T+A DAC 200

[bogi]

Although I am not T+A DAC 200 owner I would like to add my subjective opinion to the DSD softness debate. It's just an opinion so I don't ask anybody to agree or fully agree with me.

IME instrument attacks (transients) usually sound 'softer' but at the same time more detailed with direct DSD than with oversampled delta sigma PCM. IMO the reason is additional distortion which PCM output contains and which is not present in the case of direct DSD. If that wouldn't be distortion, then with DSD input I couldn't hear more fine detail. That's the only reason why I like to upsample PCM content to DSD with HQPlayer.

I think that the mentioned distortion is result of intermodulation artifacts caused by unfiltered second oversampling stage of delta sigma DACs - that's what one can bypass with direct DSD (as well as with DSD signal fed to ESS chip based DACs). I perceive that additional distortion as some additional hardness on transients. My opinion is that most of people are already used to 'PCM sound' containing such a distortion and consider it to be correct (since we are listening to it for years everywhere). IMO that's the reason why sound coming from direct DSD may be perceived as softer, even if it provides more fine detail and is more accurate. If I am right then it is not by coincidence that such a 'PCM hardness' is not present when playing from vinyl on pure analog chain.

If 'soft' would mean less dynamics and 'hard' more dynamics, then IMO DSD256 or DSD512 is not softer.

[bogi]

6 hours ago, Jud said:

- What is 'delta sigma PCM'?

I meant output of delta sigma DAC when it is fed by PCM signal. My post is not related to output of R2R DACs, since I never owned such one.

 

6 hours ago, Jud said:

- I thought all oversampling stages had to be filtered?

Every resampling creates images of audio band at multiples of target fs. Miska mentioned numerous times, that hardware oversampling of delta sigma DACs is realized in 2 stages. For the 1st one it may be possible to choose filter type. For T+A DAC200 they are FIR1, FIR2, Bezier1, Bezier2 and it is realized outside of DAC chip. High frequency content containing images of audio band is filtered in this stage. This is only up to some fs. Then the 2nd stage follows within delta sigma DAC chip which is simplified (usually sample and hold). See this Miska's post. High frequency content remains unfiltered and therefore images of audio band at multiples of last filtered target fs are coming to input of delta sigma modulator and should be filtered by analog filter behind the modulator. But that analog filter is usually not so steep that it could fully filter the nearest images, as shown many times in Miska's measurements.

For demonstration see for example 2th, 3th and 4th graph of old iFi Micro measuremens. https://audiophilestyle.com/blogs/entry/428-ifi-idsd-micro-measurements/. Images of audio band (a pair mirrored - non mirrored) appear around every multiple of 352.8k, which is the border fs between oversampling stages.

With DSD these images of audio band don't appear on measurements. The reason is that upsampling in HQPlayer filters these images digitally up to the target fs.

[bogi]

33 minutes ago, Jud said:

Yep, I had seen this and now understand exactly what you were referring to. It is of course possible to do something other than bad filtering at the 2nd or 3rd oversampling (8x oversampling is usual, so 3 doublings) though as you say the measurements appear to show it is not typical at least in those DACs for which Miska has released measurements.

No, I still meant it otherwise. Taking the iFi Micro as an example, I did not mean 2th and 3th step of 8x oversampling (3 steps of fs doubling, 1->2, 2->4, 4->8). One or more (or all three) of these steps may be skipped if DAC is fed by higher than 44.1k / 48k PCM input (up to 352.8 / 384k).

As 1st oversampling stage I meant those max. 3 steps as a whole. That stage is filtered, so it does not leave images of audio band in frequency spectrum. This 1st stage ends at some fs, for iFi it was 352.8/384k, but it may be higher with other chips like ESS. But still it is not at delta sigma stage operating frequency. Therefore 2nd oversampling stage is needed and that one is unfiltered.

Although it may be possible to substitute the on chip (I mean DAC chip) 1st stage by external solution (T+A DAC200 uses this approach), it is not possible to substitute or skip the 2nd stage. The target fs of the 2nd stage is ~ 10MHz (like 128x44.1/128*48, or 256x44.1/256*48), exact value depends on DAC chip.

Every time when Miska mentioned ZOH (Zero Order Hold) or SaH (S&H) (Sample And Hold) he meant the on chip unfiltered 2nd oversampling stage. That means simple repeating of previous sample n times - it's the easiest way to increase sample rate up to that ~10MHz range. DAC chips don't have computational power to do more on such a high sample rates.

[bogi]

3 hours ago, Nkam said:

The question now is: does the ΣΔ  modulator stage perform oversampling or not ?

That's what I called as 2nd oversampling stage in this post and the following two.

[bogi]

3 hours ago, Nkam said:

This question is a bit philosophical and the answer depends on weather you regard the output signal of the ΣΔ  modulator as a highly oversampled 1 bit digital signal (in this case it performs a kind of oversampling), or if you regard the modulators output signal as an analog signal having an average value representing the analog output value - in this case the modulator is DAC delivering an analog output voltage which only needs some averaging (analog low-pass filtering) to get rid of the unwanted high frequency noise and to deliver the wanted analog signal average.

Delta sigma modulators in current DAC chips have still digital output. That output may be yet re-arranged to n bit unary form and then converted to analog using n equally weighted elements (they are usually realized in the form of switched resistor DAC or switched capacitor DAC). So the D/A conversion itself (which in fact is a kind of low pass filtering) is implemented behind digital output of delta sigma modulator.

[bogi]

@NkamI will try to be simpler. PCM1795 (like other older Burr Brown chips) may have 1 bit modulator output (not multibit like is nowadays more typical). Miska most probably knows that exactly.

But let start it simpler with the case of 1 bit 2 level modulator output. Then such an output signal is in principle the same as DSD signal. You raised a philosophic question if DSD signal is analog or digital. It is usually considered to be digital. I mentioned also other interpretations like it is on border of analog and digital and even an opinion of @tailspn(the debates about 10 years ago) that it is in fact analog. But generally DSD is considered to be a digital signal. So I would 1 bit output of delta sigma modulator consider to be digital signal too.

DAC in narrower sense is a circuit appearing behind digital output of delta sigma modulator. You see it also on PCM1795 block diagram

 

You see that 'Current Segment DAC' is a separate block for each channel. The modulator module name 'Advanced Segment DAC Modulator' can be misleading since you find the term DAC here. The modulator block does not contain DAC. The whole DAC chip is described as Advanced Segment, Stereo Audio Digital-to-Analog Converter. So the modulator block could be also named 'Delta Sigma Modulator of Advanced Segment DAC'.

 

In my previous post I pointed to a fact that most of current D/A chips contain delta sigma modulator with multibit output. Then the technique of n equally weighted elements is used to get better output precision (SNR and dynamic range). In the case of multibit delta sigma DACs it is yet more visible that the modulator output is digital and D/A conversion happens behind it.

[bogi]

1 minute ago, Nkam said:

i didn’t know I raised a philosophical question if DSDZ is analog or digital.

You raised a philosophical question if delta sigma modulator output in DAC chip is considered to be digital or analog:

58 minutes ago, bogi said:

This question is a bit philosophical and the answer depends on weather you regard the output signal of the ΣΔ  modulator as a highly oversampled 1 bit digital signal (in this case it performs a kind of oversampling), or if you regard the modulators output signal as an analog signal having an average value representing the analog output value - in this case the modulator is DAC delivering an analog output voltage which only needs some averaging (analog low-pass filtering) to get rid of the unwanted high frequency noise and to deliver the wanted analog signal average.

I attempted to explain that 1 bit 2 level modulator output is in principle the same as DSD signal which is generally not considered to be analog. I used this to argue that delta sigma modulator output in current DAC chips has to be considered digital and that the D/A stage in narrower sense follows the modulator.

Your thinking was probably inspired by the general delta sigma modulator description you mentioned. Yes, both modulator input and output can be analog too, but in current DAC chips delta sigma modulator is implemented as digital circuit.

[bogi]

27 minutes ago, Nkam said:

i think what I was asking was if the T+A DAC since it has a PCM ΔΣ DAC , if it can upsample using external filters such as HQplayer and have the same results as a NOS dac.

Miska already answered that question here: https://audiophilestyle.com/forums/topic/63998-ta-dac-200/page/10/#comment-1235019

No matter what was done in the first stage (for example upsampling with external filters) there is no way to skip that 2nd oversampling stage with PCM input. The target fs of the 2nd oversampling stage is ΔΣ modulator operating frequency which is ~ 10MHz range.

 

With DSD input both oversampling stages (including that second one) can be skipped.

[bogi]

3 minutes ago, Nkam said:

and what happens in that stage?

I assume you are asking to that Zero Order Hold (or Sample And Hold) stage which I referred as 2nd oversampling stage.
2 things happen here:

1) No correct interpolation algorithm is used. Although the 1st stage has more impact on sound than the 2nd, the result is increased noise floor in comparison with the case when new interpolated value would be computed by a suitable algorithm.
2) Since the 2nd stage is unfiltered, repeating images of audio band at multiples of last filtered fs are coming to delta sigma modulator input and then they appear also on modulator output. See my answers to Jud one week ago. They are source of possible intermodulation distortion.

With DSD input and HQPlayer upsampling to say DSD256, HQPlayer interpolation filter algorithms are used up to the destination fs ~ 10MHz. Audio band images (which are result of every upsampling algorithm) are digitally filtered by HQPlayer so they cannot become source of intermodulation distortion.

About the graphs, ask Miska for answer.

[bogi]

12 hours ago, Nkam said:

second stage of PCM in the T+A also includes the noise shaper of something like HQplayer?

Noise shaping is result of delta sigma modulation. The term 'noise shaper' means use of delta sigma modulator. No noise shaping without delta sigma modulator. Noise shaping = delta sigma modulation.

Short answer:  With PCM input T+A DAC (like all delta sigma DACs) performs oversampling and delta sigma modulation (noise shaping) - similar process like what si done in HQPlayer when you do PCM to DSD processing.

Longer answer:

With PCM input delta sigma modulators contained in PCM1795 DAC chips are used - the picture I posted on previous page. You see here oversampling module in front of delta sigma modulator.
With DSD input in direct DSD mode delta sigma modulation (noise shaping) of PCM to DSD software converter (for example HQPlayer) is used and the modulator in DAC chip is skipped.

Note about terminology - maybe it confused you:
In this post Miska bound together 2nd oversampling stage (that zero order hold thing) and delta sigma modulator as reaction to formulation used in quoted OE333's post. Miska considered that 2nd oversampling stage to be part of 'delta sigma stage' - for the purpose of that post. But it is only matter of terminology used for a particular case. In previous discussions Miska many times mentioned two oversampling stages in front of delta sigma modulation - that's what I did in my previous posts too.

[bogi]

1 hour ago, StreamFidelity said:

By the way, the digital volume control has some advantages. No channel imbalance and no noise (unless the amplifier is noisy). And no bit loss, because the HQPlayer upconverts everything to at least 64Bit before. 6dB correspond to 1Bit.

 

That's not so easy. Noise level remains the same with digital volume control - it's the quantization noise.
Then, no bit loss is valid only for digital domain, when processing is performed at higher bit resolution than the audio content resolution. Lowering of digital volume level by 6dB approximately looks like shifting audio content one bit right, so the highest bit will be always zero and thus not used for audio content. So with every further volume lowering by 6dB you effectively shift audio content by one bit again. If you have enough bits in digital domain, no information is lost. So far OK.

 

Then, it is important to consider analog domain in the picture. Resolution of DAC is restricted to say 20 bits. Let's your audio content is 16bit recording.

16 bit recording at 0dB digital volume setting:
x ...16 bits usable for audio content
o ... remaining bits still within 20bit DAC resolution
- ... bits out of DAC resolution (not usable for audio content with given DAC)

xxxxxxxxxxxxxxxxoooo------------

When digital volume is lowered by 24 dB, 4 upper bits are no more used and 16 bit audio content is shifted to lower bits:

0000xxxxxxxxxxxxxxxx------------

You see there are no remaining bits within 20 bit DAC resolution for 16bit recording. When you would lower digital volume yet more, part of input digital audio content could not be reflected at DAC output because is would be lost below DAC noise floor.

 

With 24bit recording it is a bit more complicated. Audio content is 24 bit, but it's questionable what's the real resolution of the recording. No A/D or D/A converter is able of 24 bit resolution (although DACs are going close). But the same principle works with 24bit content too. You loose more and more bits of original recording when you output digital volume control processing result to devices with restricted resolution.

 

Digital volume control has advantages and disadvantages. It is wise to use it in a range considering your real DAC resolution. Digital volume control in player is the safe way to avoid intersample overflows in DAC. Some of you may experience that lowering digital volume level and increasing amp volume level change sound a bit. It may have more reasons. One of them is that DACs usually provide lowest distortion somewhere at -10 dB. It may be good idea to find an optimal digital volume level for your chain for example somewhere between -3 and -12dB according to your listening test and then to use analog volume control for adjustments.

 

I have HQPlayer set for volume adjustment in range from -12dB to 0dB, so I am not getting lower that -12 dB with digital volume control. Of course, I am using analog volume control of my headamp.

 

[bogi]

3 minutes ago, barrows said:

Even in a very quiet room when playing music at very high levels (say 120 dB peaks), one is only going to have, perhaps, around 12, maybe 14 bits of real, audible dynamic range.

 

I am not so much focused on loudspeaker topics since I am listening on headphones, but I am surprised by those numbers.

Maybe you mean an usual not much treated room with quite high amount of reflections, where real acoustic pressure level cannot lower so much and so quickly like in open space.

[bogi]

2 hours ago, stereousa said:

I keep hearing how good the DAC200 is when HQPlayer sends DSD streams to its USB. What about PCM playback or DSD streams from Foobar2000? Anybody has done this here?

 

It's technically of course possible. What's different is quality of PCM to DSD conversion. The difference is significant, but HQPlayer requires much more computer resources to get this result. I of course recommend HQPlayer.

 

HQPlayer Embedded is UPnP capable. That's gives possibility to use it with streaming services. Since you mentioned Tidal, you could stream Tidal content to HQPlayer Embedded from BubbleUPnP or mConnect Player running on your phone, or from JPLAY ... many possibilities.

 

If you have enough strong computer available, just start HQPlayer trial and compare it with foobar2000.

[bogi]

1 hour ago, davidv100 said:

Then obviously I understand that Z (Impedance) is frequency dependent

That's relevant for electrodynamic headphones like HD800. High impedance makes easier to pair them with headamps which don't have extra low output impedance.

 

Impedance of planars is known to be constant across frequency spectrum, so 16 Ohm impedance should not cause an issue with any headamp. Because of that constant nature, planars impedance is often not published as graph but only mentioned as a value. Here you can see for example impedance graph of my HE-500:

 

What low impedance and not quite low sensitivity headphones need is enough current.

[bogi]

25 minutes ago, Shadorne said:

True but the impedance is never exactly flat across all frequencies. Therefore a highish output impedance amp into low input impedance devices can modulate the frequency response. 

 

IMO such thinking is not needed with planars. Frequency response of headphones contains much higher level of imperfections than influence of 1 or 2% impedance deviation from the average.

I found also Arya (not the Organic one) impedance graph:

[bogi]

4 hours ago, stereousa said:

But if you were to play PCM stream on the D200 without converting to DSD, how good is the playback, compared to say Chord Hugo TT2 or Holo May KTE? Was wondering if anyone has tried or experienced the difference.

 

You first asked for a difference between foobar2000 and HQPlayer. Now you are asking for a difference between different DACs.
I will stay on foobar2000 vs HQPlayer topic. If you do no upsampling and you use exclusive audio device access, there is no difference in the stream sent to DAC except of possible differences in dithering (if any) and eventually other DSP, like volume control, applied before dithering.

 

Many people are reporting a bit different sound when playing the same content from different players. These differences could be attributed to different noise profiles generated by different players, or more generally digital audio sources. But this area is not much confirmed by measurements.

 

It's simply on you to install more software players, to buy or borrow more DACs and to do such a comparison self. Different people have different tastes. You may like a different player and DAC combination than I.

[bogi]

My post doubled by editing.

[bogi]

3 hours ago, stereousa said:

if Foobar2000 outputs DSD (PCM files converted to DSD), then will that sound better on the the DAC 200

 

In my experience yes with all DACs I tried so far. But I cannot comment on DAC200 since DACs I tried used internal upsampling of DAC chips with PCM input.
 

3 hours ago, stereousa said:

I heard that HQPlayer needs fans and GPUs.

 

Not as a strict rule. You could find here people using HQPlayer on similar Intel CPUs like yours without add on GPU, for example with HQPlayer OS, which does not contain nVidia CUDA support. And some are using also passive cooled designs. At DSD256 you could have a chance and it could sound much better than foobar2000 at DSD512. In my opinion it is worth to try - trial costs nothing.

[bogi]

2 hours ago, davidv100 said:

DAC200 is supposed to have a very good USB galvanic isolation, and a very good clock. If my understanding is correct, the downstream clock (DAC200) is master for the upstream source(the iPad, the Android tablet).

Then, with a tablet (iPad, Android), if it is well build, you are isolated from any power pollution (since its internal battery acts a buffer) - It is even better than any high-end power solutions "rebuilding the current".
So,
- you are out of any jitter issue (DAC200 clock is master of timing);

- you are out of any power pollution issue, twofold (Tablet running on battery as a buffer, galvanic isolation provided by DAC200).

 

AFAIK some parts of the above quote are simplified and others are incorrect.

 

I read about galvanic isolation between digital and analog part of DAC200. I did not read anything about galvanic isolation of DAC200 USB interface at its input side. What I observed is Miska's recommendation (few posts above) to use Intona USB isolator with DAC200. Please correct me if I am wrong.

Then, you don't seem to know much about audio source noise impact to DACs. It is not limited only to ground loops, which you cannot get with battery powered digital audio source. Phones, tablets and portable audio players, like every computer or digital device, are source of high frequency noise, which does not have character of DC causing ground loops. With portable, battery powered devices you are not free of noise impacts to USB connected downstream devices.

 

Now about convenience of use. You are thinking about iPad or Android device connected by USB cable to DAC200. Is that convenient for mobile device operation in loudspeaker setup? Will you lead 5 meter long USB cable from DAC200 to your sofa? Or do you want to move from your sofa to DAC200 every time you want to operate your battery powered device?

 

How long time you can stream from your iPad or Android device from built-in battery? Will you care to charge it every night? When it gets out of power, will you stop listening, or will you connect a charger? Isn't then your thought advantage of battery power lost?

 

2 hours ago, davidv100 said:

How can any High End Audio Brand (with tenth or hundredth employee) compete against behemoths like Apple, Samsung

 

Simply said, MP200 and similar devices are targeted to customers, which want the best possible sound and at the same time the most convenient way to use different streaming services and ways to connect other audio devices. What's OK for you my be not OK for others. Devices like MP200 are targeted to customers who understand and are interested in the added value.

[bogi]

55 minutes ago, OE333 said:

So, if it is the general opinion, that intersample overs constitute a severe problem, I will be happy to discuss this matter with my former colleagues at T+A to see if it makes sense to introduce a few dBs of additional headroom in the oversampling filters.

 

Not every album is mastered up to 0dB. I think the right approach is to adjust volume level at digital source (player) side and not to compromise SNR for example for classical music or jazz recordings playback.

 

How many people are using players/transports which don't allow volume control, like old fashion CD players? I have no idea, but on this forum people are using mostly software players, often running also room eq and/or other DSP in them.

[bogi]

On 11/23/2023 at 2:07 PM, GoldenOne said:

It'd be great if the DAC did have at least an option/setting to mitigate intersample overs.

If at all, then please an option, not fixed setting, for people with fixed 0dB level digital sources. Because for people with software players that would be redundant processing, which is better to avoid.