HiFi Rose RS250 audio & video streaming D/A preamplifier Measurements

Sidebar 3: Measurements

I measured the HiFi Rose RS250 with my Audio Precision SYS2722 system (see the January 2008 As We See It), repeating some tests with the magazine's more recent Audio Precision APx500 analyzer.

I first looked at the RS250's performance as a D/A processor. Apple's USB Prober utility identified the HiFi Rose as "RS250-DAC" from "HiFi ROSE." The USB port operated in the optimal isochronous asynchronous mode, and Apple's AudioMIDI utility revealed that the RS250 accepted 16-, 24-, and 32-bit integer data via USB sampled at all rates from 44.1kHz to 768kHz. The coaxial and TosLink S/PDIF inputs accepted data sampled at rates up to 192kHz from my MacBook Pro's optical output, but the optical input was limited to sample rates of 96kHz and below when connected to the APx500's TosLink output. Peculiarly, the RS250's S/PDIF inputs would only lock to data sampled at 48kHz when connected to the SYS2722's digital outputs.

The RS250's single-ended analog outputs can be set to variable gain or to one of several fixed maximum outputs: "100mV," "150mV," "200mV," "300mV," "500mV," "1000mV," "2000mV," or "2200mV." The output level with full-scale 1kHz data fed to the RS250's TosLink, USB, and network inputs was within a few millivolts of the nominal value, with the greatest discrepancy at "2200mV," where I measured 2.26V. (Note that this is the only fixed level available with USB data.) The maximum headphone output level with a 1kHz tone at 0dBFS was 507.3mV. The HiFi Rose's analog and headphone outputs preserved absolute polarity (ie, were noninverting) from all of the RS250's digital inputs. The analog output's source impedance was the specified 100 ohms at all audio frequencies and that from the headphone output was just 1 ohm. The RS250 should have no problem driving low-impedance headphones.

In common with other processors using ESS Sabre DAC chips, the RS250 offers a choice of reconstruction filters for PCM data fed to its S/PDIF and network inputs, though not to its USB input. These are labeled "Brick Wall" (BW), "Apodizing Fast Roll-off" (AFR), "Corrected minimum phase Fast Rolloff" (CMPFR), "Minimum phase Slow Roll-off" (MPSR), "Minimum phase Fast Roll-off" (MPFR), "Linear phase Slow Roll-off" (LPSR), and "Linear phase Fast Roll-off" (LPFR). Fig.1 shows the MPFR filter's impulse response with 44.1kHz data, a conventional minimum-phase filter with all the ringing following the single full-scale sample. (This is the default filter and the only one available with the USB input.) The impulse response of the MPSR filter was much shorter (fig.2), while that of the CMPFR filter (fig.3) had less post-ringing than MPFR along with a small amount of preringing. The BW, AFR, and LPFR filters had identical linear-phase impulse responses, with equal amounts of ringing before and after the single full-scale sample (fig.4).


Fig.1 HiFi Rose RS250, MPFR filter, impulse response (one sample at 0dBFS, 44.1kHz sampling, 4ms time window).


Fig.2 HiFi Rose RS250, MPSR filter, impulse response (one sample at 0dBFS, 44.1kHz sampling, 4ms time window).


Fig.3 HiFi Rose RS250, CMPFR filter, impulse response (one sample at 0dBFS, 44.1kHz sampling, 4ms time window).


Fig.4 HiFi Rose RS250, BW, AFR, and LPFR filters, impulse response (one sample at 0dBFS, 44.1kHz sampling, 4ms time window).

The MPFR filter with 44.1kHz white-noise data (fig.5, magenta and red traces) rolls off quickly above 20kHz. The ultrasonic rolloff is disturbed by aliased images of the audioband noise signal centered on 44.1kHz. However, as these ultrasonic images lie below –90dBFS, they will be inconsequential. The aliased image at 25kHz of a full-scale tone at 19.1kHz (cyan, blue) is suppressed by around 52dB, though a higher-level, higher-order image can be seen at 63.2kHz (2 × 44.1–25kHz). As its name suggests, the MPSR filter rolls off more slowly above the audioband (fig.6, magenta, red), but still with the low-level, aliased noise images visible. The aliased image of the full-scale tone at 19.1kHz (cyan, blue) is suppressed by just 32dB. The CMPFR, AFR, and BW filters all behaved similarly, reaching full stop-band attenuation at exactly half the sample rate, shown by the vertical green line in fig.7. The harmonics associated with the 19.1kHz tone are all very low in level.


Fig.5 HiFi Rose RS250, MPFR filter, wideband spectrum of white noise at –4dBFS (left channel red, right magenta) and 19.1kHz tone at 0dBFS (left blue, right cyan) into 100k ohms with data sampled at 44.1kHz (20dB/vertical div.).


Fig.6 HiFi Rose RS250, MPSR filter, wideband spectrum of white noise at –4dBFS (left channel red, right magenta) and 19.1kHz tone at 0dBFS (left blue, right cyan) into 100k ohms with data sampled at 44.1kHz (20dB/vertical div.).


Fig.7 HiFi Rose RS250, CMPSR filter, wideband spectrum of white noise at –4dBFS (left channel red, right magenta) and 19.1kHz tone at 0dBFS (left blue, right cyan) into 100k ohms with data sampled at 44.1kHz (20dB/vertical div.).

Fig.8 shows the frequency responses with the MPFR filter and data sampled at 44.1, 96, and 192kHz. All conform to the same shape but with a fast rolloff just below half the sample rate. Channel separation was >90dB in both directions above 3kHz but decreased at lower frequencies, reaching 62dB at 100Hz. This is unusual and suggests that the RS250's power supply has a relatively high source impedance.


Fig.8 HiFi Rose RS250, MPFR filter, frequency response at –12dBFS into 100k ohms with data sampled at: 44.1kHz (left channel green, right blue), 96kHz (left cyan, right magenta), and 192kHz (left gray, right red) (1dB/vertical div.).

When I examined the spectrum of the RS250's noisefloor, the levels of power supply–related spuriae depended on whether I had the HiFi Rose connected to my network. (My router is in the listening room, and I ran a 75' CAT-5 cable from it to the RS250 when the latter was in the test lab.) The cyan and blue traces in fig.9 show the spectrum of the noisefloor with the network connection when the RS250 was decoding 24-bit, 1kHz data sourced from my battery-powered MacBook Pro via USB; the magenta and red traces show the spectrum with the network disconnected. The 60Hz, 180Hz, and 300Hz spuriae are still very low with the network connected but drop by 9–12dB when I removed the connection. A 120Hz component visible at –113dB effectively disappears when disconnected from the network. Repeating this analysis with optical S/PDIF data (not shown) eliminated the lower-frequency supply-related spuriae, but low-level sidebands associated with the 1kHz tone at ±120Hz and ±240Hz were still present. None of this should have any influence on sound quality.


Fig.9 HiFi Rose RS250, spectrum with noise and spuriae of dithered 1kHz tone at 0dBFS with 24-bit data sourced from MacBook Pro via USB with RS250's Ethernet port connected to the network (left channel blue, right cyan) and without the RS250 connected to the network (left magenta, right red) (20dB/vertical div.).

An increase in bit depth from 16 to 24, with dithered network data representing a 1kHz tone at –90dBFS sourced from Roon, dropped the RS250's noisefloor by around 20dB (fig.10). This implies a resolution of better than 19 bits, and when I played undithered data representing a tone at exactly –90.31dBFS, the waveform was symmetrical, with negligible DC offset—25µV in the left channel, 50µV in the right—and the three DC voltage levels described by the data were free from noise (fig.11). This measurement was taken with the MPFR filter; the random noise level in the RS250's output is sufficiently low that the minimum-phase ringing is clearly visible at the bit transitions.


Fig.10 HiFi Rose RS250, spectrum with noise and spuriae of dithered 1kHz tone at –90dBFS with: 16-bit Ethernet data (left channel cyan, right magenta), 24-bit Ethernet data (left blue, right red) (20dB/vertical div.).


Fig.11 HiFi Rose RS250, MPFR filter, waveform of undithered 1kHz sinewave at –90.31dBFS, 16-bit data (left channel blue, right red).

The RS250 produced very low levels of harmonic distortion with full-scale data, challenging the resolution of my SYS2722 analyzer. I therefore used the higher-performance APx555. The result, taken with TosLink data, is shown in fig.12. The third harmonic is the highest in level at just –116dB (0.00015%) in the left channel and at –124dB (0.00006%) in the right, though the low-level sidebands at ±120Hz that I mentioned earlier are visible. This spectrum was taken into the high 100k ohm load. When I reduced the load impedance to the punishing 600 ohms, the levels of the harmonics didn't change. Intermodulation distortion with an equal mix of 19 and 20kHz tones, each lying at –6dBFS, was very low (fig.13), with the difference tone at 1kHz not visible above the noisefloor. This graph was taken with the MPFR filter; the aliased images of the primary tones can be seen. A suspicious-looking rise in the noisefloor can also be seen on either side of the 19kHz and 20kHz tones.


Fig.12 HiFi Rose RS250, 24-bit USB data, spectrum of 1kHz sinewave, DC–1kHz, at 0dBFS into 100k ohms (left channel blue, right red; linear frequency scale).


Fig.13 HiFi Rose RS250, MPFR filter, 24-bit TosLink data, HF intermodulation spectrum, DC–30kHz, 19+20kHz at 0dBFS into 100k ohms, 44.1kHz data (left channel blue, right red; linear frequency scale).

This behavior, which is most likely due to jitter with a random low-frequency spectrum, was clearly visible when I examined the RS250's rejection of word-clock jitter. Fig.14 shows the spectrum of the RS250's output when it was fed high-level, optical 16-bit J-Test data. (The USB and network inputs behaved identically.) Almost all the odd-order harmonics of the undithered low-frequency, LSB-level squarewave lie at the correct levels, indicated by the sloping green line. However, the closest pair of sidebands surrounding the high-level tone at one-quarter the sample rate are boosted in level, as is the noisefloor. Peculiarly, this sideband pair is also present with 24-bit J-Test data (fig.15). This jitter-related behavior is only present with the RS250's analog outputs. Feeding its USB output to a high-performance D/A processor while it received 16-bit J-Test data from Roon over the network resulted in a jitter-free spectrum (fig.16).


Fig.14 HiFi Rose RS250, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz: 16-bit TosLink data from MacBook Pro (left channel blue, right red). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.


Fig.15 HiFi Rose RS250, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz: 24-bit Ethernet data (left channel blue, right red). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.


Fig.16 Ayre QB-9 Twenty, high-resolution jitter spectrum of analog output signal, 11.025kHz at –6dBFS, sampled at 44.1kHz with LSB toggled at 229Hz: 16-bit USB data sourced from the HiFi Rose RS250 (left channel blue, right red). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.,

Turning to the RS250's analog inputs, the input impedance was relatively low, at 3.3k ohms, which will be a problem with source components having tubed output stages. The analog inputs offered a wide frequency response, with the output down by 3dB at 72kHz (fig.17). The maximum gain at the headphone output was –12.6dB, ie, an input of 1V resulted in an output of 234mV, while the maximum gain at the analog outputs was 0.325dB. An input of 1V was indicated as "–4dB" on the front-panel VU meters, and the input clipped at 2.2V input (fig.18). Both outputs preserved absolute polarity with analog input signals. The distortion via the analog inputs was slightly higher than it had been with the digital inputs (fig.19) but was still very low in absolute terms.


Fig.17 HiFi Rose RS250, analog input,


Fig.18 HiFi Rose RS250, analog input, THD+N (%) vs input level in V.


Fig.19 HiFi Rose RS250, analog input, spectrum of 50Hz sinewave, DC–1kHz, at 1V into 100k ohms (left channel blue, right red; linear frequency scale).

Other than the somewhat disappointing results with J-Test data (footnote 1), HiFi Rose's RS250 did well on the test bench.—John Atkinson

Footnote 1: The HiFi Rose RS150, reviewed in our sister magazine Hi-Fi News & Record Review, behaved similarly when tested for jitter rejection.
HiFi Rose
US distributor: MoFi Distribution
1811 W. Bryn Mawr Ave.
Chicago, IL 60660
(312) 841-4087

Axiom05's picture

Well, the jitter spectrum of the QB-9 Twenty looks good at least. I really should get my QB-9 DSD upgraded.

CG's picture

Good catch!

That's probably the review of the QB-9 Twenty in its entirety. After all, the Twenty is an upgrade to a no longer manufactured product.

And, yeah, you should get the upgrade. At the risk of sounding like a shill for Ayre, the upgrade is really worth the asking price, and probably a lot more. (Which I paid in full - if I'm a shill, I'm also a very bad negotiator.)

Axiom05's picture

I just upgraded the USB board in my QX-5 Twenty and WOW, I was not expecting this kind of improvement. Definitely a worthwhile upgrade. I guess we're both shills for Ayre. :-)

CG's picture

I can see why.

Compare Figure 16 to the comparable plot in the Stereophile review for the original QX-5 Twenty.

Of course, many, many people will tell you that this sort of thing is inaudible and that you are crazy. But, crazy people can be happy, too.

Jack L's picture


When one crazy enough to burn his/her easy money to be happy, why not ?

Life is short so make the best use of it.

Jacl L

CG's picture

Somewhat OT:

This seems as good a place to ask this as any.

This particular digital solution seems to perform not quite as well as some other products with regard to jitter sidebands. Is this audible? By how much? Why?

Now... Head to the bottom of this very web page and click on the button that says "hi-finews". That'll take you to a website for the magazine of the same name.

Look at a turntable review. Any turntable review.

In the Lab Report, there is a plot of what is labeled Wow and flutter. This is a spectral display of a single tone from a vinyl disc played back through the turntable under review.

Isn't that pretty much the same concept as the jitter test, at least with regard to the central tone of the J-test at 11 KHz?

So, how is one to interpret all of this? One is obviously far different than the other.

It's not obvious from either plot and associated labelling what the measurement parameters are for the spectrum analyzer. The resolution bandwidth, the video filtering, the averaging type and number of samples, the detector type, and so on are not shown for either. (To be fair, this might be explained in an article somewhere that I failed to find - my bad.) Variations there would explain a lot. But, maybe there's much more to it.

I find this all confusing. Perhaps I am the only one who does, though.

Perhaps I am overthinking all of this, too. Wouldn't be the first time.

Archimago's picture

I don't think there's much confusion here. Jitter was never all that audible as an issue. No need for audiophiles to fear this "boogeyman" in general. I posted a demo for folks to listen to years ago - just Google "Archimago Jitter Demo".

Yeah, the J-Test for a device like this is not good for modern 2021 digital especially for the 24-bit ethernet input. I still don't think it's audible in real music anyways, it's more of a reflection of the engineering that the time-domain wasn't better despite the claims of using femtoclock parts and the ESS ES9038Q2M DAC chip!

No surprise as well that turntables are comparatively inaccurate vs. digital (Google "Archimago vinyl LP fidelity" for a discussion). It's very obvious if one listens to a pure tone like 3150Hz as per HiFi News. Time-domain is poor with LP playback not just because of turntable rpm variations but also the imperfections of the vinyl itself. Again, with music we don't notice these issues as much.

BluesDog's picture

Nice article. Thanks for testing storage drives (i.e. usb thumb drive, portable HDD, etc) on this device Some of us have amassed significant size CD quality or better on storage drives. From the late 90’s and before streaming had the quality we see tday. Streaming Qobuz and Tidal are impressive but some of us aren’t quite ready to pay for yet another streaming subscription. Articles like yours (and the use of Roon) help prepare us for if that plunge ever comes.

Jack L's picture


If only HD music videos are needed (what else??), I would not bother to spend my nerves & hard-earned cash on those bells & whistles loaded 'hi-tech' streamers.

Yes, YouTube provides 24/7 tons HD & 4KUHD music shows, streamable FREE by any WiFi HD & 4KUDH blue-ray players. Let's lead a simple life & enjoy really
eye-widening & ear-thrilling HD/4KUHD music shows on our 4KUHD TVs streamed free from YouRTube. Who needs to sweat fighting with those complex 'hi-teck' remotes ???

This is exactly what yours truly have been doing since a couple years back. I simply feed the digital PCM audio signals from the coaxial O/P jack of the HD Blu-ray player to my 24bit/192KHz DAC which in turn hooked up to the linestage I/P of my design/built tube phono-preamp with its PASSIVE bypass switched on. So the HD audio signals from the DAC electrically feeds direct to the power amp !

Please note any DACs deliver adequate output voltage to drive directly any power amps to their rated full output power. Or simply feed the DAC L/R analogue audio signals direct into the power amp if it comes with built-in volume controls. The sonic is sooo rewarding..

With such DIRECT signal feed to the power amp, the HD sonic quality of so many those HD/4KUHD music shows from YouTube becomes surprisingly sooo good to make my skeptical vinyl-addicted ears up-pointing !! No kidding.

Life is short so let's make it simple.

Listening is believing

Jack L

franz2022's picture

The one year warranty from a new company is a disappointment. At a minimum it should be two years. The NAD M10 and Cambridge Audio EVO 150 are covered for three years. I've had far too many expensive electronics fail within three years of ownership. HP replaced a 4K monitor that stopped working unexpectedly after two years (covered for 3) while I was out $2k when my Noble Audio IEMs failed shortly exceeding their warranty period.