain instead, setting listening level downstream on the amp.

I tested the Centaurus main volume in both OS-4x/Preamp and Best-OS/DAC-Fixed modes to be sure - same result both times.

Takeaway: don't assume the volume knob protects you against ISP clipping. On the Centaurus specifically it doesn't; PEQ pre-gain is the control that does. Worth checking on any DAC rather than assuming, since the answer isn't consistent even within one brand.

Would be curious whether others measuring these DACs see the same.

Interesting. I suspect that how the "volume control" is implemented is not reliably defined or documented in the marketing materials for products!

You might want to have a look at how @NTTY does his rather elegant Inter Sample Over test using tones with a phase angle. Example is here:

Topping DX1 II DAC & HP Amp Review (NTTY)

Hello everyone, This is a review and measurements of the Topping DX1 II Dac and headphone amplifier. I bought it from Topping site and received it rapidly. I got it for around 120$. Topping DX1 II - Presentation This little device acts as a DAC (USB or Optical input), with line level...

www.audiosciencereview.com

 

[NTTY]

Yep, we did some exploratory research on this with the SMSL D200. This one has (digital) volume control post DAC, in analog domain. And the SMSL DO400 has digital volume control pre-DAC. The later offers full resistance to ISO, but the former offers higher resolution with volume down.
This is why I now test ISO resistance with volume down.

 

[Roland68]

Agree on RME and PEQ, and on ruling out analog-domain controls. But "preamp mode"/DSP-internal isn't one consistent answer either - both my D70 and Centaurus main volumes are that same category, and they came out opposite (D70 pre-filter, Centaurus post-filter). So knowing which category a control falls into doesn't actually tell you pre- or post-filter for a given unit - it still has to be checked per DAC, even within that category.

The volume controls of the D70 Pro and the Centaurus cannot possibly belong to the same category.
The D70 Pro uses the integrated volume control of the DAC chip, while the Centaurus doesn't have a DAC chip at all, but rather discrete R2R modules.

If you say that the volume on the Centaurus is controlled after the filter, then that suggests an analog volume control.

 

[antcollinet]

Takeaway: don't assume the volume knob protects you against ISP clipping. On the Centaurus specifically it doesn't; PEQ pre-gain is the control that does. Worth checking on any DAC rather than assuming, since the answer isn't consistent even within one brand.

Bottom line though - if you need to check this by measurement - and therefore, presumably are not able to determine by listening:

Does it really matter?

 

[Parakmiakos]

Bottom line though - if you need to check this by measurement - and therefore, presumably are not able to determine by listening:

Does it really matter?

This isn't really an audibility question - it's whether a control prevents actual digital clipping, which is a binary fact about the device whether or not you can pick it out by ear on any given track. And you couldn't reliably listen for it anyway: it's an intermittent, program-dependent artifact that only shows up on specific transients in specific material. That's exactly the kind of thing measurement exists to catch instead of guessing.

 

[Parakmiakos]

Yep, we did some exploratory research on this with the SMSL D200. This one has (digital) volume control post DAC, in analog domain. And the SMSL DO400 has digital volume control pre-DAC. The later offers full resistance to ISO, but the former offers higher resolution with volume down.
This is why I now test ISO resistance with volume down.

Thanks for those examples - D200/DO400 is a clean parallel to the D70/Centaurus case. That's several device pairs now all showing the same thing: you really can't infer this from architecture or price tier, it has to be checked per unit.

 

[rocksteady]

Yep, we did some exploratory research on this with the SMSL D200. This one has (digital) volume control post DAC, in analog domain. And the SMSL DO400 has digital volume control pre-DAC. The later offers full resistance to ISO, but the former offers higher resolution with volume down.
This is why I now test ISO resistance with volume down.

Very interesting findings. Would anybody know in which category or camp the SMSL DO200 fits into?

 

[Roland68]

Very interesting findings. Would anybody know in which category or camp the SMSL DO200 fits into?

What exactly do you mean?
It is a standard DAC with analog volume control in the analog output stage.

 

[rocksteady]

What exactly do you mean?
It is a standard DAC with analog volume control in the analog output stage.

"And the SMSL DO400 has digital volume control pre-DAC. The later offers full resistance to ISO, but the former offers higher resolution with volume down."

I meant the above. I was hoping the DO200 would feature a similar implementation to its "bigger" brother, thus offering resistance to ISOs...

 

[mocenigo]

The first CD-players had only 14 bit DAC's (Philips) with its impact to the sound. Some of the early facts from a new system are burned in the common memory and generalized forever.

That, and listening to CDs with preemphasis on players that did not properly support the feature.

 

[Roland68]

Starting from 0 dB, to what specific fixed value would the digital volume—for example, on the D200—need to be lowered to ensure full resistance to ISO?

 

[Parakmiakos]

Starting from 0 dB, to what specific fixed value would the digital volume—for example, on the D200—need to be lowered to ensure full resistance to ISO?

There isn't one number that's guaranteed correct for every track on every DAC - it depends on how far a given track's true peak sits above 0dBFS (varies by track) and how much headroom the DAC's own output stage has before it actually clips (varies by DAC).

That said, not everyone wants to measure their own library and DAC, and you don't have to. A flat -3dB is a reasonable, low-effort default that covers the large majority of real-world material without needing any of that. Hans Beekhuyzen has a good explanation of exactly this approach here:

Measuring just lets you tailor the number to your own worst-case track and your own DAC's actual headroom instead of a generic margin - useful if you want it exact, but -3dB is a perfectly sensible thing to just set and forget if you don't.

 

[danadam]

Starting from 0 dB, to what specific fixed value would the digital volume—for example, on the D200—need to be lowered to ensure full resistance to ISO?

Archimago analyzed his own library and got:
"If we want to ensure absence of intersample clipping for 99% of tracks in each of these genres based on the library, here are the dB peaks the DAC/CD player should be able to handle - sorted from genre with lowest to highest value:

Classical                  +0.79dB

Soundtracks                +1.70dB
Vocals                     +1.82dB
Country                    +1.89dB
Hard Rock/Metal            +2.03dB
Classic & Progressive Rock +2.35dB
Jazz & Blues               +2.42dB
Pop                        +2.87dB

Rap / Hip-Hop              +3.08dB
Electronica                +4.22dB

and:
"the one with the highest True Peak was from the electronic album Making Orange Things by Venetian Snares & Speedranch, called "Meta Abuse" with peaks up at 7.72dBTP!"

A look at the music library: How much DAC intersample overhead is enough? True Peak analysis of genres in an eclectic digital collection.

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

archimago.blogspot.com

 

[rocksteady]

Archimago analyzed his own library and got:
"If we want to ensure absence of intersample clipping for 99% of tracks in each of these genres based on the library, here are the dB peaks the DAC/CD player should be able to handle - sorted from genre with lowest to highest value:

Classical                  +0.79dB

Soundtracks                +1.70dB
Vocals                     +1.82dB
Country                    +1.89dB
Hard Rock/Metal            +2.03dB
Classic & Progressive Rock +2.35dB
Jazz & Blues               +2.42dB
Pop                        +2.87dB

Rap / Hip-Hop              +3.08dB
Electronica                +4.22dB

and:
"the one with the highest True Peak was from the electronic album Making Orange Things by Venetian Snares & Speedranch, called "Meta Abuse" with peaks up at 7.72dBTP!"

A look at the music library: How much DAC intersample overhead is enough? True Peak analysis of genres in an eclectic digital collection.

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

archimago.blogspot.com

The "Abuse" in Meta is an apt description!