Banzai's KM84 DIY Body & PCB kit build thread

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Unfortunately I agree with you.
I bought a pair from a previous batch, but they didn't sound anything like a real KM84.
So I sold them...
It’s unfortunate really, but any of the other 2 mics in the files, the Beyer and the Oktava although not pretending to be based on the KM84 are much more close to the KM84 sound than the Banzai.
 
I have a couple of bias related questions here. I'm trying to learn about the correlation between the THD and harmonics shown in the RTA and the view here. I first set the bias such that the waveform looked best and I felt the sine wave I was hearing in my headphone was purest. I arrived at 12.44k on the trimmer.
Then I figured I would take a look at the THD and found that if I decreased the trimmer to around 11.41k, the 2nd harmonic dropped considerably and I measured less THD. Which is more desirable? Is something flawed here in this methodology?
 

Attachments

  • 11.41k.png
    11.41k.png
    250.7 KB
  • 11.41k.jpg
    11.41k.jpg
    71.6 KB
  • 12.44k.png
    12.44k.png
    251.5 KB
  • 12.44k.jpg
    12.44k.jpg
    71.8 KB
The 2nd question is about the shape of the waveform... you can see the top of the waveform is wider than the bottom. This is true even before the signal starts clipping thru the fet. Before the FET, it's a clean sine wave straight from my computer. I double checked with my old heathkit signal generator and I get the same results. What would cause this, is it normal for this type of circuit?
 

Attachments

  • Screen Shot 2021-08-10 at 2.28.30 AM.png
    Screen Shot 2021-08-10 at 2.28.30 AM.png
    117.4 KB
Strange that all FETs need such a high value source resistor...
In my microphones I usually end up with a value around 3 to 5 K.ohm for a 2N3819 in the same configuration.
(Biased for lowest distortion.)
 
Last edited:
The 2nd question is about the shape of the waveform... you can see the top of the waveform is wider than the bottom. This is true even before the signal starts clipping thru the fet.
The waveform looks pretty distorted! I don't remember having seen such a distorted waveform at levels below clipping.
 
The waveform looks pretty distorted! I don't remember having seen such a distorted waveform at levels below clipping.
Thanks very much for your reply. These are “Fairchild” resistors that another member had gotten from eBay so I’m suspecting these may not be authentic 3918s. I have 7 of these mics and they are all showing similar issues. Does anyone have a recommendation for a source in USA of authentic transistors?
 
From what I understand, Banzai changed the brand of the FET due to inconsistencies with the first batch. The ones in this second batch need a much different bias resistor setting. I'm not really sure it makes a whole lot of difference in the sound. The FET is being used as a voltage amplifier and not a current amplifier, so it shouldn't make much difference.
 
It’s unfortunate really, but any of the other 2 mics in the files, the Beyer and the Oktava although not pretending to be based on the KM84 are much more close to the KM84 sound than the Banzai.
I'd agree with this. I did change out C4 in one of my two Banzai KM84's to a 1uF. That may provide the needed hpf the circuit needs. I'm a big fan of both the Oktava and the Beyerdynamic. The Banzai was a good covid diversion and I will keep playing with it. It may be just the right mic on some sources. Some people on other forums did pick it as their favorite... until they found it wasn't the KM84. Then the answers started to change. I do like the KM84. It's not always the best or most magical mic but it never sounds bad or wrong. The very wide polar pattern also makes off-axis recordings sound much better. I found the Maiku capsule to have a very large change when off-axis. Still, I will do more experimenting and let you all know my findings. Or, maybe I can post a round two of the comparisons.
 
The Banzai may be just the right mic on some sources.

Although I really didn't like the sound, it's not what I expect from a SDC condenser, I can understand that it could be useful for some sources.

The problem is that these kits are advertised as "Neumann KM84" and are supposed to be Neumann KM84 clones, that's why people buy the Kits, and the reality is the sound is nowhere similar to the KM84, neither to most SDC mics, then it looses all the point.

I'm quite convinced at this point that there's something terrible wrong with this project, like I said before maybe the Capsule is not good, or the transformer, the body head or a combination of these 3...

I suspect the Capsule...
 
Last edited:
I have a couple of bias related questions here. I'm trying to learn about the correlation between the THD and harmonics shown in the RTA and the view here. I first set the bias such that the waveform looked best and I felt the sine wave I was hearing in my headphone was purest. I arrived at 12.44k on the trimmer.
Then I figured I would take a look at the THD and found that if I decreased the trimmer to around 11.41k, the 2nd harmonic dropped considerably and I measured less THD. Which is more desirable? Is something flawed here in this methodology?
On my builds, to get the least distortion, my bias resistors ended up being 18k-21.5k. I think you can change out the trimmer with a 25K and dial it in better. That waveform is not symmetrical.
 
On my builds, to get the least distortion, my bias resistors ended up being 18k-21.5k. I think you can change out the trimmer with a 25K and dial it in better. That waveform is not symmetrical.
I actually managed to get symmetrical clipping with all but one of my transistors with the trimmers between 7k-12k. My main concern is the shape of the waveform after the FET. I've ordered new FETs from Small Bear Electronics and will give them a try.

For what it's worth, I put one of the mics up against my KM86 and also noticed the excessive bass in comparison to the Neumann. The high end was different too, but of course the KM86 has a very different head basket design. Once I try out some different FETs I will be sure to do a more thorough test and comparison of this build and my KM86s. Happy to share the resulting WAVs...
 
For what it's worth, I put one of the mics up against my KM86 and also noticed the excessive bass in comparison to the Neumann. The high end was different too, but of course the KM86 has a very different head basket design. Once I try out some different FETs I will be sure to do a more thorough test and comparison of this build and my KM86s. Happy to share the resulting WAVs...

please do, would love to have a listen.

The differences you found to the KM86 are consistent to the the differences with the files from McIrish either the KM84, not only excessive Low End but also completely different in the high end
 
From what I understand, Banzai changed the brand of the FET due to inconsistencies with the first batch. The ones in this second batch need a much different bias resistor setting. I'm not really sure it makes a whole lot of difference in the sound. The FET is being used as a voltage amplifier and not a current amplifier, so it shouldn't make much difference.
No. It's not possible for me to reliably source several hundred Fairchild 2N3819's for each batch.

This is the 4TH batch of this project and the 3rd year we've been doing this. So you can imagine how many hundreds of this kit are out there, and how many Fairchild FETs we already sourced.

So for this batch we went with Mouser instead. It's much easier for you as individuals to source 2 or 3 Fairchild FETs each. Something I can't do collectively for everyone.
 
Last edited:
For what it's worth, I put one of the mics up against my KM86 and also noticed the excessive bass in comparison to the Neumann. The high end was different too, but of course the KM86 has a very different head basket design. Once I try out some different FETs I will be sure to do a more thorough test and comparison of this build and my KM86s. Happy to share the resulting WAVs...
You have a balance problem. Once you fix the lowend, the highs will fix themselves.

See below.
 
For everyone building these kits:

DIY means each one of you is in charge and doing all the work yourselves. We only provide a framework for you to build on. Adjusting the circuit and tuning the final sound of your microphones, is something you need to do by yourselves. Same applies to every microphone build on this forum.

And maybe this wasn't as obvious as it should have been, but the transformers available for this project do not sound the same. None of the 3U builds will sound the same as the Cinemag builds or the AMI builds.

In practice that means you'll need to make adjustments for each transformer build. If you don't know how to adjust the circuit, or even where to start, this thread is the perfect place to figure it out together.

And this should have been clear from the beginning:
Just because there isn't any complicated wiring and you're not fiddling with power-supplies, doesn't mean this is an easy project.

Every microphone build is complex, and needs a good grasp of how every component interacts with each other, along with a bag of tricks to get everything to sit where it's supposed to. Less demanding projects like this one are a great way to learn how things work, but you still need to apply the same process as you would to much bigger builds.

For e.g., in a tube mic you can adjust the cathode bypass capacitor, the coupling capacitor and the grid resistor to tune the low-mids and lows. So how do you do it in a FET circuit, when they aren't that different? What happens to the balance of a microphone when you remove bass?


Just remember almost everything in this circuit can be adjusted. You don't need to stop at what we've provided, it's DIY. Try different capacitor materials. See if there's any difference using a FET that needs a low source resistor vs the high resistance you're using now. Remove parts and see what happens. Even try a different transformer if you can't make another one work. Because this project will reward however much effort you put in – it's those who weren't afraid to try different things who reported the best results. Again, all of this applies to every build on this forum.

Otherwise, if you want something that performs well with minimal work, then we suggest using the Cinemag 5722. This kit was designed using that transformer. But until you adjust the balance of your 3U builds, you won't know what it's capable of.
 
No. It's not possible for me to reliably source several hundred Fairchild 2N3819's for each batch.

This is the 4TH batch of this project and the 3rd year we've been doing this. So you can imagine how many hundreds of this kit are out there, and how many Fairchild FETs we already sourced.

So for this batch we went with Mouser instead. It's much easier for you as individuals to source 2 or 3 Fairchild FETs each. Something I can't do collectively for everyone.
We actually can't imagine! Something interesting I've noticed about the DIY community and the folks who put kits together: the consumers have no context for how many pieces are being sold. I've noticed it's rare on the whitepages to see a quantity cap ("I'm making 100 PCBs and then it's done").
BY NO MEANS is this a criticism: DIY means many things for many people. For example, some guitar pedals come with matched FETs and the bias resistor predetermined and sent to you ready for soldering. Plug & Play. This project is not that. Perhaps some folks who bought into this project were not clear on that fact?
 
Otherwise, if you want something that performs well with minimal work, then we suggest using the Cinemag 5722. This kit was designed using that transformer. But until you adjust the balance of your 3U builds, you won't know what it's capable of.
It's not clear if you are referencing the low end issue I was describing - but I do in fact have the Cinemag transformers. Either way, I will be trying a 1uf tantalum in C4 position connected to the x-former primary...
 
Back
Top