FET847 Microphone Project Circuit By Jonathan Burtner PCB Layout By Poctop
- Thread starter poctop
- Start date
Help Support GroupDIY Audio Forum:
0dbfs
Well-known member
This is a single FET driving a transformer. It's not going to have a huge output voltage. This is essentially the same as a U87 without the de-emphasis which is also noted for "low output".
Consider that this might be a desirable quality in some regards because for one thing less gain in the mic causes you to use more gain in your pre-amp which often is achieved by reducing negative feedback and thus simply makes the signal "more better" or my other favorite technical term "Magic-Microphone".
A PCB jumperable (or switchable) capacitive pad could be useful if you have a set of these mic's dedicated for close mic drum recording... But you could also simply solder it in place and dedicate those mic's for close mic drum duties...
Cheers!
-johann
Consider that this might be a desirable quality in some regards because for one thing less gain in the mic causes you to use more gain in your pre-amp which often is achieved by reducing negative feedback and thus simply makes the signal "more better" or my other favorite technical term "Magic-Microphone".
A PCB jumperable (or switchable) capacitive pad could be useful if you have a set of these mic's dedicated for close mic drum recording... But you could also simply solder it in place and dedicate those mic's for close mic drum duties...
Cheers!
-johann
0dbfs
Well-known member
Khron said:
The backplate, R2 (1G), and C11 all connect to form a node. This is hi-z node for DC signals and low Z for AC signals and should be floating or terminated at a teflon standoff terminal.
Hmmm... Ok. If the capsule bias is reduced =by half then we lose 6db of signal. When it's @ 48VDC then bringing it up to a "prescribed" 60VDC will provide better signal to noise ratio but not a full 6db more. This is true and 60V is what we use standard in many other mic's. Vintage U87 uses this "48V" method so we know it's a solution that worked fine during that particular run of the mic.Khron said:And besides, 1G + 100nF gives you a 100-second RC constant, which means the capsule bias voltage will only reach its maximum in about five times as long (over 6 minutes)
----
I recommend for DIY'ers to experiment with variable capsule voltage by temp modifying a tube mic using variable rear capsule voltage as the pattern select - discnect capsule front into the grid so that it's only using the rear diaphragm then adjust your pattern to increase/decrease bias voltage and see how that changes the sound... COmpare 48V, 60V, 30V, etc...
----
Not sure I understand what you did with re-working the PCB but that's great if it works. This circuit is a simple basic building block which has certainly been implemented in other mic's. The original design goal with this implementation was to make a known good high quality build available with a minimum of extra features and a BOM order from one vendor for components other than capsule, transformer, body.
A KM84 modified for LDC use with the idea that more of them in your locker is just better and you can throw one on just about anything...
You can do a lot from this starting point to modify for your needs. Did you say your PCB is single layer single sided? Will you upload the gerbers and masks so full on DIY'ers can self etch drill and modify further if desired?
Cheers and good luck to all!
-jb
Just to try to make my point of view as clear as i can, i've attached a schematic with the two variants of input circuitry.
In SDC's, mechanics / physical constraints force the construction of the capsule to have one side of it inevitably connected to ground (see the left side, as in the KM84, Schoeps, MXL603, whatever other non-electret SDC's you can think of).
So in that case, there's nothing that can be done to change the topology.
Now, many of the cheap chinese LDC's use the same Schoeps circuit, BUT in LDC's we're NOT forced to have one side of the capsule unavoidably connected to ground. I can't be 1000000% sure about it, but this, as well as the desire to skip over the input coupling (or DC blocking) cap, is likely what inspired Zapnspark to come up with his "ChinaMicMod" alteration.
I'm not sure whether it's moral or ethical to upload / publish his actual schematic of the thing here, but it's definitely over on the micbuilders Yahoo group. But i've redrawn the essentials on the right side of the attached screenshot.
Now, power supplies are pretty much universally low-impedance sources, correct? So how and why would the R2/C11 node (need to) be a high-impedance one? This is on the "back" side of the capsule, which is, in essence, a capacitor itself.
In the SDC version, with one side of the capsule to the ground, the other side, of course, needs a bias voltage across it (since the signal voltage is produced from the varying capacitance of the capsule, and without a bias voltage to modulate, you would get no signal). That then gets coupled through C3 (which also blocks the high bias voltage from "toasting" the JFET) in my schematic here, over to the gate.
But in the LDC version, since we have both sides of the capsule "free", we can apply the bias voltage (which "should" come from and be, a relatively low-impedance source) to the backplate, and use the capsule itself (which is a capacitor, remember?) AS the input DC-blocking capacitor.
It's worthy to note, though, that between the two variants of this input topology, the phase is reversed (since the bias voltage is applied to the other side of the capsule).
Is any of this making any sense whatsoever? ;D
Regarding the time constant, i never mentioned anything about 48v, or 60v, or any absolute value of the bias voltage. All i was saying was, if you read more carefully, that with that 1G / 100nF filter, it would take something in the region of 500 seconds (SIX minutes) for the capsule bias voltage to reach whatever final / maximum voltage it can.
You DO know what the RC constant is, right? http://www.referencedesigner.com/rfcal/cal_05.php
In SDC's, mechanics / physical constraints force the construction of the capsule to have one side of it inevitably connected to ground (see the left side, as in the KM84, Schoeps, MXL603, whatever other non-electret SDC's you can think of).
So in that case, there's nothing that can be done to change the topology.
Now, many of the cheap chinese LDC's use the same Schoeps circuit, BUT in LDC's we're NOT forced to have one side of the capsule unavoidably connected to ground. I can't be 1000000% sure about it, but this, as well as the desire to skip over the input coupling (or DC blocking) cap, is likely what inspired Zapnspark to come up with his "ChinaMicMod" alteration.
I'm not sure whether it's moral or ethical to upload / publish his actual schematic of the thing here, but it's definitely over on the micbuilders Yahoo group. But i've redrawn the essentials on the right side of the attached screenshot.
Now, power supplies are pretty much universally low-impedance sources, correct? So how and why would the R2/C11 node (need to) be a high-impedance one? This is on the "back" side of the capsule, which is, in essence, a capacitor itself.
In the SDC version, with one side of the capsule to the ground, the other side, of course, needs a bias voltage across it (since the signal voltage is produced from the varying capacitance of the capsule, and without a bias voltage to modulate, you would get no signal). That then gets coupled through C3 (which also blocks the high bias voltage from "toasting" the JFET) in my schematic here, over to the gate.
But in the LDC version, since we have both sides of the capsule "free", we can apply the bias voltage (which "should" come from and be, a relatively low-impedance source) to the backplate, and use the capsule itself (which is a capacitor, remember?) AS the input DC-blocking capacitor.
It's worthy to note, though, that between the two variants of this input topology, the phase is reversed (since the bias voltage is applied to the other side of the capsule).
Is any of this making any sense whatsoever? ;D
Regarding the time constant, i never mentioned anything about 48v, or 60v, or any absolute value of the bias voltage. All i was saying was, if you read more carefully, that with that 1G / 100nF filter, it would take something in the region of 500 seconds (SIX minutes) for the capsule bias voltage to reach whatever final / maximum voltage it can.
You DO know what the RC constant is, right? http://www.referencedesigner.com/rfcal/cal_05.php
Attachments
thx,
the case was about "warming up" mic and waiting for raising capsule voltage - in first test i was recording just after enabling phantom voltage wchich resulted in low gain.
I was not aware of such high value of RC contant - thanks again for clarifying this
I did my clone on veroboard, so i don't need pcb design.
One more question about C1 - if i understand it correctly it makes negative feedback - when value of this cap is lower it should give me slightly more gain and bass - it is right?
the case was about "warming up" mic and waiting for raising capsule voltage - in first test i was recording just after enabling phantom voltage wchich resulted in low gain.
I was not aware of such high value of RC contant - thanks again for clarifying this
I did my clone on veroboard, so i don't need pcb design.
One more question about C1 - if i understand it correctly it makes negative feedback - when value of this cap is lower it should give me slightly more gain and bass - it is right?
Henry Spragens did a pretty nice write-up on the KM84 circuit over on his blog:
http://audioimprov.com/AudioImprov/Mics/Entries/2015/4/23_Basic_FET_Microphone_Circuits.html
Scroll down to the second (coloured) schematic
And thanks for confirming my hunch
http://audioimprov.com/AudioImprov/Mics/Entries/2015/4/23_Basic_FET_Microphone_Circuits.html
Scroll down to the second (coloured) schematic
WrmZ said:
And thanks for confirming my hunch
WrmZ said:
I'll attach the gerbers one of these days, after i "polish" the component silkscreen and export the files, but at least for now, here's my suggestion of a Mouser BOM, with more... shall we say, "down to earth" component choices? :
1.48eu for each (!!!) of the two temporary testing capsule-replacement capacitors; same for each of the two output filter inductors- are you friggin' kidding me???
Ok, deep breath, calming down....
1.48eu for each (!!!) of the two temporary testing capsule-replacement capacitors; same for each of the two output filter inductors- are you friggin' kidding me???
Ok, deep breath, calming down....
Code:
jumper 571-1-881545-4 1pc
3-pin header 710-61300311121 1pc
68p test-capsule 810-FG18C0G1H680JNT0 2pcs
2.2k 1% 603-MFR-25FBF52-2K21 10-20pcs, hand-match
SPDT switch 642-TL36W000000 1pc
100n film caps 505-MKS2D031001A00MC 3pcs
1u output cap 505-MKS2C041001FMSSD 1pc
4.7u source cap 647-ULD1H4R7MDD1TD 1pc (option, less bass)
10u caps 667-ECA-1HM100I 2-3pcs
22u cap 667-ECA-1VM220B 1pc (option, more bass; could be used to replace C4 & C6 too)
47k resistors 603-MFR-25FBF52-47K5 2pcs (replacing the 56k R6)
10k 603-MFR-25FBF52-10K 1pc
4.7p C0G feedback 810-FG18C0G2A4R7CNT6 1pc (option)
4p mica feedback 598-CD15CD040DO3F 1pc (option)
10M resistors 279-CFR25J10M 3pc (replacing the 1G R2)
25k trimmer 858-67WR25KLF 1pc
JFET 610-2N3819 1pc
pin turret 534-11218 1pc
1n caps 505-FKS2D011001AKC00 2pcs (replacing the 330p ones)
47u inductors 542-78F470-RC 2pcs
1G resistor 279-RGP0207CHK1G0 1pc (1G R2 unnecessary)
24v zener 512-1N5252BTR 1pcKhron said:I'll attach the gerbers one of these days, after i "polish" the component silkscreen and export the files, but at least for now, here's my suggestion of a Mouser BOM, with more... shall we say, "down to earth" component choices? :
1.48eu for each (!!!) of the two temporary testing capsule-replacement capacitors; same for each of the two output filter inductors- are you friggin' kidding me???
Ok, deep breath, calming down....
Code:jumper 571-1-881545-4 1pc 3-pin header 710-61300311121 1pc 68p test-capsule 810-FG18C0G1H680JNT0 2pcs 2.2k 1% 603-MFR-25FBF52-2K21 10-20pcs, hand-match SPDT switch 642-TL36W000000 1pc 100n film caps 505-MKS2D031001A00MC 3pcs 1u output cap 505-MKS2C041001FMSSD 1pc 4.7u source cap 647-ULD1H4R7MDD1TD 1pc (option, less bass) 10u caps 667-ECA-1HM100I 2-3pcs 22u cap 667-ECA-1VM220B 1pc (option, more bass; could be used to replace C4 & C6 too) 47k resistors 603-MFR-25FBF52-47K5 2pcs (replacing the 56k R6) 10k 603-MFR-25FBF52-10K 1pc 4.7p C0G feedback 810-FG18C0G2A4R7CNT6 1pc (option) 4p mica feedback 598-CD15CD040DO3F 1pc (option) 10M resistors 279-CFR25J10M 3pc (replacing the 1G R2) 25k trimmer 858-67WR25KLF 1pc JFET 610-2N3819 1pc pin turret 534-11218 1pc 1n caps 505-FKS2D011001AKC00 2pcs (replacing the 330p ones) 47u inductors 542-78F470-RC 2pcs 1G resistor 279-RGP0207CHK1G0 1pc (1G R2 unnecessary) 24v zener 512-1N5252BTR 1pc
Added all costs down and its about 10-12 euros, depends on some personal choices and options !!
Amazing kudos khron
I think i'm pretty much done with making the board "pretty". And by "board" i mean "silkscreen" - surely the part values are enough, and component designations aren't necessary (R2, C6 etc), right?
For the record, the board doesn't even need to have that tapered shape; my BM800 boards aren't, and they fit just fine. The only catch is to not have any component leads too close to the outside edges of the board (so they don't clash with the chassis beams).
Which Gerber layers / files are required? So i know which ones to export...
[Later edit, re: component value alterations]
With 3x 10M and 3x 100nF filtering for the capsule bias voltage, it should reach its maximum in around "only" 45sec or so
The difference between R6 being 56k vs. 47k is literally minimal. I simulated a worst-case load in LTspice - a 48v source, with 4.5kohm series resistance (the 2x 6.8k in parallel inside the preamp, in series with the 2x 2.2k in the mic), and a 57k load (the 47k drain resistor plus the 10k R5 that's part of the power-filtering, a zero source resistor and the JFET being a dead short).
Stock current draw is 408uA, and 462uA with 47k. Voltage across the zener diode (ie. before R5 in the schematic) is 23.28v stock vs. 24.2 with 47k. Current through the load is 408uA stock (likely because the voltage is lower than the 24v of the zener, so it's NOT actually doing any regulating; seems to act more like a voltage clamp / limiter) vs. 424uA with 47k.
[Yet another edit, this time on the PCB]
I noticed / remembered the length (longest dimension) of the board was 52mm or so, so i moved the top half a couple ticks lower, to fit it within (ie. ever so slightly under) 50mm. Board manufacturers usually have different prices for "up to 5x5cm boards" versus "up to 5x10cm boards" Just a(nother) little cost-saving measure, with the side-effect of getting an even more compact board, and as such, shorter signal trace lengths ;D
For the record, the board doesn't even need to have that tapered shape; my BM800 boards aren't, and they fit just fine. The only catch is to not have any component leads too close to the outside edges of the board (so they don't clash with the chassis beams).
Which Gerber layers / files are required? So i know which ones to export...
0dbfs said:
[Later edit, re: component value alterations]
With 3x 10M and 3x 100nF filtering for the capsule bias voltage, it should reach its maximum in around "only" 45sec or so
The difference between R6 being 56k vs. 47k is literally minimal. I simulated a worst-case load in LTspice - a 48v source, with 4.5kohm series resistance (the 2x 6.8k in parallel inside the preamp, in series with the 2x 2.2k in the mic), and a 57k load (the 47k drain resistor plus the 10k R5 that's part of the power-filtering, a zero source resistor and the JFET being a dead short).
Stock current draw is 408uA, and 462uA with 47k. Voltage across the zener diode (ie. before R5 in the schematic) is 23.28v stock vs. 24.2 with 47k. Current through the load is 408uA stock (likely because the voltage is lower than the 24v of the zener, so it's NOT actually doing any regulating; seems to act more like a voltage clamp / limiter) vs. 424uA with 47k.
[Yet another edit, this time on the PCB]
I noticed / remembered the length (longest dimension) of the board was 52mm or so, so i moved the top half a couple ticks lower, to fit it within (ie. ever so slightly under) 50mm. Board manufacturers usually have different prices for "up to 5x5cm boards" versus "up to 5x10cm boards"
Just a(nother) little cost-saving measure, with the side-effect of getting an even more compact board, and as such, shorter signal trace lengths ;DAttachments
0dbfs said:
While taking a look around, i came across the D-U87 thread, and i was quite surprised to see this:
https://cdn.groupbuilder.com/groupdiy/u/39511/58d1402a02187.jpg
I don't see any teflon stand-offs for those 1G resistors, soooo... What's the deal there?
1G resistors, but no high-impedance nodes to worry about?
Khron said:While taking a look around, i came across the D-U87 thread, and i was quite surprised to see this:
https://cdn.groupbuilder.com/groupdiy/u/39511/58d1402a02187.jpg
I don't see any teflon stand-offs for those 1G resistors, soooo... What's the deal there?
Some mikes don't have it - works without problem, some have just teflon isolator (like for transistor heatsink mount), and some have (very clever) just legs in the air (gate and resistor + capsule wire).
Khron said:Which Gerber layers / files are required? So i know which ones to export...
... (snip)
I noticed / remembered the length (longest dimension) of the board was 52mm or so, so i moved the top half a couple ticks lower, to fit it within (ie. ever so slightly under) 50mm. Board manufacturers usually have different prices for "up to 5x5cm boards" versus "up to 5x10cm boards"
I think that we need the full zip file of the gerbers for using with board manufacturers.
For self-etching i thnik the bottom layer would be sufficient, but others with more experience could help better with that aspect.
tip: to upload gerbers on this site you would need to change the extension, to for example .pdf or .png in stead of .zip
Thanks for doing this
"Of course" i'd put them all in a zip file - i'd have to be almost insane to attach each file to a separate post ;D
Yes, that limitation's quite annoying, but good thing there are ways to "bend" that rule...
Well, since mine's a single-layer board, i'm guessing (in a physical order):
- bottom soldermask
- bottom copper
- top soldermask(?)
- top silkscreen (component values only, no designators, ie. R3, C6 etc, to not overcrowd the board - it's full of components )
And no real thanks needed - this project was recently brought (back) to my attention, and when i took a closer look at the board, i took it as a challenge to considerably improve the layout
Some might argue i am challenged, but... that's a story for another time ;D
Yes, that limitation's quite annoying, but good thing there are ways to "bend" that rule...
Well, since mine's a single-layer board, i'm guessing (in a physical order):
- bottom soldermask
- bottom copper
- top soldermask(?)
- top silkscreen (component values only, no designators, ie. R3, C6 etc, to not overcrowd the board - it's full of components
)wlinart said:I think that we need the full zip file of the gerbers for using with board manufacturers.
For self-etching i thnik the bottom layer would be sufficient, but others with more experience could help better with that aspect.
tip: to upload gerbers on this site you would need to change the extension, to for example .pdf or .png in stead of .zip
Thanks for doing this
And no real thanks needed - this project was recently brought (back) to my attention, and when i took a closer look at the board, i took it as a challenge to considerably improve the layout
Some might argue i am challenged, but... that's a story for another time ;D
Khron said:"Of course" i'd put them all in a zip file - i'd have to be almost insane to attach each file to a separate post ;D
Yes, that limitation's quite annoying, but good thing there are ways to "bend" that rule...
Well, since mine's a single-layer board, i'm guessing (in a physical order):
- bottom soldermask
- bottom copper
- top soldermask(?)
- top silkscreen (component values only, no designators, ie. R3, C6 etc, to not overcrowd the board - it's full of components
What i actually meant to write was: I think that for using a manufacturer of pcb's, like you suggested, we would need the full gerbers, with all the layers. But of course i could be wrong about that.
As far as i know, and depending on what boards you order (single or double sided), manufacturers only use / apply the layers you provide them. Some even have options for silkscreen on one side only, or both sides (which may or may not cost extra), etc.
That's why i gave that example of a layer list
For the record, Eagle (the CAD i use) has dozens of layers one can use, but i doubt that's necessary in this case ;D
That's why i gave that example of a layer list
For the record, Eagle (the CAD i use) has dozens of layers one can use, but i doubt that's necessary in this case ;D
wlinart said:
