KM84 revisited

[Chrisfromthepast]

In the meantime, I acquired a real KM84, but it needs repair. If it isn't the capsule, then I guess I will be able to fix it.

Id love to hear some off axis tests, even if the datapoint is an outlier, I love to see some data.
I haven’t been keeping up with my measurements lately, and will be performing some more sweeps soon. I can probably grab some larger speakers from rental inventory and get higher spl measurements than I was able to achieve with the Genelec.
I typically do 8” from a coaxial source and get up to 110db before limit.
L-acoustics x15”s are very flat when mounted on a pole, I’ll see how loud I can make my test rig. I could try k2 at 24” but i worry about room noise and the contour would need to be flattened in network manager before testing.
I want to measure at extreme spl because all of my km84 home builds pass the voice test, but sometimes have trouble under real world spl.
Is there a test prescription we can agree on before I contribute meaningless results?
So stoked on this final chapter to my km84 journey.

 

[jp8]

Interesting! Will you be using the 3U/CM-63/KM84++ for this test? The off-axus response of the 3U capsule is supposed to be quite good. Better than the stock CM-63 capsule.

I designed the KM84++ for a max SPL handling of 140 dB with the pad engaged and assuming a 3U cardioid capsule. That's 10 dB up compared to a classic KM84. If you want to further increase the max SPL handling, you can reduce the mic sensitivity by taking a higher value for C2 (Which can be done on any KM84 incarnation, btw. Not just the KM84++.). This is a major advantage of the the charge amplifier concept as opposed to the voltage amplifier circuit of e.g. the Schoeps circuit: Schoeps has fixed gain, whereas the KM84 gain can be tweaked by changing a single capacitor.

Jan

 

[Chrisfromthepast]

140 dB spl is probably going to be to high for me to test a full sweep. My measurement mic reaches 5% distortion at 130. 1k might be clean enough, I’ll investigate.
I’ll hopefully do a few capsules including 3u.
So, I can produce reasonable graphs for on axis, 90degrees, 180 degrees.
I can determine a max spl (for 5%thd at a target frequency of 100 hz?) 1k hz seems too high to catch transformer saturation, but thats the figure to compare to commercial mics so maybe I should do both.
Self noise: how do I measure this one? I imagine with a capacitor replacing the capsule, and then calibrate the input of the preamp to be 0db gain? Then i subtract from….
To test for electromagnetic noise, I have this plasma ball lamp that emits broadband rf junk I can bring near the DUT. Is there a more qualitative measurement approach?
For sensitivity, I can calibrate the test rig to be 94db spl and the preamp gain to be 0db, and then if I understand correctly, sensitivity is calculated from the max spl minus this figure? I usually just try to match relative sensitivity among pairs, and im not confident how to produce a measurement referenced to 1 pascal.
Thats plenty of questions for now.

 

[Chrisfromthepast]

These measurements are raw from Genelec 8331A, 8 inches from the center of the cone. She is just under limit so I don’t necessarily trust the distortion measurements.
This is the first one I built, more to come. The capsule is 3u super cardioid, which looks kind of dumb but measures well.
I have 3mm capsules I knew I wanted to use so I made the pogo pin just a tad too short to test the stock capsule. I’ll work on it.

 

[jp8]

@Chrisfromthepast , thanks for sharing the frequency charts! The 3U capsules are measuring quite flat. I've done some measurements in combination with the stock Takstar CM-60 and CM-63 bodies + head amp circuits, which should not deviate much from Takstar bodies + KM84+ or KM84++ head amp circuits, except maybe for the lowest one or two octaves.

You can't measure THD on microphones with commercially available speakers. They are likely to distort more than your microphone, especially in the lower octaves. The way I do these measurements, which is probably also done by microphone manufacturers, is to measure the THD of the head amp circuit alone, feeding the signal from the generator into the circuit with a capacitor with a value equal to the capsule capacitance. I turn up the signal level until 0.5% THD is reached. I then read the output signal from REW or the voltmeter in my analyzer. If the sensitivity of the mic is known, you can easily calculate the max SPL level from there. Example: mic sensitivity is 15 mV @ 94 dBSPL and 0.5% THD is obtained at 500 mV output level. The max SPL for his mic can be calculated as: Max_SPL = 94 + 20 * Log(500/15) = 124.5 dBSPL.

If you really want to get a good idea of the distortion of a complete microphone instead of just the head amp circuit, I would suggest a Difference Tone Distortion test with two speakers. Another project I have been meaning to do for a few years now, but I haven't gotten around to it yet.

I was planning to write an article about noise measurements on my website, but I haven't gotten around to it yet. And I'm afraid I won't get around to it for a while.

To test for electromagnetic noise, I have this plasma ball lamp that emits broadband rf junk I can bring near the DUT. Is there a more qualitative measurement approach?

Yes, there is. At least, if comparative measurements are good enough for you. I made an RF jammer to that end. You can read more about it here.

Jan

 

[Chrisfromthepast]

I have not been able to get any of my km84++ populated by pcbway to have distortion under 1% at 1k. Using the capsule voltage trimpot, I was able to match the levels of 3 mics after a few tries of measuring and adjusting the trim pot. The bias point can be set 11V or somewhere nearby to get the lowest 3rd harmonic, but the distortion figures are 2 to 3 x any other km84 type mic in my locker. Should I try a different fet?
The stock cm63 circuit is quite clean until about 135dbSPL, but even at very low levels, the distortion measurements of the km84++ have been high. They roughly match the REW file I provided.
Should I try to replace the fet with something from my km84 parts bin? Or go for another output transformer? Only 3 of the 4 aliexpress transformers worked. One had a shorted primary winding

 

[jp8]

I'm sorry you could not get the KM84++ working properly. THD levels of the KM84++ should be orders of magnitude lower than a KM84 circuit, even at signal levels corresponding with > 100 dBSPL. See here.

Before you start replacing parts on a whim, could you please answer the questions below so we can help you better?

• What does your measurement setup look like? Please provide a schematic, pictures, signal levels applied, etc. Whatever you have available.
  
  
• Did you measure the distortion by feeding a signal through a capacitor into the pogo-pin? This capacitor should have approximately the same capacity as the capsule. It is essential not to feed the signal directly into the circuit from the output of your sound card but via a capacitor. Or did you measure the distortion of the complete, finished mic recording a sweep from a speaker?
  
  
• If you measure without a capsule, are you measuring in a Faraday cage? Mains hum pickup could have saturated the circuit and caused distortion.
  
  
• Did you allow PCBWay to use parts sourced from other sources than Mouser or Digikey? I know they have been using counterfeit X2Y capacitors because the one I specified was in short supply. Those counterfeits were smaller and could not be soldered on the board. Maybe they also used other counterfeit parts? Can you read the text on the JFET?
  
  
• You said one of the Aliexpress transformers had a shorted primary winding. Maybe the others have some kind of defect as well? If you have a known good transformer from another KM84 build, you could give that one a shot. I sourced all my transformers from the Getwintech02 Store on Aliexpress and only had good experiences with them. If you have any doubts about the transformers, measure them separately outside the circuit to be sure you have a good transformer in your KM84++ build.

Jan

 

[Chrisfromthepast]

These measurements are raw from Genelec 8331A, 8 inches from the center of the cone.

THD levels of the KM84++ should be orders of magnitude lower than a KM84 circuit,

Im with you, I want to get them tip top.

What does your measurement setup look like?

Audix tm1, genelec coaxial 8331A 8inch wireless antenna used as a spacer, I output a sweep from REW at -18dbfs.
I calibrate the input gain with a 114db or 94 db tone emitting calibration tool, make sure the system has plenty of headroom in terms of the mic pre and the speaker output, and run a test as loud as possible within this clean range. Low end would require a larger rig, but the Genelec does 105 ish quite cleanly at 1k. As you said I cannot accurately measure the distortion of this speaker. But I regularly use this mic, and preamp and rew to validate if speaker systems are up to specifications (hf magnet gaps with dirt in them and stuff) and Im familiar enough with the genelec and REW’s rendering of the distortion graphs to know that its the mic itself.

If you measure without a capsule, are you measuring in a Faraday cage? Mains hum pickup could have saturated the circuit and caused distortion.

I will do a test as you describe, however, I use this equipment to verify loudspeakers and mics regularly, and the native distortion of the loudspeaker is at least an order of magnitude lower than the figures we are talking about.
When I get one of the mics below 1%, I will change my approach to directly measuring the capsule across a capacitor.
My measurement set always includes a control group measurement with the audix, which proves the test is clean, and provides a basis.

Did you allow PCBWay to use parts

I went back and forth with them a little bit to make sure they used only parts from your BOM. They were surprisingly attentive; they altered the pcb thickness to conform to your notes, and waited 3 extra weeks for the xy cap

If you have any doubts about the transformers, measure them separately outside the circuit to be sure you have a good transformer in your KM84++ build.

I’ll try a nte 10:3 and a gzt 84

 

[Chrisfromthepast]

Im familiar enough with the genelec and REW’s rendering of the distortion graphs to know that its the mic itself.

Forgive me, I was doing my tests too loudly and in too messy of a space so I created a rattle.
At 94db test SPL, and after clearing some junk, Im back under 1%.
More to come.
Oh, and the transformer, I pulled back some tape until I found the leads, nudged them, and now it tests 20 omhs primary and 200 ohms secondary. So that wouldn’t do ANYTHING, so it means I had the leads soldered together and just couldn’t see it.
My mistake. Thanks for your patience.

 

[jp8]

As far as I know, when doing THD measurements on microphones with a speaker as the source, the THD of the speaker will generally be higher than that of the microphone. To get an idea of the non-linear distortion of a microphone, it is better to do a Difference Tone Distortion (DFD) measurement. Difference tones are also disturbing at a much lower level, usually non-musical and therefore actually more interesting than THD. For DFD measurement, you can use two identical speakers driven at the same level but with a different frequency. They don't necessarily have to have low THD; because each speaker handles only one frequency, it will not create DFD distortion that would obscure the mic DFD signals, as THD from the speaker would do.

More info here:
https://www.dpamicrophones.com/mic-university/technology/the-basics-about-distortion-in-mics/

Jan

 

[Karl-0_0-]

Mouser and Digikey have run out of stock on the Johanson X2Y capacitor type 101X18N102MV4E or EMDD101G102M1GV001E, which is used in the RFI filter. Based solely on datasheet comparisons, Knowles Syfer part 1206J1000102MCTE01 should be a good replacement. I will change the BOM accordingly.

Jan

Was shopping some components for the KM84++ and noticed some more than the Johanson capacitor are out of stock. Any suggestions for replacements on these components?

Murata BLM18HD102SH1D
TE Connectivity / Holsworthy RGP0207CHK1G0

 

[jp8]

For the ferrite bead, you could use these instead:
Würth 742841160
Murata BLM18HG102SH1

They may not be quite as effective against cell phone RFI, but if that proves to be an issue, then replace them later with the BLM18HD102SH1D.

There aren't many affordable 1 GOhm resistors available from Mouser or Digikey. And most are way too big. I'd suggest to get some HVR82 series types from Aliexpress. I recently bought 2G and 5G versions from the link below. The higher the resistance, the less noise. But it will also potentially make the mic more moisture sensitive and with 5G, it can take up to a minute before the JFET bias has settled and it will pass audio. I assume most people will be perfectly happy with the noise level when using 1G resistors.

Jan


€5,39 | 10pcs HVR82MU1005 Chip Resistor 500M10M100MF10GJ5M2G33KF1G1% Ohm 0.5W Glass Glazed Non Inductive High Voltage Resistor
https://a.aliexpress.com/_EvTX0lA

 

[Karl-0_0-]

For the ferrite bead, you could use these instead:
Würth 742841160
Murata BLM18HG102SH1

They may not be quite as effective against cell phone RFI, but if that proves to be an issue, then replace them later with the BLM18HD102SH1D.

There aren't many affordable 1 GOhm resistors available from Mouser or Digikey. And most are way too big. I'd suggest to get some HVR82 series types from Aliexpress. I recently bought 2G and 5G versions from the link below. The higher the resistance, the less noise. But it will also potentially make the mic more moisture sensitive and with 5G, it can take up to a minute before the JFET bias has settled and it will pass audio. I assume most people will be perfectly happy with the noise level when using 1G resistors.

Jan


€5,39 | 10pcs HVR82MU1005 Chip Resistor 500M10M100MF10GJ5M2G33KF1G1% Ohm 0.5W Glass Glazed Non Inductive High Voltage Resistor
https://a.aliexpress.com/_EvTX0lA

Thanks for the help.

I noticed there is no 2GM in the Ali-link, "only" GF, GJ and GK. What's the difference between these?

 

[jp8]

G = Giga Ohm, as you probably guessed already.

The 2nd character indicates the tolerance. Most often used characters are:
F = 1%
J = 5%
K = 10%
M = 20%

Accuracy is not important in this case, but strangely enough I always get tempted to choose the most accurate one offered.

Jan

 

[Karl-0_0-]

G = Giga Ohm, as you probably guessed already.

The 2nd character indicates the tolerance. Most often used characters are:
F = 1%
J = 5%
K = 10%
M = 20%

Accuracy is not important in this case, but strangely enough I always get tempted to choose the most accurate one offered.

Jan

Ah, I see. So 20% is more accurate than 1%?

 

[Khron]

Ah, I see. So 20% is more accurate than 1%?

20% = the actual value will be within +/-20% of the nominal (specified) value.

Eg. 1G 20% will count as "within spec" for anything between 800meg and 1.2G.

 

[Karl-0_0-]

20% = the actual value will be within +/-20% of the nominal (specified) value.

Eg. 1G 20% will count as "within spec" for anything between 800meg and 1.2G.

That's what I though. I asked only because when I clicked the Ali-link, the most accurate option offered wasn't selected.