RME UFX III Audio Interface review

[Tovarich007]

One of the most audible aspect of amplifying small signals - like the output signal from a low sensitivity microphone when the source is far or not very loud - is noise.
In usual HiFi DACs and ADCs, we measure noise at a fixed "gain": That's what the Dynamic range value gives you.

But with a microphone preamp, this gain is not fixed.
The very goal of a microphone preamp is to amplify the signal to bring it above the ADC's noise level by a margin.
EIN (Equivalent Input Noise) helps us to understand how big this margin can be, comparing directly the noise as a voltage (usually expressed in dBu), so independently from the gain factor.

The noise is coming from 2 sources:
The source's noise : In the EIN, usually. a resistor of 150 Ohm is used as a "source".
This value is similar to a dynamic microphone impedance.
Of course, we don't want a microphone signal to mess the measurement, so we use a fix resistor instead.
The resistor itself is subject to Thermal noise (see here), so it contributes to the EIN noise.
At 20°C, a 150 Ohm resistor will have a thermal noise for a 20hz-20kHz bandwidth of 220 nVrms, or -130,92 dBu unweighted.

Then the preamplifier adds its own noise, which is what we want to evaluate.

As you may see in my measurements, the EIN at maximum gain under the same circumstances is at -128.2dBu.
If we deduct the resistor thermal noise, that's approx. 206 nVrms of noise added by the preamp itself (I compute this value and show it in the bottom right on each EIN measurement).
So less than the resistor's own noise.

As the gain decreases, this value will increase, since you'll be more and more limited by the ADC's noise.

In an ideal world, an ADC with an infinitely low noise level would not need a preamp.

Above a certain gain, the benefit of increasing the gain becomes too small, since we reach the preamp electronic's limits.

Thanks a lot for your very precise and complete answer. I've learned something pretty interesting today !

 

[Rja4000]

Excellent work!

Thanks !

Can you show a system loopback of the RME ADI-2/4 Pro SE on the AP's bench mode screen at the measurement levels you picked?

The APx trial license is expired now.

As it stands, we have no idea how much is contributed from the A/D-D/A (instrument) and how much is contributed by the A/D-D/A (DUT).

Well, I always check that.

I'm using various tricks when the raw performance of the ADI-2/4 is not sufficient or borderline.

I plan to write about what I learned in a separate thread.

 

[Blumlein 88]

I love the way you did a complete suite of meaningful measurements of the ADC side which is the important one for doing recordings. I'd like to see Amir incorporate at least the MIC gain measures referenced to some level and the EIN testing. @amirm

Excellent job here.

 

[sarieri]

If I understand your measurement setup correctly, the high spikes of H2/3 is a combination result of adi2’s ADC and UFX3’s DAC? UFX3 alone is probably having better SINAD?

 

[johny_2000]

The APx trial license is expired now.

Oh, that explains why I see so many AP chart screens in your review. You had a trial license from them.
What about the k$$$ device itself? Or does the software work with other input/output signal devices?

 

[Rja4000]

If I understand your measurement setup correctly, the high spikes of H2/3 is a combination result of adi2’s ADC and UFX3’s DAC? UFX3 alone is probably having better SINAD?

Of course, any measurement is always measuring the combination of the measurement tool and the device under test.

I think those measurements are representative of the UFX III's performances.

By the way, they match RME's specs pretty well (and they measure with an APx 555B, I've been told).

For measurements, I use the ADI-2/4 in mono, summing both outputs and both inputs, for lower noise.

If you speak about the ADC part, I have pretty accurate measurements of the RME ADI-2/4 DAC, since I'm able to measure it with a notch filter, so limiting the measurement ADC's impact to minimum.


Here is the RME ADI-2/4 Pro SE DAC SINAD vs level:

EDITED
Graph shown here before was the wrong graph (ADI-2 Pro fs R instead of ADI-2/4 Pro SE)

The graph above was measured with the E1DA APU Notch (which is limited to 10 Vrms, so I couldn't measure 24dBu range this way)

As you may read on the graph, we get over 120dB SINAD at 4V and 120dB at 6V.



If you speak about the DAC measurements part, I checked the values (except at 24dBu range) with the same notch, and that gave no difference.
I therefore used the ADI-2/4 alone from there on.

The RME ADI-2/4 Pro SE ADC in mono reaches up to 122dB SINAD.
And 120dB at 4V.

@restorer-john I think that answers your question ?


The most tricky part in this case was to get accurate low level signal.
As an example, measuring dynamic range at high gain with AES-17 method (with a signal at -60dBFS) was almost impossible.
At high gain, the EIN method gives more accurate values. And is independent of any external device.

 

[Rja4000]

Oh, that explains why I see so many AP chart screens in your review. You had a trial license from them.
What about the k$$$ device itself? Or does the software work with other input/output signal devices?

I used an APx 500 Flex trial license.
I don't own the hardware.

 

[MC_RME]

@Rja4000: I find it interesting that - looking at the DA impulse response - RME seems to have implemented a IIR "SD Sharp" filter rather than the usual "technically correct" FIR filter.

All our audio interfaces default to Short Delay Sharp filters, for lowest latency with best frequency response. I would say it's a natural choice for the application...

 

[Rja4000]

If I understand your measurement setup correctly, the high spikes of H2/3 is a combination result of adi2’s ADC and UFX3’s DAC? UFX3 alone is probably having better SINAD?

I've updated the answer above.

 

[pseudoid]

I've updated the answer above.

I think you had originally stated that just the testing phase had taken you four months.
I am inclined to think that you did not realize - at the time of posting your results - that this was to become a full-time involvement.

In case my thinking approximates your current state-of-affairs; I would like to remind you that Summer Solstice starts in just two days...

 

[johny_2000]

I used an APx 500 Flex trial license.
I don't own the hardware.

So this software works with any audio I/O devices under Windows?
I remember that some of the APx500 software settings looked like they used proprietary interfaces (high performance sine wave generator, variable frequency notch filter, bandpass filters, etc.). I don't know how these features will work on other devices.

 

[Rja4000]

So this software works with any audio I/O devices under Windows?

It runs with any device with DAC and ADC and having an ASIO driver.

Well, technically, it could use ASIO4ALL or Fkexasio to extend to other devices, but, personally, I hate those.

I remember that some of the APx500 software settings looked like they used proprietary interfaces (high performance sine wave generator, variable frequency notch filter, bandpass filters, etc.). I don't know how these features will work on other devices.

An Audioprecision hardware analyzer has some advanced hardware with serious benefits.
It includes a variable frequency low distortion generator, a variable frequency (!) analog notch filter and several preamp levels. Also, voltage may be calibrated, So it acts as a real measurement instrument.

With a "normal" interface like the one I'm using, you have to patiently setup for best performance for each and every measurement.

First, the RME ADI-2/4 Pro SE is probably the best interface available today for the task.

The RME ADI-2/4 Pro SE has excellent performances, and may automatically switch ranges, with 6dB between each range.

I have an excellent E1DA APU device, which includes a notch filter at 1kHz or 10kHz.
It is really helpful to measure DAC SINAD above 112dB or so.

For low noise signals, I use a mic preamp (the RME 12Mic-D or, now, the equally performant UFX III) to raise and measure the device under test's noise.

I calibrate the ADC or DAC voltage before each measurement with a good and calibrated DMM.

So all this allows to get measures comparable with (and sometimes better than) the AP hardware, but, as you understand, it is quite time consuming. (And really requires you to double check each step.)


And that's the deal, really: An AP hardware allows all this (and more) by just pressing a button.
Of course, for paid professionals, the hardware bill will be paid back in a few months.


One last word about the bandpass filter:
If you except the hardware, the reason it's hard to reproduce AP measurements with a PC software like REW or similar is that AP uses analog bandpass and notch filters.
REW or similar softwares usually just drop FFT bins outside of the range of interest, acting as a brickwall filter.
(This difference is especially visible for the IMD distortion.)

In the APx 500 flex software, they reproduced the frequency behaviour of those analog filters in software.
The goal is of course to produce similar results with the software and the hardware.

Understanding this, I'm now able to emulate the APx filters almost exactly with Virtins MultiInstrument, and get very close results.

But, to be complete, the APx software is checking multiple measurements for each step, until they are close enough.
This makes sweeps both more accurate and faster than what I'm currently able to do with MI.

 

[iconoclast]

Thanks for the super detailed review! One part of me would like to know how much better this is than my original UFX (the one previous to UFX II), but the other part of me definitely does not want to know.

 

[Rja4000]

Thanks for the super detailed review!

Welcome

One part of me would like to know how much better this is than my original UFX (the one previous to UFX II), but the other part of me definitely does not want to know.

Well, if the part that wants to know wins someday, you may still compare specs.
As RME specs are pretty accurate (never seen any major difference with my measurements so far), that would give you a pretty accurate picture.

 

[ZolaIII]

Not really impressed for the price. Which can get you similar capable interface and 2x 240W stereo amps with full DSP support (PEQ +FIR) and all quality of course (MOTU + Dynacord).

 

[morillon]

RME UFX III - Other measurements - Conclusion

Round Trip Latency measurement

RTL measured in ms, as a function of sampling frequency and buffer size

View attachment 375707

Totalmix

View attachment 375712

Every RME interface comes with TotalMix
This is a virtual mixer running in hardware within the interface hardware itself.
It has, of course, a control software that may run on a PC or on an iPad.

I won't go into the details.
This is far beyond the scope of this review.

But I just wanted to measure 2 things about it:

Internal headroom
I've done a simple test:
How many tracks with a 1kHz sine at 0dBFS may I sum before saturation ?
The answer is 16.
That's 20 * LOG10(16) = 24 dB headroom
(I found after this that this is exactly the value specified in the manual.)

As a comparison, my venerable Yamaha DM1000 40 channels mixer allows 8 channels, or 18dB headroom.

There is little chance the internal headroom will limit you.


EQ quality
(coming soon)


Conclusion

I think it's pretty clear at that stage that this device is really excellent.

It may not be the best for all aspects.
But what I think is the most impressive is that it's pretty close to the best in ALL area.
I couldn't even find one weak point left.
(And, believe me, I tried hard)
And for such a complex device, that's is quite an achievement !

What is even more impressive, maybe, is that it seems the very same technology, in term of basic hardware (Mic inputs, Totalmix), has been included in the lower end RME interfaces, the UFX II and 802fs.
I didn't check them, but if that's true, that makes them very interesting devices.

With RME, the good thing is that you have some kind of fool-proof check when you do your own measurements.
If your results don't match RME specs, you'd better check them twice. Because, on their side, they check everything.
My measurements are mostly better than their specs. As usual.
I found one or two exceptions, but we speak 0.5dB or so, so they are mostly irrelevant.


So, in short, for me,
This is just excellent !



Closing words

This review was quite a journey.
I started 4 months ago !
One of the most difficult part was to try to summarize relevant infiormation while not flooding the reader with just too many details.
Let me know what you think.

Thanks to everyone who made this review possible !

well done!
these complete interfaces require so much effort and rigor...
enormous!

 

[Matias]

I used an APx 500 Flex trial license.
I don't own the hardware.

Do you plan on continuing reviewing? And buying the software? If so considere a funding campaign to help with the costs.

 

[morillon]

virtins well mastered (the case of rja) is very powerful...see also multitones of pkane....the essential thing is to know the hardware differences with the AP approach and to clearly distinguish things...

 

[AdrianusG]

THANK YOU SO MUCH FOR THIS AWESOME TROVE OF INFORMATION!

AND TO AMIRIM FOR MAKING IT ALL POSSIBLE!!

WE LIVE IN THE BEST TIMES EVER FOR AUDIO!!!

I recently spent $10k on an alpine 17" screen system for my truck and am looking forward to having all 1.6G of my music downloaded on a disc soon.

Turns out when you let a subwoofer do all the heavy lifting < 500Hz and just let the mains play above that, with zero distortion cause they're not being driven past their comfort zone.

Between that and my home 120dB Wall Of Sound I'm gonna go deaf!

HEY: Currently stuck in a groove on the Scorpions, EVERY TUNE THEY MADE ROCKS IN IT'S OWN WAY, and well recorded.

Highly recommended.

Assuming you mean the German Scorpions, you are absolutely right a very under appreciated band, not quite metal like Maiden and Metallica, but their heavier songs really rocked, also great singer and musicians.

 

[AdrianusG]

Performance that was impossible a few years ago, available for about US $3000

Great job evaluating this.

Or you can buy a nice "High End" USB cable for this price, ehm... let me think, choices choices