Mic Preamp Design & Headaches
I wrote this about 5 years ago, and just came across it a couple days back. I’ve got another mic preamp circuit I’m playing with, and figured I best go back through my old notes to avoid wheel re-invention.
Why is it that the only decent preamps were made back in the 70’s…
I’ve been fighting this for almost 2 years. Hundreds of hours of messing around trying to get the right mic preamp. When I started this project, I mistakenly thought, how hard can it be. After all, its just audio. And at this point, maybe thats the issue, its just audio!
I’ve designed thousands of boards over the last 30 years, first as a hobbyist, than as a U student, and then as an engineer. Some boards were super high gain, and super low frequency, and others were GHz boards. Audio should be a calkwalk, and its not.
The biggest issue turns out to be phase shift. Noise is solvable, as is gain, and THD. One can deal with dynamic range pretty easily if the other 3 factors are handled appropriately. Where I end up getting bitten is the transfer function of phase shift, THD, and dynamic range as a function of frequency. It seems none of the commercial mic amps want to spec this… and with two good reasons. First, for those who understand it, they would be scared off. Secondly, most people don’t understand the phenomena. The problem is that whether a person understands it, or just reads it in a spec sheet, it makes little difference…. the issue is folks can HEAR it.
More on this rant later, gotta go eat
Part of the deal is the cost of labor. In the 70’s common practice was to test select parts. This usually meant matching transistors and even transformers and in a few cases, not by test gear, but by sound during the final test stages. Today, most assembly houses would have a bird if you asked them to test select, and would no bid it if such required a listening test. The 70’s often required a craftsman at the test bench, and today with its penny pinching models, most of the test techs are not at that level. They want speed and throughput, and any drive and patience they once has been driven away by bean counters. The mic preamp of the 70’s is not a commodity like today.
In addition, one would like to think that non-discrete solutions such as the TI INA163 upgrade to the SSM2017 would be an even better solution that the TL074 of years past. In fact, the SSM2017 is still often designed in, even though the INA163 has better specs. There is a lot that goes on in chasing ones tail in audio. A proven pcb layout is hard to beat, and to rip up/redo a SSM2017 layout may well equate to multiple spins to get equivalent performance in the INA163, and even more spins to get the mfgrs speced performance out of it. Remember, we’re not just chasing audio performance, but in commercial gear, we also need to meet EMC standards, which often times make for really messy compromises between the audio and RF domains.
One other thing that comes to mind is servo control of the ground, and DC coupling the audio to minimize phase error. Walt Jung had a number of papers on this in the early 90’s. On the outset, this seems a better solution than transformer coupling and its high cost and space requirements. On the other hand, a properly engineered C-Core transformer is near impossible to out perform spec wise… so it all comes down to economics, and one can buy a whole ton of op amps and discretes rather than a $12 C-Core at 10K/yr. Perhaps the saddest part of this, is economics often dictates letting the phase transfer functions slide all over the place. For most apps, if the other parameters are likewise under control, if the phase varies, its not all that big of a deal… but then again this is the 2006 not the 1970’s.