MEASUREMENTS: Melody Onix SP3 Mk. II (~2006) Class AB Tube Integrated Amplifier. (And it isn't "diminishing returns" on sound quality, often there are simply "no returns".)

Recently I was curious about trying a tube amp in my system and my dad who's into his vintage gear including some older tube devices offered me this amp to test out. It was time to take up the offer and have a listen plus put some tube gear on the test bench for a peek at the performance.

As you can see, for today's post, we have the Melody Onix SP3 Mark II which I believe was first released in 2004 for US$1000. This one I believe was bought in 2006. It's a relatively compact unit (12" x 8" x 13" or so), 56lb beast designed by Melody Australia, made in China. Over the years, there have been various reviews on this product from 6Moons (Best of 2005 list), Audioholics, Home Theater Review, High Fidelity Review among others. Here's the Melody company website. I think this model was discontinued by around 2010.

I've always liked the look of this box. There's a cage that normally protects fingers from the four large power tubes and individual smaller covers (you can see 3 on the right removed) that protects the 6 flanking tubes. Turn off the lights at night and it all looks very cool.

Here are a few shots of the machine:

There's a nice, firm, 24-step volume knob attenuated by discrete resistors as the only front control. Single blue, non-distracting low brightness power LED.

As you can see, there are six 5-way binding posts where you can select the output for 4Ω or 8Ω loads. Vacuum tubes are high voltage, high impedance, low current devices. Output transformers are needed to convert that into lower voltage, lower output impedance in order to effectively drive our low-impedance speakers. The different output "taps" allows for better matching in order to optimize the power.

We also see two single-ended RCA inputs for "CD" and "AUX" with the stereo channels split widely across each side of the unit.

The titanium gray enclosure is quite attractive. My dad uses this amp as a secondary device so although it's about 15 years old, the number of hours of use remains low. He also did some tube-rolling over the years, so for testing purposes, the original tubes were reinserted and bias was recalibrated to 1.15V as recommended (you can see the holes on the side for measuring bias and calibration). The stock tubes are 4x5881 (large power tubes), 2x12AX7 (preamp), 2x6922 (phase inverter), and 2x12AU7 (driver) - check out this page for more information about the SP3 including schematics (and a good thread on AudioCircle).

On the right side (when facing forward) of the device is the input selector - CD/AUX. On the left side, we see the ON/OFF switch.

I. Basic Amplifier Characteristics

As you can see, this is an integrated amplifier with its own volume control. For my testing, I'll just use the Topping D10 DAC as signal source (standard 2Vrms output, more than enough resolution). The general testing set-up looks like this:

Raspberry Pi "Touch" / Intel NUC --> USB --> Topping D10 --> shielded RCA --> Melody Onix SP3 II --> 12AWG speaker cables --> 4/8Ω power resistors or speaker load 

Speaker cables tapped for measurements  --> Linear Audio AutoRanger Mk II (1V nominal setting used) --> RME ADI-2 Pro FS ADC --> Intel NUC

If anyone is wondering, I'm using a hospital-grade 14AWG IEC power cable connected directly to the wall outlet for testing (I have not found normal gauge power cables to be inadequate).

As usual, I'll be using the excellent Linear Audio AutoRanger for attenuation of high voltage output to present to the RME for measurements. Before testing, I let the amplifier warm up for 1 hour with some music playing into the 8Ω load, volume set to 1W out with 0dBFS 1kHz sine. Here's the test bench with a few things hooked up letting the amplifier warm up:

A glass of wine while measuring IS allowed ;-). Notice the 300W power resistors, both 4Ω and 8Ω sets on the table.

Normally for monoblocks, I have a look at the amplifier gain. I did not bother for this integrated amp. The 2V level from the Topping D10 was plenty to drive the output way beyond clipping.

First, let's look at the amplifier damping factor:

Notice that this is a bit different from my usual in that I wanted to give you a sense of the difference between the 4 and 8Ω outputs. In the middle, in blue and yellow are the damping values when you use a 4Ω load with the 4Ω tap and likewise 8Ω load with 8Ω tap. Notice that the damping factor is low with this amp, between 3-4. As you would expect, if we use an 8Ω load connected to the 4Ω tap, we achieve better damping up around 6 (red), and conversely, a 4Ω load to the 8Ω tap results in a poor damping factor of <2 (green) - not advised!

Damping factor is a reflection of the relative impedance of the load with the output impedance of the amplifier. If we look at a measure of frequency response, the greater the damping factor, the more resistant the amplifier is to the reactive properties of the load (speaker), and generally the better the amplifier can keep the frequency response flat.

Let's demonstrate this...

In blue, we have the frequency response of a flat 4Ω load connected to the 4Ω output where we see the native frequency response of the amplifier. Notice that left and right channels are very well balanced with the bias calibrated properly, there's a slight -0.5dB roll-off at 10kHz and -1.75dB by 20kHz.

In red and green, I've connected the Sony SS-H1600 speaker (rated 8Ω nominal impedance, see here for impedance measurement) to either the 4 or 8Ω outputs. Notice that due to higher damping when using the 4Ω tap (red), we see less frequency response fluctuation which is a "good thing" if we're aiming for flatter, more neutral playback. As usual, subjectively, listeners will report different preferences. The SS-H1600 speaker connected to the 8Ω tap (green) will have bass humps at 55 and 130Hz, dipping quickly into 350Hz, then rolls up into 3kHz before dipping again into 10kHz. This kind of "tone control" with >2dB variations across the audible spectrum might be preferred by some listeners. Since the SS-H1600 speaker is actually more of a 6-8Ω load, lower impedance 4Ω speakers likely will create even more variance in the frequency response than this.

II. Single-Tone Harmonic Distortion and Noise

This amplifier is rated as 38Wpc but I'm not sure what the conditions were for this specification. As usual, let's for now focus on my standard 4Ω load and check out the distortion across a frequency sweep at 1W (2Vrms into 4Ω load, 4Ω tap):

I put a cursor at -40dB which would be 1% and at 1kHz if you want to look at the legend and exact amounts of distortion for 2nd to 9th harmonic. As you can see, the lower order 2nd and 3rd harmonics predominate throughout. Except at lower frequencies <80Hz, 3rd harmonic is slightly higher than 2nd.

Here are some detailed THD(+N) FFTs with the 4Ω load (connected to 4Ω tap) - both channels driven for your consideration:

Notice that I'm using 950Hz as the fundamental frequency. Because I'm doing this through a USB system, sometimes I'll see some 8kHz USB PHY noise spilling into the measurements which I don't want as part of the THD result (ie. 8kHz is the 8th harmonic at 1kHz).

Over all, not bad. Into 4Ω, the amplifier is able to maintain <1% THD+N at least to 25W. Nonetheless, we're not looking at any award-winning low distortion vs. power results which we'll look at in more detail below.

For completeness, we can also look at an 8Ω load into the 8Ω tap using similar output voltages, again, both channels driven:

You might be wondering why I have such strange output voltages like 2.72V. That's basically because I used the 24-step voltage control while the DAC played a -6dBFS signal and these are the levels that came out from the volume control.

Okay, so let's combine these (and other) readings to create some THD(+N) vs. Power graphs:

As you can see, I have 3 curves up there. The first 2 are self-explanatory; straightforward 4Ω load into 4Ω tap, 8Ω load into 8Ω tap. I assume the claims of "38W" of power might be for 4Ω load, single channel driven, 1% THD. In my tests, I can confirm up to 30W into 4Ω, 1% THD with both channels driven. With an 8Ω load into 8Ω tap, this amplifier can achieve about 19W, both channels driven into 1% THD.

Unfortunately, this is a rather noisy amplifier. The noise floor elevation is easily seen on the RME ADC's digital meter when it's supposed to be "silent" during testing. If we look at the ~0.85V 950Hz FFT against the Hypex NC252MP at 1V, we can see that the Melody has many extraneous tones plus the noise floor is around -95dB on average compared to around -120dB on the Hypex.

With a 4Ω load, THD(+N) floats around -60dB (0.1%) which I've generally used as my threshold of "high fidelity" for amplifiers until about 5.75V or just over 8W; after which the THD+N gradually rises well over that 0.1% threshold.

The third THD vs. power graph is to help answer the question: "What if I plug an 8Ω load into a 4Ω tap"? Since we already know noise is relatively high, let's just overlay the 8Ω-load THD (exclude the +N result) vs. power graphs plugged into either the 8Ω or 4Ω tap:

So what we see here is that if you have an 8Ω load, plugging it into the 4Ω tap results in lower distortion up to about 1W. Beyond which, for at least a little bit, the 8Ω tap performed better although the difference was minimal. Above 5W, it actually looks like the 4Ω tap maintained lower distortion into 20W. Distortion picks up quickly thereafter.

Basically, I'm not able to find an advantage of using the 8Ω tap here with a straight 8Ω load when just examining the THD value. But wait... Let's look at the detailed FFT plots:

So, despite the 8Ω/8Ω not really achieving better THD(+N) mainly due to the strong underlying harmonics, notice that the overall noise level is better with the 8Ω tap (green highlight) as we increase the power above 20W. We can easily see this visually with what looks like varying amounts of intermodulation sidebands. How audible this is in the face of high harmonic distortion is unclear to me.

The bottom line at least with this Melody Onix is that with an 8Ω load, the 4Ω tap will provide better damping (as discussed above), and will provide lower harmonic distortion. Noise and intermodulation may be worse so it's recommended to listen for oneself. Maybe there will be a power advantage with higher impedance loads >8Ω into that 8Ω tap than what I found here.

As a 2-channel device, let's have a quick peek at the crosstalk using a simple test signal with a 4kHz tone on the left, and 300Hz on the right and look for "seepage" across channels:

Looks like lots of higher frequency 4kHz seepage from left to right channel whereas little low-frequency stuff at 300Hz from right into left. Taking the average, I get a crosstalk average of -56dB. Obviously not the best; the Onkyo TX-NR1009 measured -69dB on this, and the Hypex NC252MP got -78dB. Of course more testing can be done but this is probably a reasonable estimate without getting too involved; right-left separation certainly sounds good enough with music playback in my sound room.

III. Multi-Tone Testing: Intermodulation Distortion and Triple-Tone TD+N

Okay, let's have a look at the various IMD tests at my standard 2V/1W into 4Ω, both channels driven:

Basically, we see a consistent -50 to -53dB IMD distortion amount. It's a tube amp, so can't expect very low distortion.

Let's see what the TIM ("Transient InterModulation") distortion test signal looks like. As described previously, this test consists of a 1kHz square wave with a 12kHz sine inserted, created at 192kHz samplerate (96kHz bandwidth), measured again at 2V into 4Ω load (4Ω tap):

We see distortion at the -80dBFS level which is pretty high. Typically, audiophiles talk about TIM being a problem with solid state amps but a test like this applied to a tube device also yields anomalies. For comparison, a "typical" high-fidelity Class AB solid state amplifier like the Emotiva XPA-1L monoblock (which granted as a pair would be about 20-30% more expensive when new than the Onix), has this distortion level below -100dBFS. Using these same test conditions, the affordable Class D Hypex NC252MP had distortion levels down at -120dB; basically into the noise floor.

Finally, let's look at the Triple-Tone Distortion and Noise measurement which I use as a measure of the distortion for my scoring system. Here's the TD+N at 1W which is 2V into 4Ω (4Ω tap, both channels driven), both right and left channels measured:

Notice that the right channel has slightly higher distortion than the left. On average we're looking at -57dB for 2V TD+N from this amplifier.

IV. Square Wave and Wideband Noise

Here's the "non-aliasing" 24/384, 1kHz square wave at around 2.5V through the oscilloscope:

Slight insignificant rolling of the leading edge probably indicative of the slight roll-off in the frequency response as shown above. Otherwise, the 2 channels are well balanced with the right channel (CH2) just slightly lower in amplitude by 0.03dB.

For completeness, here's the wideband low-res FFT to 600kHz:

The intent here is just to make sure we don't have much ultrasonic noise coming from the amplifier like with Class D amps. As expected, no problem here, notice the 93kHz signal has rolled-off quite a bit compared to the 5kHz peak.

V. Subjective Impressions

Okay, so here's the Melody Onix SP3 in my sound room for a few evenings of music listening prior to performing the measurements above:

As you can see, it's not large but significantly larger and way heavier than my DIY Hypex NC252MP Class D amplifier hidden behind the Melody (you can make out the box's black top). I connected my Paradigm Signature S8 v.3s to the amp into the 4Ω taps. No issue with driving these speakers (rated as 92dB/W/m in room) to louder-than-reference levels.

How shall I describe the sound? Well, let's start with female vocals. The system sounded gorgeous with this amp on Melody Gardot "Baby I'm A Fool" (from My One And Only Thrill). For sure this is not a thin sounding audio component. The voice is projected in a more forward manner with great mid and upper mid definition (Gardot has that signature tight vibrato). Sibilance is well controlled. Vocal separation from background instrumentation is very good.

I like to close my eyes when I play Camille's "Le Festin" from the Ratatouille soundtrack with her little solo intro. When done well, she sounds like she's singing front-and-center and you might even imagine the breath itself emanating from that space. Yup, the amp can deliver the effect through my speakers.

Soundstage on a challenging track like "One Day More" (Les Miserables 1987 Broadway cast) presents a nice virtual view of the stage. No problem with tight vocal placement such as when singers seem to be standing beside each other - like Marius and Cosette singing "Tomorrow you'll be worlds away/And yet with you, my world has started..." Nice. There is however a limitation which is when multiple voices are singing, there's a bit of congestion, not as resolving as something like the Hypex NCore in that way.

While we're still on the topic of soundstaging, I like how this amp sounds "meaty" with bass and low-mids. For example the bass at the start of Pentangle's "Let No Man Steal Your Thyme" smoothly panning from left to right had a beautiful tonality and the song maintains a wide left-right separation of the string instruments throughout.

Speaking of "meaty" sound, I think the Run DMC + Aerosmith collaboration on "Walk This Way" (Raising Hell) is my favourite from the '80s. Great sounding beats with a nice "propulsion", "drive". Clean vocals. Good bass even with the subs turned off through this amp.

I played some Taco "Puttin' On The Ritz" (After Eight, 1982) for fun. This is an early '80s synthpop/disco production which can sound thin, glaring, and harsh. Of course, even a tube amp can't do wonders to the source material but it sounded less objectionable than "flat" frequency response amps adding to my impression that this amp rolled off the highs a little bit.

Finally, I needed to have a listen to some acoustic music - I fired up Grisman/Taylor's Tone Poems II. Very nice. The instruments were more forward and there seemed to be a very pleasant accentuation of the mids with good dynamics. There was satisfying "impact" one could hear with each strum and pluck on the strings.

So far, everything seems great, but there was clearly one downside that I was not happy about. The noise level was clearly higher than my other amps and can be distracting during silent parts or between tracks in my low ambient noise system. During the daytime if there's traffic noise or my kids making sounds upstairs it might not be a problem. Likewise, I suspect if this were upstairs in my living room system, I would not notice. In the evening when the world is dark and streets silent (in the midst of a pandemic!), this audible noise floor is simply distracting and not something I could be satisfied with. I checked that this was not a ground loop and using my power conditioner made no difference. It sounds like a combination of upper-mid buzzing with 60Hz hum components and came from both speakers evenly; didn't matter if I used 8 or 4Ω output tap. I did not see any mention of this in other reviews of this product. In my experience, a silent noise floor is not one of the strengths with tube amps in general (even very expensive ones I've heard in audio shows), so this did not come as a surprise.

BTW: I think it's important that reviewers tell us what the ambient noise level in the room is when listening. A guy reviewing in a downtown condo might never truly appreciate the audibly elevated noise level of a device like this for example.

VI. Conclusions

So, how did the Melody Onix SP3 II do on the 4Ω "AMOAR" Composite Score?

Connected to a 4Ω load, the Melody Onix SP3 II is a low-damping factor amplifier (average 2.9x, notice the >1dB swing in frequency response with the Sony speaker load), capable of modest resolution with around 0.1% (-60dB) distortion using a triple-tone signal at 2V output level (1W). It is able to produce up to 5.8V or around 8.5W with 0.1%THD distortion which should be adequate for reasonably high efficiency speakers in a moderate sized room (note that I'm being generous with this amp since it tends to hover around 0.3-0.1% quite a bit). The amplifier rolls off the high frequencies a bit, about -0.5dB at 10kHz and -1.7dB by 20kHz. While not captured in this summary graphic, the noise floor is a bit higher than most amps and can be distracting if your sound room is very quiet.

Objectively, compared to other amplifiers I've tested, it's tempting to compare this to the Pass Amp Camp Amp 1.1 measured last year. Notice the low damping factor of the ACA and how this affects the frequency response when connected to the Sony speaker load. Overall, the Melody is a more capable amplifier with more power and significantly better Triple-Tone Distortion Factor (Pass ACA got a score of -30dB, and could not achieve <0.1% THD+N into 4Ω). Well, I suppose the Pass ACA has achieved its goal of being a solid state design that emulates the performance characteristics of a tube amp. ;-) One significant positive of the Pass ACA is that the background noise was not as high in my system compared to the Melody.

Although this is considered to be a "budget" made-in-China tube amp costing <US$1000 released 15 years ago, I suspect in terms of objective performance, an equivalent tube amplifier these days would perform similarly given maturity of tube designs. As one may expect, vacuum tube based amplifiers just don't measure all that well compared to modern Class AB solid state amplifiers or good Class D amplifiers as a generalization, nor are tube amps typically able to provide as much power with similar size/weight/price. Yeah, sure, maybe tube amps "clip" more graciously, but honestly, who wants to listen for long at levels that are regularly pushing against the power envelope of the machine?!

Despite the limited performance, tube amps do have their charms - be it esthetics ("cool" looking, and gorgeous "warmth" and glow!) or some hard-to-pin-down "euphonic" quality (perhaps related to harmonic distortion amounts or the effect of low damping factor). Furthermore, while I measured the amplifier with its stock tubes in place, this device invites the hobbyist to try out "tube rolling" should he/she desire - could be fun I suppose.

Gratuitous nighttime tube glow close-up shot.

Based on the results, I recommend trying the 4Ω output even if you know you have a higher impedance speaker. It provides for a higher damping factor, slightly lower distortion, and at least with the 8Ω power resistor load, I didn't notice any significant power limitation up to about 20W which probably would be adequate for most users, beyond which distortion rises significantly anyways.

On the whole I liked the sound but the noticeable higher background noise level in my quiet soundroom was a bit more than I'm used to. In comparison, it's quite easy to hear the cleaner, tighter, more resolving sound from the Hypex NCore Class D rising out of a background of silence. As an obsessive audiophile, I could not "unhear" the higher noise floor during late night listening sessions.

One of the age-old debates in audiophilia is the question of "tubes vs. solid state". Honestly, as a guy who aims for transparency as the goal for my reference system, coloration is not what I'm after regardless of whether it's tube or solid state. Amplifiers are supposed to amplify the sound that comes from the source, and I don't think it's the amp's business to be changing the signal being fed to my speakers. That's a personal and philosophical choice although there are instances where I might be more open to euphonic coloration (for example we'll talk about "fun" sounding headphones in the days ahead ;-). 

Obviously, I don't consider this $1000 amp to be the best of what tubes can offer but I do think the "meaty" and at times even "voluptuously" mids especially with female vocals are part of the tube "magic" some talk about and love.

--------------------

Recently, I saw this review of the US$10,000 (base model without XLR or phono stage) VAC Sigma 170i iQ in Stereophile and also this more reasonably priced £1699 (US$2300) Line Magnetic LM-34IA in Hi-Fi News.

As per the measurements, we're looking at similar levels of performance among many of these tube amps despite the price differentials. Notice it's almost useless taking tube amp manufacturer claims of watts at face value unless verified with tests!

I know, subjective reviewers will say all kinds of wonderful things they "heard", but we're looking at middling power levels typically <50Wpc, moderate distortion, and low damping factor with most Class A/B tube amps compared to solid state counterparts. Arguably, this little Melody amp performed even better in some areas (like even lower distortion) than the other two. For the VAC, I guess it's cool to have a faceplate "covered in multicoat black gloss lacquer with either gold or silver metal flakes" - I'm sure that goes well with the tube glow. Given up to a 10x price differential with the VAC and 2.5x with the Line Magnetic, it would be a great test of "diminishing returns" if we could blind-test the 3 amps with the same pair of speakers! Of course, nobody wants to do this and magazine writers will typically go one step further and argue that blind listening tests are worthless. Anyhow, down the road I'll show you measurements of some even more expensive tube stuff that are well regarded among audiophiles and discuss what these things really sound like. ;-)

To end off, notice that the term "diminishing returns" has become common lately in discussions... Whether it's Steve Guttenberg not believing it exists, or John Darko acknowledging but not really considering "value" all that well. I want to make it clear that I believe the concept of diminished returns obviously is a factor in audiophile goods, especially the "high end". While people can buy whatever they want, if there are truly sound quality gains to be had, fantastic! I'm all for better performance even if it's minor and costs an extra $10k.

A sports car enthusiast can say their Porsche 911 GT2 RS with a top speed of 211mph costing MSRP $290k is "worth it" compared to the 911 Carrera with a top speed of 182mph at MSRP $99k. At least we can appreciate the increase in price actually comes with the objective measurable performance in the form of upgrades in the engine, transmission, body, tires, etc. for speed. Furthermore, we can acknowledge the "non-utilitarian benefits" like gaining the admiration of the guys in the car club even though a top speed difference of +16% but paid with +193% in price clearly is a demonstration of "diminished returns"! Do we in general have at least an objective correlation between "sound quality" and price with things like expensive amplifiers which audiophiles insist they can hear? Maybe, maybe not (and I would argue it's more "not"). At least we can say objectively, amplifiers do measure differently if not necessarily "better".

What really is puzzling to me is not so much "diminishing returns" for devices that measure differently. It's when magazines, reviewers, and audiophiles claim that there are "diminishing returns" on sound quality at higher prices when there are in fact NO RETURNS ON SOUND QUALITY objectively. If there is any gain to be had, it's only on "non-utilitarian" factors of luxury. These categories of products IMO are the true audio snake oil. I would consider expensive cables, bit-perfect expensive digital streamers, jazzed up bejeweled luxury products that's all about "bling" as examples and there are "classes" of snake oilery as previously discussed.

As usual, I think it's important to be wise about purchases especially when it demands a significant amount of faith.

Have a good Easter 2021 weekend everyone. Hope you're enjoying the music!