HOLO Audio Spring DAC - R2R DSD512

[SwissBear]

Hi everyone,

 

I’m very interested in this DAC. Before making the plunge, I would like to know if some of you have an experience of feeding an SPDIF or I2S flow into it, clocked with a high precision clock like the Mutec Ref-10, and noticed that the DAC was using the external clock signal instead of its internal one.
Thanks.

[SwissBear]

1 minute ago, wouterk said:

Hi,

 

Mine is the original Spring DAC - Level 2.

Initially I used it with Singxer SU-1 - USB in, I2S out.

I also use the ISO Regen, with a linear PSU to clean up the USB signal.

I decided to take the Singxer out, as this gave me a cleaner sound as it removes one element in the chain. The new version of the DAC has better isolation on the USB, so it should perform better 'as is'.

Thanks @wouterk. In my case, the idea would not so much be to "galvanicaly" isolate the DAC, but to provide it with a better clock signal, which is the goal of the Ref-10 https://mutec-net.com/product_ref10_se120.php

[SwissBear]

On 12/2/2020 at 3:20 PM, SwissBear said:

Hi everyone,

 

I’m very interested in this DAC. Before making the plunge, I would like to know if some of you have an experience of feeding an SPDIF or I2S flow into it, clocked with a high precision clock like the Mutec Ref-10, and noticed that the DAC was using the external clock signal instead of its internal one.
Thanks.

Thanks again to @wouterkfor answering my question.

 

I have gone through most of the thread an I have noted the following:
- At the beginning (say Rev1 of the Holo Spring), some users have attempted to feed the DAC through its I2S interface, mostly with Singxer SU-1, and notices improvement in the sound quality. This was partly attributed to better galvanic isolation and partly to better clocks in the SU-1 than in the DAC
- From the beginning, some other users, @Miska in particular, have challenged this observation, publishing measurements of the jitter of the Spring in Rev 1 and saying that such improvements, if measurable, would be hardly noticeable
- @scan80269 has made an interesting experiment which tended to prove that the I2S input was using the clock signal sent by the I2S flow and not by the clock of the DAC
- Since Rev2 of the Spring, a new USB interface has been released, together with better clocks, and this has led to measurements of improvements of Rev 2 vs Rev 1. Since this release, the majority of users has reported having dropped their I2S connexion, in favour of direct USB connexion.
- I have noted that the last measurements published with a Singer SU-1 connexion were reporting a higher noise floor with the interface than without it, which does not favour this kind of feed.
- I have read that there was a sweet spot for using this DAC and this was in NOS, with input sampling frequencies of PCM 1.536 MHz or DSD 256 when upsampled via HQPlayer
- I have not read of any attempt to listen/measure to installations where the Spring would be fed by a more sophisticated USB/I2S interface, which would be clocked by a high precision clock like the Mutec Ref-10
It my understsnding correct ?
Questions:
1. Is there a link between the oversampling via HQPlayer and the reduction of jitter at the output of the DAC ?
2. Some users are reporting an even better experience with the Holo May than with Spring. Is this understandable from a technical/engineering PoV, in particular due to the fact that May is a dual mono DAC ? Under circumstances when listening these boxes at home is not feasible, is it worth the price difference ?

 

[SwissBear]

34 minutes ago, barrows said:

I would not speculate that "dual mono" construction has much, if at all, to do with any improvement of the May over the Spring.  Although the argument, technically, for true and complete dual mono construction is sound, in practice, I have not found dual mono to actually produce better sonics: I have made DIY DACS with both full dual mono construction (fully separate power supplies for the L and R channels) and what I would call "pseudo dual mono" construction (shared transformer for L and R channels with separately regulated power rails), and as long as the capacity and impedance of the supply is kept low, their is no discernible difference in sonics between the two.

I would expect the improvements in the May to be mostly derived form better power supply design in general, and better analog stage design, as well as, perhaps, subtle improvements added to the conversion stages (most designers will learn something from iteration to iteration and incorporate subtle improvements with any new layout of a given design).

In any case, the May can be seen to perform better than the Spring, in a measurable senses nd in reported sonics, so for those looking for the best version of Holo Audio's approach to DAC design, the May is the clear choice.

Thank you @barrows for this documented response :-) 

[SwissBear]

15 hours ago, SwissBear said:

Questions:

1. Is there a link between the oversampling via HQPlayer and the reduction of jitter at the output of the DAC ?

@Miska: Hi Jussi,

I observed that the jitter measurements you are posting are made with a reasonable level of upsampling in HQPlayer (PCM352, DSD 256 or DSD 512), and that you are sometimes precising that the measurements are made with your workstation, as an indication that there is no tuning on the music server side. Am I right ?

I tend to infer from it that HQPlayer upsampling has a positive effect on jitter reduction at the DAC, even when feeding the DAC via its USB input.

Could you kindly share with us some explanations about this ?

Thanks,

Pierre

[SwissBear]

46 minutes ago, Miska said:

 

Yes, that's the case.

 

 

There are likely multiple reasons. One is moving the processing out of the box and performing it asynchronously. Another is that generally higher rates lower the jitter because timing contribution of individual samples is less, the clock divisors are smaller and the distribution becomes different. Running actual conversion stage at NOS 44.1k is absolute worst case in many ways.

 

I'm not sure what is meant by the "even" part, because USB is best interface the DAC has. :)

 

Thanks a lot for your answer, which makes much sense to me :-)

[SwissBear]

6 hours ago, wanta911 said:

 

I just tried updating the firmware with 31.40, I followed all the instructions with level changes to get into configuration mode but it didn't work when I executed the bat file.

 

File not found

Start updating firmware with and 'xprog' is not recognized as an internal or external command, operable program or batch file.

Update firmware failed.

Please make sure you have extracted the zip file you downloaded before trying to update and you launch the update process from the extracted folder.

[SwissBear]

@Miska

Hello Jussi,

I have extensively measured the output of my Holo Spring 3 lately, with an ADI-2 Pro and REW.
I have been amazed to witness, with these measurements, the incredible behavior of the DAC when fed with your DSD 256 upsampled flow, even when the DAC was connected directly via USB to the MAC. This was a big surprize for me as I expected some nasty EM pollution to flow from the Mac Mini to the DAC, which did not seem to happen. So kudos to you for this incredible perfomance, and also to Holo for the incredible galvanic isolation !

So I dared to listen to it in this simple configuration, which I had never done, and I was very pleased.
What surprized me too was the lack of additonal power shown when feeding this DAC with DSD 256/512 over NAA. As if the simple setup of the Mac Mini connected via USB to the Holo Spring 3 was able to squeeze the very best out of this DAC.
Is this in accordance with your own measurements ? Have I missed anything in my process ?
 

[SwissBear]

On 8/23/2023 at 10:07 PM, Miska said:

NAA is still useful when you want small fanless endpoint and locate the HQPlayer server in a different room, which is the case for me. So I have the UP Gateway running as a NAA, hidden behind Spring 3. While big HQPlayer server with relatively loud fans is running in corner of my office. So the loud fans on the server won't disturb listening...

 

Hi, 

 

Yes, I am fully aware of the benefits of your network infrastructure that I value everyday having 2 HQPe Linux machines. Kudos also for that.

 

One other thing I noticed in my measurements, and which I would like to understand more when you have time to comment, is that distorsion produced by the Holo Spring 3 seems to be a monotonic function of upsampling resolution (see attached).

 

 

 

 

I have read that HQPlayer is upsampling the redbook format 64 times in DSD064 prior to sending to modulators and 512 times in DSD512. 64fs is the upsampling resolution used in the Tambaqui if I am not mistaken, which is already huge.

 

In this context, how do you evaluate the best compromise between a potential decrease of quantisation noise, which would be brought by higher resolution, and the increase of distorsion which is measurable ?

 

Also, have you found, in the course of your measurements, DAC technologies which behave better in the field of distorsion when using high sampling frequencies like DSD512 ? Are T+A 200 or Teac 701n any better at this ?

 

Thank you.

[SwissBear]

11 minutes ago, Miska said:

 

I don't get as drastic differences (with ASDM7ECv2):

Thank you . I was comparing DSD64 to DSD512.

 

11 minutes ago, Miska said:

It is up to balancing. Level of ultrasonic noise drop compared to distortion increase is something like every doubling of sampling rate increases distortion some 3 - 6 dB and ultrasonic noise moves to twice higher frequency and drops some 20 dB. If you use for example class-D amplifier, you may want to focus on reducing ultrasonic noise levels. At DSD512 the ultrasonic noise level drops below analog noise floor.

Thank you. This ultrasonic noise level was the part missing for my understanding.

Grateful for your time and explanation.

 

[SwissBear]

4 hours ago, Miska said:

 

Your DSD512 results are just worse than what I've got. Which modulator did you use?

 

 

ASDM7EC-Super

[SwissBear]

I checked:

• the influence of modulators on distorsions, and it seems to be less of an adjustment variable than upsampling frequency in this respect
• the subject of ultrasonic noise. According to his claims (https://hometheaterhifi.com/editorial/oped/secrets-qa-with-john-siau-of-benchmark-media-systems/), it seems that John Siau had taken the subject of ultrasonic noise into consideration when designing the Benchmark AHB2, which would reduce the vulnerability of my system to these problems. In case you have a different opinion, I would appreciate if you can comment on this.

Thanks again for your help.

[SwissBear]

6 hours ago, Miska said:

 

Here's what I've got today with DSD512 and ASDM7EC-super:

 

I would use DSD256 in your case, as it fits all PCM hires available today. But up to you of course!

 

Pretty big difference indeed. Thanks a lot for checking. It’s worth for me trying to understand where the differences lie.

I sincerely appreciate your support 🙏

[SwissBear]

On 8/29/2023 at 12:26 AM, Miska said:

 

Here's what I've got today with DSD512 and ASDM7EC-super:

 

I would use DSD256 in your case, as it fits all PCM hires available today. But up to you of course!

 

 

Can I ask the level of HQPe output you are using when making your measurements?

[SwissBear]

23 hours ago, Miska said:

 

Same as listening volume so -3 dBFS. Source files in most cases are 0 dBFS 32-bit TPDF dithered 44.1k PCM (for some tests like TIM, higher rates are needed though). So the source files are always the same and will never cause inter-sample overs, no matter what I'm measuring, PCM or DSD.

 

Before poly-sinc-gauss group I used poly-sinc-ext2. But now either poly-sinc-gauss-long or poly-sinc-gauss-xl.

 

All my three Spring's are L2 versions. And audio band measurements are done from balanced outputs. In above case over 5m long "microphone" XLR cable, since I didn't want to move the devices... But usually I use 1.5m long XLR cable. Wide band measurements (5+ MHz) are done from single-ended outputs using RG58 coaxial cable.

 

 

I corrected my "volume adjustment" mistake. I am now fully on the same page as you are. DSD256 definitely is the sweet spot for me. Thanks again for your help.

[SwissBear]

23 minutes ago, Zauurx said:

The next question (for me). 😉

Do I keep my A26 (DSD direct in 256) connected to my Benchmark or do I try a Cyan in 512 ? (or 1024 but I'll have to find an external NAA and perhaps lose the gain of the streamer integrated into the A26).
@Miska... I'm waiting for the Cyan measurements and tests !

 

The answer to this question is, IMO, more emotional than technical. 
Technicaly, the Cyan 2 seems to be at the level of the Spring 3 in terms of SINAD on the DSD channel. 
So this could be a real DSD giant killer.

But as you have witnessed with your A26 and DSD256 modulators, the techno offered by Jussi already delivers a perfect reproduction of music at this stage. I am not sure you will get anything more with the DSD512 modulators given that you have your Benchmark amp. At least I have not been able to witness such benefits in the same configuration.

 

Would the DSD1024 modulators provide you with more emotions, a better engagement to music, to the price of a least perfect technical reproduction ?
This is what happened to me one week ago when I discovered them.

 

If I may, it's probably better to wait for this Holo Cyan to be available in Europe, with the return possibilities offered by EU laws, and judge by yourself listening to it... than to make a decision based on measurements which will not tell you anything...

[SwissBear]

21 hours ago, Miska said:

DACs do sound different, despite similar SINAD numbers. For example Holo and T+A do sound different, while both perform extremely well in terms of SINAD figures.

 

Differences in conversion stage and analog filter/buffer stage implementations do make a difference. And those differences are more complex than what a simple SINAD figure can express.

 

 

Do you expect the Holo Cyan to sound differently from Holo Spring and Holo May, knowing that they share the same lineage, and that they display similar SINAD according to the sellers' claims ?