[lairpost]

@GoldenOne Have you compared any DDCs with the May to see if the I2S input outperforms USB direct from computer/renderer etc?

I'm thinking of a Denafrips Hermes. I don't need all the I/O of the Gaia and I don't think the Iris would be a step up. I'm kind of waiting for someone to make a DDC that outputs 1.536MHz PCM. I find some that do DSD1024 but not 1.5M PCM which is odd since the bandwidth is effectively the same for each.

 

[Triode User]

You also own HQ Player don´t you?
Have you ever made any direct comparisons between its max capacity settings and Mscaler/Dave with any of the dacs discussed here?

yea I do own HQP. The only dac I have used it with has been Dave. I concluded that I preferred the MScaler and I could not find any HQP settings that worked for me despite many many people saying how much they like HQP.

I mainly got HQP because it was recommended for use in a non upsampling, non filter pass through mode as a player for the recently launched PGGB off line resampling software. (PGGB is sort of like HQP on steroids!). However I moved on to the MPD player which I felt was better in that role.

 

[Triode User]

I read other comments with interest, but I am really puzzled with a second paragraph. Re: details in the individual notes and instruments: it is actually a strong attribute of of Delta-Sigma converters like Dave.

Is Dave delta sigma?

The DAVE is not a silicon-based DAC (ie Delta-Sigma) even though it oversamples and while it is a discrete DAC and uses resistors, it is not an R2R DAC either. Based on Rob's postings, the most specific categorization that I can provide the DAVE is that it is a single-ended FPGA 20-element Pulse Array DAC and it is the only DAC of its kind.

 

[GoldenSound]

@GoldenOne Have you compared any DDCs with the May to see if the I2S input outperforms USB direct from computer/renderer etc?

I'm thinking of a Denafrips Hermes. I don't need all the I/O of the Gaia and I don't think the Iris would be a step up. I'm kind of waiting for someone to make a DDC that outputs 1.536MHz PCM. I find some that do DSD1024 but not 1.5M PCM which is odd since the bandwidth is effectively the same for each.

I have yes. Both objectively and subjectively.

Subjectively, I find that with the PLL off, it's kinda the same as any other DAC, digital source makes a difference for sure, though i've actually ended up preferring the USB to many of the DDCs i've tried. With some like the pi2aes simply being a little different but not clearly better/worse.

With the PLL on though.... the differences between all digital sources are so incredibly minor that I have not been able to tell them apart in a blind test and really would not recommend spending much on a DDC other than a decent streamer in order to get away from having a PC in the chain (ifi iGalvanic also works wonders for this). It just puts EVERYTHING on such a level playing field it's nuts. There have been times where I've THOUGHT I heard a difference, but then if I tried to pick it out blind I just couldn't.
Even optical from the PC the PLL takes it from this:

to this:

Literally, it makes straight up PC optical better than denafrips hermes I2S:

The USB itself is also quite frankly nuts, jitter is almost entirely below -170dB and I'm basically limited by the accuracy of my ADC not the DAC. (those close in spurs might actually be the result of the ADC not the DAC, would need a notch filter to check)

This is the upgraded USB module found in the Level 2/3. I don't know how the spring 2 or May L1 usb would perform. The USB also has the advantage of no PLL locking time like there is on the other inputs.

So honestly, my advice would be that the May has such an incredible PLL and USB input (which also has full galvanic isolation) that honestly a DDC just isn't needed.

 

[nagi8404]

One more question (and thanks for your help). How did you find the soundstage compared between the May and the MScaler+Dave combo. I particularly enjoy soundstage depth and almost holographic nature of MScaler + Chord dac. While other dacs seem to do soundstage width very well, I'm not sure about depth and the almost 3D/holographic effect. I am just starting out and don't have much experience with different Dacs.

Likewise can't speak for the May since I don't have one, but compared with delta-sigma DACs that I own (SMSL SU-8 and RME ADI-2) the DAVE is amazing at producing holographic sound. When I listen to classical music on the other two DACs, music sounds as if they are a 2 dimensional flat plane inside your head while with the DAVE, music just surrounds your head. The difference is like listening to a recording vs in the recording.

 

[GoldenSound]

Is Dave delta sigma?

Yes.
It does have some differences from other delta sigma dacs, but it is still a delta sigma process.

It modulates to a 1-bit level (delta sigma) and then outputs via a 'pulse array' of 1-bit converters.

 

[lairpost]

I have yes. Both objectively and subjectively.


So honestly, my advice would be that the May has such an incredible PLL and USB input (which also has full galvanic isolation) that honestly a DDC just isn't needed.

Thanks, I forgot you did do a review of the Hermes et al. A buddy did some recordings that show a difference with the May fed by USB vs. Matrix XSpidf-2 I2S. The IS2 sounded fuller and the vocals were more present. His USB path though is Nucleus direct USB to May.

I have the Intona 7055-C USB box (already had it before the May) that I believe does the same or similar job as the iFi. I also don't want to give up 1.536MHz PCM.

Thanks for your comments and time. I'll stick with USB with Intona box and Intona Ultimate cable (both acquired used at a great price).

 

[sajunky]

Something you might find interesting is it turns out that denafrips dacs might not actually be NOS after all.

I have the ares 2 in for review at the moment. And initially I was surprised both because I really didn't like the nos mode on it much, preferring OS. But also because the changes between os and nos were very different in sound to on the may.

I recorded the impulse response and.... Yeah
.... This ain't NOS.


Basically the 'nos' mode is just linearly interpolating from sample to sample, ie:it is still oversampling.
Rather than what actual nos would do which is adjust to the value of the new sample once it arrives, hold, and then adjust to the value of the next sample once it arrives etc etc.

Denafrips is instead moving in a line directly toward the next sample.

Interesting, as ASR reviewer didn't detect any abnormal behaviour. I can't properly interprete your photo, as it doesn't show sampling boundaries. What is dv/dt of the slope and averaging period?

It is interesting finding, many people say that Ares NOS is not much different much from OS slow roll off filter. While you are busy on the Ares review, I'd like to point you to the Russian forum. A guy with an ordinary osciloscope found that Holo Springs NOS is not really a pure NOS mode. A response to the sine signal looks fine only up to the 500Hz. Everything from 1kHz and above shows a stepping behaviour with approx. 6-bits resolution. Within of each 6-bits boundaries the output is randomised with a 16th discrete values. these numbers seem to be changing with a frequency. It shows a really wierd picture after reconstruction filter (the first screenshot of given frequency). Taking a shot directly from a DAC shows a randomised discrete values (the second screenshot with a sticker on top right). OS mode do not show such abnormal stepping behaviour.

100Hz

1kHz

8kHz

Test signal 16/44.1kHz 10 sec each. More screenshots and discussion here:
https://dastereo.ru/t/fizika-i-lirika-r2r-matriczy/3916/114

 

[GoldenSound]

Amir at audiosciencereview didn't test the NOS mode. He couldn't figure out how to change the filter setting so he just used OS for the review

In regards to the russian stuff, it's really hard to tell what's going on there because of the blur. But the stepping behaviour is simply because it's non-oversampling. It doesn't imply a reduction in resolution. And the "step size" will simply vary depending on frequency and sample rate. And I'm not getting whatever the issue is he's showing in that last pic.

This is a 1khz tone for example:

You can see that because it's running NOS, meaning there is no ultrasonic attenuation other than the analog LPF/reconstruction, there are the ultrasonic aliased tones way out past 40khz. As you play a higher frequency tone, these imaged tones get closer to the audible band.


So a 15khz tone:

You can see how the multiple tones superimpose more clearly by zooming out a bit like in this first pic.

The aliased tone is now just under 30khz. Still inaudible but because it's lower frequency and closer to the frequency of the fundamental, it's more visible on the waveform/scope view, and people get worried about it.

It's important to note though that audible band performance is completely un-affected here. It's just that the imposed ultrasonic tones look concerning on the scope view.
In fact the may measures quite frankly insane. With a true dynamic range of over 140dB (Even the mola mola tambaqui can't break 130dB) and THD+N/SINAD of 118dB.....so much for R2R dacs measuring badly

So yes, the Holo dacs are indeed true NOS. And the stepping behaviour is simply because of that. You can't have NOS and a 'visibly' ideal sine because they're contradictory. In order to reconstruct a sine that looks perfect in the scope view you'd need to apply ultrasonic filtering/attenuation, ie: you'd need to oversample.
It shouldn't be a concern as it doesn't affect any audible band content. But if it does concern you, the holo dacs also have oversampling available:

 

[kkrazik2008]

Yes.
It does have some differences from other delta sigma dacs, but it is still a delta sigma process.

It modulates to a 1-bit level (delta sigma) and then outputs via a 'pulse array' of 1-bit converters.

Wait, what?
I thought all Chord DACs used FPGA to process and convert.
Are Schiit Multi-bit also not R2R?
Sometimes, these differences are confusing….

 

[GoldenSound]

Wait, what?
I thought all Chord DACs used FPGA to process and convert.
Are Schiit Multi-bit also not R2R?
Sometimes, these differences are confusing….

An FPGA is a configurable/programmable IC basically and can be used to do various things. But there are to my knowledge not any dacs that are actually outputting directly from the FPGA itself. (Other than I think maybe the PS Audio dac?)

Chord uses the FPGA to oversample and do the modulation to 1-bit, but the actual analog output is done with discrete hardware as it's better suited for the task (and also because they use their proprietary pulse array design).

FPGA's can be used for all sorts of things. In R2R dacs they control the resistor ladders

In dacs like the Audiobyte Black Dragon they do the oversampling/modulation but then have two AKM chips do the actual analog output.

And in chord the FPGA does the oversampling/modulation and controls the pulse array.

An FPGA is a very flexible device that allows the manufacturer to create their own digital processing methods. BUT, there are limitations in terms of how ideal they are as an actual 1-bit final output device. Which is why even when the FPGA is outputting the 1-bit information like in chord, dCS and a lot of other products, other discrete hardware (Or other DAC chips) are used to do what could almost be considered 'reclocking' and cleanup in a way.

Schiit Multibit is genuine R2R. They use multibit chips from Analog Devices which are R2R.

Though they are oversampling only and cannot run NOS.
Subjectively I'd also say that the schiit multibit dacs definitely have their own distinct house sound. They sound very different to both other R2R dacs like denafrips/holo, as well as other Analog Devices based dacs like Bricasti. (The chip in the bricasti isn't actually a full multibit chip but still)

Also, the term "Multibit" can and often is used in different ways. It can technically describe any DAC chip which converts at anything other than a 1-bit level. Which actually is basically all modern DAC chips from AKM/ESS/Burr Brown. So be careful as some companies are using somewhat misleading marketing descriptions in this regard at the moment.

This description for example is arguably completely untrue:

 

[Thenewguy007]

I have yes. Both objectively and subjectively.

Subjectively, I find that with the PLL off, it's kinda the same as any other DAC, digital source makes a difference for sure, though i've actually ended up preferring the USB to many of the DDCs i've tried. With some like the pi2aes simply being a little different but not clearly better/worse.

With the PLL on though.... the differences between all digital sources are so incredibly minor that I have not been able to tell them apart in a blind test

Is the PLL on by default or do you have to switch it manually when the May first arrived?

 

[kkrazik2008]

An FPGA is a configurable/programmable IC basically and can be used to do various things. But there are to my knowledge not any dacs that are actually outputting directly from the FPGA itself. (Other than I think maybe the PS Audio dac?)

Chord uses the FPGA to oversample and do the modulation to 1-bit, but the actual analog output is done with discrete hardware as it's better suited for the task (and also because they use their proprietary pulse array design).

FPGA's can be used for all sorts of things. In R2R dacs they control the resistor ladders


In dacs like the Audiobyte Black Dragon they do the oversampling/modulation but then have two AKM chips do the actual analog output.


And in chord the FPGA does the oversampling/modulation and controls the pulse array.



An FPGA is a very flexible device that allows the manufacturer to create their own digital processing methods. BUT, there are limitations in terms of how ideal they are as an actual 1-bit final output device. Which is why even when the FPGA is outputting the 1-bit information like in chord, dCS and a lot of other products, other discrete hardware (Or other DAC chips) are used to do what could almost be considered 'reclocking' and cleanup in a way.

Schiit Multibit is genuine R2R. They use multibit chips from Analog Devices which are R2R.

Though they are oversampling only and cannot run NOS.
Subjectively I'd also say that the schiit multibit dacs definitely have their own distinct house sound. They sound very different to both other R2R dacs like denafrips/holo, as well as other Analog Devices based dacs like Bricasti. (The chip in the bricasti isn't actually a full multibit chip but still)

Also, the term "Multibit" can and often is used in different ways. It can technically describe any DAC chip which converts at anything other than a 1-bit level. Which actually is basically all modern DAC chips from AKM/ESS/Burr Brown. So be careful as some companies are using somewhat misleading marketing descriptions in this regard at the moment.

This description for example is arguably completely untrue:

@GoldenOne you rock!
This is one of the most concise explanation in laymen terms for the subtle complexity and differences, had no clue the FPGA was only part of the equatio. I had always understood that to be another option and DACs could fall into three buckets; Delta-Sigma, R2R, or FPGA. The whole NOS and OS, I am only beginning to understand as I assumed this was part of the digital to analog conversion.
Many thanks, and keep those videos coming and fighting the big fights!

 

[GoldenSound]

Is the PLL on by default or do you have to switch it manually when the May first arrived?

It's on by default I believe.

You can tell if it's on by the locking time.
If when you first switch to one of the non-USB inputs, or switch sample rate on those inputs there is a few seconds of time where the DAC says "LOCKING" then the PLL is on.

@GoldenOne you rock!
This is one of the most concise explanation in laymen terms for the subtle complexity and differences, had no clue the FPGA was only part of the equatio. I had always understood that to be another option and DACs could fall into three buckets; Delta-Sigma, R2R, or FPGA. The whole NOS and OS, I am only beginning to understand as I assumed this was part of the digital to analog conversion.
Many thanks, and keep those videos coming and fighting the big fights!

Glad it helped!
I'm actually working on a series of videos which will hopefully be something akin to "A foolproof guide to DACs" which will aim to explain how every step in the D/A conversion process works in a simple to understand manner.

In terms of the main categories of DACs though, I'd probably lay it out like this:


1) R2R
This is the only true "Native" PCM converter. It takes a PCM sample and converts each sample as presented using a resistor ladder. This means it can also operate without any kind of further processing or oversampling to the signal, as it can just convert the samples natively. Rather than needing to process/alter the information to make it compatible with the converter.
R2R dacs get more accurate by having more precise resistor values.


2) 1-Bit Delta Sigma
R2R dacs are difficult and expensive to make to a good level of accuracy. In order to address this, a clever workaround was created. Delta sigma, which is actually basically a switch. It can only be on or off. It can't output any voltage you choose like an R2R dac can. The trick is that if you flip that switch on/off fast enough you can 'average out' the value to the result you want. (massive oversimplification of course but still )

And so it takes the PCM information. Oversamples, and then modulates down to a very high sample rate 1-bit (on/off) data stream.
This 1-bit PDM (pulse density modulated) signal is put through an analog low-pass filter and the result is the analog signal.
(Fun fact: This is why DSD is POTENTIALLY better than PCM on a delta sigma dac. It skips the oversampling/modulation stages because the info is already 1-bit! It can just be put through the low pass filter and you're done)
1-Bit delta sigma dacs get more accurate by oversampling to and modulating to 1-bit at higher rates. The maths is that they trade precision of information for speed of information. And so the faster the final 1-bit PDM stream the better.

3) Multilevel delta sigma
This is actually what MOST consumer dacs nowadays are. Making a discrete R2R or switched resistor dac accurate to even 16 bits is REALLY difficult. But 3 bits? 6 bits? That ain't too hard.
And so these are sort of a mix of the two designs. They still do delta sigma processing, oversampling and then modulating the signal down to a lower bit depth. But they modulate to normally 6 bit, instead of 1. And then have a fast switched-resistor dac to convert the 6-bit information.
This avoids the challenges of either needing to have extremely fast 1-bit converters (or more often the processing power needed for good, fast modulators), or extremely precise 16/24 bit converters.
Most of these dacs CAN convert 1-bit info if needed though (like when being fed DSD).

There are lots of variations in the designs and even the three main companies (AKM/ESS/Burr brown) who all use the third category, still do things differently.
But personally I'd probably class the above as the three 'main' DAC categories.
FPGAs are a bit confusing, but only cause they are inherently so flexible and used for all sorts of different things. I wouldn't at all categorise FPGA dacs as their own thing though as they are almost always used as part of one of the above 3 designs. But it is normally a nice thing to see as it means the manufacturer is doing proprietary stuff rather than just using an off-the-shelf dac chip

(Also just to avoid starting arguments, I don't think any one type of DAC is "best", there are pros and cons, good and bad designs, and reasons to use or not to use all of them. If there was one 'best' everyone would be using it)

 

[sajunky]

@kkrazik2008. It is a Chord and dCS who try to confuse users, many others too, I suspect a Holo Audio is on the list.

Schiit Multibit is genuine R2R. They use multibit chips from Analog Devices which are R2R.

Though they are oversampling only and cannot run NOS.
Subjectively I'd also say that the schiit multibit dacs definitely have their own distinct house sound. They sound very different to both other R2R dacs like denafrips/holo, as well as other Analog Devices based dacs like Bricasti. (The chip in the bricasti isn't actually a full multibit chip but still)

Also, the term "Multibit" can and often is used in different ways. It can technically describe any DAC chip which converts at anything other than a 1-bit level. Which actually is basically all modern DAC chips from AKM/ESS/Burr Brown. So be careful as some companies are using somewhat misleading marketing descriptions in this regard at the moment.

Actually Schiit DACs are not pure R2R. The same is with Holo Audio. There are more classic ladder types with two practical textbook types and combination of some other types. The following will explain a difference.

1. R2R. A ladder consist of equal value R and 2xR resistors in a specific layout. Examples are Burr Brown PCM63, PCM1704 chipDACs or discrete implementation from Denafrips and Audio GD.

2. Binary weighted. A ladder consists of series resistors where each resistor has a double value of the previous one.

3. Segmented. A portion of the ladder is R2R and other is some other type. The most known chipdac segmented architecture is TDA1541, a later version is TDA1387. It is a type of the ladder used in Schiit multibit DACs. By example Bifrost 2 and Gunmir use Analog Devices chip AD5781, Yuggdrasil AD5791. Holo Audio DACs are also examples of a discrete implementation segmented architecture.

Segmented architecture is relaxing requirements for the tolerance of resistors, it makes a DAC to look better in DSP eyes (measurements), but does it sound better? Not really. A typical example is a TDA1541, it has never proven to sound better than PCM63 which is a true laser trimmed R2R.

There are two reasons for the inferior sound of segmented implementation. One is a temptation to reduce costs, the other is that segmented ladder do not respond equally to a fast transients. These dynamic effects cannot be measured, but our ears are very sophisticated devices and process sound differently than a current lab equipment. Holo Audio commit a bigger offense. For a pure marketting reason they are adding a separate compensation ladder that breaks a dynamic response even further. It is why I am suspecting a digital preprocessing is in place to deal with the issue (see my previous post).

You are right a term 'Multibit' frequently is used in different ways, I would say - incorrectly. A Schiit use is correct, their ladders are not true R2R, so they call it 'Multibit'. An example of incorrect use is for a modern Delta-Sigma DACs implementation, either chips or discrete. Details of chips DACs are hidden under NDA, Ares II DSD decoding is described as "6-bit DSD (32 steps FIR Filters)". It means that DSP is required to feed each resistor, as opposed to a traditional combination logic in ladder DACs. A correct term is a multistream single-bit DSD processing, or just "multistream processing", "multistream DSD".