Triple Bypass Part-2
Measurements of the DL151, ASP800, and X32 signal paths
As mentioned earlier, the Midas DL151 has three 8-preamp boards that are built with the same general dimensions as the M32 and X32 preamp board. As far as I can tell, that’s coincidental with the fact that 8 side‑by‑side XLR’s occupy a width that, in turn, sets the board’s minimum dimension along that edge. From the (sorry for the blurry cellphone pix) photos of the interior of the DL151 and the X32, it’s obvious that these boards only share dimensions.
The I/O connections, capacitor count and positioning, and a few other things show that they’re substantially different. Behringer leveraged the respect for the Midas brand by clearly stating Midas-Designed on the X32 product line, but there wasn’t (and still isn’t fully) enough trust in Behringer’s reformation to prevent people from thinking and writing the worst (and wrong) interpretations of that plain statement. I’ve not done a component-level comparison of parts, beyond confirming the presence of the little square CS5368 converter on each brand of board.
As the pictures along the left side of this article show, there are valuable tidbits to be learned about a preamp by just looking at the noise floor FFT. For example, the ART TubeOpto 8 image shows 120Hz and 240Hz harmonics leaking out of the general power supply or tube power circuit into the conversion circuitry. It’s not horrible, in that the peaks of those leaks top out at about -110dBFS, but since none of the other products show this profile, it’s safe to say that the power supply could be better. If this profile was shown on a 120dBFS Y-Axis, the only things poking up from silence would be those two frequencies. Since I only have this one example of a TubeOpto, I’ll not make any more generalizations.
The only other channel under test that had some spikes above -120dBFS was the TRS input of the Behringer X32. There are two spikes, one at 6kHz and another at 12kHz, and the overall level of the ‘grass’ is 12dB or so ‘taller’ than the DL151, ASP800 and the X32’s XLR inputs. In contrast, the DL151, ASP800 and the X32 XLR had nothing to say above -120dBFS, and my graphs would have been quite (revealingly) boring indeed if I’d stuck to the 120dBFS Y-Axis scale.
Next, we must explore what happens in the presence of a signal. The point of checking the pre-existing noise floor is to determine whether our external preamps or other connected devices will be compromised in some way by spurious signals already within the X32 or DL151. As you can see, just as we trust an Audient ASP800 to get our sound into the digital domain, the DL151/M32 is equally worthy of that trust. And truth be told, the X32 at 0-gain is not far behind. If you look closely at the Voxengo SPAN display for the DL151 Noise Floor and X32 Noise Floor pictures, you’ll see SPAN’s reported RMS values as ‑114dBFS and ‑112dBFS respectively.
For test signals, I used two different signal levels, and three different target levels within the DAW. A ‑18dBFS test sine wave at 992Hz was generated in Adobe Audition under Yosemite. This was sent from the Bus1-XLR1 output of the X32 mixer and routed at 0-gain through three destinations: the DL151, the ASP800’s line input with pad-on, and into the TRS input of the X32. This provided a ‑18dBFS (peak) signal back into the DAW. Using my eyeballs, it looks like about ‑22dBFS RMS. I know that’s on the low side as a target RMS figure, but I wanted to be consistent with the signal levels I used in the first set of articles.
A second set of measurements used the same sine wave at ‑48dBFS, with two different gain levels applied. One was 30dB of gain to be consistent with the first set of articles and show the expected ‑18dBFS level. The other was 42dB of gain to get a ‑6dBFS peak level in the DAW.
As you can see in the 30dB gain and 42dB gain figures, the noise floor that we so ardently pursued with high-dollar electronics is obliterated by the Johnson-Nyquist Noise that accompanies all preamp gain. Although we strive to overcome the physics of materials at human-survivable temperatures, there’s no getting around the molecular noise chorus when amplifying small signals. It’s also a fact that we can only improve the Johnson Noise problem by about 5dB by going someplace super cold, like the dark side of the moon at minus 153-degrees Celsius, minus 243-degrees Fahrenheit!
This depressing reality is immediately evident as you turn up the gain in real time, and watch Voxengo SPAN paint a rising green tide of noise ripples across the whole audio spectrum. In a great sense, what we could readily see in the absence of a signal is now enjoining us to listen more carefully because what happens between the notes is as important as what happens to each note.
As you go through the 30dB and 42dB gain graphs, the visual differences become less distinct in the snapshots because the grassy noise surface is a dancing spectrum of harmonics and Johnson Noise. The slope of the trailing edges into the relative silence of the noise floor cannot be seen with the test signal, and it’s in this graceful management of the musical edges that a preamp earns its keep.
Another way to look at this is akin to EPA mileage testing. You can tell quite a lot about what to expect within the discipline of the test rollers in the mileage lab. You can get a sense of the cruise range, and by checking the torque figures, a sense of the acceleration capability. That’s the equivalent of what we can learn with sine wave testing. We don’t learn the engine’s interaction with the accelerator pedal under hot, cold, uphill, and downshift conditions. Equally, the sine wave doesn’t tell us enough about level/frequency-dependent harmonics, and dynamic microphone interactions. These distinguishing characteristics are what makes an active marketplace for a wide selection of preamp models.
My personal strategy has been to get ‘variable color’ preamps like the ASP800 Retro Channel, and the Focusrite Liquid Saffire 56. What I may give up in ultimate capability, I gain in flexibility for matching the intended flavor to the instrument and/or singer.
It's quite a task to wire, calibrate, loop, snip, annotate, convert, upload, format and chronicle these experiments. I hope you can make good use of this data in discussions with your friends and enemies.
The other 9 test images are in the Part-3 page that follows.
- Ted Gary of TedLand
April 26, 2017