Graphs and Corruption
A set of visuals that show the behavior of the XLR inputs on the X32
To summarize a boatload of information, few things are effective as a well-designed graph. However, there are times when a mathematically accurate graph might not convey the essentials of what needs to be understood. This is one of those times.
As you know, the harmonics of the 992 Hz (1H) sine wave signal are going to be discrete integer multiples of the fundamental. Therefore 2H, 3H, 4H, etc., are valid X-axis values as harmonics. Anything in-between the integer multiples are inharmonic. Someday when I start evaluating compressor and limiter plug-ins, we’ll see some inharmonic values. Right now, sticking to the hardware realm, any non-integer multiples of the fundamental are spikes from the noise floor.
I chose a line graph for displaying the harmonic profiles at various input levels because it was easier for my eye to follow the relative level of a single color across the X-axis range. The line is continuous between the 2H, 3H, 4H, values, but there really isn’t any signal there. When I graphed it as a bar graph, I found it exhausting to try to pay attention to which color represented which level, and evaluate each cluster at an H-value, one at a time.
The line graph shows how the harmonic series gets smaller and smaller as the input level descends from -1.8dBFS (effectively -0.04dBFS for a sine wave) down to the optimal -18dBFS and onward to the rather nice -24dBFS. While building the graphs , I made some strategic decisions about the data values I entered into Excel. When the distortion values were -120dB or even quieter, the value recorded for that field is -120dB. First of all, it makes for a neater graph. But the understanding should be that all values shown as -120dB are effectively silent, and we need not expand the vertical axis to take readings in the ‘grass’ of all the grunge down there is the silent deep of the converters’ Johnson (and power supply) noise. In addition to the line graph, individual screenshots from the playback bus were graphed using the free VST plug-in called Voxengo SPAN. Feel free to view them in the Gallery. (Man, I’m going to have to organize that Gallery section.)
What’s called the ‘system gain’ or ‘process gain’ of 1.76dB that’s common A/D systems handling sine waves jumped out of the bushes and bit me! At -1.8 dB all was well with a nice disciplined lineup of harmonics. A mere 0.1 dB higher produced a world of harmonic hash that’s impossible for my little neat harmonic profile graphs to capture. Look at the [Figure-2] -1.7dB picture from SPAN to see what I mean.
Since we’re talking about A/D converters, it’s important to note that the 32 XLR channels in the X32 use the CS5368 converters, while the Aux channels with the TRS inputs use the CS42438. The CS42438 has 6 A/D converters, and its specifications are not as impressive as the CS5368 used for the XLR connections of the X32. When I initially installed my X32, I had gone through some pains to connect my external preamps to the TRS inputs because I was following the Internet-based crowd knowledge about ‘bypassing the preamps’ to avoid their heinous colorations. If you read enough forum posts, you’ll hear that refrain about how important it is to avoid running ‘a preamp through a preamp’ because bad things must naturally ensue. The set of measurements that I’ve done in this series shows that there are no bad things.
The Voxengo SPAN has enormous flexibility. I opted to use the same settings that are part of a very long thread on the Gearslutz.com forum (it’s a much nicer place than the name would suggest) because that would make future comparisons much easier for everyone. The thread is called. “Lets do It: The Ultimate Plug-in Analysis Thread”, and it’s part of the Music Computers subset in that website. That particular thread was started in August 2009, and at the time of this writing, has over 2600 posts. As part of full disclosure, I am MediaGary on the Gearslutz forum. Right now, I have over 300 posts in the website, but I have not participated in that specific thread. Now, back to business.
I have also shown no interest in recording past the 11th harmonic even when the SPAN tool shows there to be readable values. In general, the contributions to the tone of the music border on zero because the signals up there are so weak. The series of pictures for values starting at -3.0dBFS and descending through -6 dBFS, -12 dBFS, and -18 dBFS show a heartening trend of ever-lower distortion products falling well into the “don’t care “ range of values. Boring (but good) graphs begin at -18dBFS and continue downward through -24, -30, -36, -42, -48, and -54. This is the behavior of a well-behaved 24-bit converter providing 19 to 20 bits of signal, just as all the math and signal theory predict that it should. These measurements support a resounding “yes” to the question of whether the X32 is sufficiently transparent to be used as a line level digital mixer for external preamps.
Now, for a dip in the pool of outrageous solutions, I have a comment about the practicalities and usefulness of 32-bit converters. The X32 preamps have a delightfully high 11K-ohm input impedance. When it’s time to use my ribbon microphones on the drum set, I’ll have the clean combination of low gain requirements (historically about +11dB for the snare ribbon) and at the top frequencies, the ribbon mic will easily dance with the high input impedance of the preamp and the wide 19-bit effective dynamic range of the converters. Some of the drums will have the benefit of the tube colorations from the ART TubeOpto 8; its input impedance is 6.4k-ohms.
We know that 32-bit converters from a few chip makers exist right now. By my calculations, an 11K-ohm input along with its associated Johnson noise floor, with a 32-bit dynamic range (192 dB) gives us peak requirement approaching +81 dBu. That’s about 8700 volts RMS, or about 25000 volts peak-to-peak. Therefore, the voltage rails from the power supply have to be in the neighborhood of +13KV and -13KV. I wouldn’t hold my breath waiting for the bus-powered USB version of a real 32-bit capable interface.
As we will see in the following articles that profile the behavior of the (ever-changing) preamp pool at TedLand Studio, we’ll see that the noise floor and distortion products (both desirable and undesirable) are far stronger than anything that we’ll see contributed from a properly gain staged X32. Again, the upshot is: run your X32 with 0 gain with a target level of -18dBFS into your DAW and enjoy the flavors that your external devices are able to deliver. There’s nothing in the X32’s signal path to get in the way of what the preamps and DI boxes are able to give. But remember, NEVER go above -1.8dBFS with a sine wave.
- Ted Gary of TedLand
May 31, 2016