Tale of two USB isolators

Guessing on which forum to post this test.

Anyway, looking to see WHY a couple different DACs sound different. This is not a debate if or what, just a measured test.

One thought was how well they rejected USB power or signal noise. So I bought a $12 e-bay special galvanic isolator and the $80 Linear Technology demo board for the LTM2884 module.

Just looking at the 5V line going into the isolator and the 5V coming out:
Cheap board: 5 to 5.5 mV input. 5.5 to 6mV output. Isolated maybe, but not very clean!

LT board, same 5 to 5.5mV input, but .9 mv steady out. Measured with my Fluke as my scope died so can't show fancy pictures of what the noise looks like. Same cable.

Not making claims on audibility, but just objective measures of one aspect.

Curious, my Focusrite I wanted to use for a loopback test does not get recognized with either board, but my other DACs do. Need to figure that out so I can do distortion analysis of the DAC, and not just of the Focusrite. I thought not enough current, but the LT board 5V does not droop, so don't think that is it.

My conclusion: I make no claims about magic sonic improvements, but the LT module works and the Chinese one does not as far as the power line goes. It has what it claims in an AMD chip on it, ADUM3160. IF it is real, it does not do very well on the power side.

More tests to follow if I can get any that are not masked by the Focus rite as the DACS are both cleaner than it is. Just tested with a poured USB hub, and still the PC does not recognize the box. Odd. Going to contact Focus rite.

Note, the AMD chip lists @ $3, the LT module lists @ 40. Maybe a reason :)
 
mV of AC measured on the 5V line, input and output. Sorry if not clear. Pretty much says whatever the cheap thing is using for a regulator stinks.
If this has any bearing on the DAC performance probably depends entirely how they implement it. As it happens, both my "good" DACs are line powered so I suspect it is totally irrelevant for me. ( Schiit Asgard and JDS Atom DAC+. ) Every objective data point is good background.

I actually bought these when I thought I was going to use a USB scope.
 
That seems to be the same design as the cheap one I have. Same chip at least. Or maybe a real one?

The Modius uses a USB isolator internally as a prime feature of their Unison interface. I wonder if cascading isolators cause too much propagation delay in the handshake? Just a thought. I'll be curious what Focusrite says.
JS ( JDS) suggested their USB implementations is quite robust so it is not a problem and does not need one.

The problem with tests as Arch' did is just as he says, most don't have a problem so they don't need one anyway. A more valid test would be to inject known levels of hash and see how different DACs perform with and without. He is using a superb, I mean no one can touch, DAC for his test and I would upset if I had one and it was not taken care of inside. I would be more interested if he did the test again with a Dragonfly and crappy laptop. Would any artifacts added by noise exceed -110 dB or so?

As I said, I bought these for a digital scope isolation. It's smart to physically protect the PC in case of a miss-step with the probe. Not for some magic cure for DAC problems. THOSE I have traced to something that aggravates my speakers in the 3100 range. The Schiit and Topping did more than the JDS. Old Muse Wolfson from 20 years ago was even worse. None of them seem to have the "glare" problem I hear through speakers when I listen through my headphones. It is a mystery. My best guess is high order harmonics and IM with how the tweeter reacts in the breakup zone, doppler distortion. As I am happy with the JDS, my next step is to build another set of speakers. Got my eye on the CSS drivers, but a couple SB and Peerless got my eye too.

I tired to discuss it over on that "objective site" but if you do not believe all DACS and all amplifiers sound the same, you get called a liar and thrown off. Can't figure out, if one believes they are all the same, why do they participate in a forum whose purpose is to define the differences objectively and get so upset when one is interested in additional tests or interpretation of tests that could relate to what we subjectively hear?
 

NicMac

Member
2009-02-26 8:34 pm
Just looking at the 5V line going into the isolator and the 5V coming out:
Cheap board: 5 to 5.5 mV input. 5.5 to 6mV output. Isolated maybe, but not very clean!
It is a galvanic USB isolator, not a DC voltage regulator, so why do you expect less AC on the output?
The only USB isolator I have tried that works consistently across my devices (including bus powered USB scopes) is the Intona USB Isolator. It is expensive, but you get what you pay for.
 
Yea, it has an A and B plug on it.
I expect that from e-bay sellers, not Linear Technology.

It also explains why a couple companies adding isolators in their DAC are using ethernet transformers, not dedicated USB. Plus they don't need the DC-DC portion. It is also an interesting read on LT about all the other devices that should be part of a USB receiver to make it robust. And that does not count the better DACs doing their clock thing. So, again, it is more to do with implementation than chip brand.

Playing around with it using my partially working remaining scope input, laptop to the Atom, just probing the lines, not testing output:

My takeaway, if
If you have a noise problem, these things work to reduce the hash. Basic filtering because they are a transformer.
If you have a noise problem, best go with a line powered DAC
If you have a noise problem, the better DACs probably take care of it anyway
If you have a ground loop, yea these will break it, but best to fix the root of the problem

Rethinking a USB scope. I really don't want a project, but I guess the only way to get what I want is to build a 100 gain, AC/DC protected amplifier to front it. Probably more than a perf-board for the noise levels I need. Darn consumer electronics are so much better than when I was a lab rat, it is hard to imagine. Preamplifiers for $150, in a box, with power supply and remote, and .00001% distortion, noise below 120dB. Really, this is hard to phantom. Of course the scope amp only needs to be about 1%, but noise has to be down there. 1M input so that is a noise challenge. Darn Johnson. Need to look at the scope probes better. Could be I could do 100K and be OK. Scope probes are 100 Ohm 1X, and 9M 10X
 
That seems to be the same design as the cheap one I have. Same chip at least. Or maybe a real one?

The Modius uses a USB isolator internally as a prime feature of their Unison interface. I wonder if cascading isolators cause too much propagation delay in the handshake? Just a thought. I'll be curious what Focusrite says.
JS ( JDS) suggested their USB implementations is quite robust so it is not a problem and does not need one.
The Modius has Unison USB but does not feature galvanic isolation. Only the Schiit DAC's above it in the line are galvanically isolated.
The problem with tests as Arch' did is just as he says, most don't have a problem so they don't need one anyway. A more valid test would be to inject known levels of hash and see how different DACs perform with and without. He is using a superb, I mean no one can touch, DAC for his test and I would upset if I had one and it was not taken care of inside. I would be more interested if he did the test again with a Dragonfly and crappy laptop. Would any artifacts added by noise exceed -110 dB or so?

My Dell laptop, RPi 3B+ and QNAP NAS all benefit from cutting the 5V USB power. To do this I use an iDefender, dual head USB cable or the USB Isolator, with an external PSU. I've used these in my desktop system, with a number of inexpensive DAC's and always heard an improvement. These tweaks are not needed, nor desirable, in my two better systems.