Which one to buy: UpTone USB REGEN or Intona USB Isolator?

[intona]

I am glad about your positive feedback.

 

My name is Daniel and I am CEO of Intona Technology, Germany and also the head developer of the USB 2.0 Hi-Speed Isolator.

 

Indeed, we were not aware of the post by Phasure and didn't get any direct feedback by him so far. Our accounting says that he has serial no #1 of mid September and we made some changes to the firmware since then - maybe this or maybe your thoughts about power consumption of his DAC caused the issue.

 

There are some audio-specific things I discussed already with tubesound by mail. With his permission, I will quote this here again to the public.

 

> 1. Would the isolator improve the sound quality?

 

The intended purpose is complete isolation of both data and power lines. Resulting sound quality improvement is the effect of breaking all the noise coming from the host computer.

 

As this is a true industrial product, we focus primarly professional applications like automation systems, medical and research labs (even the computer manufacturer with the fruit has one).

But of course, as this product is designed for low noise (research labs need this to gain dynamic range vom nanovolts to kilovolts) and perfect signal integrity, we understand your demand.

We have also some pro-audio and pro-video (4k/high frame rate) customers and they successfully solved noise coupling issues completely.

 

> 2. Would the standard version of isolator meet my needs? In other words, the industrial version could be an over kill for me.

 

The 1kV (standard) version will be sufficient for your application. Even in a classical ground loop situation, there are no potential differences that would require higher isolation to get rid of the loop. Of course, the industrial version has even better connectors and all parts are certified for extended temperature and, finally, every unit has to pass an isolation test with 2,500 volts for some seconds.

 

> 3. Does the isolator deal with common mode noise? I currently have an *********** USB Filter which reduces the common-mode noise in USB cable. Would your isolator make *********** redundant? In other words, would the galvanic isolation filter out common-mode noise?

 

The *********** seems to be audiophile stuff. I don't want to say anything bad about this but nobody would ever use something like that in the industry to get better SNR or the like. Guess why.

Using galvanic isolation is more like wireless. There is absolutely no contact with any computer signal.

As there is 1. a complete new and absolutely silent ground and 2. all power regulators are linear, there is no common noise coming from the isolator. All common noise from the computer is being ripped by the isolation barrier.

While developing the isolator, we took special attention to the noise coming from its data and power lines. We managed to keep the noise as low as possible. See attached plot (http://intona.eu/pub/7054_fft_output.pdf) showing measured output noise. That is, what the power supply of your DAC will see. Your DAC has a PSRR of like > 90dB which you can subtract down from these values.

 

> 4. There's no off-the-shelf USB isolation chip which supports high speed. I suppose you did some FPGA application inside the isolator. Does the isolator do reclocking and DC-DC conversion?

 

Yes, it has two FPGAs built in and it does reclocking. It also has a proprietary DC-DC converter followed by a ADP125 ultra-low-noise linear voltage regulator.

Also be aware that power output of the isolator is limited to 300mA with no voltage drop and 500mA max with voltage drop to 4.5V. This may be a gereneral issue with USB-powered devices.

Further, some non-professional users reported that they need to add a hub before or after the isolator. On subsequent tests in our lab, our SDA 'scope showed us clearly that computers with non-certified USB ports have some really big issues that our isolator cannot solve directly. This happens rarely and mostly with very cheap computer hardware. And as said, can be solved using a good hub.

 

> 5. Then the quality of the crystal oscillator is important too. Are you using, say, Crystek oscillators?

 

It is a SiLabs MEMS oscillator, with measured Jitter around 2ps. This is, because 1. we have some customers that need mechanicly rugged hardware because they use our isolators at repeated >2g acceleration 24h/7d - crystals will fail here - and 2. power consumption of MEMS is 1/10 of crystals. As all voltage regulators are linear in the isolator, we had to pay attention to overall efficiency.

 

Ahem, this clocking thing on USB is an ambivalent issue that I recognize on many forums. All transceiver and hub chips have their own PLLs inside and they all rely on internal RC oscillators to derive the bit clock. We didn't see any change on eye diagrams of the USB bus whether the transceivers where clocked with +-500ps Jittter or +-5ps Jitter on our LeCroy SDA. Anyway, we took special attention to all design aspects and price economy was never a subject.

 

> 6. I chose your isolator over ***** in that I believe your design is superior and it has more intellectual value. ***** claims it can reduce 8KHz packet noise and does impedance matching. I'm wondering whether your isolator also addresses these two issues. BTW, do the two issues matter for USB audio?

 

Our design is completely impedance controlled, also for the tight rules of USB 2.0 which is 90 Ohms for diff/odd and 45 even. We are experts in high speed circuit design with many years of experience.

 

I am aware of this packet noise issue in the audiophiles world. I must say, i never heard about that in the professional world. Correct designed ADC/DAC does not complain or suffer about this. I guess this IS a big issue in the audiophiles world because many hobby designers fail in high speed circuit design, which is also true for analog (audio stage).

 

Okay, nevertheless, we measured this 8 kHz thing early in design stage and did optimize accordingly. See attached power supply plot (http://intona.eu/pub/7054_fft_downstream.pdf) for full downstream utilization using a RME Fireface UC, 8 channels at 192 kHz. There is simply nothing to see at 8 kHz (< 145dBu). So, this is what they wanted with ***** but failed.

Of course, this 8 kHz thing may appear again in a badly designed device after its USB transceiver.

 

Here is a cite from myself to a customer some weeks ago:

+-+-+

At Intona, our business nature is Professional Audio. Thanks to our fully equippped audio lab, we had the chance to take special attention keeping the noise of the power supply output as low as possible in the 20-20k range and we also optimized current flow using a high-speed current sensor probe that resulted in the probably industrial-wide lowest 8 kHz packet noise. We also did ensure the compatiblity to UAC 2.0 devices. Lastly, both XMOS-based and RME interfaces are tested and found 100% compatible to our high speed isolator product as it supports the full bandwidth of 480 MBits.

+-+-+

 

> 7. does your Isolator has a "break-in" period? I know capacitors need time to "loosen up".

 

Due to the industrial target for this product, we use very high quality and long life capacitors. In case of electrolytics, these are Panasonic FP Industrial and in case of multi-layer ceramics, these are high-tech parts from including Samsung Electro-Mechanics America dedicated for high speed signalling. We also always use thin-film resistors in our linear regulators. From industry-view, there will (hopefully!) no change in terms of break-in for years. But from audiophiles view, these will (hopefully!) smooth out for your ears.

 

Thanks again to tubesound for his very good questions.

 

-Daniel

[intona]

So, this is what they wanted with ***** but failed.

 

Apologize, this is not my native language. I do not want to discredit anybody. This is not my intention at all.

This is meant in sense of: they failed because of the technically not available ground isolation.

Of course - using any filter, hub or similar may or may not improve any issues. This must be found out by the user individually.

 

-Daniel

[intona]

In use, should the Intona device be closer to the source or the DAC?

 

If a DAC does not need the 5V line, should it be cut-off at the source, or does it not matter? (My DAC does not, and I block it at the server.)

 

Do not cut-off the 5V line. Most devices use this as VBUS detection and they will probably to not work without. The 5V coming out of our isolator is as clean as it would come from a linear wall wart.

 

In gerneral, look at the contacts in your USB cables' connectors. They become dark with time, especially when pulled off and on a lot. Gold plating is very thin and may scrape off. This leads to higher impedance and affects signal quality. USB is prone to errors here because of its nature alternating differential and single ended signals.

Maybe I will do some eye diagram measurements with different cable lengths and aging connectors when there is some time left. This may show some insight from a scientific point of view.

 

-Daniel

[intona]

This afternoon was an exercise in frustration in transmitting through the Intona Isolator to a Mutec MC-1.2 on OSX and Windows. The Mutec MC-1.2 is either a USB Powered or mains powered USB to AES3/S/PDIF/Toslink converter. It requires a driver for Windows, but no driver required for OSX. For these tests, only USB power was used and multiple reboots to ensure reliability, or until something broke, which in this case didn't take long.

 

Previously OSX worked, until the MacMini is rebooted, and the Mutec is gone from the system once the machine is ready to go. Not the outcome I had expected. The Intona flashes a steady 5Hz LED meaning, there's no traffic. AudioMidi detects nothing. Added a simple USB 2 hub on the OUTPUT of the Intona, and the Mutec is there again, but there's no sound from HQPlayer. Start again.

Tried a few different USB cables from the Generic to the exotic without rebooting and all worked with a USB stick, but a few worked with the Mutec, but no consistency as to which cable was rock steady. Proof of the pudding is a reboot, and no cables worked properly, even the USB hub trick didn't work either again, must have been lucky.

 

OK, move on to Windows, see if that works somehow previously didn't at all.

 

Windows is a little more sensitive to cabling than OSX. Tried Oyaide Continental and wouldn't work, but Nordost Blue Heaven worked OK as far as the Intona was concerned with rapid LED flashing. That joy was short lived though, cause there's no output to listen to music. Device Manager reported the ! and USBview also showed !. The mouse and keyboard are on PS2 ports, so the only other USB device was a hub on the Dell monitor with nothing plugged into the ports.

 

I'm finding the Intona Isolator not fit for purpose in that it does not pass UAC2, if it does, it's a one time only, whatever you do, don't reboot the computer.

 

Of course, the isolator can handle UAC2.

The reported issues are handshaking and/or EOP-glitch problems. We have seen some glitches on older XMOS firmware some time ago which confused the isolators high-speed-handshaking procedure. We will re-visit this and I am sure that this can (and will) be fixed.

 

I have the industrial version of the Intona. From PeterSt recommendations, when I ordered it, I asked not to have the led blinking when playing music. Daniel of Intona graciously agreed to do this modification, so the led stays steady (no blinking) once Windows is loaded.

 

I haven't encountered any problem since I got it last Wednesday, but the Phasure NOS1, while requiring the USB handshake (this was at least true at the time I was using a Sotm USB 3 PCIe card, that I now have replaced with a Silverstone PCIe card), does not need the USB vbus at all after. Maybe that explains why I don't have any problem of the sort...

 

IIRC, Daniel talks about the potential problem of not having sufficient amperage and/or voltage when requiring the USB Vbus ?

 

This Intona device is intriguing: for a good while there was no such USB isolator for the highest speeds.

 

If you are going the route of optimising by having no blinking LED, assuming there was a circuit causing its own digititis there, why not turn out the LED completely?

 

Maximum output current is 500mA with a voltage drop to 4.5V. At 300mA, there is no voltage drop. If your device needs more, you have to use a powered hub or a power-injection cable.

If unsure, you may ask Intona before buying and we will investigate if your device can be supplied by the isolator directly.

 

About the blinking: the often quoted customer complained about this because of two points. First, this was simply too nervous for his eyes to look at. Second, there was a reasonable objection of interference to the USB signal by the LED current. Don't confuse with his first test on serial number #1, which did not work for him because of a bug in the packet re-timer.

 

The LED was a hot topic of debate while development as we knew that our customers do not want that additional noise by switching indicators on and off. So, we did following on the isolated side:

 

- We routed the traces to the LED as a differential pair from the FPGA. No current (and return current) can cross between the LED and the USB power/data lines or any other signal. The FPGA is powered by its own linear regulator. Further, this is decoupled by several 10µ ceramic high speed 150µ low-ESR caps. The phase of the regulator's feedback loop was adjusted carefully using VNA to ensure stability while handling the remaining load transients.

 

- We checked current flow in FEM simulations and further optimized the layout. This is a electro magnetic simulation of the circuit board which is mandatory for high speed designs.

 

- We checked again using a high speed current probe for any unwanted current crossing. This way, we can even make return currents of the GND layer of a working unit visible.

 

- We selected a 2mA low current LED and also set the output slew rate of the FPGA to "QuietIO". The FPGA allows between three slopes: Fast, Slow and QuietIO. Lastly is fairly gentle and usually not suitable for data transmission.

 

After this, we can really say that there is no interference by the LED at all. Some customers measure micro-vibrations of ball bearings. They would have lots of trouble with an 14-Hz-interference (or any harmonics) if there was a LED-current-on-USB issue - but there is none. So whether the LED is blinking or not, expect lowest noise possible from the isolator's output.

 

So, there remains the looking at the nervous blinking. This is an intuitive indicator of what's the device doing in an industrial environment. Just for the comfort of the said customer, we swapped the blinking mode of "high-speed" with the mode of "failure", which is "always on".

 

Best regards,

Daniel