EMU 0404 Distortion Analyzer

N.B.: Recently, I have seen reports of people who have managed to get the EMU working on Windows 10 using a revamped driver, so you may want to do some research if you still have one.

I noticed an article on the DIY Audio Forum about how to do distortion analysis using REW software with a Focusrite Scarlett Solo USB interface. REW stands for Room EQ Wizard and it has a distortion module. The Solo is an inexpensive audio interface, consisting of A/D and D/A converters with a USB port for computer interfacing.  I did try using my Macbook Pro's internal converters in loopback mode, but results were mediocre. At the same time, I discovered a new EMU driver effort by programmer Wouter1 on GitHub for macOS. I tried the driver with the EMU connected to my Macbook and it worked! I installed REW and discovered I could get good distortion readings using the EMU and the new driver using my Macbook Pro.

The diagram below shows the setup I settled on after trying several variations. I discovered that using the balanced input on the EMU gave the lowest distortion readings with the EMU in loopback mode; that is, eliminating the DEVICE UNDER TEST and connecting the LEFT LINE OUT using a balanced connection directly to the RIGHT input using an XLR connector. This gave me a baseline for the inherent noise and distortion of the EMU. BTW, the reason for using the RIGHT input is that the EMU was known to have a noisier LEFT input due to the internal PCB wiring. Initially, I was using the original EMU power supply (a 5 Volt "wall wart") to test, and I noticed a 120 Hz bump in the FFT (Fast Fourier Transform) display. I eliminated this by replacing the wall wart with a 5 Volt Li-ion battery supply, allowing the EMU to completely float. 

The chart below shows the FFT of the EMU in loopback mode and powered by battery. The internal REW signal generator is used to output a 1kHz sine wave. The level has to be adjusted to around -15 dB to avoid overloading the EMU inputs and causing added distortion. The inputs are very sensitive. The THD (Total Harmonic Distortion) is only 0.00017%, which puts it in the same class as $1000+ analyzers. Unfortunately, the noise figure is a fairly significant 0.002%, and I was not able to reduce this further. However, the overall performance is very good and more than sufficient for a hobbyist test bench.

Next steps: assemble a very low distortion signal generator to eliminate the influence of the EMU D/A converter. I will probably use something similar to this design by Jim WIlliams or this design by Vojtěch Janásek. They are similar in that they use Wein bridge designs with low distortion op-amps and CDS photocoupler stabilization. I'll probably look at the recently introduced OPA1656 opamp which has really good performance.

*****UPDATE***** 

See this project for a low distortion oscillator I have built - August 2021.

The input circuitry of the EMU is interesting. It uses the clever differential bootstrap technique that Bill Whitlock (of Jensen Transformer fame) patented and licensed to THATCORP, They created an IC version of his principle. It results in very high common-mode input impedance and better CMRR. Also, you can see why the XLR inputs are better - the 5532 input buffers are bypassed when you don't plug in phone plugs. You can see the reverse-facing bootstrap opamp in the differential section.

I have forgotten where I found this schematic - I think someone reverse-engineered while doing mods to the EMU.