The Pain in the Gain
All preamps have gain to offer, but it comes at a price
On the cruise ship of life, preamps get to ride on the top deck. Their gain figures are directly stated, the input impedance is carefully noted so the microphone feeding it can have its best chance at glory, and the output connections are guaranteed to have a well-behaved load that doesn’t get wonky with reactance (changing impedance with frequency).
Power amplifiers get to ride in steerage. At the consumer level, the users will wander about, confused about what levels to send to the balanced inputs, and the outputs are going into all kinds of exotic reactive jungles, populated with multi-way passive crossovers that themselves are interacting with the magnetic circuits of the speaker drivers. At the professional level, in clubs and outdoor festivals, the power amps face a Mad Max experience, connected to weapons-grade subwoofer cabinets, only to be rewarded thereafter with a trip in a truck.
What they do have in common is gain. Preamp gain is notably wide ranging. Preamp specifications routinely offer 50dB, 60dB, 75dB and even 85dB of gain. Power amps generally have a smaller range, although it’s not nearly as directly stated. A good set of power amp specifications will point out the necessary input voltage for an amount of output power, but will leave it to the reader to convert that into a voltage gain. Let me save you some trouble with an example.
I have three Alesis RA500 amplifiers. Their job is to run the room speaker combination that has twin 15-inch subwoofers, a twin 30-inch line array of 5.25-inch wide range drivers, and a pair of 30-inch ribbon tweeters. The RA500 specifications say that each channel can deliver 250 watts into 4-ohms at 1-kHz with 0.02% THD+N, using a 1.23 volt (+4dBu) input. The spec sheet stops there. We have the basic information to determine the gain, but it’s only because Ohm’s Law is friendly and direct. Power = e^2/R, so the voltage that gets us 250 watts is the square root of 1000. Ergo, 31.6 volts RMS into a 4-ohm load gives 250 watts. That 1.23 volt input was multiplied by 25.7 to reach 31.6 volts, so the gain is 28.2 dB. For completeness’ sake, the RA300 offers 25.6 dB of gain, and the RA150 checks in at 23 dB gain.
I went back to the soulution-audio.com website to check on the world’s greatest DAC’s, preamps, and amplifiers for the consumer market. [See the Peregrinations about Preamps article for more info.] They’re very forthcoming with the gain specifications of the amps. The model 701 mono block amp offers 26 dB and 32 dB of gain depending on its mode. By the way, by my standards, a company that wants to sell me a $40,000 preamp should not have typos in the spec sheets…but that’s just me.
Notice the clustering of gain values around the 30dB mark. I have captured many readings of my preamps collection at 30dB of gain. The D/A output of the X32 had been set at a -48dBFS level, and 30dB of gain through the preamps was used to provide -18dBFS level into the A/D. For the preamps, 30dB gain is the equivalent of the evening gown competition. Their magnificent gain capabilities will be on display, while any potential flaws that can occur at higher gain values are completely hidden by the flowing fabrics and flattering patterns. Even if there is a carbuncle under that wispy ode to textile technology, we’ll never know at 30dB of gain. On the other hand, for the power amps, 30dB gain is sensible shoes and casual Friday … because if you’re a power amp, every day is casual Friday.
Take note that during the testing of the preamps in this series, there’s no attempt push them into clipping. I’ve read many complaints by people abhor the un-graceful way their mic preamps behave ‘way up in the penultimate 0.1dB of their gain capability. While guitar amplifiers live to be clipped, overloaded, seared, and otherwise abused, microphone preamps aren’t destined for that kind of work. Clipping is the equivalent of spinning off the road in a car, or stalling a climbing airplane outside of its operating envelope. Some car designs can take a crash better than others, and some airplanes recover more easily from awkward airflow, but in general it’s better to keep the shiny side up, and the horizon in the normal orientation. When the patient said, “Doctor, it hurts when I do this.”, the doctor replied, “Don’t do that!”
I took a short tour of some high-end professional preamp websites, and I was surprised at the initial results. At the BAE (baeaudio.com) website, none of the products had anything beyond rudimentary specifications, although BAE was justifiably proud of their EIN values which they quoted as an un-weighted value on a 300khz bandwidth (-110dBu). That was brave.
The folks at API (apiaudio.com) didn’t say much beyond what you’d find in a Hi-Fi magazine, but redeemed the situation by stating the actual metered noise (-95dBu rather than the EIN value which is a bigger, marketable, number) but practically every workmanlike preamp on the planet can get within 5dB of the theoretical maximum EIN of -131dBu. The people at Millennia (mil-media.com) really impressed me by quoting their preamp specifications at a specific gain value of +35dB. I consider that approach to be of terrific value because that’s where our gain knobs are likely to be during daily usage.
Here’s an example. If you have a typical condenser microphone with a -36dBV sensitivity at 94dB SPL, its 16-millivolt output would be raised to by a factor of 63 with +36dB of gain to almost exactly 1-volt, which in turn, will clock in right around -20dBFS depending on how closely you’ve calibrated your A/D inputs to be +4dBu equal to -18dBFS. A singer working a microphone at 8 inches won’t be pushing very hard to achieve 94dB SPL, so +30dB to +36dB is a place where I generally find my preamps to be set. During my testing, the knob calibration on my Golden Age Pre-73’s turned out to be nicely accurate, and the 30dB to 40dB range of gain seemed to be typical of a lot of projects. More or less is needed depend on the singer, the instrument, and how close or far the artistic considerations demand placement of the microphones. Obviously, microphones on guitar cabinets don’t need much gain, and at the very bottom of my gain requirements is the preamp that captures the ribbon microphone on the side of the snare drum. Only +11dB is needed there on the snare.
Here’s the central point of this multi-article exercise: The series of graphs in figures 16, and 20 through 30 show how the various preamps at my disposal are able to manage the -48dBFS signal presented to them, and bring it up to the standard operating level of -18dBFS. It’s notable that the noise floor of all of the preamps is about the same as they work at the 30dB gain level. It’s also notable that the floor for most of them is about -102 dBFS, or fully 18dBFS (up to 25dBFS depending on your line through the 'grass') higher than the native noise floor of the 0-dB pass-through offered by the Behringer X32’s preamp-A/D combo. The implications are huge.
The first implication is that all of my intent to buy the Midas M32 for its better noise specifications is a waste of money. It would be a different story if I intended to use the Midas preamps; they measure wonderfully, about 10dB better than the X32 preamps in many important ways. Elements like TIM (transient intermodulation distortion) and freedom from harmful harmonic artifacts in the M32 are measurably better than the X32 in independent testing published in Germany. Nevertheless, all of that stuff is academic when running the X32 preamps at zero gain. They simply don’t contribute anything to the sound character of the attached preamp, and any potential effects are so far down in the sound scale, that it’s almost laughable to attempt to elevate it to be a serious issue.
The next implication is that many enjoyable preamps simply look horrible on a spectrum analyzer. I just bought an Aphex 207 preamp for my guitar closet amp. The Aphex is in pristine condition. I ran it on the test suite at 30dB gain, and was appalled to see that it looked worse in every way than any and all of my preamps. It could be why the API, BAE and Millennia companies are reluctant (or simply smart not) to offer this test data as part of their marketing. It’s a bit like getting new leather shoes: They look marvelous and feel good, but you really don’t want to inspect the processing chemistry of leather. Somewhere along line, the smell of sodium sulfide for hair removal will stop you in your tracks. That stink can knock a buzzard off a fence at 30 paces.
Transparency isn’t all that it’s cracked up to be. There’s a long online discussion among engineers that want to most transparently archive the output of an open reel tape machine. The elephant in the room is that nobody is suggesting that another tape machine could be used as an archive device for the first tape machine. That’s because the beauty of what’s in/on the tape are the non-linear effects of the transformer hysteresis, tape saturation, bias frequency, head profile, tape asperity, and a bunch of other stuff that a follow-on tape machine would only ruin. Do it once, and make it sweet, but a second go ‘round isn’t a good formula. It wouldn’t look very nice on a spectrum analyzer either.
The third implication is that there’s pain in all hardware gain. All of the microphone preamps in the test have the same Johnson Noise problem, and nobody beats the basic physics. If the noise floor is -120 dBFS, and one adds +30 dB of marvelous, lovely, perfect gain, then the new noise floor is -90dBFS. That gain can come with or without harmonics, but the goal is to get it without additional noise. If that +30dB of gain came with another +14dB of noise beyond what was already there, then the preamp isn’t helping the recording as much as it should.
If you remember the dB addition arithmetic of uncorrelated sound sources, you’ll recall that even a full 64 tracks that started at -120dBFS will yield a noise floor of -101dBFS when added together. That’s not bad. Despite the high track count, -101dBFS is still significantly quieter than the noise contribution of the gain from one real world hardware preamp.
There is still hope that I can somehow justify at least buying the external DL16 stage snake made by Midas. It can complement the S16 that I already own. I am awaiting a set of measurements from a kind soul that agreed to run my testing protocol through his M32. If the noise floor measurements turn out to be more than 10dB quieter than the measurements I’m getting in my X32, I’ll ask Santa to get a DL16 for me. That will take some of the pain out of the gain.
- Ted Gary of TedLand
June 12, 2016