HOLO Audio MAY DAC

[Miska]

7 hours ago, reeve_mike said:

  

@Miska I note that you use the May's maximum PCM rate but a somewhat lower than maximum DSD rate.  Would you be willing to share the reason for sticking to DSD256?

 

 

I don't use Holo Audio DACs in PCM mode for anything else than testing.

 

But the DSD mode performs best at DSD256 which is also maximum you can do with current hardware and EC modulators.

 

[Miska]

23 minutes ago, damascato said:

Hey guys, while waiting for a Usbridge, I tried to connect the May via USB to my MacBook Pro (with an usbA—>C adapter).

 

When putting the Tidal app in exclusive mode, the volume becomes all of a sudden incredibly low so that to have a decent listening level I need to crank up my amp almost to max (which is not what usually does, it’s a Hegel 190).

 

when I had the May connected vis coax to a usbridge this never happened.

 

any ideas/tip?

 

Your May has volume control?

 

[Miska]

5 minutes ago, damascato said:

i don't really think that the May has any volume control.

 

I was just wondering if there was a pre-amp module variant of it like there is for Spring 3.

 

In any case, if there's a digital volume control or similar, it could be controlled through USB and have low default value... So go to sound preferences, or Audio & MIDI Settings and check if there's some volume control...

 

Technically USB Audio Class has such capability, while S/PDIF doesn't.

 

[Miska]

On 8/6/2021 at 3:15 AM, WilliamWykeham said:

"theoretically it can do ... 3.072Mhz", what are the points in the audio chain preventing the theoretical limit from being reached today.  

 

USB Audio Class...

[Miska]

17 hours ago, camott said:

Sure but it’s interesting that most of the hype about the May DAC is around it’s performance with NOS PCM, via the highly engineered R2R ladder and FPGA error correction. If you are just using DSD why spend all that money on engineering that you aren’t using?

 

Same way if you are using just the R2R PCM mode and not the DSD side, why spend all that money on engineering that you aren't using?

 

Reason is that leaving out the PCM side wouldn't make the device so much cheaper or simpler. Since you'd still need the same USB interface, the same PSU's, the DSD conversion side, same analog stages, same case work, etc... You would just leave out bunch of SMD resistors and that's all.

 

Quote

In DSD mode how much better is it than a much cheaper RME or topping DAC? (I don’t know the answer myself).

 

You could ask this same question about PCM mode.

 

[Miska]

15 hours ago, barrows said:

Note to anyone doing these kinds of tests: many DACs mute their outputs when no signal is present, so one cannot test the noise floor accurately with no signal.  Make sure to be giving the DAC a signal to be sure the output is not muted.  Of course if you are going to be testing this at full gain of your system, be absolutely sure that your signal is digital silence!

 

And for example sending digital silence is not enough, since DAC chips like ESS Sabre detect this and engage mute (they mute on any repetitive same sample value). To properly test noise, you need to send dithered silence (and make sure it matches the DAC's resolution, so that your dither doesn't disappear somewhere on the way due to truncation).

 

[Miska]

NOS PCM at 44.1k just produces horrible amount of distortion, of course if one likes this, it is another thing. But it is not accurate rendering of the original signal. It will sound "smooth" because of the 3 dB high frequency roll-off that is by-product of running D/A conversion at 44.1k.

[Miska]

10 hours ago, luisma said:

Went to 30.12 to be able to do PCM 1536, it plays at like half speed when upsampling PCM and distorted, I am assuming this is because of older gen 6th gen Intel renderer NAA device.

 

Anyone care to recommend what to use as NAA to do PCM 1536? sometimes small cheap and simple?

 

What do you run NAA on? NAA OS image?

 

[Miska]

2 hours ago, BrownMagic said:

My mac mini m1 is unable to do DSD 256, ASD7EC. 

 

Mine can, so likely there is some difference in settings? What filter are you trying to use? For example poly-sinc-ext2 should work fine.

 

[Miska]

11 hours ago, Cogito said:

Sinc L is the best filter in HQP IMO. It works fine for pcm output, but for converting pcm 44.1 to DSD 512, it is choking on my system powered by Ryzen 9 5900x.  Jussi says its because Sinc L is suing 32 mil taps in that configuration.  

 

Now that I calculated I got 64M taps, but anyway processing that many taps at 20+ MHz sampling rate is quite a lot of processing work. It may work on a very powerful GPU though.

 

[Miska]

7 hours ago, pis99 said:

Could you point me to an embedded version that I can buy to work with powerful GPU from my son's gaming desktop? It is I7-10700F/32G/Z590/RTX3090 with wired ethernet on the motherboard. I know nothing about linux but do want to try embedded version. I already have desktop version running on Mac M1. 

 

You need Ubuntu Server 20.04 LTS for HQPlayer Embedded with CUDA support (Nvidia GPU). And install latest Nvidia drivers on it. It is a bit easier to setup Ubuntu Desktop 20.04 LTS and HQPlayer Desktop if you are not familiar with Linux. You can have Windows + Linux dual-boot on the same machine, I have such setup for example on my work laptop.

 

You could also check the performance on Windows 10, if the computer already has Windows. With latest Nvidia drivers you can check the CUDA performance with RTX3090.

 

[Miska]

4 hours ago, luisma said:

Right now Audiolinux, I was using Ubuntu in the past, I can use Ubuntu just I found easier to run AL on a USB stick.

 

I would recommend to try the same with my NAA OS or HQPlayer OS image. You can easily write it on USB stick and boot from it, no need to install anything.

[Miska]

51 minutes ago, luisma said:

Yes I know but which hardware USB will support the PCM 1536 on the May..... ???

 

Intel, with suitable drivers. Since you are using Intel hardware, you should try with my OS.

 

[Miska]

17 hours ago, pis99 said:

Maybe a stupid question but can I buy your embedded HQP OS that is bootable from a USB sticker with CUDA support? I try to not engage with my son's gaming computer and hoping nothing changed with system HD.

 

No, HQPlayer OS doesn't support CUDA. CUDA (for Nvidia) and ROCm (for AMD) is only supported on Ubuntu.

 

[Miska]

If the USB interface has digital volume, you can use HQPlayer OS as NAA and perform following:

 

1. Login as "root" (no password)
2. Run "aplay -l" and take note of card number of your DAC
3. Run "alsamixer -c N" where "N" is your card number
4. Adjust possible volume control with arrow keys
5. Hit ESC to exit
6. Run "alsactl store" to store the settings for subsequent boots

 

This way you can get around with interfaces that have volume control and you want to set fixed volume.

 

In addition, HQPlayer Embedded supports optimized hardware + software volume combination where hardware volume is used primarily and software volume secondarily. But this makes sense only when DAC has actual USB controllable analog volume control (instead of digital). Setting this up is outside of this context, but it is documented in HQPlayer Embedded readme.txt

 

[Miska]

1 minute ago, michaelvv said:

A ssh access would be a major upgrade to hqplayer embedded image :-)

 

It is not supposed to be necessary to access console on the image. However, I have included some small tools for console access to deal with special cases.

 

And remote root access without strong random passwords is a security risk.

 

[Miska]

2 hours ago, MichaelHiFi said:

And about that driver?

 

Windows PC?

 

You need to have ASIO backend selected in HQPlayer and the Holo Audio driver as output device.

 

In Roon audio settings make sure you have only HQPlayer enabled, and make sure the Holo Audio driver is not enabled there.

 

[Miska]

1 hour ago, simonklp said:

My feeling is that the SQ from the DSD side of May may not be as good as that of T+A DAC 8 DSD.

 

At least technically, DSD side of Spring 3 is at least as good as the PCM side...

 

[Miska]

1 hour ago, 87mpi said:

This is the specs of analog output:

MAY: 

48k NOS PCM:

THD+N 0.00017% @1K(-115dB)

DNR 130dB

2.9Vrms (RCA) , 5.8Vrms (XLR)

128X DSD:

THD+N 0.00025% @1K(-112dB)

DNR 115dB

1.45Vrms (RCA), 2.9Vrms (XLR)

 

SPRING 3: 

PCM: THD+N 0.00032% @1K(-110dB)

DNR 127dB

DSD: THD+N 0.00035% @1K(-109dB)

DNR 115dB

Same output voltage 

 

The DNR difference between pcm and dsd is 15dB (or 12dB for S3) 

Could it be for this reason that many prefer PCM on Holo dacs? 

 

However, 48k NOS PCM has imaging distortion in tens of % range... So especially high frequency transients are totally off and incorrect.

 

I don't know how they measure those figures, since I get different results.

 

Here's my Spring 3 THD measurements 16x PCM vs DSD256:

 

And IMD test results, for the same:

 

[Miska]

55 minutes ago, firedog said:

But in both sets of graphs it looks like the noise is mostly 120 db down. Is it audible at all?

 

Sometimes. It is quite complex topic... So I'd like to avoid sweeping statements about audibility one way or other.

 

These are not the only measurable differences between PCM and DSD sides either. To me, those sides sound different. And I personally prefer DSD side.

 

[Miska]

5 minutes ago, Cogito said:

Sounds at -120dB are not audible by themselves, but Rob Watts of Chord explains that as the noise floor is lowered the depth of sound stage is increased. See video below.

 

I don't agree with Rob Watts in general with most things. Not on this either.

 

For me, subjectively, DSD side has more authority, weight and kick (as I've explained earlier, with a bit different words). Without becoming congested as easily. But I wouldn't state it is about noise floor level as such.

 

[Miska]

I have not noticed any burn-in stuff, the Spring's I've had for years haven't changed at all over time. Always sounded the same...

 

Which I consider a good thing, I'd hate gear that keeps drifting around...

 

[Miska]

41 minutes ago, Ozan Bolat said:

I like the idea there that if the idea is to keep PCM PCM then the chosen filter should be as neutral as possible : is that a correct assumption regarding Closed Form M vs often preferred/quoted M and L? 

 

No, it is not correct assumption. Apodizing filter is needed to correct issues in PCM source content, filter choice generally doesn't differ whether you output PCM or SDM.

 

I personally don't prefer the filters you mention. I mostly stick to poly-sinc family. Sometimes sinc-S or sinc-Mx may be useful though.

 

56 minutes ago, Ozan Bolat said:

(with tracks without apodizing issues)

 

Unfortunately not many of such exist in RedBook format...

 

Using an apodizing filter on tracks without apodizing issues is equivalent of using non-apodizing version of the same filter.

 

It is good to remember that at least 90% of content recorded in 1990 - 2000 is result of SDM -> PCM conversion. And practically 100% of content recorded since 2000 is result of SDM -> PCM conversion. So you won't find much "pure" PCM content anyway. There's a lot of "PCM" recording gear using for example TI/BB PCM4202/PCM4204 ADC chip, meaning that the recording started it's life as DSD128 and got converted to PCM using the on-chip DSD-to-PCM conversion. Many of praised "PCM" recordings are made with such converters. These on-chip conversions (or software rate down-conversions later in mastering) leave their fingerprint on the data, that you can deal with using apodizing filters.

 

[Miska]

13 minutes ago, firedog said:

I just listened to a 24/96 version of that track and it sounds like finger snaps to me. Acc'd to the Wikipedia album page, it's finger snaps. Interestingly, when I use HQP to play it back in DSD 256 7EC, I'd say it sounds less like finger snaps (less fleshy, more hard/sharp), but I'd still  identify it as finger snaps if you asked me. 

 

Same upsampling filter in both cases? Do you notice a difference for example between poly-sinc-short-mp(-2s) and poly-sinc-gauss-long?

 

[Miska]

If you don't use same upsampling filter for both PCM and SDM cases, you are not actually comparing PCM vs SDM, or the modulator, but something else.