Saturday 15 April 2023

MEASUREMENTS: Fosi Audio TB10D [Upgraded Version] stereo amplifier (TI TPA3255 Class D). Ethernet is inherently "isolated".


Not much time this week to play with the toys as I'll be heading off on vacation for a few weeks. However, I did receive the Fosi Audio TB10D [Updated Version] (<US$80) above recently that I wanted to use as an inexpensive "utility" amplifier for various measurements and experiments. It's also based on the TI TPA3255 Class D chip amplifier; same as the AOSHIDA A7 recently reviewed.

As you can see in the unboxed picture above, there are no frills, just the amplifier and a 32V/5A power supply (same specs as AOSHIDA).

This is the "Updated Version" with improvements as discussed by the Cheap Audio Man back in late 2022. Presumably this update was released around September or October 2022. I see ASR measured this amplifier in September 2022 - not sure which version that was. I purchased this amp from the usual retail channels.

Due to the time constraints, I'll focus more on "core" measurements and compare especially to the AOSHIDA.

"Flat" frequency response is for the treble and bass knobs pointing up at 12:00 position.

A closer look at the front above. It's quite a compact metal box measuring about 10.5 x 15.5 x 4 cm, and weighs just over 500g. You can see that it has a power switch, small red LED below the power icon, "TREBLE" and "BASS" control knobs, as well as the larger volume control to the right. I think it's unfortunate that there's no tactile articulated notch for the treble and bass knobs to let the user know where the "flat" setting would be.

And the rear:

Standard 5-way speaker cable binding posts; not the most robust quality but it'll work fine. Single-ended RCA inputs left, and power barrel connector on the right.

As usual, I had a listen to the amplifier prior to testing to form a subjective opinion then put this on the test bench.

Feeding a low 0.05Vrms signal in and increasing the volume knob to the max tells me this amp allows up to +30dB gain which should be more than enough for any DAC or preamp to push the amplifier to clipping levels.

Here's a peek at what 1kHz sine and bandlimited non-ringing square waves looks like through the oscilloscope:

Right and left channels basically equivalent, a well channel-balanced amp (as you may recall, the AOSHIDA A7 has independent channel knobs; good for channel balance control but could be a hassle to match). As usual with these Class D switching amps, notice the switching noise embedded in the signal. We can examine this using a wide bandwidth FFT on the oscilloscope:

Switching noise is primarily at 400kHz and 600kHz; I looked out to 1MHz and didn't see anything beyond 600kHz. As usual, these will be filtered out through the audio system.

Here's the damping factor graph. Note that I've increased this to 8 measurement points as an evolution of the measurements to cover the audible frequencies better. This will not significantly affect comparison with previous results:

Calculated average damping factor of 28x into 4Ω across the audible frequencies. This is a bit lower than the AOSHIDA A7 but follows the same general shape with increasing output impedance above 2.5kHz. At 20kHz, output impedance is up above 1.3Ω.

I put the treble and bass controls at 50% or "12:00" position which reflects "flat" frequency response to have a look into a 4Ω load (2Vrms output level):

Not bad, at 50% (best I could estimate) on those treble and bass knobs, I see a wee bit more bass energy than treble but of course this can be fine adjusted. Notice phase relatively flat and at 0°, which means that the absolute polarity is maintained (unlike the inverting AOSHIDA).

Given the damping factor calculated above, we should see reasonable control of a speaker load, so here's a comparison with standard 4/8Ω loads and the Sony SS-H1600 (ported, rated as 6Ω) bookshelf speaker which I've used as a convenient "standard" over the years:

As expected, due to the lower damping factor compared to the A7, there's a little more variation in the frequency response when presented with a reactive load like the Sony speakers. Still very minor fluctuations until above 2kHz where the output impedance starts to rise. Like the AOSHIDA, it looks like this amp aims to be flat measured into 4Ω.

So let's have a look now at the magnitude of the variations using the treble and bass knobs as compared to "flat":

A fair amount of range on the bass controls if you want that "megabass" sound and the +7dB adjust on the treble end can spruce up any "dull" sounding speakers or tame treble harshness of the speakers/playback material.

Okay, let's move on to distortion and power measurements...

I usual put my "high-fidelity" threshold at 0.1% THD+N (-60dB). Into 4Ω, this little amp could deliver quite a healthy 16.8V (70.5W), and into 8Ω up to 18.9V (45W) before hitting 0.1% distortion. Nice.

A quick look at THD vs. Frequency at 2Vrms looks fine, I checked a few other voltage levels below clipping and didn't see anything of concern:

Let's have a look at the 1kHz 2V into 4Ω THD+N FFT:

That's good in that both channels measure equivalently with THD+N -78dB. Notice the 3rd order harmonic is the highest. Unlike the AOSHIDA which tends to have more 2nd order product, this Fosi Audio tends to show more 3rd and 5th odd-order harmonics. We can see this tendency comparing the THD vs. Voltage graph between the two amps.

Something else to notice is that the noise floor increases into the bass frequencies. This is not as clean as the AOSHIDA which has a flatter and lower noise level (N -82dB) reflected in a better overall 2Vrms into 4Ω THD+N result of -82dB.

And let's do the 2Vrms Triple-Tone TD+N which includes intermodulation products; a simple test I use as a summary distortion amount in my ratings:

A pretty clean Triple-Tone -80dB TD+N for both left and right channels.

Finally then, here's my AMOAR summary graphic for the Fosi Audio TB10D [Updated]:

Note the Y-axis is just 4dB, accentuating the appearance of frequency response variation.

Clearly, lots of similarities between this and the AOSHIDA A7 previously reviewed. To be expected, given that they're both based on the TI TPA3255 Class D chip amp. The Fosi Audio has lower damping factor (higher output impedance), a wee bit more distortion with 3rd order harmonic predominance, and hence just a slightly lower power rating at 0.1% THD distortion. Presumably some of these differences are related to the addition of the treble+bass tone controls. At a typical output level of 1W into 4Ω, distortion level was basically the same as the AOSHIDA at -80dB TD+N on the Triple-Tone signal. 

So how does this sound? I had a listen to this amplifier in my sound room prior to these measurements and I thought it sounded quite good. Similar to the AOSHIDA, about the same power capability (later confirmed with measurements of course). Subjectively, I wondered if it didn't sound as smooth and lacked some "air" when listening to tracks like "Make Us Stronger" from Part Of The Dream. Likewise, the guitar work on The Last of Us: Season 1 soundtrack (2023, DR9) might not have been as clean with the attacks or smoothness of the decays. Otherwise bass impact was good and treble well extended. Note that this is listening with my Paradigm Reference Signature S8 speakers which cost >100x the price of this little amp when I bought the pair new, fed by the Sabaj A20d 2022 DAC which costs >4x the price so I'd say I am still quite impressed given the low cost of the Fosi Audio amp!

Of course, the above is just my subjective experience without any quick AB switching or ensuring the same output levels between amps. To be honest, I never put a huge amount of faith anyways in my own subjective impressions unless I do more formal listening tests; much too easy to be biased even just knowing that the AOSHIDA A7 is more than twice the price. Furthermore, I can tell that my hearing ability changes day-to-day depending on stress level, allergies, colds in the winter, etc... So I have to be honest with myself (and you) when it comes to subjective confidence in what I write. Anyhow, I think many music lovers / audiophiles would be happy with this sound quality especially if used in the bedroom or living room rather than a more "perfectionist" sound room with low ambient noise level.

A couple of parting criticisms that would have made this little amp even nicer if addressed:

1. There are no markings or articulated tactile notches to know when the treble and bass knobs are "flat". This would have been an obvious touch.

2. I noticed more of an inrush "pop" when turning this amplifier on than the soft "click" I hear with something like the Hypex nCore NC252MP, and the AOSHIDA A7 is essentially silent. I don't think it'll damage the speakers but it was a bit louder than I'd like to hear.

Bottom line: the Fosi Audio TB10D [Upgraded Version] is a fine, inexpensive, no nonsense, Class D amplifier currently available for less than US$80. It has bass and treble tone controls and might provide as much fidelity and power (over 70W into 4Ω and 45W into 8Ω, 0.1% THD+N) as you might need with your speakers/room. Like the AOSHIDA A7, the stock power supply can be switched out for something more powerful/cleaner if needed.


To end, I just want to remind everyone that ethernet uses differential signaling and is also "isolated" by design. The ethernet IEEE 802.3 standard specifies 1.5kVrms (2.1kV peak, usually ~60 seconds) isolation. This is done with a coupling transformer between the cable coming into the box and the Ethernet PHY chip. The picture above is of a typical 1Gbit LAN 1:1 transformer module rated at 1.5kV isolation as used inside typical ethernet boxes. These things are cheap and cost something like $0.50 each, even cheaper in bulk to the manufacturer. Heck, have a peek at any of your ethernet switches or network cards like this old translucent-boxed switch and you'll see multiple of these isolation transformers inside (exact number depends on the parts and how many ports you have):

Notice the multiple "FUN-JIN" PT20-1404C isolation transformers situated between the ports and Tamarack TC3097 multiport ethernet chip. 

Just as with isolator devices for USB (that works well when needed), it's interesting to see companies like iFi are coming out now with their iFi LAN iSilencer ethernet "filter". As usual with companies like iFi (or AudioQuest), these devices come with minimal technical specifications (see brochure, I noticed in one of the pictures the use of a Pulse isolation transformer on the board which could go up to 6kV). There are no measurements to show what noise is being suppressed ("silenced"), the specs say nothing about amount of voltage isolation the device is capable of; unlike say this Tripp Lite medical-grade ethernet isolator advertised as providing 16kV protection. Furthermore, the advertised "zero-jitter memory buffer" is more than likely of no significance to sound quality because the data will be buffered in other ways by your streamer device and DAC anyways. Jitter in the analogue output is not determined by whether the asynchronous ethernet system has jitter or not, but the clock inside the DAC. So ultimately, the only thing this isolator can do is maybe act as a kind of "surge suppressor" that can take a hit to protect your streamer or network device in the event of a lightning strike. I'll leave you to decide whether spending the money represents good value. As usual, I'm sure there will be people who claim to "hear" a difference.

[There are other devices out there like this EverStar MI-310 marketed to audiophiles and health care apparently.]

These days we see all kinds of companies selling very expensive low-volume items like $100k speakers and $20k DACs. Some of this stuff probably isn't really worth the asking price when it comes to actual audio fidelity (what I've called Class C "Audio Jewelry" assuming they're well engineered). Then there are companies like iFi that will sell relatively inexpensive doohickeys like this, typically around the US$100 price-point with no evidence that it "works" to improve sound quality (Class B "Unnecessary Audio Accessory"). They'll make their profits based on sales volume and probably will keep coming out with devices like these of questionable benefit as part of the business model so long as audiophiles keep paying the company. [The AudioQuest JitterBug is a classic example of this low-cost, high volume, questionable benefit sales model. Audiophile fuses are another example although MSRP can be in the hundreds of dollars!]

A number of years ago, I ran a test comparing noise floor from a Raspberry Pi wired with ethernet or communicating through WiFi powered using a lithium battery. The only difference I found was a very subtle 60Hz hum below -120dBFS when the ethernet cable was plugged in. So unless there's something else we think yet another doohickey hanging off the computer/switch port can do to improve music playback, demand that the companies show the benefit(s)! Who knows, maybe the iFi LAN iSilencer would have filtered that hum. ;-)

In other news, Michael Fremer believes the US$3900 Silent Angel Bonn NX ethernet switch improves sound and is awesome presumably because of lower noise and jitter of course. Yup, this same reviewer who penned the kiss-of-death "MQA-CD bests everything" article. How strange to have the "analogue guy" review something like this. With any luck, let's also hope this positive review dooms this class of "audiophile" snake oilish product. Kudos to team extreme subjective!

Stay real audiophiles. I hope you're enjoying the music! I'm off on vacation for a bit... See y'all around.

[Probably last reminder: The 16-bit vs. 24-bit Blind Listening Test and Audiophile Survey closing off May 1. Add your "voice" if not done so already.]

No comments:

Post a Comment