Adum4166 USB isolator

It turns out that the 66 differs from 65 in that peripheral side with clock can be powered from clean supply.

No.
In any case, clock is not powered directly from Vbus, there is LD (5v->3.3v).
USB requirements to the clock are not very strong, it is not the audio clock.

but then we need additional isolated 5V source for PC side to cut VBUS from PC

You will get even worst results. So, no - PC side should be powered from PC. Isolated side, should be powered from the connected device, or also from the PC side through isolated DC-DC.


In case of 3166/4166 and the secondary is powered from the connected device, there is no clock.

Alex.
 
  • Like
Reactions: Anatolii_A
Hello,
Here's the circuit I'm planning to build.
It will be placed between my streamer and my DAC.

If you have any comments, they're welcome.

Capture d’écran 2025-06-03 114016.png
 
Hello Altor,

You say you're using TI's ISOUSB211 chip.
Do you have any reasons to prefer this chip over the ADuMx16x chips? (For audio applications).

I have a simple criterion. I'll be the one soldering these chips onto the PCB. I can solder correctly up to a spacing of 0.95 mm.
Below that, my soldering is unreliable.
 
If you have any comments, they're welcome.
If you reserve space for the crystal and 2 capacitors also on pins 5-6, you will be able to try both 65 and 66 chips.

You say you're using TI's ISOUSB211 chip.
Do you have any reasons to prefer this chip over the ADuMx16x chips? (For audio applications).
TI seems more reliable to me, although a little more expensive.
I've tried it with several DACs (not mine, since my DACs already have isolation on the I2S bus between the USB part and the DAC part), and I use it all the time with the Cosmos ADC to prevent ground loops during measurements.

With AD sometimes you had to unplug and plug the device back in to get it to work.
Also my customers prefer TI.

I can solder correctly up to a spacing of 0.95 mm.

AD and TI have the same pitch - 0.65mm.
Use mini-wave soldering tip (BCM2, in case of T12) and don't skimp on flux 🙂
I did it many times 🙂

You can also order PCBA at JLCPB, they have 3165/66 at stock.

Alex.
 
Yes, the 3165/6 is in an SSOP package and the 4165/6 is in an SOIC package.
But you were right to insist, as I only just discovered it.
I instinctively went for the 41 without looking at the 31 in detail.

I also just discovered that the 31 is significantly cheaper than the 41. Without your intervention, I might have been tempted to go for the 31.

The difference in isolation is explained by the difference in package size.
So I'll stick with the 41, but I'll keep your idea of the dual layout so I can test the 4165 vs. the 4166.
 
  • Like
Reactions: altor
I will keep you updated on the comparison between a cable without an isolator, a cable with a 4166 isolator, and a cable with a 4165 isolator.

Hi Alren,

I'm mainly interested in your opinion about the difference between 65/66.

Alex.


P.S. The difference between "with isolator" and without " - I know". I actually made it specifically for the Cosmos ADC, since it has no isolation and when connecting it and the DUT to the PC at the same time, sometimes there are ground loops (mostly because JTAG/XTAG connection).
And this is very clearly visible with a spectrum analyzer.
(My own measuring ADC, Olivine-2 has, like all my DACs, an I2S isolation inside, Cosmos ADC not, but it allows to measure smaller distortions).
 
There is now a version of the Cosmos ADC with built in isolator. I suggested at one time but it was rejected as unnecessary - a year later or so it was there. Didn't help me so I had to get one from altor 🙂

//
 
  • Like
Reactions: altor
Hello Icsaszar,

First, I need to finish the prototypes and verify that they provide a real audible gain.

Then, I designed these PCBs using an old software program that doesn't generate Gerber files. It only generates a PDF image of each side of the PCB. A small local manufacturer will be manufacturing the PCBs for me.

However, I would be happy to provide all the elements I have (schematic, PCB images, BOM, etc.) so that a professional can generate these Gerber files and allow as many people as possible to make this USB isolator.
 
  • Like
Reactions: lcsaszar
Hello everyone,

I've completed and tested the first USB isolator, based on the ADuM4166 chip. In this circuit, the clock is on the output side.
Compared to a normal USB cable, the advantage of the USB isolator is audible.
It's not a revolution, but some details are clear enough to be heard.

First of all, the bass is cleaner and more punchy.
This also applies to struck strings, like those on a piano.
Vibrated strings, like those on a cello, are magnificent. You can feel the vibration, and the notes never stop fading.

I had set the filter on my R26 DAC to "Fast" for more detail. With the USB isolator, I returned it to "Middle," while still achieving more finesse.

@altor,
I made the second circuit with the ADuM4165 chip, that is, with the clock on the input side. It works, and I'll test it today.
I'll keep you posted on the results.
 
Compared to the schematic I posted a few messages above, I made two changes.
  • I didn't use the RM DC/DC converter, but the ROL type from the same manufacturer. It has the same pinout and footprint, but it has a power output of 0.5W instead of 0.25W for the RM.
  • I used the ECS-240-8-33B2 crystal instead of the CX3225. The ECS's stability is even better.
 
  • Like
Reactions: altor
I made the second circuit with the ADuM4165 chip, that is, with the clock on the input side. It works, and I'll test it today.
I'll keep you posted on the results.

I made the second circuit with the ADuM4165 chip, that is, with the clock on the input side. It works, and I'll test it today.
I'll keep you posted on the results.

Thanks, it will be interesting.
My PCB is also allowed to put either xx65 or xx66, but I did not try xx66 (I didn't buy them).

. The ECS's stability is even better.

In this case it doesn't matter.
For USB application almost any oscillator will work, at least in the room conditions.

Here is the new schematic

My AD chip circuit is the same, only I have a connector for external 5VDC on the secondary.
And a special circuit with an optocoupler, that completely turn off the DC-DC converter on the primary.

(And as I already said - I now use only TI chips. With them, the circuit is a little more complicated and expensive, but IMHO works better)

Alex.
 
@altor

I did some listening with both the ADuM4166 and 4165 circuits.

Both give a better result than a single cable.

But I find the 4165 a little less precise, a little less detailed, a little less punchy on the notes struck, a little too "round".

I prefer listening to the USB isolator based on the ADuM4166 chip, with the clock on the output side.

I would have liked to use the TI chips, but the chips are too small and I couldn't solder them by hand. A 0.95 mm spacing is the best I can solder.

If a soldered kit exists, I'd be happy to try it.