From the moment the Stage I ES9038 upgrades were finished, immediately, I realized I forgot to do something. That being, these performance tests (prior to upgrading the stock DAC to Stage I). When it came time to do the Stage II upgrades, I didn’t forget that again!
I did make one mistake, with regard to the testing below. However, this mistake was extremely valuable for the SS2590 review. I’m positive this would have happened there instead.
The DSP Problem:
If you use foobar2000 to produce test signals (I use DSD or High-Res FLAC files), you need to turn off all DSP effects, but I forgot this. To try and make it a somewhat fair comparison, use the same EQ during the entire thing. I did this, except in the Pink Noise test below. I was curious just how much the EQ on Stage 1, could push Pink Noise away from the expected trend (Pink Noise amplitude = 1/frequency)
Stage 2 | HL’s Spin on ES9038
I also had frequency response tables ready to go, but the EQ skewed the hell out of these graphs. They aren’t even worth sharing, except maybe the Stage II graph with the EQ off. I’ll post that soon, as that test will be used throughout the SS2590 review (on all op-amps).
Note:
All images shown below are using the frequency domain (FFT), instead of the time domain. The FFT allows us to see waveforms that are much more useful and they give us a clear idea of performance, at a glance.
Instead of seeing seconds across the X-axis, you will see raw frequencies (span, scale, center and offset are all adjustable on FFT), with amplitudes (on the Y-axis) in either dBmW or dBVRMS.
Tech note: the horizontal scale on your input channel has a drastic effect on how the image appears. I was using at least 100mS/div and slow acquisition mode with rectangular windowing.
Above you can see the Pink Noise test. Pink Noise produces a full spectrum of frequencies, with amplitudes that decrease inversely proportional to the input frequency (as you go up in hertz, amplitude goes down). In other words, the curve/waveform should decrease exponentially.
The Stage I image looks more like White Noise (equal amplitude across the spectrum) and this is due to the EQ, which was still enabled.
Looking at Stage II, we see a far more ideal output, but I did disable the EQ for this one test. The rest of the tests use the same EQ settings throughout.
Comparing Harmonic Suppression with FFT
These images are the easiest to look at and see clear changes in DAC performance. One of the biggest performance features found in modern DACs is their ability to intelligently suppress harmonic distortion.
Important Note on Scale: When I was sharing some of my first tests with Andrew Sparks of Sparkos Labs, he pointed out my scale was odd. He is right, since I was using dBmW instead of dBVRMS. I couldn’t change them all to dBVrms, or the comparison would have be useless. Take note of the unit of load, in each of the graphs. You have to multiply my dBVRMS values by the square root of 2, to get the more common dBVp-p .
In general, harmonics are multiples of the input signal; if you have a 1kHz signal, then you’d expect to see harmonics at 2kHz, 3kHz, 4kHz, etc. The amount by which the DAC can suppress harmonics, directly translates to THD performance. This is because THD is calculated based on the amplitude of all the harmonics versus the amplitude of the fundamental input frequency. While I didn’t directly do THD measurements, you can see there are clear improvements here!
Comparing FIR Filters:
I selected 20kHz/div and 50kHz center, to show you what happens at the IIR point.
FIR filters are something that many people use, but have no idea what they are really changing. Therefore, I decided to do a quick run-down (visually) on the oscilloscope, showing how it changes the output of the FFT.
Essentially, the DAC uses different FIR algorithms to shape the frequencies up to the IIR frequency (IIR = 50kHz for these tests). Beyond the frequency set for the IIR frequency, IIR filtering is used.
Wrap-Up:
This was good practice for the upcoming SS2590 discrete op-amp showdown! I definitely won’t make the mistake of having different DSP settings again. If it was possible to fix this, I would have. However, it would literally require me to put the old parts back into the DAC, which would be insane to do. I apologize about not having a better comparison, where all foobar’s DSP settings were off. That is what we do at HL, learn from our mistakes!
In the final stage of upgrades (Stage III), I will be looking at replacing the stock diodes with MUR equivalents, assuming this is ideal. First, I need to do some research on the topic. I guess I’m also going to replace the 4 main AC filtering capacitors, we’ll see. The final things are the film capacitors, but I am not sure I am going to bother. I need to do some research on performance numbers between the models I have on hand versus the WIMA 5 series installed. I do have tons of WIMA 10 series, so it might be worth it.
Note: I am purchasing very few of these parts, most of them I already had on hand.
References:
https://community.sw.siemens.com/s/article/introduction-to-filters-fir-versus-iir#:~:text=These%20filter%20types%20can%20be,produce%20an%20arbitrary%20shaped%20filter.&text=Two%20classes%20of%20digital%20filters,Infinite%20Impulse%20Response%20(IIR).&text=This%20makes%20the%20IIR%20filter%20a%20’recursive’%20function.
https://www.mikroe.com/ebooks/digital-filter-design/introduction-iir-filter
https://en.wikipedia.org/wiki/Finite_impulse_response#:~:text=In%20signal%20processing%2C%20a%20finite,to%20zero%20in%20finite%20time.
https://en.wikipedia.org/wiki/Pink_noise