Unbalanced line inputs to E2x2 - poor performance

[celstark]

I'm looking to use a new Topping E2x2 as a measurement rig as its ADC is quite solid. When I do a loopback test on the unit itself, going from its balanced TRS line output to its balanced TRS line input, I can get nice, expected results. THD is down -104 dB or so in quick testing in REW and MultiTone and life is good (yes, odds are its DAC is holding this back and yes, I know to make sure the phantom power is off before connecting).

When I try to use an external, unbalanced DAC as the source, however, things fall apart. I've got a cable setup to go from the RCA's hot to the TRS tip and its gnd to go to either both the ring and sleeve or just the ring (doing a ground lift). No matter what, THD shoots up to -85 dB or so with lots of 2nd harmonic. I've tried two different DACs so far, so I doubt it's a faulty DAC. Cables are all just a few feet on the desk.

What am I missing here? I'd have thought one of those two wiring setups would work.

 

[Sokel]

Can you post charts?
Noise at check?
Level?

What's its input impedance?

When we first heard about these interfaces we expected the usual from Topping.
Far from it, even the loopback you mention is 2005 performance.

 

[celstark]

Sure -- here's a run in MultiTone I'd done earlier. It seems the balance of H2 and H3 is a bit variable, based on the output level vs. input level. I've not mapped it out yet and am just now getting back into this after literally decades of not measuring anything but an occasional speaker (true story -- the last test rig I used was an original AP System One back when that was state of the art).

OK, now here's a run in REW showing the loopback on a full freq sweep. Note how the THD line is around 15 dB

Here's a quick RTA + distortion on the loopack. Key here, BTW, is that I've set and fixed the input level to be the same as with the Schiit Modi I'll show next, so yes, performance can be a bit better on the loopback than this (we're at about 0.5 dB from full):

Here's using the unbalanced output from the Schiit as the source. Where is that THD line at 30 dB coming from?

Here's REW's spectrum on the Schiit unbalanced input:

No adjustments, MultiTone's panel on the Loopback:

and on the Schiit unbalanced:

So, we're up here as well by 15 dB. The Modi is supposed to have a better DAC

 

[celstark]

FWIW, Topping spec's the input impedance on the line in at 9k. Not ideal, but not super low. Oddly, putting an Apple Dongle in instead of the Modi+ and I do quite a bit (10dB) better.

 

[Sokel]

Better use these settings at REW, so we can align:

You can choose dBFS units at the upper-left corner of the chart.

Multitone's Modi chart seems awfully contaminated with mains noise or ground loops, you have to take care of that.

 

[celstark]

Thanks -- similar results here with those settings. FWIW, it's 90 Hz there, not 60 Hz, so mains noise likely isn't it. But, yes, I'm going to hunt around and see what may be driving that. It's in both channels. This 90Hz is at -104 dBFS and survives unscathed switching the DAC's inputs. Pulling the RCA and the noise floor drops with an expected 60Hz showing up at -130 dBFS.

 

[celstark]

FWIW, that 90Hz shows up on multiple devices and is tied to the sample rate. 44.1k = 90 Hz, 48=100, 88=180, and 96k is 200 Hz. There's no sign of it when running the loopback balanced on the E2x2. THD is at -104 or on the loopback. But, I replicated that series of peaks across several devices and it's clearly tied to the sample rate when running an external unbalanced device into it. I've got a DX5-II on the way here which will provide a better reference and will provide both balanced and unbalanced outputs.

 

[half_dog]

Usually when evaluating an unbalanced device, I do 3 things that gives better results:
-Toggle on the instrument button.
- Try to leave the DUT "floating". I used to do this using a laptop on battery, but now I use a HS-1 USB isolator with external power supply (if necessary).
-Not always, but sometimes use a Topping L30 as buffer and/or volume control. This interface has a better THD value when the input is around -15~ -12dBFS; but the noise gonna be higher.

 

[AnalogSteph]

Oddly, putting an Apple Dongle in instead of the Modi+ and I do quite a bit (10dB) better.

I would suspect a combination of lower audio levels and some mutual cancellation of dongle DAC and input amp distortion.

Increased distortion on unbalanced input is not unheard of... the THAT626x preamps used by the 4th gen Scarlett 2i2 or MOTU M2 seem to exhibit a fair bit of it, for example. These things are not necessarily optimized for lowest common-mode distortion (even-order harmonics are generally expected to be cancelled out), so performance can be a crapshoot in that regard. Being measurement-grade in all respects is not typically an objective for a lower-midrange audio interface, even if you may get lucky. If you need a Cosmos ADC, you're better off buying one of those. Otherwise you want to be looking for models with an extra line input entirely bypassing the mic preamp (like Scarlett 4i4 or MOTU M4 have), or one with internal supply rails decidedly above +/-5 V and e.g. PGA2500 input amps (like an Ultralite).

- Try to leave the DUT "floating". I used to do this using a laptop on battery, but now I use a HS-1 USB isolator with external power supply (if necessary).

Keeping ground loops out of your test setup is always a good idea (the kind of galvanic isolation provided by an AP is not a standard feature after all). That being said, wiring RCA to 1/4" R + S ought to be a good bit better than nothing in case of shared grounds, even if using a USB isolator gives cleaner results.

 

[celstark]

Usually when evaluating an unbalanced device, I do 3 things that gives better results:
-Toggle on the instrument button.
- Try to leave the DUT "floating". I used to do this using a laptop on battery, but now I use a HS-1 USB isolator with external power supply (if necessary).
-Not always, but sometimes use a Topping L30 as buffer and/or volume control. This interface has a better THD value when the input is around -15~ -12dBFS; but the noise gonna be higher.

Can you expand on that a bit? I thought maybe the L30 had balanced 4.4 outputs and you were using it to do the conversion, but it seems to be single ended. You're using it though to isolate as a buffer?

I'm still exploring a bit with the INST vs without and various cable wiring setups. I've now got a nice collection of plugs with screw posts that let me vary what's floating, etc. These also let me dump a resistor on easily to simulate a load.

 

[celstark]

OK, @half_dog and @AnalogSteph got me thinking here (which may be a dangerous thing)... If I go down this rabbit hole, a different ADC (e.g., Cosmos) would certainly be a solution. It seems something like the SMSL H300 would be as well, with its balanced and unbalanced inputs and outputs (along with headphone outputs of course). The Topping A30 Pro seems to be a similar setup.

I don't actually have a need here other than to scratch some curiosity itches. The performance of both those units by paralleling loads of op-amps is just staggering. If these can be used as, in effect, the input stage, it's an interesting solution.

 

[Sokel]

OK, @half_dog and @AnalogSteph got me thinking here (which may be a dangerous thing)... If I go down this rabbit hole, a different ADC (e.g., Cosmos) would certainly be a solution. It seems something like the SMSL H300 would be as well, with its balanced and unbalanced inputs and outputs (along with headphone outputs of course). The Topping A30 Pro seems to be a similar setup.

I don't actually have a need here other than to scratch some curiosity itches. The performance of both those units by paralleling loads of op-amps is just staggering. If these can be used as, in effect, the input stage, it's an interesting solution.

miniDSP ADEPT is also a solution, check it out.

 

[celstark]

miniDSP ADEPT is also a solution, check it out.

Another very cool product -- we do live in amazing times... Fine, I'm in my 50s and started in this hobby in my early teens, so it's been a hot minute. But, back then, my dad's AP SystemOne didn't have specs that match what we can get for a few hundred bucks. The ADEPT there is $299. The A30 Pro is currently $215. Heck, the E2x2 was something like $130. Crazy sauce! (Again, giving away my age).

 

[Sokel]

Another very cool product -- we do live in amazing times... Fine, I'm in my 50s and started in this hobby in my early teens, so it's been a hot minute. But, back then, my dad's AP SystemOne didn't have specs that match what we can get for a few hundred bucks. The ADEPT there is $299. The A30 Pro is currently $215. Heck, the E2x2 was something like $130. Crazy sauce! (Again, giving away my age).

I have one that I'm testing for a month or so and I haven't managed to break it,.
And trust me I'm good at finding flaws.

It may not dive at the abyss of -130dB THD+N as a Cosmos can do but its higher input impedance, etc comes handy measuring just about anything with the right gear before it.
That's apart from it's mainly functionality of course.

 

[celstark]

OK, @Sokel wins the prize for the day. I grabbed a Topping HS02 before the holidays set in and I'm now at > 100 dB on my unbalanced -- really matching the balanced pretty well, albeit with a bit of 60 Hz mains still in there on the quick tests. So, the unbalanced performance was being hit a lot by ground loop noise. Testing a little Schitt Fulla's pre-out, I get to -107 dB THD and -101 THD+N. I'm sure there may be a bit more in there after optimizing levels, but this is way better than I'd been.

FWIW, a touch better performance (perhaps) with the isolator on the DAC than on the E2x2 and best performance in the INST mode (aka unbalanced TS, 1-meg impedance, rated -107 dB THD+N). I'm not looking to publish anything and, frankly, when the distortion components are 110 dB or more down, I'm not hearing that

 

[AnalogSteph]

What is the cabling you're using now? If it's still wired unbalanced to tip and ring, input impedance alone may be enough to (largely) balance the signal amplitudes, especially when only wiring up one channel. Something to keep in mind.

Now if distortion remains that low with dumb unbalanced to TS cabling (T = signal, R = S = ground), then it actually was compromised by the ground loop. That's not unheard of.

 

[celstark]

What is the cabling you're using now? If it's still wired unbalanced to tip and ring, input impedance alone may be enough to (largely) balance the signal amplitudes, especially when only wiring up one channel. Something to keep in mind.

Now if distortion remains that low with dumb unbalanced to TS cabling (T = signal, R = S = ground), then it actually was compromised by the ground loop. That's not unheard of.

That graph is using the Fulla's 3.5mm stereo pre out. I've got a mocked up cable using connectors w/screw terminals. 3.5 Tip -> 1.4" Tip; 3.5 sleeve -> 1/4" sleeve; Shield -> 1/4" sleeve. Internally, when I hit the Instrument mode, it assumes I've got a TS input. But, that's what was used in the above plot

I get similar results (have actually hit -109 dB THD on both of these) running the E2x2 in the balanced line in mode (non-INST) with 3.5 T-> 1/4" T; 3.5 S-> 1/4" R; Shield -> 1/4" S as well as when tying the R&S together. Same cables w/o the isolator and it shows all that garbage below the 1k fundamental -- garbage that moved regularly with the sample rate.

When the DX5-II comes in with the same device having both balanced and unbalanced I/O, I'll have a cleaner picture, but my goal here isn't to try to detect which device with a SINAD > 110 is "better" than another. Heck, I remember when a THD of 0.01% was thought clearly inaudible and that's -80dB. But, my CamDAC project? What's its actual noise? That little dongle? How's it doing? What's that passive x-over really doing? Did I screw it up? Lots of great applications for this now that I'm getting pretty clean results.

 

[celstark]

Quick update, as I've gotten some time with the Dx5-II as a source.

Running unbalanced out from the DX5 via just a cable conversion, if I allow myself to ride at -6 dBFS, I can do quite well here (screencap). This has the E2x2's gain set at the minimum. Increasing the input gain on the E2x2 takes the THD to -100.9, THD+N to -100 and the SNR to 108 dB.

In the same setup, running from the DX5's balanced output (gain at min on E2x2), we're cleaner as expected:

So, all in all -- the USB isolator was the key to me getting decent unbalanced performance. The E2x2's giving good, but not outstanding performance. At the $130 I got it for, I still think it's quite impressive and clearing my thresholds for performance.