KM84+ and KM84++ Head Amp Circuits

Introduction

Before I go into detail about the KM84+ and KM84++ head amp circuits, let me first tell you about how I arrived at these designs.

To start with, there was a request from my son to make some good general-purpose SDC microphones, mainly for live sound applications. And they shouldn’t cost him a fortune. I decided I would build several different types, transformer and transformerless, and with different capsules. The transformers and capsules had to be generally available and of good quality, but not necessarily the absolute top of Schoeps and Neumann. For the bodies, I would use cheap SDCs as donors. I also figured that I would need two well-known references that are generally known as “good” to compare my designs against. Two references, because I divide head amp circuits somewhat artificially into two worlds: one “colored” with transformer and with relatively high Harmonic Distortion, the other “transparent” with the lowest possible distortion and flat frequency response. Any other circuit I might think of after that, I would measure against these references. As a “transparant” circuit, I decided I would take a generic Schoeps circuit, which most cheap SDCs have. Possibly slightly modified, as described with the Takstar CM-60 and CM-63 microphones. The other “colored” one should be the classic KM84 circuit, but because I didn’t have a microphone yet with such a circuit, I decided I would build one myself. However, since I am a big fan of SMT components and the PCBs that were offered for DIY KM84 projects were only made for through-hole components, I decided to design a PCB for it myself.

At more or less the same time that I started my KM84 build, there was this thread started by joulupukki on GroupDIY, where he posted about cell phone interference issues he had with his DIY KM84 builds. At that time, I didn’t actually realize that this circuit was so sensitive to RFI issues. But because I considered good RFI rejection to be one of the key features of a good head amp design, I decided to jump in and pick up the challenge of finding a solution. More than 200 posts and several PCB iterations later, the issue seemed to be resolved, and I had created a KM84 design with very effective RFI rejection. With this exercise finished, I had my KM84 reference design, and Boyd Timothy was happy to have some very useful KM84 clones. (Click on the picture below to open his Youtube channel.)

When listening to Boyd’s recordings, I often found it difficult to tell one microphone from the other. And I was more often wrong than right in guessing which microphone belonged to which recording. But fortunately for me, I was not the only one, and opinions about what could be heard also differed widely and were sometimes even contradictory. Mind you, any differences you might be able to hear, I think are mostly due to the combination of the capsule and back chamber and vent design and to a lesser extent the head amp circuit design.

The KM84+ head amp circuit

The KM84+ circuit basically follows the classic Neumann KM84 design, as described in detail here. However, I added a few interesting features to the KM84 circuit that I genuinely consider an improvement. Hence the “+” suffix. As already mentioned in the Introduction, the most important feature added is the RFI filter. And because the CM-60 has a low-cut switch instead of the pad switch found on the original KM84, I added that to the circuit. An interesting feature of this low-cut filter is that it has a jumper-selectable roll-off slope. You can either select 12 dB/Oct, or select 6 dB/Oct (well, sort of; depends on a resistor). The steep filter can be used to prevent low-frequency rumble muddling the sound. Or worst case, causing it to saturate the transformer and increase THD or intermodulation products. The shallow roll-off can be useful to reduce the proximity effect. In an Alctron T-02A incarnation of this circuit, I added a pad switch to the circuit. (This circuit will be described here later, once I have tested and verified the prototype PCBA.)

Now let’s have a look at the circuit as applied in the CM-60.

If you know the KM84 circuit, the schematic should look familiar to you. Compared to the original KM84 circuit, I have made some changes for various reasons. In short, they are:

1. As explained in the KM84 circuit analysis, there is a small noise benefit to be expected when using a larger capacitor value for C1. That’s what I did. A 1 nF should be OK too, but I wouldn’t take smaller. There is nothing to gain by making it smaller. Especially because the capacitance of the 3U capsule is larger than the KM84 capsule capacitance, you also want to have a larger value than the 470pF from the original circuit.
   
   
2. Because the circuit mainly consists of SMT components, the JFET can no longer be the good old 2N3819 but a more contemporary MMBFJ309LT1G from onsemi. Of all the JFETs I tested, this one gave the highest gain and lowest noise.
   
   If you want to remove C2 and want two matched microphones, you’ll want to handpick JFETs and match the gm within ~5% for less than 0.5 dB gain difference between the two microphones. With a closed-loop gain of ~7x for the KM84+ circuit, a ~15% gm mismatch would yield the same 0.5 dB gain difference. I didn’t measure the gm of a batch of MMBFJ309LT1G JFETs, but I’ve built maybe 10 PCBAs now, and the gain differences measured less than a few tenths of dBs. So I’d expect that within a batch (or reel) the gm to be within a sufficiently small bandwidth to not have to bother about it.
   
   
3. Instead of hand-picking a JFET and its Source resistor (R3), I made the Source resistor adjustable. If you use a good, reliable brand like Bourns, there is nothing wrong with using a trimmer pot. They can be as stable as metal film resistors and allow you to try different bias options, e.g. for maximum output swing or minimum THD.
   
   
4. Capsule bias resistor R1 and Gate bias resistor R2 are 1GΩ in the circuit diagram depicted. This is a standard value for such resistors, but if you have adequate moisture protection, you can safely use larger values. This will yield a few dB better SNR. However, higher values can be quite expensive or difficult to find. But on Aliexpress, you can find decent-quality 2GΩ, 3GΩ, or 5GΩ resistors at a reasonable price.
   
   
5. The most important improvement to the circuit is the RFI filter. L1 blocks RF currents from flowing through the circuit ground. Beads L2 and L3 do the same for the head amp circuit. C11 is a Johanson Dielectrics X2Y capacitor, which suppresses Common Mode interference and directs the Common Mode currents to chassis ground through the XLR pin 1 to chassis connection. This connection must be as short as possible for efficient filtering. An example of what I consider a good XLR insert can be found in the Takstar CM-63. A picture can be found in the CM-60/CM-63 description. I bought it here on AliExpress. And of course, the PCB layout must be designed as if it were an RF circuit.
   
   Next to the L1/L2/L3/C11 RFI filter, I added another ferrite bead L4 in the JFET output to the transformer. This helped to reduce the RFI even better. Later, when working on the KM84++, I noticed it could be replaced with a 470R resistor having almost the same RFI suppression. Due to the 7:1 reduction of the transformer, the noise contribution of this resistor is negligible.
   
   
6. Lastly, since the CM-60 has a low-cut switch instead of the pad switch of the original KM84, I also added this feature to the circuit. By adding resistor R13, the roll-off slope of the low-cut filter can be reduced to ~6dB/Oct.

Let’s start with the last item and show the Frequency Response charts of the KM84+ head amp circuit, i.e. without the capsule. What you see below are the graphs of the two KM84+ PCBAs that were used in Boyd Timothy’s Blind Tests.

At first sight, these seem like three graphs, but the FR graphs match within 0.1 dB. And that is without hand-picking selected components. With the low-cut filter off, the FR is almost flat down to 20 Hz. With the jumper set to -6 dB/Oct, the -3 dB cut-off point is at 168 Hz. The slope is not exactly -6 dB/Oct, most notably at higher frequencies. For not too close-miking, this could be the right slope, but you might want to adjust the slope to your taste by experimenting with different C9 and R13 values.

Without a resistor parallel to C9, there is a very small amount of peaking at ~300 Hz, which I considered acceptable. The roll-off slopes down with 12 dB/Oct, effectively reducing stage rumble and handling noises without affecting mid-range instruments and voices too much.

I have measured some key performance data, which are listed below. They were measured at 1 kHz, and fed to the circuit input through a 47 pF capacitor that simulates a capsule capacitance.

How should we interpret these data?

• Gain is self-explanatory. It is assumed the transformer is not saturating.
• An input signal of 100 mV can be expected from the capsule in the 105-115 dBSPL range, which any SDC should be able to handle without much distortion.
• The maximum input level specified at 0.5% THD is reached at an input level that can be expected in the 115-125 dBSPL range. This is quite similar to the KM84 specification, so the KM84+ is on par with the original KM84.
• The Equivalent Input Noise (EIN) of the circuit does not include the capsule noise, so it does not represent the EIN of the complete microphone. We want it to be lower than the self-noise from the capsule, but at least to be less than an EIN of 15 dBA of what Neumann considers the maximum of a very good EIN. Ignoring the capsule self-noise, and assuming a 10 mV sensitivity at 94 dBSPL, we calculate the SNR as: SNR = 20*Log(10E-3/1.37E-6) = 77.2 dB. To get the EIN of the microphone, we subtract this value from the 94 dBSPL reference level, which yields EIN = 94-77.2 = 16.8 dBA. This alone no longer matches with a microphone that could be considered “Very good” according to Neumann’s standards but is good enough for most purposes. So if we aim for something exceptionally good, then there is some room for improvement here.

<<<<  Placeholder for measurement data >>>>

For several SDC bodies, a.o. the Takstar CM-60 and Alctron T-02A, I have designed KM84+ PCBs that can be ordered through PCBWay, where you will also find the Schematic, BOM and build instructions. You can check availability and find the links here.

An improved KM84+ circuit: introducing the KM84++

After designing the KM84+ circuit, I got inspired to make further improvements. Before discussing these improvements, the circuit diagram below shows the circuit concept, which I simply called KM84++ because of the still recognizable KM84 design, but with a few pluses added.

Although I had already shown in my description of the KM84 Head Amp circuit that the classic KM84 circuit is basically well thought out, I felt that improvements were possible, especially concerning the Signal-To-Noise ratio and distortion. However, these improvements are only feasible if we let go of the concept of a JFET, set to a low bias current so that the phantom voltage can be used as the capsule bias voltage. A JFET produces less noise and distorts less when it is set to a higher current. But to allow a higher bias current while at the same time keeping a high polarization voltage, we need to add a capsule voltage generator. That’s exactly what I did at the KM84++ circuit. For the generator, I use a CMOS oscillator with a charge pump voltage multiplier, as I described here.

As explained in the KM84 circuit description, the output impedance of the JFET stage, coupling capacitor, and transformer inductance constitute a resonant circuit. In the KM84 circuit, there is little or no peaking of this resonant circuit due to the high output impedance of the JFET stage. But because of the higher JFET bias current in the KM84++ design, its output impedance will be lower and a different transformer would be needed. However, I was not able to find a suitable transformer that would fit into an SDC body and that would match this circuit better than the KM84 7:1 transformer. So I ended up using the same transformer as in the KM84+. This is the ASTDS T-8 model. With this transformer, there would now be too little damping, resulting in an LF resonant peak. Whether this is acceptable, or maybe even desired, depends on the capsule used and personal preference. Let’s for now assume that we do not want this peak and add some kind of damping to tame the peak.

One solution to flatten out the peak could be adding a resistor of ~ 6k or so in series with the transformer input. But that would add noise and it would increase transformer distortion, nullifying the improvements that we made by increasing the bias current. So we have to find another ruse to add more damping… To this end, I divided the coupling capacitor into two capacitors C4 and C9. C4 acts as the DC blocking capacitor and has a value much higher than C9. Capacitor C9 is the actual coupling capacitor determining the LF cut-off frequency. When we now shunt this capacitor with resistor R13, we introduce the required damping to the circuit again. This “trick” would not work without C4, because without it, a DC current through R13 would flow from the JFET to through the transformer, upsetting both the JFET stage and the transformer. A stepped simulation below shows the effect of R13 on reducing the peak. I ended up using a 10k resistor, which, when measured in the actual circuit, gives a near-flat frequency response. With other transformers than the ASTDS T-8, or if you prefer more or less damping, you may want to try other values to voice the LF response to your taste.

Another difference between the KM84+ and the KM84++ that you may have noticed is R6. In the KM84+, this is a ferrite bead in this position (L4). However, when using a bead in the KM84++ circuit, the circuit produced funny high-pitched noises and distortions that were probably caused by RF oscillations. When I jumpered L4, the funny noises went away. But without an impedance in this position, RF rejection became worse. With L4 replaced by a 470 Ohm resistor, RF rejection was almost on par with the KM84+ circuit, while maintaining low circuit noise. A 470 Ohm resistor would seem like a serious noise adder, but thanks to the 7:1 attenuation by the transformer, the noise contribution is equal to that of a ~10 Ohm resistor.

Just as for the KM84+, I measured the same key performance data and listed it below.

And just as for the KM84+, I will explain how to interpret the data.

• The gain of this circuit has been set lower than the KM84+ because due to the higher capsule voltage, the capsule sensitivity has increased by 3.4 dB. 
• As can be seen, THD for this circuit is significantly lower than the KM84+
• The circuit can also handle much higher input levels before reaching 0.5% THD. When this circuit is combined with a 3U cardioid capsule, and with C = 8.2 pF, the microphone can handle up to 121 dBSPL with the pad switch off. And a maximum of 141 dBSPL with the pad switch on. 

<<<<  Placeholder for measurement data >>>>

For several SDC bodies, a.o. the Takstar CM-63 and Alctron T-02A, I have designed KM84++ PCBs that can be ordered through PCBWay, where you will also find the Schematic, BOM, and build instructions. You can check availability and find the links here.