Poking around inside a Swissonic UA-2x2

[AnalogSteph]

Can a 2x2, 24/96 capable (UAC1) audio interface for a mere 65€ from Thomann's house brand be any good? Well, I just happened to need something to take care of an XLR condenser mic so I could finally liberate the Audient EVO 4 from my office PC and use it for portable measurements again, and I was curious about this model (the 102/108 dB ADC/DAC specs didn't ring a bell), so I placed an order...

First impressions: That's a chunky boi - the thick steel chassis makes this unit extraordinarily heavy and sturdy-feeling. (925 grams or literally just over 2 pounds - yowza!) The pushbuttons are a bit wobbly, but overall it feels like a solid piece of kit. The wood imitation side pieces are unsurprisingly plastic.

Sorry Dave, I did turn it on before taking it apart:
Both inputs including phantom power work fine with my t.bone SC400 LDC. HD600s get plenty loud enough on playback (and show no hiss there even with volume maxed), but mic monitoring could definitely use some more gain. Input noise level at minimum gain suggests a dynamic range of around CD spec. (Audacity's level meter says about -89/90 dB unwtd at 48 kHz and -92 dB A-wtd.) I have yet to test in loopback, silly me forgot to order some TRS cables and I may have to borrow my monitors' TRS --> XLR cables again.
The main monitor outputs have substantially more oomph than my ALC1200 onboard audio, even disregarding my speaker EQ and its pre-cut and a bit of Behringer HD400 insertion loss - I can really dig into the noise floor of dynamic classical recordings. Plugging the interface into USB generates a nasty spike on the main outputs, while unplugging creates a much less dramatic thump that is affected by monitor volume.
Prompted by my findings inside, I performed a dynamic mic test as well, digging out my 600 ohm Samson Q2U - I found it's usable but not the last word in input noise. I reckon my Mackie 402VLZ4 mixer (on ALC1200 line-in) is about 5 dB better, which does not bode well for more critical models like the notorious SM7B. EIN is likely to be about -120 dBu(A). At least it still sounds OK otherwise, not like e.g. the ESI Neva line which basically delivers telephone quality at high gain (no, I'm not kidding).

So what's inside?
After removing both side pieces, three more screws hold half the steel chassis in place, which slides off towards the front. The insides are a 2-layer construction with board to board pin type interconnects between two boards with all parts on just one side. There is a fair bit of open space on the bigger main PCB in particular,
So I start looking for the part that interests me the most, the audio codec. What do I see staring at me - a freaking Wolfson Micro logo. Sure enough, it's a trusty WM8776 codec (WM8776S / 05BSMEG). A common choice for soundcards in the mid-2000s, but I haven't seen one used in new gear in donkey's years. They were only discontinued by Cirrus Logic a few years ago though, last time buy was in summer of 2022. The DAC digital filter has a bit too much periodic ripple for my tastes, while the ADC is quite good for a consumer job. Dynamic range is better on the DAC side though.

The USB interface chip is a little square 48-pin thing hiding underneath a sticker saying "ESI018". Probably one of the things that were popular before XMOS took over the world. An 8-pin (T)SSOP job nearby is probably an EEPROM (ATH008 / 08CMPH / 2008QCF).
Between them is another SO-8 chip labeled 501MLF / 100205PY / 2117, no idea what that is (that's the most recent date code though).
There are a few JRC NJM4580 opamps dotted throughout the unit, looks like NJM4580Vs in SSOP-8. There are two next to the outputs alone, which suggests there is an actual balanced line driver circuit. That would certainly explain the powerful main output. Unfortunately it looks like 4580s with not much else around them are also used on the input side, it would certainly be consistent with mic input results. So file this one under "cheap audio interface good for condenser mics".

There is a little 12-pin MSOP (?) labeled SHY / TI 051 / CE12 which I presume is a headphone driver. Anyone got a clue what that is?
There is one prominent DC/DC converter inside the unit, presumably for +48V, with a 3- or 4-pin SOT23 labeled ?S52B. I very much assume the analog section is run on +5V single-supply, as par for the course in this class.

Next to the USB (type B) socket there is an SOT223, presumably an LDO (3.3 V?), labeled (logo) EH16A / 104TI2.

Armed with the WM8776S lead, I stumbled across this really quickly:

ESI U22 XT — недорогой USB звуковой интерфейс для интернет-вещания | ProSound

prosound.ixbt.com

That's pretty much the little brother to our unit. The family resemblance inside and out is unmistakeable, mechanical construction seems quite similar. So I presume our USB chip underneath the telling "ESI018" sticker is also going to be a TE7022L, matching the general theme of "old faithful". The choice of sample rates (44.1, 48, 96 kHz - TE7022L doesn't do 88.2) would certainly support that, the Windows control panel software looks very similar as well.

vs.

(BTW, U22 XT pricing seems to generally have dropped to around 40€ now. Not a half-bad deal. Mind you, the Steinberg UR12 is being blown out for 44€ at this time, a CS4270 deal that makes decent use of this part. While still something that prefers condenser mics, that's more than just podcast grade already. Unbalanced output only though, even if converting that is not too much of a problem. Unbalanced circuitry on +/-5 V supplies internally, so half a step up from single supply jobs, especially on the input side.)

Speaking of software, version 2.0 installed on Windows 10 without a hitch and seems to be working fine. You can also download version 1.1 for older Windows versions which fixed some nasty bugs in the original 1.0 release (like bluescreens when unplugging). The driver files are bearing a copyright notice of (C) 2023 ESI Audiotechnik GmbH, further confirming the ESI connection.

During reassembly I lined the insides of the side pieces with Al foil in an effort to make the unit more RF-tight (not that I had noticed any issues, mind you, but it can't hurt; I didn't go so far as to be scraping off the paint where panels are touching). I don't think it had any issues before, but I certainly can't impress it with RF now (either 4G LTE or WiFi).

You will not be surprised to learn that the unit pretty much does not warm up at all in operation. Not a whole lot going on in there. It's not a Focusrite Forte for sure. I won't be surprised if it doesn't even draw 200 mA.

 

[AnalogSteph]

Between them is another SO-8 chip labeled 501MLF / 100205PY / 2117, no idea what that is (that's the most recent date code though).

OK, Google solved that one, it's a clock generator (Renesas ICS501). Assuming it's fed with 12 MHz from the crystal next to the USB chip, it could be used to generate 24, 36, 37.5, 48, 60, 72, 75, or 96 MHz. I wonder what they are using that one for, none of these frequencies look particularly ideal.

USB Device Tree Viewer identifies the USB VID as "Ringway Tech (JiangSu) Co., Ltd.":

Ringway Tech (jiangsu) Co., Ltd. - Musical Instruments, Sheet Music, Music Production - FOB Business Directory

ringway was founded as an ic designer in 1994, after 19 years development, today’s ringway is one of the main designer and manufacturer of the professional digital musical instruments in china.we are making digital piano, electronic keyboard, midi controller, electronic drum, double manual organ, music teaching system and midi interface products. the current plant covers a land of 47,000 square meters,and a building space of 80,000 square meters. we employ over 1000 people in china with additional engineering take place in the united states, eu,africa,south korea and hong kong.

That would explain who makes these in China. I won't be surprised one bit if more of their products turn up in Thomann's keys department.

The audio endpoint descriptor contains

bmAttributes             : 0x09 (TransferType=Isochronous  SyncType=Adaptive  EndpointType=Data)

which would be in agreement with the TE7022L datasheet. Async USB would have been asking a bit much, I guess.

There is a little 12-pin MSOP (?) labeled SHY / TI 051 / CE12 which I presume is a headphone driver. Anyone got a clue what that is?

The Kajagoogoo part may in fact be a TPS65133 split rail converter in 12-pin WSON. I'll be... Screwdriver, where are you?
...
Sure enough, I'm finding +5.1/-4.9 V supplies. Now that's not something you routinely encounter in a sub-100€ interface. Remember the Behringer UMC2xxHD series is all single supply!

So this thing has both split supplies and a balanced output driver? No wonder the monitor output has such a lot of oomph. Then there's probably a pair of 4580s hiding away somewhere that are driving the headphone output, like in the U22 XL. Yep, I can spot two near the headphone volume pot on the upper PCB, with board space for two more.

On the downside, I just found a driver bug/quirk - recording in 96 kHz via WASAPI gives me truncated 16-bit audio samples only. I think this is because as the TE7022L datasheet mentions, you cannot do 24/96 full duplex over USB 1.1 Full Speed (12 Mbit/s) bandwidth, so either playback or recording has to be switched to 16 bits. It seems recording is getting the short straw in this case, and there is no way of explicitly setting output to 16 bits (or temporarily disabling it) instead. I may need to try my luck with ASIO.

Best-case input noise level I've seen so far is about -94 dB(A) in 44.1/48 with gain set to minimum, input shorted (input gain in control panel @ 0 dB, 20 kHz 48 dB lowpass + A-weighting EQ). A fair bit of 1/f noise on there, not sure what the deal is with that (pretty much would have to be the ADC, which contains a fairly extensive input stage with PGA and stuff).

 

[MCH]

Looks identical to the subzero ai12, other than the full size xlr at the rear

I wonder where can I get one but with my name, in big letters on the top as well.

 

[AnalogSteph]

Good call - there are some unpopulated footprints for XLR jacks alongside associated components on the board.

That's a Gear4music store brand, isn't it? So a pretty similar OEM deal then.

The drivers - again - are hinting at ESI. I think that's pretty much the modus operandi for them, cooperate with talented hardware developers in the Far East (originally South Korea, now also China) and write the drivers.

I've had a look at the SZ-AI2 manual, and that actually contains some specs, notably:

Impedance . . . . . . . . . 10kOhm line input, 3.3kOhm microphone input, 500kOhm Hi-Z input
Max Input Level. . . . . . +20dBu line input, +4dBu microphone input, +12dBu Hi-Z input
Max Output Level. . . . +4dBu

Their accuracy may be up for debate, of course. You never know what has been populated differently either.

I've measured output now by generating near enough 0 dBFS 1 kHz tones and consulting my multimeter (UT71C, TrueRMS to 100 kHz). The results: I'm getting 2.28 Vrms on the monitor outputs, split evenly across hot and cold (so balanced driver confirmed). The headphone output delivers a tad over 2.3 Vrms on each channel. So approximately +9.5 dBu for both outputs. A more than decent value for a super budget audio interface. You may remember how the UMC204HD was quite unhappy at +3 dBu already. And IMHO the monitor out could very much go even louder if you insisted, after all Focusrite got +15.5 dBu out of a very similar setup. It's not exactly super necessary though.

 

[AnalogSteph]

Unit has now been dragged to the office. Win7 was throwing a hissy-fit about the v1.1 driver "not being signed" for some reason, but generic USB audio works. Selectable playback formats are up to 24/96, but recording only offers 24 bits up to 48 kHz and 16 bits at 96 kHz as previously noticed. So that would seem to be a firmware rather than driver limit.

I think my office mic Marantz MPM-1000 may be defective, it seems to have no more output than the dynamic I was testing with the other day and things are noticeably noisy. Poking around inside the mic revealed nothing of note, and it seems phantom power seems to be making it to the circuit board fine (about +34.3 V on hot and cold out of 47.8 V from the interface unloaded, suggesting about a 4 mA current draw). That's about as much testing as my tools and eyesight would allow. OK, this'll get dragged home and compared to my SC400 of nominally about the same sensitivity.

Now I get why the MPM1000 can be made so inexpensively, all the electronics are on a small circular PCB underneath the basket, with the capsule poking through. The void underneath is just filled with PET foam with a slit for the cable.

EDIT: update from home, MPM1000 sounds fine here although it is roughly 9 dB quieter than the SC400... that's not unrealistic, the SC400 spec translates to -32.6 dBV/Pa (so ~5.5 dB more) and it's got a 32 mm capsule vs. 18 mm so a more pronounced proximity effect is expected (mouth-mic distance being ~15 cm). If I put a 300 Hz 12 dB highpass on the audio, the difference seems to reduce to about 6 dB.

As a last-ditch effort, I'll be dragging along another XLR cable tomorrow, or maybe two, and maybe the SC400 just in case. If it's not the cable either, all I can conclude is that possibly the phantom power supply in the UA-2x2 is noisy and the MPM1000 doesn't particularly like that.

EDIT 2: Back in the office, armed with misc. XLR cables and my Q2U. Cable didn't make a difference. I'd say achievable noise level almost is better with the dynamic mic than with the MPM1000. The hell?!

 

[AnalogSteph]

I was not happy with condenser mic noise levels on this unit. Ordered a Steinberg UR12 Black instead - hard to go wrong at half off (44€), I thought. That thing is super basic with only one mic input and unbalanced RCA outputs only, and build is a good bit less massive, but you know what? The MPM1000 is much, much quieter like that. Mission accomplished. The USB cable feels much nicer, too.

Now that being said, it isn't all sunshine and roses. While the preamp seems to have very good input CMRR, indicated by the fact that you can turn off phantom power and the loudest thing recorded is the sound of the switch itself being flipped (I think the trick is that there's momentary muting), there is a permanent dull low-level knocking / thumping sound at about 188 bpm no matter what I do, with slight ~60 Hz hum (note: mains is 50 Hz around here). Is this why they're being sold so cheaply while the regular UR12 remains at 89€? I can live with this issue as it would not impede video calls and noise reduction seems to take care of it quite effectively, but I imagine recording music may be a different story. It seems to be related to phantom power either way, as it disappears without it. BTW, phantom power shuts down unusually quickly, in no more than a split second.

Here's some noise spectra from both interfaces with the MPM1000:
UA-2x2:

UR12 Black:

The UA-2x2 noise looks suspiciously like a zener reference or something. In any case we don't even seem to be approaching preamp / ADC noise floor until >20 kHz. The UR12 looks a lot more like what you would expect from a condenser mic.

Who knew that there could be so many pitfalls just trying to record a condenser mic?

 

[AnalogSteph]

There is one prominent DC/DC converter inside the unit, presumably for +48V, with a 3- or 4-pin SOT23 labeled ?S52B.

This might be a LM2733, though that part is only rated to 40 V... but there's an appnote on how to use a coupled inductor to extend the voltage range, so maybe L1 is one of those?

EDIT: I tried looking at the phantom power voltage using my trusty scope, no dice (only having a 10:1 is not helping matters), but my moldymeter (UNI-T UT71C, TrueRMS rated to 100 kHz) in AC mode showed about 0.0032 Vrms (3.2 mVrms). That doesn't seem like a lot, but if you consider that self-noise for the MPM-1000 comes out to only about -114 dBV(A) (2 µV(A)) with a substantial 1/f² component, it means that required microphone circuit PSRR would be at least 60-70 dB. That's asking quite a bit from what's generally a zener regulator with some RC filtering.
Unfortunately giving the multimeter a shot on the Mackie 402 VLZ4 phantom power gives much the same result, so we may be looking at the limits of the multimeter's circuitry here. I tried twisting the probes and everything.

 

[AnalogSteph]

Updates on the UA-2x2:
1. I noticed that just having the interface plugged in causes not-insubstantial CPU loading (something like close to 20% on one core, or a delta of ~10 W here) - I guess the control panel is watching it like a hawk or something. I'll be going with generic UAC drivers then (and ASIO4All if need be). No such issues there.
2. Major D'oh moment -
Inputs full gain, phantom power off:

Turn phantom power on, and the input settles on this after a while:

So that's my excessive noise floor right there.
Both inputs' noise levels are also behaving quite differently in time as phantom power is being turned on/off.

So it seems it might just be a duff input then? I could see a leaky coupling cap of dubious quality as being a definite possibility in this class (or perhaps just some flux residue in the area). And of course it had to be the left channel, the one you need for videoconferencing. It's always the left channel. Bloody Murphy.

EDIT:
Tried something else with the multimeter - a current measurement in the mA range.
Good channel - R:
2 to 1 - 6.174 mA
3 to 1 - 6.174 mA

Bad channel - L:
2 to 1 - 6.174 mA
3 to 1 - 6.154 mA

While technically still well within spec, the difference of about 20 µA in the bad channel is intriguing.

Among my poking and prodding, the bad channel now seems to have calmed down a bit, averaging a tad over -42 dB. That's still 24 dB worse than the good one though.

 

[AnalogSteph]

Dragged the UR12 Black home as well this week. I'm not seeing that pulsating noise on any of my condenser mics here (t.bone SC400, SC140, Samson CL8). Great - a USB power issue on my office rig then, perhaps? An external USB charger is something I hadn't tried (feeding the Micro-B aux power socket from the same computer didn't make a difference whatsoever).

Either way, my mixer's input still seems to be noticeably quieter than the UR12, particularly on the SC400... less white noise. It's not night and day though. You wouldn't think that this would be the case with broadband noise at about -118 dBu unweighted at 48 kHz (assuming the 0 dBu max input and +54 dB max gain specs are accurate and staring at FFT analysis in Audacity). When using the old CL8, its rather mediocre self-noise (think a tad over 20 dB SPL(A)) does remain safely above preamp levels across the audible range, and much the same applies to the SC140 SDC which is reasonably hot. Lesson learned: Modern low-noise LDCs do appreciate good (or at least non-mediocre) EIN performance.

Maximum mic input dynamic range on the UR12 at min gain seems about 3-4 dB lower than for the good input on the UA-2x2... about 91-92 dB unweighted at 48 kHz. The Rch line input fares better, at about 98 dB. That's probably ~102ish dB post A-weighting @ 20 kHz BW, approaching CS4270 specs.

If you thought that @Julian Krause's UR22 mkII results were mediocre, well yeah - the UR12 seems to be a step down even from that.

In any case this has only reaffirmed my belief that my bodgy combo of a Mackie 402VLZ4 mixer (cost me about as much as the UR12 used, pre-pandemic) with ALC1200 onboard audio can compete with audio interfaces up to around the 100€ mark.

 

[AnalogSteph]

Dragged the UR12 Black home as well this week. I'm not seeing that pulsating noise on any of my condenser mics here (t.bone SC400, SC140, Samson CL8). Great - a USB power issue on my office rig then, perhaps? An external USB charger is something I hadn't tried (feeding the Micro-B aux power socket from the same computer didn't make a difference whatsoever).

Let's correct this. Using a Samsung charger from my old Galaxy A3 (2014):

The knocking sound is in fact gone, and things look pretty much textbook now. (Finally. Man, what an ordeal.) MPM1000 noise seems to dip below UR12 noise a tad over 4 kHz - the white noise floor at -103ish dB is preamp noise.

The good input on the UA-2x2 should be looking much the same, just with everything several dB lower as it doesn't seem to have input sensitivity quite as high.

Testing the UR12 mic input at 192 kHz reveals a well-extended frequency response even at maximum gain, with a -1 dB point exceeding 46 kHz. Actual analog response seems to drop very slowly - a spot check at 0.40 fs (76.8 kHz) shows a noise floor about 8.5 dB down, with about 7.2 dB of that already being expected from the ADC's digital filter. Clearly there's no shortage of GBW in the preamp. Noise levels aside, it should sound good. (*)

The spike at 69.6 kHz presumably is power supply noise.

*) I somehow doubt they are using the classic Yamaha D-PRE circuit in this (which is basically the ESP P66 topology and has been used by Behringer and others for ages)... the input transistors would have to be complete rubbish in terms of rbb'.