Measurements (again)
Measurements (again)
Measurements and Charge Amplifiers
Friday, December 25, 2015
Back to building mics. I’ve been experimenting recently with capacitive negative feedback, commonly referred to as a charge amplifier. It was used primarily by Neumann in the 1960s when they introduced FET mics like the KM84, U87, U89, and so on. I built a T-47 from Microphone-Parts.com and a pair of mics using RK-12 capsules and transformer-less charge amplifiers (TL/C for short). So here are a bunch of measurements with comments.
First, distortion measured in the same mic, same pseudo-capsule with 3 different circuits. In this case, the RK-47 capsule is replaced by a 68pF NPO capacitor and a 10:1 (20dB) resistor pad to feed in the signal generator’s 1 KHz sine wave from the interface’s line output.
Microphone measurements continued...
The 3 circuits are a Schoeps type FET + 2 PNP outputs, a TL/C type with a K596 FET + PNP phase splitter + PNP outputs, and a TL/C with a J305 FET. Each is outputting 100mV RMS, or 0.28V p/p. at 1KHz. The main point of interest is the harmonics at 2 & 3KHz. This is a high enough level to cause a K596 without feedback to distort heavily or clip. In a TL/C circuit, second harmonic is down more than 40dB or under 1%.
In a charge amp the incoming charge is cancelled by an equal charge fed back through a cap from the output. Charge is voltage X capacity, so the smaller the feedback capacitor, the higher the output voltage must be to match a given input charge. In other words, the feedback cap sets the gain. Since the charges cancel, there is no net voltage at the input. The input stays at ground potential. This virtual ground affects a couple of things. It means that charge variation from the capsule travels as current rather than voltage as in most mics. When voltage changes, stray capacitance has to be charged along with the FET’s gate, and it’s important to keep such stray capacitance as small as possible. That’s why we choose very low input capacitance FETs and eliminate wiring between gate and capsule where possible. With the charge amp, voltage doesn’t change. Therefore, stray capacitance within and around the capsule doesn’t absorb any of the charge generated by sound. The only things which affect sensitivity are the polarizing voltage and delta in capacity of the capsule. Gain is not capsule capacitance divided by feedback capacitance as you and I might suppose by similarity to resistive feedback networks. If you have to locate a capsule away from the amplifier, try a charge amplifier topology.
The question came up whether working into a virtual short circuit adds electrical damping to the capsule’s mechanical / acoustical damping. Speakers and dynamic mics are certainly electrically damped. In many speakers the majority of damping is electrical, and damping factor is an important spec for power amps. That’s one reason I was measuring TL/C vs Schoeps, whose extremely high input impedance provides no damping at all. I expected to see peaks better controlled with a TL/C circuit.
TL/C J305
TL/C K596
Schoeps J305
Here we have an RK-99 electret S (green) and C (yellow) measured 3 cm from a 1.5 inch speaker. We see a large bass boost due to proximity. The C amp has about 6dB gain and the S amp has none. Ignore below 100 Hz. The little 1.5” speaker has no output down there. No evidence of additional damping, though. If there is electrical damping, it’s small compared to acoustic damping within the capsule.
Impressively close match, considering I had taken both capsules apart and “tuned” them.
Here are two RK-12 mics, serial #1868 is TL/C (green) and #1196 (yellow) is a Super Schoeps with J305 FET right at the capsule. The TL/C looks a bit flatter over all. Is it due to damping?
No, if we check the sheets that came with the capsules, we see the capsules are just different, and no, they weren’t sold as a matched pair. These were purchased a couple of years apart. Notice there are 2 curves for each capsule - front & back. Matched pairs from Microphone-Parts have been very closely matched in my experience.
There are a bunch more measurements of the latest pair of TL/C mics, and I’ll put them in the next installment.
Having converted one RK-99 to TL/C, I converted its mate with these results:
A word here about how I measure frequency response. When arranging the Mic Under Test and speaker, there is a tradeoff between up close where room effects are largely swamped by the direct speaker output, and farther away where proximity effects don’t mess up things. I’ve found a workable distance is around 25 cm, or 10 inches between speaker and mics. Sound absorbent foam or pillows behind the speaker, along the sides, on the desk top, and behind the mic help somewhat, but certainly don’t create a real anechoic environment.
I have a couple of Dayton EMM-6 calibrated measurement mics and a Behringer ECM 8000 which compared to the rest of my setup can be considered flat. I take one of the calibrated mics and put it alongside as close as possible to the front diaphragm of the M.U.T. Then using the dual FFT function in FEAT, I measure the MUT with the calibrated mic as reference. It works really well if the two mics have the same pattern. However most mics I measure are cardioid, and measurement mics are omnidirectional, so they pick up more room ambience and reverberation. Omni mics also have no proximity effect, and directional mics do, as seen with the RK-99s. Any sound can be used to excite the mics, but a constant white or pink noise source from FEAT’s signal generator usually works better than chirps or music.
As a check that comparison with a reference mic gives consistent results, here are measurements of an RK-12 TL/C mic, using the same calibrated reference mic, but on different days and using two speakers with much different individual frequency responses. The yellow curve is a 5 inch full range speaker and the green curve is a Tannoy Reveal 502 powered monitor. The Tannoy’s tweeter puts out more above 10K, resulting in less “grass” in the measurement, ie better S/N ratio. Otherwise, the curves are pretty close.