Which DACs bypass digital filtering?

[audiofool]

On 12/29/2023 at 2:51 PM, Miska said:

 

With for example Topping/SMSL devices this needs to be verified by measurements. It only takes one small bug in the firmware for the feature not work. And this is not about just bypassing digital filtering, but also modulator. IOW, whether the DAC is bit-perfect.

 

So maybe the topic should say "Which DACs are bit-perfect".

 

 

I have a cheap Topping DX1.  It sounds better with DSD input but impossible to tell if it is bit-perfect.  It is very stable with pcm/dsd/sample rate changes.  48x DSD rates even work if you run 44x DSD first BUT will output full volume noise if you run pcm first!! - a dangerous bug!

 

 

[bogi]

11 hours ago, audiofool said:

48x DSD rates even work if you run 44x DSD first BUT will output full volume noise if you run pcm first!! - a dangerous bug!

Check if 48x DSD does not play on slightly lowered tones, see https://audiophilestyle.com/forums/topic/67918-smsl-d300-native-dsd-dac-review-with-measurements/?do=findComment&comment=1263804

 

[Miska]

I have now also tested/measured Cayin RU6, and get unexpected results with it. So I don't know what I should think about that one...

At least it doesn't behave like a regular R2R DAC would.

 

[bogi]

19 hours ago, Miska said:

At least it doesn't behave like a regular R2R DAC would.

Interesting. Do your measurements at least basically correspond to those from l7audiolab?
https://www-l7audiolab-com.translate.goog/f/cayin-ru6/?_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en

[Miska]

2 hours ago, bogi said:

Interesting. Do your measurements at least basically correspond to those from l7audiolab?
https://www-l7audiolab-com.translate.goog/f/cayin-ru6/?_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en

 

Somewhat similar yes.

 

I would expect flat noise floor from R2R though...

 

 

 

With DAC Bits set to 16 and using 352.8/384k rate with LNS15 noise-shaper, the low level linearity becomes fairly OK.

 

[ekpln]

Hello everyone,

 

I received my SMSL D300 recently and I think its sound is improving day by day, or my ears are getting used to it :) Anyway, I'm trying DSD direct mode along with the PCM upsampled data. While examining the datasheet of the Rohm chip, I noticed something. From 352.8kHz, FIR Bypass mode is enabled and filter is not selectable.

 

 

Apart form that, I was examining BD34301EKV Evaluation Board User's guide, and I saw something more interesting. On this board, there ise a rotary switch and you can change the filters, input frequency etc. So as you can see, Mode B and Mode C are FIR bypass mode, according to the table 13 above as well.  More interestingly, with this sampling rates, there is no over sampling rate is pointed.

 

 

Actually, I couldn't find anything about non-oversampling mode on the datasheet. But depends on the official evaluation board guide, over sampling rates weren't pointed. 

 

In this case, can we talk about a PCM NOS mode above 352kHz?

 

 

[Miska]

1 hour ago, ekpln said:

Actually, I couldn't find anything about non-oversampling mode on the datasheet. But depends on the official evaluation board guide, over sampling rates weren't pointed. 

 

In this case, can we talk about a PCM NOS mode above 352kHz?

 

There's an on-chip digital filter that can do up to 352.8/384k rates, then each sample is copied 8, 16 or 32 times for the modulator (S/H). This leads for example to total 8 x 16 = 128x oversampling rate. See register address 40. Default is 8x and thus the modulator would be running at 64x rate.

 

[ekpln]

On 2/16/2024 at 12:23 AM, Miska said:

 

There's an on-chip digital filter that can do up to 352.8/384k rates, then each sample is copied 8, 16 or 32 times for the modulator (S/H). This leads for example to total 8 x 16 = 128x oversampling rate. See register address 40. Default is 8x and thus the modulator would be running at 64x rate.

 

That part was not unclear for me, thank you. So at least, 8x oversampling occurs for the PCM input and actually we don't know the configuration that SMSL engineers did.

 

In this case, the documentation of the evaluation board seems wrong about the oversampling rate.

[bogi]

On 2/15/2024 at 8:56 PM, ekpln said:

I received my SMSL D300 recently and I think its sound is improving day by day, or my ears are getting used to it :) Anyway, I'm trying DSD direct mode along with the PCM upsampled data.

 

IMO you get best sound from D300 when upsampling with HQPlayer to DSD256. DSD512 brings higher noise floor.

Sound at DSD256 is IMO excellent - detailed but not fatigue, not harsh, not bright, neutral, rather relaxed.

Compared to my previous Topping E50: Cleaner sound, more detail provided in effortless way.
Older units with MCU firmware 1.0 (those don't show MCU version in RESET menu) are also capable of 48k based DSD rates up to 24.576 MHz DSD512. Units with newer MCU firmware are capable only of 44.1k based DSD rates.

 

I have a set of measurements performed with E1DA Cosmos ADC    by @mrba

AFAIK not directly DAC output but RCA output of Topping PRE90 preamp at 0dB setting was measured.

 

An example measurement for DSD256 from 44.1k PCM, poly-sinc-hb-m, ASMD5EC-super 512+ fs, DSD filter 13k, HPC enabled.

 

An example DSD512 measurement shows higher noise floor:

 

44.1k to 352.8k, sinc-MG, NS5, slow roll-off, HPC enabled:

[woshifeng3627]

 

Isn't Gustard R26 equipped with a NOS switch? It seems that both PCM and DSD can enable NOS, but why are they not listed in the PCM or DSD list above?

[semente]

On 2/29/2024 at 8:50 AM, woshifeng3627 said:

 

 

Isn't Gustard R26 equipped with a NOS switch? It seems that both PCM and DSD can enable NOS, but why are they not listed in the PCM or DSD list above?

I seem to recall reading about teething problems when converting DSD.

Not sure if they've been addressed...

[woshifeng3627]

I consulted Singxer customer service for the latest version of the SDA-6 PRO2 DAC, which also uses the AKM 4191+4499EX chip. Similar to the GUSTARD A26, the difference is that Singxer customer service told me that their SDA-6 PRO2 DAC can use the DSD bypass channel on DSD512, which means that DSD64~512 can use the DSD bypass channel and support 48K x512. The reason is that they have more comprehensive development of AKM chips. Isn't it true that the AKM4191+4499EX combination chip DAC can only go through the DSD bypass channel under DSD128-256?

[dht4me]

It has been mentioned but not added to the list but Audio Note DACs all are NOS with no digital filtering other than relying on the IV transformer's FR and the OPT FR impact. I have upgraded the input receiver twice, first to the wm8804 which gave me 192k USB capability and now I have HDMI- i2s which gives me 768k native capability however I seem to prefer 353k or 705k  upscaled with most material. 

[cpcat]

On 7/24/2023 at 1:53 PM, semente said:

 

Thanks Matt. I am keeping valve DACs off the list, any other models to join the R1 NOS?

This is an old post, but I can’t find an explanation for why anywhere?

 

Something against tubes? 😀

[woshifeng3627]

Has anyone tested whether the latest version of SDA-6 PRO2 DAC can support DSD512 under DSD direct connection?? Because it also uses the AKM4191+4499EX combination, the official customer service said it is supported! I don't believe it now! It also supports 48xDSD, which is completely different from my A26.

[copy_of_a]

17 minutes ago, woshifeng3627 said:

the official customer service said it is supported!

Singxer is a zero marketing and no bullshit company with a strong track record in Xmos development etc.

So if you are not refering to the customer service of a reseller but to Singxer company itself (Leter Chen) I'd expect the information to be accurate.

[woshifeng3627]

7 minutes ago, copy_of_a said:

Singxer is a zero marketing and no bullshit company with a strong track record in Xmos development etc.

So if you are not refering to the customer service of a reseller but to Singxer company itself (Leter Chen) I'd expect the information to be accurate.

 

Miska just analyzed again that 4499EX cannot work on DSD64 or DSD512 or above, and can only work perfectly under DSD128 or DSD256. This is a chip issue! I tested it on A26 and it's the same!

[copy_of_a]

18 minutes ago, woshifeng3627 said:

 

Miska just analyzed again that 4499EX cannot work on DSD64 or DSD512 or above, and can only work perfectly under DSD128 or DSD256. This is a chip issue! I tested it on A26 and it's the same!

That might be correct. But I suspect you are mixing "native" and "direct" DSD.

While the new 4499EX might only play native DSD128 and DSD256 the DSD path connection to the DAC can still be direct (non DoP) for other rates.

[bogi]

22 minutes ago, copy_of_a said:

That might be correct. But I suspect you are mixing "native" and "direct" DSD.

While the new 4499EX might only play native DSD128 and DSD256 the DSD path connection to the DAC can still be direct (non DoP) for other rates.

This chipset supports DSD64 to DSD1024 native DSD input, but according to Miska direct DSD can work only at DSD128 and DSD256. Most of implementations are buggy and not working really as direct DSD. Exceptions are Gustard *26 DACs and Audalytic AH90 with non official firmware published by Gustard person on head-fi, and SMSL D-6 (again a fixed firmware, not the original one). AFAIK no manufacturer succeded without Miska's help with direct DSD with this chip combo.

According to Miska the datasheet is incorrect. DAC producers would need how to verify direct DSD by measurements. But they dont know what to search for in such a measurtement, they have troubles with much more basic things - like they don't know to generate higher rate DSD test signals. IMO Miska could provide such for public download, otherwise it's hard to expect that strugling manufacturers will be able to implement functional DSD direct with such a beast combo. Then the result is like SMSL DL300 showing direct DSD in menu but according to Miska it's not working too (Miska told he communicated it directly with SMSL and they promised something...).

[woshifeng3627]

2 hours ago, copy_of_a said:

这可能是正确的。但我怀疑你混合了“原生”和“直接”DSD。

虽然新的 4499EX 可能只能播放本机 DSD128 和 DSD256，但对于其他速率，与 DAC 的 DSD 路径连接仍然可以是直接的（非 DoP）。

Perhaps you didn't understand what I meant. What I was saying was that A26 is in DSD direct mode, which cannot play DSD512 rate audio normally and is accompanied by a lot of noise. SDA-6 PRO2 DAC enables DSD NOS (DSD Direct Mode), but can play DSD512, even DSD1024, including 48x DSD, which cannot produce normal sound in A26 DSD Direct Mode! Singxer developers said they are familiar with AKM and better at developing AKM 4191+4499EX, which is why they are able to play DSD512 and 48x DSD audio in DSD direct mode.
Now my question is, since there is a problem with the 4499EX chip, why is there not uniform performance in DSD direct mode? At the same time, why is it possible to play 48x DSD and DSD512 when turning off DSD direct mode? This design is a bit contradictory. AKM engineers probably have a short circuit in their minds. Providing such an immature DA converter is completely incompatible with AKM4191. At least in terms of specifications, AKM4191 is 64 bit and supports PCM1.5M and DSD1024, but its performance is halved in DSD direct mode! I originally thought that AKM's new combination chip could support DSD1024 like HOLO's R2R, but the result was a bit funny!

[airdog0988]

Topping is about to launch the D900 DAC and D90 discrete DAC, both are 1 bit discrete DACs.

[bogi]

4 minutes ago, airdog0988 said:

both are 1 bit discrete DACs

That doesn't guarantee that they will feature direct DSD capability. Topping is Topping - they didn't care about it till now. We will see if they really want to enter audiophile market.

[woshifeng3627]

On 2/16/2024 at 5:23 AM, Miska said:

 

有一个片上数字滤波器，可以执行高达 352.8/384k 的速率，然后为调制器 （S/H） 复制每个样本 8、16 或 32 次。例如，这导致总计 8 x 16 = 128 倍过采样率。请参阅注册地址 40。默认值为 8x，因此调制器将以 64x 速率运行。

 

I just tried the DSD direct mode of A26, and surprisingly, it can play 48 times the DSD, which is 24.576M, but I need to turn off the adaptive mode! Unify upsampling to 48 times DSD, so there is no noise! I don't know if you have tested this mode! Is it a true DSD through mode. DSD filter is still set to wide

 

 

[bogi]

2 hours ago, bogi said:

That might be correct. But I suspect you are mixing "native" and "direct" DSD.

No, usual terminology 5-10 years ago was native/DoP about DAC USB input and "direct DSD path" as a term used in many DAC datasheets for a path bypassing oversampling, volume control and modulator. As an example CS4398 chip:

 

It happened in the last years that people started to mix the words native and direct, making many times hard to understand the context in discussion.

 

[bogi]

17 minutes ago, woshifeng3627 said:

it can play 48 times the DSD, which is 24.576M, but I need to turn off the adaptive mode! Unify upsampling to 48 times DSD, so there is no noise! I don't know if you have tested this mode! Is it a true DSD through mode.

If 48k based DSD rates are available, direct DSD should behave the same for 44.1k and 48k based DSD rates.
In relation to direct DSD path capability 48k based DSD rates bring nothing more than 44.1k based rates.

The adaptive rate checkbox is intended to be used to lower CPU/GPU load by avoiding cross family conversions (44.1->48 or 48->44.1).