7055-C noise measurements

[Daniel (Intona)]

I was asked several times if I can show facts about residual power supply noise of the 7055 USB isolator series.

Well, measurements during the development phase were promising and now that production of the 7055 becomes mature, we are can carry on doing some serious characterization.

Accurately measuring band-limited noise at ultra low levels can be a can of worms. The only devices known to me that are able to achieve this are the Audio Precision System Two series and its successor, the APx555. This gear is really a point of trust.

We happen to have a System Two in our lab and in the following I will show some impressive results.
I will just post screen dumps to make clear that I did not have cooked the numbers.

[Daniel (Intona)]

Some words about how things are measured and what they indicate.

I am going to measure residual AC ripple on the power supply lines at the USB connector. These are called +VBUS and GND, having a DC potential of 5 volts. The USB device may or may not use these lines to power itself partly or completely. It definitely has to use these lines to detect the presence of the USB host.

Measurements are always band-limited. The Amplitude RMS value is nothing less than the real AC voltage within, say 20 Hz to 80 kHz. You can read it the same way as you read a multimeter. So reading 1 mV RMS is definitely 1 millivolt in sum between 20 Hz to 80 kHz.
But beware of reading the FFT plot graphics the same way. These are partly windowed (thus scaled) and only good for comparisons at the same conditions (like FFT size, averages, Window, analog pre-filters, scaling). This is the reason why I always posted complete screen shots. So you can understand the setup.

The only way to plot the noise regardless of FFT parameters (thus making it truly comparable), is calculating the noise density. This shows dBV/Root Hz on the y-axis and does not scale with FFT size or anything other parameter.

[Daniel (Intona)]

To make a proper, shielded connection, a special USB-to-XLR tap probe has been build:

The data lines D+/D- are untouched, the VBUS is tapped and tied to XLR pin 2, GND to pin 3. The shielding is wrapped, soldered all around to copper tape and to pin 1 and the shell. Of course, there is some isolation below the copper tape.

[Daniel (Intona)]

This way, the power supply lines of any USB device can be tapped without breaking the shielding:

[Daniel (Intona)]

I have two USB host candidates available here which are known to me very well. The first one is my iMac (27" 5k 2019 model) that I use for common development tasks. The second one is one of our CAD-workstations running Windows 7. It is equipped with the highly professional Rugged Science UE-1004 4-port USB 3.0 PCIe card.

These are professional machines and I expect low ripple out of the box. I will later pick the worse one for further tests.

[Daniel (Intona)]

Here is the noise profile of the Rugged Science USB card of the Windows machine (plain output, no isolation).

3.5 mV is pretty good!

Please note the red rectangles. The "Amplitude" is the RMS AC value that we are interested in. The "BW" shows the measurement bandwidth for this value.

[Daniel (Intona)]

Btw: the higher the bandwidth, the higher the noise of course.

But that "> 500 kHz" on the System Two is inaccurate. I measured its "-3dB bandwidth" using the signal generator of a Picoscope that was lying around:

Actually 430 kHz.
So, the System Two offers selectable measurement bandwidths of 20 kHz, 80 kHz and 430 kHz. This can be mixed up easily, so always have a look on the "BW" setting when reading the amplitude.

[Daniel (Intona)]

Now let's have a look on to the noise profile of the iMac.
5.5 mV, not bad either.

The iMac has the higher noise, so I pick this machine for further measurement.

[Daniel (Intona)]

Now here is the noise profile of the Intona 7055-C connected to the iMac.
5.9 µV at 430 kHz bandwidth.

[Daniel (Intona)]

The bandwidth of the FFT plot is limited to about 80 kHz on this machine. So let's set the measurement bandwidth for the AC reading to 80 kHz, too.
1.9 µV.

[Daniel (Intona)]

At 80 kHz BW, the noise of the 7055-C is just 0.4 µV above the noise floor of the System Two measurement system:

(Inputs of measurement system shorted.)

[Daniel (Intona)]

At 22 kHz BW, the 7055-C has a residual idle noise of stunning 890 Nanovolts.

Which is just 200 nV above the noise floor of the measurement system.

[Daniel (Intona)]

Now let's connect a random, externally powered DAC and play an audio stream (UAC 2.0 isochronous, 48 kHz samplerate, 10 channels utilized with 1 kHz sine tone 0 dBFS).

iMac to DAC without isolator.
7.9 mV @ 430 kHz BW:

iMac to 7055-C to DAC.
40 µV @ 430 kHz BW:

40 µV including data! This is 20 µV less then the previous 7054 model just in idle mode.
Btw, you can convert dBu levels here: //intona.eu/db#voltage_dbu

[Daniel (Intona)]

If you are not familiar with logarithmic scales, I will show you this time the linear scale. That way you can see how dramatic the difference is.

This is quite hard to show because there are multiple magnitudes of difference.

Without isolator, at a scale of 150 µV (and pretty much overflowed):

With 7055-C, at the same linear scale:

[Daniel (Intona)]

Summary: I have shown that using the Intona 7055-C isolator, ripple and noise that come from the USB host is drastically reduced down to the microvolts range.

[Daniel (Intona)]

If you have any questions or want me to measure again some specific detail, just follow up on this thread.

[Daniel (Intona)]

As I told above, truely comparable FFT plots need to reflect the noise density in root Hz.
Luckily, there is a script available at Audio Precision's website that makes exactly that. There is also a very good explanation about interpreting noise in FFT data:
[External Link Removed for Guests]

Well, here you can see that the residual noise of the 7055-C is just in par to the noise of the measurement system.

(This is by design a linear frequency scale and power is dB vs 1 volt)

[Minimalist]

How does the noise of the 7055-B compare with the -C? I'm curious because the USB 2.0 is plenty fast for most audio data and the designer of the PS Audio DS DAC claims USB 3.0 ports are far noisier than 2.0 because of the higher clock speeds. (I own the 7054, which provides a marked audible improvement in my system.)

Also, do the 7055 reclock the signal?

Thanks.

[Daniel (Intona)]

How does the noise of the 7055-B compare with the -C?

The 7055-B is based on the same linear regulator as the 7055-C. Because the 7055-C needs to deliver more current for USB 3.0, it has two of those regulators in parallel. And doing things in parallel reduces noise. The measured noise is 2.5/3/6 µV at 22/80/430 kHz bandwidth. Measured under same conditions as above. A bit more than the 7055-C but still incredible low.
Btw, that high current cannot be delivered by the 7055-B due to lack of available transformers with 5.000 V safety isolation. But then there is the AUX port...

the designer of the PS Audio DS DAC claims USB 3.0 ports are far noisier than 2.0 because of the higher clock speeds

I would not say in general that noise of USB 3.0 ports is higher. The extra "super speed" lines on the 3.0 ports are disabled when only 2.0 is used.

Also, do the 7055 reclock the signal?

Yes. The 7055 series uses the same FPGA-based isolation technique than the 7054 as described here:
//intona.eu/en/support/answer/reclock

[Minimalist]

Thanks. So, everything being equal, when playing audio via USB 2.0 the 7055-C should be ever-so-slightly better/quieter than the -B?

[Daniel (Intona)]

Well, the difference in noise level of 0.0000016 V RMS on the USB VBUS cannot be propagated to the audio outputs of any decent DAC (which has its own filtering/isolation plus the usual PSRR).

[Minimalist]

Ah, interesting point. My DS DAC does not even use power from the VBUS, but mainly benefits from ground isolation. In this case, do you think there would be any improvement over the 7054 with the 7055? Thanks again.

[Minimalist]

Further, how do the jitter measurements compare between the 7054 and 7055-B?

[Daniel (Intona)]

Further, how do the jitter measurements compare between the 7054 and 7055-B?

Don't confuse USB data jitter with jitter of MCK inside your DAC. They have no relationship to each other.

About the USB data jitter, these measurements using our LeCroy SDA-6020 are still valid for 7055:
//intona.eu/en/support/answer/1245

High jitter on the USB data bus causes drop outs, glitches or data abort. It causes not any coloration in sound.

[oslo]

What happens measurement wise if you use the aux port to power the USB 3.0 SuperSpeed Isolator 5kV instead of using the USB supplied power? Would the choice of psu supplying power to the aux port mean anything for measurements? Do you recommend a certain psu for the aux port?

[Daniel (Intona)]

What happens measurement wise if you use the aux port to power the USB 3.0 SuperSpeed Isolator 5kV instead of using the USB supplied power? Would the choice of psu supplying power to the aux port mean anything for measurements?

Good question! The AUX port powers the USB device directly (not the isolator). The internal regulation will be disabled and VBUS noise is a matter of the external supply then.

Do you recommend a certain psu for the aux port?

Don't have measured any psu yet. Of course, linear regulated supplies are preferable for low noise.
I guess it is hard to find one that measures better than the built-in regulation. I always recommend using the AUX supply only when the host or isolator cannot deliver the power that the device demands.

[Minimalist]

Haven't heard it myself, but folks on the PS Audio forum are big fans of the Uptone UltraCap LPS-1.2 power supply for the Matrix X-spdif2, if that helps.

What happens measurement wise if you use the aux port to power the USB 3.0 SuperSpeed Isolator 5kV instead of using the USB supplied power? Would the choice of psu supplying power to the aux port mean anything for measurements? Do you recommend a certain psu for the aux port?

[Daniel (Intona)]

Nice, but I don't make any statements without having serious measurements on this. Just quoting values out of the datasheet of single components that are just a part of the whole system ain't my style.

[intona_fan33]

Fantastic thread. Great engineering

[cloudezman]

Now here is the noise profile of the Intona 7055-C connected to the iMac.
5.9 µV at 430 kHz bandwidth.

imac_7055c_dbu.PNG

Hello Daniel

It is mentioned in the 7055-c Datasheet (English language) that,
Noise output @BW 20/80/500kHz
7055B: 2.5 / 3 / 6μV
7055C: 0.9 / 1.6 / 8.5μV
7055D: 2.5 / 3 / 7μV
The performance of noise output of 7055B and 7055D are better than that of 7055C at 500kHz.
Is it recorded correctly?

[External Link Removed for Guests]

[Daniel (Intona)]

This is recoreded correctly.
While 7055-B and 7055-D have an identical linear regulation circuit, the 7055-C is tweaked to get lower noise at audio frequencies.
Everything has a price.

[cloudezman]

This is recoreded correctly.
While 7055-B and 7055-D have an identical linear regulation circuit, the 7055-C is tweaked to get lower noise at audio frequencies.
Everything has a price.

Thanks for the detailed explanation!

[Softsounds]

There is a new review with measurements here.
[External Link Removed for Guests]

It also talks about a version 2.03 firmware. What’s the difference between version 1 and is there any upgrade program?