Upsampling - HQPlayer & The Chord M-Scaler

Martin Virgo

Introduction

This article has been a LONG time in the writing. This has been for two principal reasons: Determining how much technical detail should be included; and, software issues, of which more later.

Its purpose is to compare and contrast the Chord M-Scaler and HQPlayer, two very different ways of upsampling.

The problem with writing technical, or technical-lite, articles is that things can become rabbit holes damned quickly.

For instance, just look at that title. What is upsampling? Perhaps it should be oversampling? What’s the difference?

For the record: Have a look for yourself. If I commit to any answer that is a simple sentence a raft of ‘experts’ will appear to say why my approximation is inaccurate.

I have chosen to use the word upsampling as a stand-in for any form of sampling & frequency change of an input data stream within this article. My interest is whether ‘it’ sounds better.

Upsampling is something that many of us will have toyed with for many years. A cheap and easy way to an upgrade? The many methodologies I have tried over those years have all made a difference; the results have been mixed.

In the last couple of years, I have been using the Chord M-Scaler, which is a hardware upsampler. This has had a consistent effect, regardless of the DAC architecture. It resolves musical information and places it in a wider and deeper sound-stage. The unit currently costs £3,495.

Due to its dual BNC output the M-Scaler full-fat output can only be accepted by a Chord DAC, which is a shame. The addition of a USB output would have been welcome. However, Rob Watts, the designer and Chord digital guru, would probably state that the M-Scaler was designed to work primarily with his DACs; using an approach that is unique to Chord.

HQPlayer, a software upscaler, has been under constant development by Jussi Laako for many years. It has always been a HIGHLY granular way of digging into the upsampling methodologies, and so does not have the most user-friendly or straightforward user interface. However, does it work as well as the Chord M-Scaler?

For this exercise I decided to use Roon. This has been making strides as a friendly user front end onto local and online music libraries. To assist in making this available to users it has developed software interfaces that find and make available endpoints for music replay. It also has digital signal processing (DSP) to handle the audio in a variety of ways, including its own version of upsampling, via its MUSE functionality. This all runs over the bespoke Roon network music streaming protocols.

To additionally stir the miasma that is digital processing the Chord M-Scaler, HQPlayer and Roon upsampling can be combined, as well as used on their own.

I discussed this with our erstwhile editor, George, who thought there might well be an article of interest here.

A further personal boost to this idea was my review of the Synthesis Roma DC69 DAC. I was much taken with the Roma, which was further improved when fed from the Chord M-Scaler. The M-Scaler output had to be constrained to 192k, using the SPDIF coax output. However, the Roma’s USB input could take signals of up to 384k; while this was not supported by the M-Scaler it was by the ever-flexible HQPlayer.

One stark contrast in this comparison is cost. However, HQPlayer has its own set of complex requirements, depending on the use case. As a minimum you will need a capable server on which it can run it heavy mathematics. Further, if you wish to run that computer at a distance from the audio equipment you will need a box that runs the HQPlayer network client, and outputs an audio-compatible stream; such as the Sonore Rendus.

With this in mind I contacted Martin Smith of Audiostore and arranged to borrow a freshly built Prestige Server, together with a Sonore ultraRendu and Uptone SMPS. My Prestige server continues to do sterling service but doesn’t have the grunt to start doing full-bore HQPlayer upsampling.

Jussi Laako, the author, owner and maintainer of HQPlayer, kindly supplied me with a temporary key for his software.

My heartfelt thanks to both of these gentlemen.

I then ran a series of tests. These involved the Chord Qutest and Synthesis Roma DC69 DACs. I used the Audiostore Prestige Ultra server both locally to the HiFi and remotely. The remote setup necessarily made use of the Sonore ultraRendu in the playback chain.

In each of the server setups I ran through using each upsampling technology as a standalone, and then in combination.

MANY hours, days and weeks later I finished the task and had voluminous notes. How on earth was I to communicate this with you, dear reader?

Attempt One

My initial idea was to write the article as a technical introduction and travelogue, writing up each step and what I heard; covering the issues and solutions.

Frankly the resulting report was highly unsatisfactory. The technical side was hard to judge, how much detail was needed? How readable was it for our readership, and as important, were the simplifications misleading?

I felt the result was as stodge-like as thick porridge. It was sleep inducing and difficult to see the wood for the trees. Time for a re-think.

Attempt Two

I cut down a LOT of the detail. The result, a slightly runnier & simplified stodge; and was still opaque.

Attempt Three

THIS article. I have tried to explain the difficulties of writing something that will retain your interest, and give you a feel for the essential findings that I made.

As such I have:

• Simplified the theory further;
• Introduced the testing system;
• Removed the detailed listening notes, replacing them with a series of questions & answers;
• Supplied a Qutest conclusion (server local and remote);
• Supplied a Synthesis Roma conclusion (server local and remote); and
• An overall conclusion.

I have also supplied two Appendices:

• USB & SRC-DX; and
• HQPlayer settings.

The aim has been to pique your interest and encourage you to experiment with your system.

A BIT of Theory

As the Chord M-Scaler handles PCM data streams I ignored DSD for this article.

Put simply, digital music is recorded in discrete packets multiple times per second, the rate is quoted in kHz per second, e.g. 44.1kHz.

Each packet will capture an amount of data, e.g. 16 bits.

More data can be captured at each sampling, e.g. 24 bit rather than a measly (sic) 16 bit.

With each additional ‘bit’ the quantity of captured data is doubled, so 17 bit captures twice the amount of 16 bit.

The implication is that increasing the data captured, and/or the frequency with which that data is captured, improves the ultimate sound quality; this is not a given in my experience.

The intention of high definition audio is to improve the conversion of the discrete data back into a continuous waveform. However, if we have two very close data points can we mathematically create the ‘missing’ data points? This is what upsampling provides.

This data must be converted back into a continuous analogue waveform, via a digital to analogue converter (DAC).

If you want to do the deeper dive then please start by investigating the following links:

This is only a starter for ten!

Today in our systems there are potentially a range of upsampling possibilities, including:

• Software
  Such as Roon and HQPlayer;
• Hardware
  Such as the Chord M-Scaler or suitably equipped digital data converters (DDC); and
• Your DAC
  Yes, I know that this is hardware, but I thought it was worth an independent entry. All DACs inherently upsample during reconstruction, and some do further upsampling, including Benchmark & Ferrum.

There has been much talk and experimentation comparing and contrasting different approaches, such as HERE.

DACs have a number of different architectures: R2R ladder dacs; FPGA dacs; and, chip-based dacs. These architectures may be mixed, using an FPGA for a level of signal ‘shaping’ as well as a chip, which may be R2R.

For this exercise I used two DACs: The Chord Qutest, an FPGA DAC; and, a Synthesis Roma, a chip-based DAC using an AKM chip. However, PLEASE remember that any DAC is far more than the base DAC processing unit.

Testing System

I do not like having computers near my HiFi as they generate noise. However, I thought that you, dear reader, would want to know how upsampling worked having the computer direct into the DAC.

As you can see, I did not end up using the supplied Uptone SMPS. I will say that this worked very well with the Sonore ultraRendu, in part due to fine design work by John Swenson. Unusually this SMPS does link to the mains earth via its three pin mains lead.

While the sound quality via the Uptone SMPS was good it led me to revisit some past experiences and so I set up a ZeroZone LPSU which I own, this outputs up to six amps at 8.1 volts. I then used two LT3045 in parallel to supply up to two amps of very clean power to the ultraRendu; for another lift in digital performance.

Martin built and supplied one of their new Audiostore Prestige Ultra servers. This is a BIG step up in terms of computing power. Speaking to Martin is always interesting.

Please note: This is NOT a review of the components. I have primarily used the Sonore ultraRendu & Audiostore Prestige Ultra, but I have only used a fraction of their capabilities, and in only a single context. These are products I have used over a number of years and have a healthy regard for both.

HQPlayer Software

As good and well maintained as HQPlayer IS, it has an extra complication a normal hardware solution, it has to be used within a software stack. This means there is a level of complexity & dependancy.

The installation and use of HQPlayer is simple, when using the Prestige Nova operating system. This is a fork of the Small Green Computer SonicOrbiter software, managed by Audiostore. It is all managed via a web graphical interface.

However, if something does go awry trying to diagnose the issue may be less than simple; although Jussi will be on hand and can be found actively on various forums.

In the case of this article, my initial run with HQPlayer was full of issues, in terms of stability and usability; although it was possible to get a view of the sound quality.

My feeling was that the issues were not in HQPlayer, and I believe this proved to be the case.

When updates arrived in the surrounding systems, I installed them. This solved the issues I was having.

Rather than re-run the listening tests I chose to remove the usability comments and insert this section.

Test Tracks: The Good, The Bad and the Bright

The tracks were selected to allow:

• Familiar Qobuz sourced tracks known to highlight different parts of the audio spectrum;
• Different bit depths and frequencies to highlight issues in the reproduction stack;
• Comparison of older and modern tracks, with their different mastering priorities;
• How problem tracks were presented.

Qobuz Playlist: https://open.qobuz.com/playlist/18005717

Problem Files: https://open.qobuz.com/playlist/12623970

Audiostore Prestige Server Local Installation Notes

Chord Qutest

I connected the USB cable to one of the two USB 2.0 Type A connectors on the Audiostore Prestige Ultra, and this to the Qutest USB 2.0 Type B input. Having already set up HQPlayer on the Prestige and within Roon it was immediately available.

I tested the server direct into the DAC and via the SRC-DX, allowing the use of the dual BNC DAC inputs.

Testing: Key Questions and Answers

I will NOT be publishing my listening notes. They were large and, of necessity, highly repetitious. In order to give a more digestible form I have asked what I believe are some key questions, and then answered them.

Does Upsampling Work?

Technically, undoubtedly. Musically? Now I will say the answer is yes. I was less convinced a decade ago, where I would have said that my experience was that I could hear the effect and preferred it with certain musical genres.

My advice: Institute a change, if it is not objectionable then live with it for a day, then switch back. Give yourself time to acclimatise.

If you use Roon then you have the ability to quickly set things up in MUSE and listen for yourself.

My advice: Adjust the MUSE upsampling multiples of the underlying base rate, e.g. 4x44.1 = 174.6kHz, 4x48 = 192kHz. This will be limited by the specification of your DAC.

I have found all three upsampling methodologies ‘work’, but depending on your system you may find that the change helps with an issue, whilst reducing enjoyment in another area, you will need to decide what works for you.

Can I mix and match upsampling methodologies?

Yes. If you have these to hand then at least TRY them. Let YOUR ears decide.

HQPlayer or Chord M-Scaler?

For some this might be the BIG question. The cost bar of entry is undoubtedly lower for HQPlayer, but do remember that you WILL have to include the cost of the licence AND a good quality server, and the cost of a remote client host if you want the very best sound quality. Both the server and remote client will benefit from a better-quality power supply.

In terms of whether HQPlayer will best the Chord M-Scaler, I have found that this depends upon a number of factors, including: the target DAC; choice of network components; and, the choice of power supplies.

To my ear they are in the same ball-park in their efficacy.

How stable is HQPlayer?

The Chord M-Scaler is a fit and forget solution. It simply does it stuff. HQPlayer is a software solution, and so it is dependent upon a stack of underlying hardware and software. This means that as things are updated you may well need to poke around under the hood to keep things up and functioning.

How difficult is HQPlayer?

How technically competent are you? If reasonably, then I would say you will find it a straight-forward.

How can HQPlayer be used?

In these tests I used it directly, on the server and into a local DAC. I also used it remotely, from the server, in the loft, to a remote client (Sonore ultaRendu) and on into a target DAC. In these configurations I used Roon.

HQPlayer can also be used to pre-process files, and just stream the saved result. These files are quite large.

Should I have the server local or remote?

My experience in these tests was that having the server remotely and streaming via the Sonore ultraRendu gave the best results.

When using the server direct into a Chord DAC I borrowed a SRC-DX, usb to dual-bnc widget. I found this raised the sound quality of the local server to within a whisker of the remote server plus ultraRendu. A great value for money proposition.

I will be exploring the SRC-DX further.

How big were the differences between upsampling methodologies?

The Chord M-Scaler and HQPlayer consistently had the most positive benefits, with Roon upsampling coming in behind them.

What worked best for the Chord Qutest?

Surprisingly, in these tests I found that HQPlayer had the best results with the M-Scaler JUST behind.

What worked best for the Synthesis Roma DC69?

My hope was that the higher input frequency of the USB input on the Roma would lead to a step up in performance via HQPlayer. In the event, the M-Scaler was the better performer in my tests with the Synthesis DAC.

Chord Qutest Conclusion
Audiostore Prestige Server non-local, with Sonore ultraRendu

I am always bowled over by how good this DAC is, improve the feed into it and it ups its game. One issue I have had with it over the years has been its mid-weight and timbre. Well, with this exercise these area were improved.

HQPlayer proved its worth. The best sound quality I heard was with HQ-Player, alone; but the delta between it and the Chord M-Scaler was very small.

The Chord M-Scaler was far more of a fit and forget solution. However, that doesn’t mean that you can be careless about its power or upstream feed, rubbish in/rubbish out.

Roon upscaling was the weakest of the options available. If you use Roon then I would still recommend having a dabble via MUSE.

Adding Roon or HQ-Player upscaling before the M-Scaler leads to subtle changes, almost on an album by album basis.

Audiostore Prestige Server local, NO Sonore ultraRendu

The sound was not to my liking and so I fitted an iFi iSilencer; this may have improved things by a tiny margin. What improved things by a large margin was upping the output frequency rate. I ran with 705kHz and 768kHz, both of which worked well. The best results I obtained were with the SRC-DX interface box, giving access to the Chord Qutest Dual BNC inputs.

The sound quality order echoed what I had found using the Sonore ultraRendu, however, the sound quality sans ultraRendu was not as good; although it got VERY close with the SRC-DX. There was always a feel of a very slight digital edge.

What do I mean by digital edge? The effect is to give an increased sense of detail, but to highlight sibilance and generally make the listener a bit less relaxed. This can be hard to identify until it has been pointed out, as it can sound like an improvement; hearing some things that before were further back in the mix.

I discussed this with Martin from Audiostore and it may be that an appropriately specified linear power supply for the Prestige server might well provide an improvement. Unfortunately, my 19v SBooster only provides a paltry 1.75 amps, which might have corrupted the boot drive on startup.

To get the system to work without HQPlayer I had to unload the server, otherwise the direct USB connectivity failed to be seen by Roon. In the event loading and unloading HQPlayer was a matter of a few clicks and worked well.

Roon upsampling was as solid as a rock and delivered its improvements, either direct to the Chord Qutest or via the Chord M-Scaler.

Whenever the Chord M-Scaler was included in the chain it continued to deliver its improvements, but they came somewhat more subtle when hooked directly to the server.

Synthesis Roma Conclusion
Server non-local, with Sonore ultraRendu

The Synthesis Roma 69DC DAC produces music through my system that I find insightful and communicative. My fondest hope was that by using HQPlayer I could ring more upscaled goodness from its USB input.

In the event I preferred the Chord M-Scaler, even lobotomised to 192kHz via the coaxial output, this just sang.

Audiostore Prestige Server local, NO Sonore ultraRendu

Setting up HQPlayer was straightforward, but again bypassing it required it to be unloaded from the Audiostore Prestige Ultra server.

The Synthesis Roma 69DC DAC worked well in direct partnership with the Audiostore Prestige Ultra, with or without HQPlayer. Any issues in this duo were sonically well handled, so that any lack of definition in the bass or any sibilance in the vocals fails nicely, the result is still musical and not distracting.

Both HQPlayer and the Chord M-Scaler gave excellent results, either on their own or in partnership with upsampling by Roon. As the output from the Chord M-Scaler was SPDIF the frequency was capped at 192kHz, this limited the upsampling that could be achieved.

Overall Conclusion

Please remember that this is NOT a review of the parts used within these tests. The comments I will make are within the context of the test system and are NOT my opinion of the ultimate fidelity of either HQPlayer or the Chord M-Scaler.

I have achieved better sound quality in my system using the Chord M-Scaler than that I heard with these digital front ends, which doesn’t mean that I couldn’t leapfrog that sound quality by using an alternate system based upon HQPlayer.

There is a plethora of settings within HQPlayer that I only briefly explored.

The Chord M-Scaler was consistent and as solid as a rock; if a rather expensive one.

Both these solutions are playing in the same ball-park sonically within the context of this system.

If money is not a limiting factor my suspicion is that these two upsampling solutions will appeal to two rather different groups of users. In motoring there are people who like nothing better than rolling up their sleeves and getting into the engine of their car, others just want to start the car and GO. In this simile, HQPlayer is for the engineer types.

IF you are happy to sit with a computer at your elbow and dig into the playback stack occasionally, then HQPlayer does superb things for an excellent price. However, this isn’t a free lunch. In this setup, I had an excellent server. Audiostore has extensive experience in this field. Also you will need a good endpoint, such as the Sonore ultraRendu, and an excellent power supply or two.

HQPlayer is a fabulous piece of software which achieves excellent results. As such there are many ways in which it can be used, not just the Roon route that I chose here. Some even pre-process the files using PGGC, a Hitchhiker Guide to the Galaxy reference.

Many people are using HQPlayer and thoroughly enjoying the results, having spent some time listening through these tests I can understand their enthusiasm. I would just advise you to be prepared to roll up your sleeves and get a deeper understanding of audio software.

It is undoubtedly worth your time and trouble to explore upsampling solutions and listening to the results. Trust your ears.

Finally, my thanks once again to Jussi, of HQPlayer, and Martin at Audiostore.

Appendices

USB Games

Amongst the discussions I had with respect to HQPlayer instability, there was a suggestion that this might be due to the Chord USB interface. A possible solution was suggested in the SRC-DX USB to Dual BNC interface, HERE.

An owner of this well-made widget leant me theirs. In the event it did raise the sound quality of using the Prestige Ultra server locally to within a whisker of that achieved remotely using the Sonore ultraRendu, and so gets a recommendation from me.

As a result of our conversation Martin at Audiostore now stocks and sells the SRC-DX.

HQPlayer Settings

Initially I ran with the default settings.

Having done some online research I changed the Dither to LNS15. One recommendation was to change the 1x Filter and Nx Filter to Sinc-M; this was an improvement to my ears with both the Chord Qutest and the Synthesis Roma 69DC DAC.

When using the Sonore ultraRendu into the Chord Qutest I set HQPlayer to output 705kHz. This bypasses any internal Qutest upsampling.

I worked my way through 352kHz, 705kHz and 768kHz; all of which worked well.

With the Synthesis Roma I set HQPlayer to output 352kHz.