Roon Sound Quality vs. Pure Music/iTunes

That’s fine, so long as you don’t confuse your likes with fidelity.


That’s fine, so long as you don’t confuse your likes with fidelity.

+1. He’s running a TOTL DAC and really shouldn’t be applying external up-sampling and/or filters, IMO.

Frankly, I’ve done similar comparisons with Audirvana+ and Roon into my Mid-Fi DAC (Rega) + headphone rig; I think Roon sounds fantastic and does a really good job of stable delivery of bits.

I have often wondered why people buy a really good digital system then spend loads of money and time trying to make it sound like a vinyl/tape system. isn’t it easier just to buy a turntable or reel to reel ?

Just to make sure we’re all on the same wavelength, all digital audio involves filtering. At the analog to digital stage an anti-aliasing filter is used and at the digital to analog stage a reconstruction filter is used. The reconstruction filter is a low pass filter and in Redbook it has to approximate a steep “brick wall” shape between 20kHz and 22kHz. Although it only takes away inaudible high frequencies, all the signal passes through the reconstruction filter and can be affected by it.

Designing a steep brick wall filter is problematic. It can introduce artifacts seen on oscilloscopes as ringing or pre-ringing, meaning minor oscillations around transients. Recordings of solo piano are said to be useful when listening for such artifacts. Some people like the sound of such artifacts, but most hi-fi enthusiasts would prefer to do without them.

The benefit of upsampling is that it provides more headroom for better reconstruction filtering using gentler slopes that do not create unwanted artifacts. All upsamplers will apply such filters, there is no point to upsampling without doing so.

Since electronics is built of compromises there are choices to be made in such filters, including whether to prioritise minimising group delay or ringing associated with transients. Many DACs now offer a choice of filters for users to try.

So filtering is necessary and upsampling enables a broader choice of filters. None of those filters are designed to add or take away anything (except inaudible high frequencies) but each of them will introduce different types of artifacts or distortion compared to the original waveform. Our job is to judge between such compromises.

1 Like


Thanks for the informative post. It’s good to see someone willing to take the time to elucidate, rather than questioning a person’s preferences and decisions.

I don’t profess to know much all the ins and outs of digital playback; learning every day, which is also how I like it.

When I first set this rig up, I tried a few combinations and then settled in with what sounded best to my ears. Now I’m looking forward to getting back to the rig to try different settings and possibilities based on what I’ve read here, and I expect more favorable results.

I’ve enjoyed the convenience of digital playback, having relegated my CD collection to the closet. Vinyl? Not for me; I’ve been there. Tape? No thanks. Roon brings the best of streaming combined seamlessly with my library. Sometimes one just needs guidance as to the best options.

1 Like

Between DSD upsampling with HQP and fidelity, you can keep fidelity.

Taking a relatively low resolution signal such as Redbook audio and carefully interpolating samples between the existing samples to increase the effective resolution is the kind of “distortion” I can get behind. I imagine a crude drawing on graph paper with medium sized blocks transposed onto another piece of graph paper with much smaller blocks.

Anyway, the purpose is not to make the original data sound “better”, but rather to feed your DAC a signal that it can convert optimally. Which leads to the other point that makes this discussion moot, in my opinion. All but a very small number of DAC’s upsample lower resolution signals internally anyway. Digital filtering is far less invasive with higher resolution data. The popular ESS Sabre chip (used in the Auralic Vega and many other premium DACs) also converts all PCM to a DSD-like SDM signal before conversion to analog. To achieve state of the art performance without this kind of DSP is possible, but it’s difficult and expensive and I have no idea whether the end result is any better for having remained “pure”.

I wonder how many people who prefer their data stream to remain bit-perfect realize that it’s being molested inside the DAC before being converted to analog? :grinning: The developer of HQP posits that it is better to handle the DSP in the server, bypassing the DSP in the DAC, which has a tiny fraction of the processing power found in a modern server. Plus, using the computer, the DSP is configurable and upgradable. Rob_Hanson has enjoyed the results with PureMusic and many of us have had a positive experience with HQP, which is why there is so much interest in Roon/HQP integration.

Purity of the signal seems like an obviously desirable ideal, but I feel that view is oversimplified. Upsampling externally in the server generally bypasses the internal processing the DAC, yielding better sound because the DAC is doing less harm to the signal.

Agree with that Keith, but I don’t see upsampling and fidelity as necessarily opposed. Upsampling is digital signal processing usually intended to get closer to the signal, which is how I view fidelity.

Also I have to confess to a fondness for tape. A friend came over to dinner last Saturday and we had a great time listening to “Axis Bold as Love” on tape. Sure there was hiss and I’ve no doubt digital is higher fidelity, but it was enormous fun.

I completely agree.

I was addressing the widely held opinion anything other than “bit-perfect” playback is a distortion or coloration of the sound.

That idea makes complete sense superficially, so I was trying suggest that there’s more to all of that than meets the eye.

IMHO converting 44.1 to 192 is up sampling that does not maintain ‘fidelity’ of the original bits. A 4x integer up-sample from 44.1 to 176.4 would be better. Not sure I have seen this distinction in the comments above.

I’m not suggesting a user couldn’t prefer a non integer conversion.

6 posts were split to a new topic: Digital Audio - Frequency and Time Domain Requirements for Audio Reproduction

I suspect everyone here knows this.

What is sometimes missing from the PC CPU vs. DSP in the DAC discussion is that higher end DACs usually implement their filters within FPGAs, which are software upgradable, prior to sending to the DAC.

In this environment, I am of the opinion that all of that extra CPU horsepower does little more than generate electrical noise, because the FPGAs and ASICs are, by design, “fast enough” for the task at hand.

I realize that this puts the customer at the mercy of the DAC vendor to upgrade/update filters via a software update, but I personally would rather leave that task to people who really know what they are doing (i.e. the DAC vendor), rather than making a bunch of incremental adjustments to things like pre-ringing, slope, et al., and then relying on sighted comparisons. If you really enjoy that kind of tweaking & experimenting, then by all means, it’s a hobby and have fun.

My Rega DAC has 5x2 sets of filters that may be selected from the front panel and the differences are really, really subtle, even listening with resolving headphones.


The use of FPGAs in DACs is relatively new, but is certainly catching on and your points are entirely reasonable as FPGAs turn this discussion on its head.

However, the salient point in my opinion is that nearly all of us are listening to altered data streams. That’s a fact that the “bit-perfect” advocates seem to dispute. But once you accept that fact, then different methods of upsampling (DAC chip, FPGA, HQP, JRiver, foobar, PureMusic) will sound different. I prefer to experiment (especially since the software can be tried at no cost) rather than decide that one method is likely to be best based on any objective theory - such as “all of that extra CPU horsepower does little more than generate electrical noise”.

1 Like

I am in agreement with you.

But let me state this: in my armchair experience, what seems to provide the biggest impact with respect to improving sound quality, all things being at a certain quality level to begin with, is mitigating electrical noise.

This is why I am a believer in using purpose built rendering machines at the DAC and separating the CPU intensive work into a machine somewhere else on the network (optical or wireless isolated, if I’m being OCD).

Of course, this fits nicely with the RoonServer / HQP processing -> Roon Remote -> DAC model.

A post was merged into an existing topic: Digital Audio - Frequency and Time Domain Requirements for Audio Reproduction

That makes sense to me. I suspect that you have more experience than I, but I had that same concern when I was building my system.

I chose a unique DAC that does no upampling and in fact doesn’t even have a DAC chip (Lampizator DSD Only DAC). So it was mandatory for me to perform the upsampling and conversion of all PCM to DSD on the server.

To mitigate the concern about the effect of all the extra processing on the server, I use jPlay to create a dual server system. The first PC does all of the heavy lifting while the second, which is connected to the DAC, loafs in “hibernate” mode. Initially, I questioned how “damage” done in PC 1 could be undone by adding another component to the chain, but I tried it anyway. I’m enjoying truly extraordinary sound. It’s the best I’ve ever heard… which… doesn’t mean a lot objectively because who knows what I’ve heard :smile: , but I would love to be able to realistically compare my system to others. Anyway, I don’t have the “itch” to improve my sound anymore. As a fellow audiophile, you know that’s a major achievement!

I recognize that your arrangement offers the same benefits with greater flexibility, but I’m okay with having only one room with the ultimate in sound quality while the remotes offer excellent but more down-to-earth performance.

1 Like

A followup. First, thanks to all for a lively and informative post.

Having returned home from a trip, I had the chance to test-drive HQP last night. I ran with the default settings – pretty much – and fed it 16 44.1 files from my library.

I must say that I was impressed, and the SQ was right up there with PureMusic/iTunes, if not better.

If Roon and HQP play well together when the combination is released, as I’m sure they will, I’ll be a lifer.

Now if only there was a way to decipher the settings in HQP for best effect on my system. Not being familiar with all the terms, the option settings are rather overwhelming.

Hi Rob,

In addition to the manual found in the HQP folder there are a couple of CA threads that are very helpful with HQP:

Geoffrey Armstrong’s Kickstart guide

Jussi/Miska’s guide to the settings

Impressions of settings by CA members


I tried HQP once, more than a year before I actually started using it. Initially, I was completely confused by its interface and all the settings seemed very complex, so I went back to using JRiver. But I kept hearing raves about it, so I dove back in.

My experience has been that the 4 drop down lists in the main UI are by far the most important. To oversimplify things to an extreme:

  • The first 2 allow you to choose between HQP’s upsampling algorithms. There are no right or wrong choices here, it’s a matter of what sounds best to you - and which of them your server is powerful enough to run. You can read about others preferences in the links @andybob provided above.
  • The last 2 are simple. They select the output sampling rate and whether the output should be PCM or DSD.

Of course there is much more to know, but for me, the above comprises the “80%” in the 80/20 rule.

Thanks @andybob I look forward to checking out those pointers you’ve provided.

@k6davis Keith, though I haven’t had the opportunity to check out the above links, I’d be interested in what settings you’ve chosen. And… should I be outputting PCM or DSD?

Thanks. Still learnin’.