MXL to KM84 project troubleshooting.

[SparkleBear]

Hey all...

Been working on a bunch of fun MXL circuit conversions to classic fet and tube transformer designs. Right now I am working through the km84 build.

Everything looks good with the voltages. Trimmed 10v at the fet drain, getting 23.4v at the kathode of D1, seeing 46.5v at the r6 r10 junction... I didn't use a polarized cap for the output decoupler... but a film wima.

But the audio is very low level and very thin sounding. Recorded with u3 omni capsule (original stock capsule sounds even worse lol) through apollo x4 with 35db gain on preamp. Here is a recording and some photos/pics of the design. Anything stick out to anyone?

https://www.dropbox.com/s/97yqjhu7e0zpyw3/KM84 thin sounding.mp3?dl=0








Thanks for any support, friends.

-Eric James

 

[Khron]

Are you 100% sure the transformer is connected the right way around?

If the primary and secondary were reversed, 1uF into a low impedance would explain the thinness of the sound. The low impedance would be dragging down the level coming off the high-ish impedance drain connection quite drastically, despite the then step-up of the transformer ratio.

 

[SparkleBear]

I am certain that the primary is wired properly. Red and brown are p+/-, orange yellow are secondaries.

Could a bad fet or a bad cap between the two standoffs be the culprit?

 

[Khron]

Measure the resistances of the primary and secondary. Colours mean nothing, electrically.

But sure, might as well swap the fet, if you have spares. Just remember to recheck the biasing.

 

[SparkleBear]

I did measure the resistances. Low resistance is towards the XLR and high resistance is towards the circuit.

Is it ok off the bat to just bias the fet to 10v at the drain? And when its all close then do the signal injection biasing?

I just went back and reflowed every joint and replaced the 470pf cap incase i flooded it with ipa accidentally somehow...

Still thin and crappy sounding ;/

I'll swap the FET out tomorrow. Dang.

 

[Khron]

If your meter can go high enough, measure the two 10meg resistors in the capsule bias circuit too. Or parallel a couple of 10k or something more "pedestrian" with them, at least for testing.

 

[lightningbefore]

Dumb semi-related transformer question - can you measure the transformer resistance of the wires out of circuit? I don't know what wire is what on this toroidal, and am not sure where my DMM wires should connect or what the resulting values imply other than two should be signal, one is ground, one should be +

 

[Khron]

Dumb semi-related transformer question - can you measure the transformer resistance of the wires out of circuit? I don't know what wire is what on this toroidal, and am not sure where my DMM wires should connect or what the resulting values imply other than two should be signal, one is ground, one should be +

Any idea what that transformer was meant / intended for?

 

[SparkleBear]

If your meter can go high enough, measure the two 10meg resistors in the capsule bias circuit too. Or parallel a couple of 10k or something more "pedestrian" with them, at least for testing.

What am I looking out for? Just resistance values? or voltages? I've been combing over the circuit all day. Can't exactly find out what is wrong with it. It matches the original schematic as far as I can tell...

Gate, drain and source are correct... capsule is going to the correct standoff... Replaced the 470p with film incase I accidentally soaked the styroflex one with IPA when cleaning...

Would incorrect bias make the sound super thin?? I didn't scope it out but its set to 10v.

I did realize that my first fet was installed backwards.. (gate was correct but drain and source were flipped) But the second fet was installed correctly and the sound is still thin. Is it possible that reversing the pinout could have damaged any compnents downstream? I doubt it but I've never really worked with fet circuits before. DOH

 

[Khron]

Well, by "measure the two 10meg resistors", i meant measure the resistance of those two resistors (if your multimeter can measure that high). Otherwise i would've asked for voltage readings.

But by the sound of your question, sounds like you didn't yet try the second part of that (soldering a couple of 10k's in parallel with those 10megs, "just in case" they're not ok).

Also, depending on what resistance value the trimmer needs to be set to (the higher the Idss of the JFET, the lower the resistance you'll need), to get the proper drain voltage, that 4.7uF source bypass cap might've ended up as too-low a value.
In that position, that capacitor starts "boosting the gain", in essence, starting from a frequency around 1/(2*pi*R*C) ; pay attention to the multipliers (uF - microfarads - 10^-6).

Do you happen to have access to an oscilloscope, by any chance? Failing that, you can even MacGyver a probe to go into the high-impedance (instrument) input on your audio interface...

 

[SparkleBear]

Found something else wrong...



With my pin lifting trickery I had a mislabeled net. C2 was not connected to the drain!!!

I ran a bus wire between the c2 pad and pin 3 (drain) of the FET. However, the audio still is unbearably thin...

https://www.dropbox.com/s/k55yg09fqlbl0ua/Km84 still thin.mp3?dl=0
The mystery is still open... ;P

 

[Khron]

Ouch... Maybe it's just me, but that's quite the oversight on the Kicad side. One would think you'd want the unrouted nets would be more obviously highlighted(?).

More importantly though, does connecting C2 to the drain, fix the issue?

 

[SparkleBear]

Well, by "measure the two 10meg resistors", i meant measure the resistance of those two resistors (if your multimeter can measure that high). Otherwise i would've asked for voltage readings.

But by the sound of your question, sounds like you didn't yet try the second part of that (soldering a couple of 10k's in parallel with those 10megs, "just in case" they're not ok).

Also, depending on what resistance value the trimmer needs to be set to (the higher the Idss of the JFET, the lower the resistance you'll need), to get the proper drain voltage, that 4.7uF source bypass cap might've ended up as too-low a value.
In that position, that capacitor starts "boosting the gain", in essence, starting from a frequency around 1/(2*pi*R*C) ; pay attention to the multipliers (uF - microfarads - 10^-6).

Do you happen to have access to an oscilloscope, by any chance? Failing that, you can even MacGyver a probe to go into the high-impedance (instrument) input on your audio interface...

Yeah my fluke can measure... they are both measuring 10m resistance.

I think with all of the microphone work I've been doing it would be worth getting a meter that can go to G ohms and can meter voltage in the high z sections too.. Do you have a recommendation?

I have a little 10 mhz pocket scope I got from amazon a few years ago. I could probably also find an instrument cable to sacrifice to make a probe if this one doesn't do the job.

What is the purpose of the parallel 10k resistors? Im still not fully understanding the implementation...

Thanks again for the support mate.

 

[Khron]

You still didn't answer the more important question - did connecting C2 where it was supposed to, fix the "thin sound" issue???

And with resistors being as cheap as they are, i wouldn't shell out big bucks on a special meter that i'll use once or twice a year. Easier and quicker just to swap one out, isn't it?

The purpose of paralleling 10k's there is to make sure the capsule bias supply is fine. Those 10megs are there because 1) there's like zero current running through them (and according to Ohm's law, thus zero voltage drop across them), and 2) that enables the use of smaller capacitor values to still get a nice low cut-off frequency for the resulting RC low-pass filters.

If either are, for whatever reason, unwell, 10k's are easily-verifiable values that will work well enough for testing purposes (where noise levels aren't necessarily THAT critical). And since the stock ones are so high value, parallelling such a much-lower value will rule out their influence (since electrons will take the path of least resistance).

 

[SparkleBear]

You still didn't answer the more important question - did connecting C2 where it was supposed to, fix the "thin sound" issue???

Connecting C2 to drain did NOT thicken up the signal. Its still VERY thin and zingy... ouch.

And with resistors being as cheap as they are, i wouldn't shell out big bucks on a special meter that i'll use once or twice a year. Easier and quicker just to swap one out, isn't it?

The purpose of paralleling 10k's there is to make sure the capsule bias supply is fine. Those 10megs are there because 1) there's like zero current running through them (and according to Ohm's law, thus zero voltage drop across them), and 2) that enables the use of smaller capacitor values to still get a nice low cut-off frequency for the resulting RC low-pass filters.

I see...

If either are, for whatever reason, unwell, 10k's are easily-verifiable values that will work well enough for testing purposes (where noise levels aren't necessarily THAT critical). And since the stock ones are so high value, parallelling such a much-lower value will rule out their influence (since electrons will take the path of least resistance).

I could give that a shot.



The Kicad unrouted net is my fault because I am deleting pads and making edge cuts in place of the pads for the standoffs. I should've been extra diligent with those connections. Unfortunately, c2 to drain didn't fix my issue.


Hmmmm.... Maybe swap out the transformer just to check?

 

[lightningbefore]

Any idea what that transformer was meant / intended for?

Ah my bad, for a C12 circuit.

 

[Khron]

Ah my bad, for a C12 circuit.

Could've considered starting your own thread for this, but in such a circuit, the primary will be the greater-resistance one (hundreds of ohms, or even Kohms), and the secondary will be the lower-resistance one (tens of ohms).

 

[Khron]

Connecting C2 to drain did NOT thicken up the signal. Its still VERY thin and zingy... ouch.

.....

Hmmmm.... Maybe swap out the transformer just to check?

C2 is a form of feedback that, for the most part, "only" reduces the gain. Should have zero influence on the frequency response.

https://audioimprov.com/AudioImprov/Mics/Entries/2015/12/20_Modding_a_BM-800_Mic.html
Sure, swapping out the transformer would rule out that as a culprit. Got any spare capsules too, though?

 

[SparkleBear]

I have several capsules including the stock electret, and 3u Omni. They all sound brittle. I’ll try a new transformer tomorrow...

 

[SparkleBear]

Oh goodness... i finally was able to test a different transformer in there. It solved the tinny sound issue. So either there is something wrong with the transformer. I am going to stick the other transformer in there again just to see... but here is a test recording of the mic. The fet is only roughly biased to the 10-11v using a DMM. Will do the optimum fet bias procedure soon...

https://www.dropbox.com/s/stufi5padqkpgoo/DIY km84.mp3?dl=0
Shucks, that was a lot of mental gymnastics for such a simple fix.

Turns out that the first transformer (cm5722) that was installed had a bad primary winding.