Best DAC for less than 2500 euros, New or used ?

[Miska]

4 hours ago, Ozan Bolat said:

One exciting thing would be to listen to PCM with R2R and to DSD with DSD module but I think i read @Miskauses his Denafrips and Holo as DSD only while many May owners seem to listen PCM PCM.

What is the status?

 

If PCM with ladder then there would be the matter of pure NOS vs OS in HQP : some seem to prefer pure NOS even with 44.1 but I think I read that @Miskasays that if I do filters to correct room/speakers in HQP I have to at least double the initial FR.

Is that correct? Is it better to not have HQP so everything PCM is NOS ?

 

Point of using such NOS DACs with HQPlayer is that it let's bit-perfect reproduction of data from HQPlayer algorithms.

 

There are clear technical reasons for using OS and SDM (DSD).

 

But when used correctly, HQPlayer can also linearize R2R like the Holo Audio ones, when used with PCM.

 

4 hours ago, Ozan Bolat said:

If OS in HQP  is there a consensus that max (1536) is best ?

 

Yes, it is quite clear from objective point of view.

 

[Miska]

40 minutes ago, Ozan Bolat said:

It is critical point for me between T+A and Holo : if using OS SDM for everything is better then Spring loses advantage if i can find T+A DSD8 cheaper (since I assume PCM part is better and more consistent in Spring with R2R than in T+A with SD chips)

 

Well, the two are still different DACs.

 

Quote

So, I'd be happy and probably others if you could elaborate on "There are clear technical reasons for using OS and SDM (DSD)." when the source is PCM and the DAC is R2R

 

In short, this has been discussed many times over the years here.

 

SDM DACs were created for a reason, due to various problems of R2R. With R2R you need steeper analog reconstruction filter because settling time to within +-½ LSB limits possible OS rate. Meaning incomplete reconstruction due to images leaking through and analog filter causing phase shifts in audio band. In addition precision of R2R ladder limits low level linearity of the converter which is due to physical problems of creating high precision resistors. Although this is common problem in SDM DAC chips too, but for different reason - because their DSP processing power is too limited.

 

SDM allows you to use for example 256x OS filters and sampling rate, without settling time problems. This is enabled by advances in DSP processing power. Allowing you to get around both the OS rate, settling time and low level linearity issues.

 

Quote

"But when used correctly, HQPlayer can also linearize R2R like the Holo Audio ones, when used with PCM." is completely mysterious and beyond my understanding so I'd be happy, and probably others,  if you could please elaborate

 

I have documented this extensively in the Holo threads. If you set output bits in HQPlayer to 20 and select a suitable noise shaper combined with highest (705.6k+) output sampling rates, you can linearize the R2R ladder without losing any dynamic range. This will significantly reduce distortion of low level signals when using R2R.

 

Here's Spring 2 playing 1 kHz -120 dBFS tone with 24-bit TPDF dithered output without noise shaping:

 

Here's Spring 2 playing the same tone, at 1.4112M but now with NS5 noise shaper and 20-bit output:

You can see all the distortion components have disappeared.

 

I just test and measure DACs to find out how to improve their technical performance with HQPlayer's DSP.

 

[Miska]

12 minutes ago, Ozan Bolat said:

NS5, NS9 or LNS15 all the same ?

 

Similar results, if you use the highest sampling rates and 20 bit. Differences are more in the ultrasonic band when used at lower rates like 352.8/384k or such. And some differences in audio band at those lower rates when used at lower word lengths (16 bit and less).

 

[Miska]

37 minutes ago, Zauurx said:

A difference in load distribution on the cores ? (NS5 vs LNS15)

 

No, the load from those noise shapers is so low at these low PCM rates that it is barely noticeable in first place.

 

[Miska]

Also one thing where DSD256 excels compared to PCM is jitter performance. So far jitter in the DAC output seems to be almost invariably lower with DSD inputs than with PCM.

 

[Miska]

17 hours ago, Ozan Bolat said:

then, would'nt the Holo Cyan DSD be the perfect DAC ? Why they stopped it? comments seem to suggest there were issues with it and that it was only so so but wouldn't a May reduced to DSD at Spring price point be the absolute steal ?

 

I have Cyan DSD as well. It doesn't perform as well with DSD as Spring does. Did they stop it? What issues? Likely Cyan PCM too?

 

I cannot know why some manufacturer does something with their products. Price difference to the Spring wasn't huge anyway.

 

[Miska]

10 hours ago, Ozan Bolat said:

too bad ! sure HQP + a May level DSD only DAC would be a good thing. Does it exist in another brand ?

 

At least in form of variants of my DSC1 design. But I don't think having or not having PCM support in May makes much difference in terms of price.

 

[Miska]

40 minutes ago, Ozan Bolat said:

Do you infer that I should go the extra mile and get a May rather than a Spring even if I am to use it DSD only in the end ?

 

I don't have answer to that, but I'd say Spring vs May doesn't change whether you use PCM or DSD.

 

[Miska]

15 minutes ago, Zauurx said:

@Miska If I summarize... (I have a Pegasus): PCM 1.4112 in 20bits is the most linear but DSD 256 has a lower jitter.

 

20-bits or less with correct type of noise shaping is most linear for Holo Audio DACs when in PCM mode. It is way to make the PCM mode work better. Not related to DSD.

 

[Miska]

On 8/6/2021 at 9:50 AM, Ozan Bolat said:

@Superdadnever chimed in . Maybe @Miska could share his rationals to order his Spring 3 without the Tinatis USB ?

 

because he has a @Superdad's thing or similar ? because of using a NAA ?

 

I would need to have a reason to do so, but based on performance of USB module in Spring 2, I'd expect the regular USB module in Spring 3 to be at least on same level. I've had L2 of Spring 1 and 2 before and Spring 3 L2 was already expensive for my limited R&D budget.

 

[Miska]

11 minutes ago, Ozan Bolat said:

Does NAA and/or add ons such as Uptone's Regen help fill the eventual gap with Titanis module?

 

I don't use devices like Regen. My networks are based on Hewlett-Packard Enterprise / Aruba and Cisco networking gear.

 

To comment about "gap with Titanis module", I would need to know what the "Titanis module" has different and have to see measurements from the DAC output demonstrating this difference.