[dericchan1]

I have a pair of the Western Electric too. Right now I have a pair of Tesla 6CC42 running. The original Tesla before JJ bought the rights. They are super tubes as well. Right up there with the WE but much less expensive.

I agree with you on how good the iDSD PRO is. I spent an entire evening going back and forth between the iFi and the DSC2. Don't ask me to pick one lol. I will say that I am enjoying the Signalyst the most right now.

Yes, I currently have the idsd pro in my bedroom with a class A EL84 tube amp with a pair of focal Aria speakers. I played the 2xHD version of jazz at the pawnshop series dsd128, it sounded so good I woke up a few times at night wondering to myself “this is not real”

 

[Failed Engineer]

I would also love to hear a DSC DAC but would need to commission one. Haven’t known of anyone to do them.

 

[bogi]

Some DSD DACs indeed provide direct paths and use a filter similar to what is in the Signalyst DAC. iFi, T+A, Holo Audio, and SOME implementations of AKM that allow for the bypass to be turned on.

Sorry but you are mixing oversampoling (interpolation) filters and delta sigma modulation with D/A stage.
D/A stage of delta sigma DACs (that are all direct DSD capable DACs) is implemented as a FIR low pass filter.

When we speak about bypassing DAC digital processing with direct DSD path, it is about bypassing oversampling filters and modulators, of course not the D/A stage. One cannot bypass the D/A stage itself since it is the most important and the only mandatory part of a DAC device. D/A stage is of course always implemented by hardware since you don't get anything analog in digital domain.

When different direct DSD capable DAC devices mention DSD filters like 40 kHz, 60kHz,... it is analog filter setting, applied only in the case of DSD input.
When different delta sigma DAC devices mention PCM filters (like sharp, slow, ...), it is about digital oversampling filter setting. Of course, analog filter is used with PCM input too, just that there is no user option, since oversampling is fully under DAC chip control.

These two cases may cause confusion. Direct DSD is not about skipping anything what touches analog world. Analog DSD filters are part of D/A stage.

 

[xuyunwei001]

我在当铺系列 dsd128 播放了 2xHD 版本的爵士乐，听起来太棒了，我在晚上醒来好几次，心里想“这不是真的”：笑脸：,祝贺你！！是的，我目前的卧室里有 idsd pro，配有 A 级 EL84 电子管放大器和一对焦点 Aria 扬声器。我在当铺系列 dsd128 播放了 2xHD 版本的爵士乐，听起来太棒了，我在晚上醒来好几次，心里想“这不是真的”

 

[MLGrado]

Sorry but you are mixing oversampoling (interpolation) filters and delta sigma modulation with D/A stage.
D/A stage of delta sigma DACs (that are all direct DSD capable DACs) is implemented as a FIR low pass filter.

When we speak about bypassing DAC digital processing with direct DSD path, it is about bypassing oversampling filters and modulators, of course not the D/A stage. One cannot bypass the D/A stage itself since it is the most important and the only mandatory part of a DAC device. D/A stage is of course always implemented by hardware since you don't get anything analog in digital domain.

When different direct DSD capable DAC devices mention DSD filters like 40 kHz, 60kHz,... it is analog filter setting, applied only in the case of DSD input.
When different delta sigma DAC devices mention PCM filters (like sharp, slow, ...), it is about digital oversampling filter setting. Of course, analog filter is used with PCM input too, just that there is no user option, since oversampling is fully under DAC chip control.

These two cases may cause confusion. Direct DSD is not about skipping anything what touches analog world. Analog DSD filters are part of D/A stage.

Sorry if it seemed that way, but I indeed was talking about bypassing any DSP and Modulation. Of course you can't bypass the D/A stage. All the native DSD DACs I mentioned use the same mechanism basically. Biggest difference in most cases is length of delay line (number of taps) and some use weighted elements, others equal elements.

1-bit signal untouched by any oversampling fiters or DSM, through shift register, to switches via some kind of dynamic element matching, summed into analog. That is the DAC part, and is all that is needed for DSD. The analog filter IS the DAC. This is exactly why I am using the DSC2 DAC. Its a analog 32 tap/33level CIC filter realized with discrete parts. The filter is the DAC.

Theoretically the DSC2 and other DACS like it could easily convert a multi-bit DSM output too, (of course most already do this internally)

But if we actually had a multi-bit DSD "source format". Wouldn't that be cool? If DSD was 32 levels at speeds 128x or 256x as the delivery format. Yes, it would be data and bandwidth intensive, which is why we don't have it. But, one can always dream..

Perhaps I just misread what you were objecting to. I just skimmed the post.

 

[Kelvin will]

Blessings to all of you for the information! I just got into using HQP and its a massive rabbit hole of info

 

[jlaako]

But if we actually had a multi-bit DSD "source format". Wouldn't that be cool? If DSD was 32 levels at speeds 128x or 256x as the delivery format. Yes, it would be data and bandwidth intensive, which is why we don't have it. But, one can always dream..

At one point I had such experimental one, using 8-bit data format (U8). And technically HQPlayer can support such any day.

But it is not necessarily any better, because you immediately introduce some problems familiar from PCM.

We can actually do better with 1-bit streams than multi-bit, as long as the DAC implementation is correct. I know how to make DSC much better, but that project had certain set of goals and requirements that it met. Remember that things like ESS DAC chips and Mola-Mola DACs are also 1-bit stream to the conversion stage.

AKM moved from switched capacitor designs to switched resistors in their recent top of the line. Not sure if it improved anything... They knew how to do SCF well, now they need to learn how to do that with resistor arrays. I think this has more to do with market demand than technical reasons.

 

[bogi]

Perhaps I just misread what you were objecting to. I just skimmed the post.

Previously I didn't get the meaning of your formulation. Thanks for making it clear for me.

 

[Ghoostknight]

AKM moved from switched capacitor designs to switched resistors in their recent top of the line. Not sure if it improved anything... They knew how to do SCF well, now they need to learn how to do that with resistor arrays. I think this has more to do with market demand than technical reasons.

were you able to compare the AK4499 + AK4491 combo yourself?

never heared any akm dac, so im still trying to decide if the ak4499 is worth it, tho many reviewers like it

 

[camrector]

were you able to compare the AK4499 + AK4491 combo yourself?

never heared any akm dac, so im still trying to decide if the ak4499 is worth it, tho many reviewers like it

Gustard A26 is a fantastic sounding dsd direct dac with built in NAA.

 

[jlaako]

were you able to compare the AK4499 + AK4491 combo yourself?

never heared any akm dac, so im still trying to decide if the ak4499 is worth it, tho many reviewers like it

You mean AK4191? I have Gustard A26 for example. And SMSL DL300.

After Gustard fixed the A26 firmware so that DSD Direct works correctly, it is pretty nice DAC, especially through ASDM7EC-light modulator at DSD256. I also have SMSL M400.

But the earlier AK4490/AK4493 work really nice in DSD Direct, and objectively not any worse. Especially my two ADI-2 Pro's with AK4490 are very good.

I also have the EVGA NU Audio with AK4493 and DSD support and it works really nicely.

 

[MLGrado]

At one point I had such experimental one, using 8-bit data format (U8). And technically HQPlayer can support such any day.

But it is not necessarily any better, because you immediately introduce some problems familiar from PCM.

We can actually do better with 1-bit streams than multi-bit, as long as the DAC implementation is correct. I know how to make DSC much better, but that project had certain set of goals and requirements that it met. Remember that things like ESS DAC chips and Mola-Mola DACs are also 1-bit stream to the conversion stage.

AKM moved from switched capacitor designs to switched resistors in their recent top of the line. Not sure if it improved anything... They knew how to do SCF well, now they need to learn how to do that with resistor arrays. I think this has more to do with market demand than technical reasons.

U8 as in unsigned PCM?

I knew Mola-Mola used a 1-bit PWM final stage after a 32bit/3.125Mhz DSP. I was under the impression that the ESS output stage was still multi-bit DSM.


Yes, I have played around with the latest AKM chipset. 4191+4499. It is confusing because there are two different 4499 chips now. The very latest of which must be paired with the 4191. 4191 has the digital filters and modulators. 4499 has the converter elements only.

The implementation I tested (SMSL D400 Pro) had some problems in bypass mode. The 'Sound Color' settings interacted directly with the 'Wide' and 'Narrow' DSD filters, creating lots of audible artifacts/noise. On a bit of a tangent, I took some measurements and both the Wide and Narrow filter modes via bypass mode seem to cutoff around 70khz at DSD64 rate. That is just an approximation. Narrow rolls-off faster. Wide rolls-off much much slower.


Also, it was very picky. 48x rate based DSD signals played at the 44.1x rate.. some clock issues with bypass. It was like a tape playing slow. As expected, not using bypass mode, there was no sound at all with 48x DSD rates. Also, the DSD filters non-bypass mode pre-modulator were identical. Even set to 'wide', the roll-off started at 19khz and stopband was reached around just 30khz. Now, all of that could be bugs in the SMSL implementation.

But I will say this. As far as DSD via bypass mode, the RME with AK4493 SCF measures far superior than the AK4191-4499 resistor based filter I tested. Again, it could be a case of implementation.

As far as SMSL goes, the cheaper D300 is higher on my list. The one with the ROHM DAC that uses a 1-bit converter for all DSD. I love that little thing.

 

[shwnwllms]

This is likely similar to what is going on with some DACs featuring MQA decoding. I have not done extensive testing, but I suspect there are cases where it could be messing with the data when it shouldn't.



You could try playing Qobuz with just HQPlayer Client for comparison. In that case you know there's no MQA or similar sitting on the path.

From mathematical perspective, both Roon (with all DSP disabled), and HQPlayer playing the same content from Qobuz should have same result.

Finally got around to trying this, with MQA decoding turned off in Roon (and all DSP disabled) I would be hard pressed to tell the difference between Roon or HQP Client playing the same file from Qobuz.

With MQA decoding turned on (which I believe is the default state IIRC) I always thought HQP Client sounded better

I’m sure for critical listening I’ll continue to use HQP Client for piece of mind but this makes me think there is something else going on with the MQA decoding setting in Roon (even for non MQA material)

/ cc @EMINENT

 

[camrector]

U8 as in unsigned PCM?

I knew Mola-Mola used a 1-bit PWM final stage after a 32bit/3.125Mhz DSP. I was under the impression that the ESS output stage was still multi-bit DSM.


Yes, I have played around with the latest AKM chipset. 4191+4499. It is confusing because there are two different 4499 chips now. The very latest of which must be paired with the 4191. 4191 has the digital filters and modulators. 4499 has the converter elements only.




The implementation I tested (SMSL D400 Pro) had some problems in bypass mode. The 'Sound Color' settings interacted directly with the 'Wide' and 'Narrow' DSD filters, creating lots of audible artifacts/noise. On a bit of a tangent, I took some measurements and both the Wide and Narrow filter modes via bypass mode seem to cutoff around 70khz at DSD64 rate. That is just an approximation. Narrow rolls-off faster. Wide rolls-off much much slower.


Also, it was very picky. 48x rate based DSD signals played at the 44.1x rate.. some clock issues with bypass. It was like a tape playing slow. As expected, not using bypass mode, there was no sound at all with 48x DSD rates. Also, the DSD filters non-bypass mode pre-modulator were identical. Even set to 'wide', the roll-off started at 19khz and stopband was reached around just 30khz. Now, all of that could be bugs in the SMSL implementation.

But I will say this. As far as DSD via bypass mode, the RME with AK4493 SCF measures far superior than the AK4191-4499 resistor based filter I tested. Again, it could be a case of implementation.

As far as SMSL goes, the cheaper D300 is higher on my list. The one with the ROHM DAC that uses a 1-bit converter for all DSD. I love that little thing.

A26 has the same chip set 4499ex/4191
But Gustard has updated the firmware to support DSD direct after Jussi noticed/measured similar behavior to what you describe.
Still advised to only use 44.1 family rates only does DSD128/256 natively.

 

[jlaako]

U8 as in unsigned PCM?

Yes, that is pretty unused overall, so it was good choice for the purpose.

Now, all of that could be bugs in the SMSL implementation.

Some may be, but many of those are just bugs in AK4191. Or poor documentation from their side (which at least seems to be the case). Now it is just like ESS chips, which have always been like that.

But I will say this. As far as DSD via bypass mode, the RME with AK4493 SCF measures far superior than the AK4191-4499 resistor based filter I tested. Again, it could be a case of implementation.

Yeah, partially due to implementation. But I don't think there is technical reason why this new chipset would be any better than AK4490 or AK4493... Not at least until they figure out all the things they've figured out for SCF practical implementation over the past decades. Cirrus Logic is another SCF based DAC chip vendor. While TI/BB and ESS are resistor based. But I wouldn't say that either one is clearly better than the other.

As far as SMSL goes, the cheaper D300 is higher on my list. The one with the ROHM DAC that uses a 1-bit converter for all DSD. I love that little thing.

Unfortunately mine got bricked on a firmware update. I shouldn't have attempted that... I can reflash it all day long, but it doesn't wake up anymore.