USB sound card review: ESI U24 XL

Audio performance measured

Abstract

I plan to do THD measurement of the audio equipment I design using REW and a sound card. I have the ESI U24 XL at home, which I bought years ago for a different purpose. In this article I present my findings evaluating the fitness of this sound card for doing THD measurements of audio equipment. I also found out something about the general audio performance metrics of the sound card.

Here is a short spoiler: The renowned audio measurement equipment manufacturer Audio Precision supports sound cards as analog measurement interface, but also explains that this is only suitable in case the device being measured has significant distortion, like loudspeakers for example. For measuring power amplifiers that have very low distortion, it may be difficult to find a suitable sound card. Nonetheless, I wanted to know how good or bad the sound card I have at home actually is. This lead to a lengthy and detailed investigation.

Interface

Measurement setup

For the measurement, the sound card was connected to my notebook using a standard USB cable.
The sound card was tested using REW 5.19 running on Debian buster. Sound card was set to 48kHz / 24Bit.
Each analog output was wired to the corresponding analog input using a short cable.
In between this connection, I would later place the DUT, that means my audio circuitry like amplifiers, processing the audio signal. I would then measure the return signal using the sound cards input.

Noise floor

Noise floor is roughly between -120dB and -125dB for most of the audio spectrum. No anomalies here except the peaks at 1kHz, 2kHz and 3kHz, which likely are artifacts of the USB interface. For normal audio use, -120dB noise floor is okay. For precise measurements, this is too high. I would prefer at least -140dB noise floor, which in theory should be possible with 24Bit. Note that the noise floor is unrealistic for real audio applications, but results from FFT gain mostly.

The noise flor does not only appear lower due to FFT gain, but also due to bandwidth limiting of the noise measurement. This is how noise looks like in reality recorded on a 100MHz oscilloscope:

Above image shows the full spectrum noise floor of the analog output. I believe this image was taken prior to improvement of the device. Units are 10mV per division and 10µs per division.

Above image shows the full spectrum noise floor of the analog output together with a sine wave. Units are 500mV per division and 200µs per division.

Note that it is perfectly normal that Delta-Sigma DACs like the one inside the ESI interface output a lot of high frequency noise. The high frequency noise however should be filtered after the DAC. ESI didn't implement any meaningful filter except an EMI filter right at the analog output to partially contain the high frequency noise that the U24 XL generates. Also note that the measurement of the sine wave was done after improvement of the U24 XL, which included adding a filter ahead of the analog output amplifier.

Also note that I haven't measured a different DACs output this way so it is difficult to say whether the level of high frequency noise at the analog output is excessive or not.

The high frequency noise is far outside the human hearing limits, but still may adversely affect the sound in a subtle way.

THD

THD is a standard measurement for audio equipment. In order to see whether this sound card could be suitable to measure any other equipment THD, I measured the level of THD of the sound card using its own analog output and input. So the THD shown here is a sum of both the output THD and the input THD.

THD 1000Hz unloaded

Above screenshots show total THD of the output and input at different levels, but unloaded (because the analog input does not present a significant load to the analog output). The uppermost plot shows THD at pretty high level. Each plot below has the level reduced by 10dB. At high level, THD is significant, reduced by 10dB, THD goes down considerably, but signal to noise ratio worsens and reducing the level by 10dB again does not significantly improve THD, but further worsens signal to noise ratio instead. Overall, the spectrum looks acceptable with 2nd harmonic above 3rd harmonic and all higher harmonics near the noise floor.

THD 1200Hz unloaded

Measuring THD at 1kHz is industry standard, but USB devices may have artifacts at this frequency. In order to distinguish those artifacts from the signal, I repeated the measurement at 1200Hz:

THD 30Hz unloaded

Here are the results at low frequency unloaded:

THD 1200Hz headphone load

This is the THD of the same setup, but with the output loaded by a 43Ω resistor. I had to increase the level in order to compensate the attenuation that is due to the high output resistance (100Ω) of the headphone output. The 43Ω load is somewhere in between my 55Ω headphone and a 32Ω headphone:

First surprise is that the USB artifact at 1kHz is gone and second surprise is that distortion is not as bad as I expected. While higher order harmonics (4th, 5th and 6th) are above the noise floor now, they are much lower in amplitude than the lower order ones, which is a good sign.

Device information

Specification from ESI Audio (Rev. 1.1):

USB Controller

  • Chipset: ES7018L
  • Sample rate: 32kHz, 44.1kHz, 48kHz
  • USB1.1 and USB audio class 1.0 supported
  • USB2.0 Full speed compliant
  • I2S CODEC Interface, 2in/2out, Built in IEC60958 professional S/PDIF TX & RX
  • How it identifies on the USB: 0a92:00d1 EGO SYStems, Inc.

Line Input

  • Type: Unbalanced 1/4"
  • Maximum input level: +4.7dBu
  • Frequency response: 20Hz - 20kHz, +/- 0.02dB
  • Input peak LED: +3.7dBu
  • Input Impedance: 10kOhm
  • ADC:
    • Type: 24Bit, 192kHz
    • Dynamic range: 102dB @ -60dBFS A-weighted
    • S/(N+D) ratio: -92dB (@ 0dBFS, measurement method)
    • THD+N dB (A) -90dB (1kHz @ 0dBFs)

Analog Output

  • Type: Unbalanced 1/4"
  • Maximum output level: +6.9 dBu
  • Frequency response: 20Hz - 20kHz, +/- 0.02dB
  • DAC:
    • Type: 24Bit, 192kHz
    • Dynamic range: 108dB @ -60dBFS A-weighted
    • S/(N+D) ratio: -97dB (@ 0dBFS, measurement method)
    • THD+N dB (A) -97dB (1kHz @ 0dBFs)
    • Interchannel isolation: 100dB

Digital Input and Output

  • Maximum 24Bit / 48kHz
  • Type: Coaxial & Optical
  • Format: IEC-958 Consumer (S/PDIF)

Linux compatibility

What I have tested successfully on Debian Buster:

  • Analog Stereo Duplex
  • Digital Stereo Duplex (IEC958) using the coax interface

What KDE System Settings offers, but I didn't test:

  • Analog Stereo Output + Digital Stereo (IEC958) Input
  • Digital Stereo (IEC958) Output + Analog Stereo Input

PCB Assembly

Analog input and output performance in detail

Here is some information about what I found out about the PCB assembly:

  • Mine has the PCB Rev 1.1
  • The analog audio input is built around two NJM4580 operational amplifiers. It seems, one op-amp is used for signal processing, the other op-amp for the clipping indicator.
  • The analog audio output is built around a single NJM4556AD operational amplifier.
  • The USB streaming controller is the TE7022L from Tenor.
  • DAC and ADC is WM8776 from Wolfson Microelectronics (now Cirrus Logic).

Analog Input

The analog input is reasonably well designed.

For a full review of the analog input circuit details, please refer to my article detailing the potential improvement of the ESI U24 XL USB sound card.

Analog Output

The analog output works as line output, but is seriously flawed as headphone output. Both frequency and response and distortion are unacceptable for a headphone output.

Frequency response
ESI U24 XL analog output frequency response loaded with 43R resistor
Being connected to a line input, the frequency response is okay. In case a headphone with a low impedance, maybe around 50Ω is connected, the frequency response is terrible with a substantial lack of bass. Expect roughly 1dB loss at 70Hz, -2dB at 45Hz, -3dB at 45Hz and more than -5dB at 20Hz. This is a completely unnecessary problem that stems from saving a few cents worth of components and therefore could have been avoided easily. Luckily it is also pretty easy to fix for anybody who is willing to invest less than 1€ worth of electronic components and is able to use a soldering iron.
Inter-modulation distortion

Above image shows the 60Hz and 7kHz IMD of the analog output unloaded (i.e. connected to the analog input).

Above image shows the 60Hz and 7kHz IMD of the analog output loaded using a 43Ω resistor to mimic a headphone load:

Noise

The analog output also is very noisy. I measured more than 20mV of noise here. The FFT spectrum plots may indicate a low noise level, but that is due to FFT gain and not true in reality. Noise in FFT appears much lower than it actually is.

Circuit details and improvement

For a full review of the analog output circuit details, please refer to my article detailing the potential improvement of the ESI U24 XL USB sound card.

Conclusion

Overall, the sound card appears to be a reasonable choice for normal audio applications. Selling for roughly 75€ nowadays, this budget friendly sound card offers many different interface options and this is the reason why I bought it back then. I bought mine for 100€ in year 2013.

From a build quality point of view, I cannot see any flaws. The chassis is nice and durable and the assembly of the electronics inside is state of the art manufacturing quality.

Using the sound card for doing measurements is not a good idea due to high distortion and poor signal to noise ratio of the sound card. This is why I'm still investigation whether the performance could be improved by modification and I'm also looking for a sound card with better technical specification for doing measurements. I had a look at the Focusrite Scarlett 2i2 gen3 as well for this purpose, but didn't find this suitable either.

I still like the sound card because it is small, has many interface options and works well with Linux. For listening music using headphones, I prefer the sound card over the one built into my notebook. I guess this is because the ESI sound card features a somewhat competent headphone driver, while my notebook output might be too weak for driving my big headphones. The U24 XL headphone output has a terrible frequency response into low impedance headphones seriously lacking low frequency output. Many years I thought that my headphones have weak bass performance, but found out now that the headphones are fine and the sound card is the problem. I fixed mine using components that cost a few cents and now the headphone output works fine.

The analog part could have been done much better. Some of the potential improvement would not cost a fortune, but this sound card is designed with budget in mind resulting in some questionable design decisions. I detail my findings on the analog circuitry and improvement of the ESI U24 XL USB sound card in another article.

I'm not sure however whether spending more money on another sound card would buy a better product. Somehow it seems that most of the audio products either fall into the budget segment or the high end segment with a extremely large gap in between. The Focusrite Scarlett for example is another budget sound card and although being at least 50% more expensive, didn't measure significantly better and I decided that it is not worth the extra money.

Revision history

This article is surprisingly popular compared to the other pages of my website. This is why I revised it mid of year 2021 adding my point of view on the analog circuitry, which cluttered the review too much and so I decided to move the analog circuitry review to an extra article about improvement of the ESI U24 XL USB sound card. The original article was written mid of year 2020 and was more positive about the sound card. Detailed investigation of the circuitry made me change my mind about the quality of the product.