As I’ve posted in the past, my audio measurement setup is built around an EMU 0404 USB sound card. Its a fairly old device, its driver is old too. On the other hand, you can get it for almost nothing on eBay, and it has excellent measurable performance for the price. It is good for 0.001% THD at 1KHz without any modifications. With some help, its front-end is good enough for even 0.0001% THD measurement, as I’ve showed in this post. However, as you increase the frequency, the distortion will grow, as you’d expect. Additionally, if you look inside the box, there are quite a few parts there that make you wonder “how good can it be if I put a few extra $ into it?”. That’s exactly what I wanted to find out. I didn’t want to spend much time, nor funds, as I was happy with the performance I was getting. This was mostly for fun, and the results are shown in this post.
After an initial look inside the 0404 USB unit, it seems that there are many things that could be improved if you don’t mind spending a few $, and more importantly, hours of work. There are no low noise supply regulators, routing of some signals seems strange, the “balanced” outputs arent exactly balanced in structure, etc. All of these will take quite a bit of time to modify properly. However, one thing that seems very easy and tempting to try is replacing the operational amplifiers (Op-Amps). There are quite a few of these, all of which are either 15532 or 2068, neither of which is too impressive. At the output there are 3 devices (and more on the headphones output path), while at the input there are 4 for each channel. I didn’t trace out the circuit, but it seems like at least one of these is only used for the high impedance input and will be bypassed for the lower impedance XLR input. I aim to check this when I get the time to make an adapter for a loop-back for TRS-XLR, as it might allow lower THD+N measurement limits.
I couldn’t resist, and had to check what the limits of the current devices are, and if they could be improved on the cheap. The limits I’ve seen so far are ~0.0008% THD at 1KHz with optimal signal level (approximately -10dBFS). However, as the test frequency is swept (using STEPS), it can be seen that the THD figure starts degrading at ~1.5KHz, and gets gradually worse with frequency, and approaches ~0.008% at 14KHz. This means that reliably measuring distortion of <0.01% at 10Khz or over isn’t really possible. Next I’ve looked for an appropriate replacement. One could consider the OPA1612 from TI, which is a wonderful Op-Amp. However, I was looking for a cheaper solution, as I was aiming at replacing 7 parts, 3 for the output, and 4 for the right channel input (left is unused in my setup, and I want it unaltered as a fail-safe). Luckily, there are some excellent devices that could be had for much less. I chose TI’s (National in the past) LM4562. This is a great Op-Amp, with wide BW, low distortion, and low noise, and it is very cheap. Getting 7 of these for the 0404 USB will only set you back 10$.
I went ahead and replaced these devices with the LM4562, cleaned the board with IPA, and put it all back together to check what affect it had. I should point out that I am using a loop-back test only, so it is difficult to isolate the contribution of the output vs. the input on the overall performance using this method. To try and isolate the contributions of the output/input stages, I’ve repeated this measurement for the Left channel too. The Left channel didn’t have its input stage modified, so while some differences between the two channels are expected, it will help get an estimate of the affect this mod has on each stage. The STEPS measurement results for THD vs. frequency are shown in figure 2 for the Right (fully modified) channel.
It is quite interesting to see these results, as at lower frequency they are unchanged from the results I’ve seen with the original parts in place. However, the most noticeable difference is what happen once you go over 1KHz. The 3rd harmonic now only starts rising at ~7KHz, which is 5X better than it was with the original parts, and therefore extends the measurable THD limit at higher frequencies by 5X at a total cost of 10$ and 15 minutes worth of work.
Figure 3 shows the same figure for the Left (only output modified) channel. It is interesting to see that while it is better than the unmodified measurement, its not quite as good as the fully modified channel. Therefore, it seems that the difference in performance is somewhat more significant for the input stage, yet both will have to be modified to enjoy the full extent of the performance upgrade.
To answer the question in the title of this post, is replacing the Op-Amps worth the effort? The answer for me is maybe. Its cheap, easy, quick, and has measurable improvement in performance. However, considering that the limits of the 0404 USB with its original parts are sufficient for most DIY needs, I’m not sure its a “must have”. If you need extra linearity at higher frequencies, its a viable solution. Otherwise, it probably not worth the hassle.
Added on Aug 2, 2019:
I have repeated the measurement with a TRS->XLR cable, to check if the lower impedance XLR input will show different distortion, and I see no significant change. The amplitude shows up as being ~1dB higher on this input, but THD is almost unchanged, with only a very minimal improvement at high frequencies, which can easily be neglected. This is true for both channels of my 0404, both modded and un-modded input.