Topping D30Pro Review (Balanced DAC) | Audio Science Review (ASR) Forum
Amir measured -135 dB 2nd and -138 db 3rd. This DAC supposedly uses 4x CS43198. Noise shaping corner is north of 40 kHz, which is maybe the only thing that limits use as a signal generator for mesurements
I didn't really have Crystal on my shortlist for the very lowest distortion DACs. Is the chip this good or did they implement a super nifty circuit?
Edit: the data sheet of the chip does not look all that stellar. -115 dB THD+N typ, -109 max with 24 bit data. With 16 bit, only -94 dB typ, 88 max. What's going on about the poor 16 bit performance? A TDA1541A-S1 was better than that. Can't be noise, so is it distortion?
Amir measured -135 dB 2nd and -138 db 3rd. This DAC supposedly uses 4x CS43198. Noise shaping corner is north of 40 kHz, which is maybe the only thing that limits use as a signal generator for mesurements
I didn't really have Crystal on my shortlist for the very lowest distortion DACs. Is the chip this good or did they implement a super nifty circuit?
Edit: the data sheet of the chip does not look all that stellar. -115 dB THD+N typ, -109 max with 24 bit data. With 16 bit, only -94 dB typ, 88 max. What's going on about the poor 16 bit performance? A TDA1541A-S1 was better than that. Can't be noise, so is it distortion?
Last edited:
Impressive 24/32 bit performance.
That's odd with the poor 16-bit performance. It sorta looks like they blew the math in an interpolation circuit. It would be interesting to compare against other DACs.
Tom
That's odd with the poor 16-bit performance. It sorta looks like they blew the math in an interpolation circuit. It would be interesting to compare against other DACs.
Tom
Wolfs measurements look even better than Amir's:
Measurements of Topping D30Pro DAC | L7Audiolab
Just comparing among Amir's reviews:
The SMSL M400 with AK4499 is 5 dB better on 3rd than Amir's the D30 Pro, but slightly poorer on 2nd and completely looses out on 4th and 5th. The Toppings DX7 with ESS is in a similar league. The Toppings D90 with AK4499 does not come close to either.
Measurements of Topping D30Pro DAC | L7Audiolab
Just comparing among Amir's reviews:
The SMSL M400 with AK4499 is 5 dB better on 3rd than Amir's the D30 Pro, but slightly poorer on 2nd and completely looses out on 4th and 5th. The Toppings DX7 with ESS is in a similar league. The Toppings D90 with AK4499 does not come close to either.
Last edited:
I independently stumbled across these yesterday, too.
I was already wondering where CS43198 based DACs are, given that the chip specs are nothing less than first-rate and it's a SOTA converter design.
The D30 Pro uses 4 of these? That explains why dynamic range is >131 dB(A) then, I don't think I've ever seen a single CS43130 hit much more than 125 dB so the chip specs may be slightly generous. Wouldn't be a first for Cirrus Logic, I think their idea of "typical" is more in line with other consumer-level semiconductor manufacturers than, say, AD. (Read: You're probably not going to see this in practice.)
Not sure what's up with the poor 16-bit performance but honestly, you wouldn't be feeding a DAC on this performance level with 16-bit data anyway.
Last edited:
I don't think there is any issue with the 16-bit performance. Check the datasheet carefully and also the older CS4398 datasheet, done in a similar style. The CS43198 datasheet shows Minimum values that are all 6 dB lower than the Typical values. It could just be a datasheet error where someone subtracted 6 from all the numbers for that column without any consideration. The Typical values are pretty much what you would expect and match the CS4398 exactly, except there is no Minimum shown for CS4398. The CS4398 is in a lot of pro products and measures very well. If there were some problem I'm pretty sure it would have been noticed by now.
Even if there was some problem, you could feed in 16-bit audio as the 16 MSBs of a 32-bit word anyway.
Even if there was some problem, you could feed in 16-bit audio as the 16 MSBs of a 32-bit word anyway.
Be very carful attributing accuracy to those ultra low distortion numbers. When the distortion get close to the residual of the analyzer the phase alignment of the harmonics can throw them way off. Its back to a passive notch to start to have real numbers. They are all well beyond anything that would be a conceivable issue for an audio signal.
Yeah, you are right to be suspicious of anything approaching -140 dB. I think Jens has seen similar numbers out of the evaluation board with a different test setup, although maybe not quite as good. Still, pretty impressive for a chip that consumes 26mW in normal operation. Although, the D30Pro apparently uses 4 of them.
I was going to buy some of these from DigiKey for a personal project whenever I get some free time, but I notice that the QFN parts are out of stock everywhere now, thanks AKM 🙁. I personally don't have a problem with the BGA, but the pitch requires an HDI board and that gets expensive.
I was going to buy some of these from DigiKey for a personal project whenever I get some free time, but I notice that the QFN parts are out of stock everywhere now, thanks AKM 🙁. I personally don't have a problem with the BGA, but the pitch requires an HDI board and that gets expensive.
Last edited:
Re 16 bit performance, the theoretical limit is 98.1 dB.
The venerable TDA1541A (non S1, non S2) was specified at -95 dB typ, -90 dB min, S1 and S2 were better. PCM56-PK was -94dB typ, -92 dB max. I'm never sure if those old test methods and spec sheets were with dither or not.
Re accuracy of those extremely low numbers, I don't think it is about phase alignment. It's a question of the nonlinearity of the amplitude transfer function, which is going to be a polynomial (y = c1*x + c2 * x^2 + c3 * x^3 + ....). If the signs of the nonlinearity coefficients of DAC and ADC are opposite, they can cancel, giving us unrealistically low distortion peaks.
By inverting the signal, you should be able to quantify this effect for the odd coefficients. I wonder why this is not a standard part of testing?
BTW, I don't think it's an issue for any kind of audio signal, but for testing, we would like to be better than the audio signal 🙂
The venerable TDA1541A (non S1, non S2) was specified at -95 dB typ, -90 dB min, S1 and S2 were better. PCM56-PK was -94dB typ, -92 dB max. I'm never sure if those old test methods and spec sheets were with dither or not.
Re accuracy of those extremely low numbers, I don't think it is about phase alignment. It's a question of the nonlinearity of the amplitude transfer function, which is going to be a polynomial (y = c1*x + c2 * x^2 + c3 * x^3 + ....). If the signs of the nonlinearity coefficients of DAC and ADC are opposite, they can cancel, giving us unrealistically low distortion peaks.
By inverting the signal, you should be able to quantify this effect for the odd coefficients. I wonder why this is not a standard part of testing?
BTW, I don't think it's an issue for any kind of audio signal, but for testing, we would like to be better than the audio signal 🙂
Last edited:
I’m going to go out on a limb and say it’s a datasheet / test setup artifact and that Cirrus probably knows how to get 16 good bits out of their flagship DAC that has -115 dB THD+N or whatever it is exactly.
Yes, I just noticed that the unweigthed 16 bit DR of 94 typ, 88 min is exactly the same as the 16 bit THD+N of -94 (-88). So we can safely conclude this is noise limited, not THD limited.
As for CS43130 vs. CS43198, they are exactly the same in all DR specs (including 16 bit), but the CS43198 has taken 24 bit THD+N from 108 to 115 dB, so the CS43130 was THD limited, and the CS43198 is less so. Still, comparing 115 THD+N to 127 dB of unweighted DR, one would expect much poorer THD performance than what was demonstrated on the Topping D30 pro.
As for CS43130 vs. CS43198, they are exactly the same in all DR specs (including 16 bit), but the CS43198 has taken 24 bit THD+N from 108 to 115 dB, so the CS43130 was THD limited, and the CS43198 is less so. Still, comparing 115 THD+N to 127 dB of unweighted DR, one would expect much poorer THD performance than what was demonstrated on the Topping D30 pro.
The static transfer polynomial produces harmonics only at fixed phases (H2 at +/-90°, depending on the c2*x^2 coefficient sign, H3 at +/-180°, etc). But dynamic charateristics further rotate the distortion phases, as nicely seen e.g. in REW distortion table which lists also harmonics phases. As Demian said the measured harmonic behind the ADC is a vector sum of all the same-order distortions along the chain. The vectors' angles are just as important as their magnitudes. I agree with him that measuring distortions comparable to internal distortions of the analyzer is very unreliable, as magnitude of the vector sum can be anything between 0 and sum of their magnitudes. I am not sure internal H2/H3 of an AP are significantly below -135dB.
It appears the CS43131, which has the same DR specs as the 130 and 198 and the same THD specs as the 198, i.e.
127 unweighted DR and -115 dB THD+N at 0 dB (-97 at -20, -67 at -60) also measured great.
(anybody figured out the difference between the 131 and 198 chips? - I haven't yet)
魅族HiFi解码耳放与魅族HiFi解码耳放Pro评测 - 知乎
The Meizu Hifi dongle measures -124 dB 2nd and -141 dB 3rd and nothing much else at 2 V, 300 R single ended (!!).
The Hifi Pro has an added OPA1622 headphone driver. The 300 R measurement is at 11 kHz, so not comparable, but at 1 V / 32 R, it clocks - 119 dB 2nd and 3rd.
Noise looks good (< -150 dB and flat) for either dongle.
Amir measured
for the non-pro, -123 dB 2nd and 3rd and terrible noise at 2 V, no load
for the pro, he got -130 / -130 and better noise but with something that looks like noise shaping beginning at 4 kHz (??)
- for the Hifi pro
This guy, 0dB, had nice looking flat noise plots, but more in -140 dB territory, maybe because of what binning hi used. His distortion numbers may be too high because he was using a 525 analyzer as opposed to 555 for the other two.
노트10 이어폰 잭 2탄! 메이주 HIFI DAC vs. HIFI DAC PRO - 영디비
at no load, 2 V
-120 dB 2nd, 118 dB 3rd for the Pro
- 136 2nd, -117 dB 3rd for the non-Pro
Amir suspected some of the noise he saw was injected through the USB port of his computer (the other two were measured on Android phones). Sample to sample variation may also be high in these dongles, but it serves to show that the CS43131 has much better THD performance than to be expected from its data sheet.
127 unweighted DR and -115 dB THD+N at 0 dB (-97 at -20, -67 at -60) also measured great.
(anybody figured out the difference between the 131 and 198 chips? - I haven't yet)
魅族HiFi解码耳放与魅族HiFi解码耳放Pro评测 - 知乎
The Meizu Hifi dongle measures -124 dB 2nd and -141 dB 3rd and nothing much else at 2 V, 300 R single ended (!!).
The Hifi Pro has an added OPA1622 headphone driver. The 300 R measurement is at 11 kHz, so not comparable, but at 1 V / 32 R, it clocks - 119 dB 2nd and 3rd.
Noise looks good (< -150 dB and flat) for either dongle.
Amir measured
for the non-pro, -123 dB 2nd and 3rd and terrible noise at 2 V, no load
for the pro, he got -130 / -130 and better noise but with something that looks like noise shaping beginning at 4 kHz (??)
- for the Hifi pro
This guy, 0dB, had nice looking flat noise plots, but more in -140 dB territory, maybe because of what binning hi used. His distortion numbers may be too high because he was using a 525 analyzer as opposed to 555 for the other two.
노트10 이어폰 잭 2탄! 메이주 HIFI DAC vs. HIFI DAC PRO - 영디비
at no load, 2 V
-120 dB 2nd, 118 dB 3rd for the Pro
- 136 2nd, -117 dB 3rd for the non-Pro
Amir suspected some of the noise he saw was injected through the USB port of his computer (the other two were measured on Android phones). Sample to sample variation may also be high in these dongles, but it serves to show that the CS43131 has much better THD performance than to be expected from its data sheet.
Last edited:
The 131 is specified to drive headphones down to 32 ohm. The 198 is not.
But whether there is a real difference, I don't know. Could perhaps be the same chip?
But whether there is a real difference, I don't know. Could perhaps be the same chip?
I have been working with cs43198 (on a RPI DAC) for the past 6 months and completed my design . We get H2 at -125 and H3 at - 128 unweighted 1Khz 1.7Vrms ..in master mode and single ended .
More than happy to share info on the design.
I can post FFT and Jtest if anyone wants to see them..but yes after H2 and H3 there is not much distortion .
More than happy to share info on the design.
I can post FFT and Jtest if anyone wants to see them..but yes after H2 and H3 there is not much distortion .
Yes, please do post. Problem seems to be that the diy solderable version is sold out after the AKM fire, and BGA is not nice for us...
Interesting, is there any evidence of the static transfer function actually being much frequency dependent at 1 kHz?
Am I off in thinking that varying the phase between source and analyzer should be able to catch compensation effects?
The CS43198 is a next-generation, low-power audio DAC with pseudodifferential stereo analog output, while the CS43131 combines the DAC with a high-fidelity headphone amplifier and integrated high performance, ground-centered stereo headphone outputs. Both DACs deliver up to 2 VRMS output and feature headphone detection, mono mode support, integrated PLL, and Direct Stream Digital (DSD) path support. The devices' integrated PLL enables maximum clocking flexibility, while Popguard® technology eliminates output transients in power-down or power-up events.
Hmm, but the CS43198 also has headphone detection. There does not seem to be a penalty for the higher drive capability of the CS43131. Free lunch?
Hmm, but the CS43198 also has headphone detection. There does not seem to be a penalty for the higher drive capability of the CS43131. Free lunch?
The AK4493, which has inferior specs to both CS43131 and 198 in DR and THD+N, at least gives us numbers for DR(A weighted): 123 dB at 24 bit, 118 dB at 20 bit and 101 dB at 16 bit. The CS43131 and 198 giveus DR weighted at 16 bit 97 dB and 94 dB unweighted. So for some reason, while the CS chips do everything else better at 24 bits, they become too noisy on 16 bit (at least, according to their data sheet).