DIY RF Condenser Mics

[shot]

So I finished my mic!
In fact, I've made two mics.

I've tried 2N5457 before J113 got delivered and it sounds same as J113.
Since I've socketed all the transistors, I had a chance to try different transistors in it. For Q2 and Q3 it didn't matter what I've installed. At first I went with matched pair od BC557C, and after that I tried 557B, 560B, 560B. Matched and unmatched. It just didn't produce any different result. For Q1 I didn't have BC549 so instead I went with BC547C. I've tried 547B, 550B, 550C and they all did work but there was audible difference in low mid frequency range. It was subtle, but definitely noticeable. I settled with 547C that sounded the best to me. Totally subjective here.

I noticed that two capsules that I bought on aliexpress were very different from each other. Actually I would dare to say that first capsule is almost defective! I have installed it in the first mic and I was annoyed how it produced intermittent noise, sounding as if I was rubbing head grille of the mic. The level was low-ish and a bit dark. But when I closed the mic body noise dissapeared. That's odd since this mic doesn't have problems with shielding since it's not super high impedance. Anyway, I thought that this dark quiet signal is the sound of this mic. But after I populated second pcb and assembled the second mic, that one blew me away. First of all it was louder and brighter. It sounds very different from the sample Rogs posted on his performance page. This one is open without any kind of compression, contrary to Rogs example that sounds a bit like radio DJ's mic processing! And this mic measured 5.4db louder than the first one.
After that I populated new pcb and installed it in the first mic - since I convinced myself that I did something wrong on the first board. But to my surprise it sounds just the same. So it must be due to capsule.

So I guess I'll order more capsules and match two mics since they can easily be tuned to act as a matched pair. But considering how cheap parts and donor body is, I bet I'll even order another body so I can have third mic.
And I'm not saying that my first mic sounds bad - I can see it on some bass cab or some loud sources like amp or drums.

This mic is a real refreshment after spending big bucks building brand name mics and hunting rare or expensive parts for them.  Just few weeks ago I built FET847 with chinese capsule and this mic sounds almost as good. I'll have to do a serious test in my studio when I find some spare time to compare them since on my voice they sound very very close!

Big up to Rogs, Khron and all of you guys for this little victory! Humans win!!



Luka

 

[rogs]

Thanks for posting this report Luka.... It's good to read that you've had some successful results..

I too have discovered that there can be quite a difference between capsules - even between those that are supposed to be the same type!

It's also useful to read that you have been able to try  out different semiconductors, and have discovered that - for the most part -  you are not restricted to very specific types.  The ones chosen for my schematic are all available from one UK  source - CPC -  which tend to have good prices and low delivery costs, hence my choice.

There are still some tweaks I'm trying out, to see if there are further mods to suggest, but it's good to know that the  present circuit has now been successfully completed by someone else!

 

[shot]

I went ahead to test RF.AMX mic even more!
Came to my studio and I recorded myself talking. Then as a comparison I also did two more recordings into FET847 and U87 clone.
It's all in this link:
https://1drv.ms/u/s!AuvQ2JGkf8-IhtlfDhBG4ML_kFFyzQ
All files are 48kHz/24bit WAV packed in 70mb ZIP file
(I'm not native english speaker so excuise my accent)

Rogs, if you find it interesting you can host it on your page. I'll also try to have this file hosted on my OneDrive as long as I can.

Bottom line - after more testing I find RF.AMX seriously good!



Luka

 

[rogs]

Thanks for posting these samples Luka ... 
I've uploaded the zip file to my server, and added a link to it among the audio samples at the bottom of the audio performance web site page here:  http://www.amx.jp137.com/index-perform.html

Really interesting to hear how the mic sounds on speech with a different voice  from mine!  I think it stands up pretty well against the other mics you've recorded samples of as well....
The hi-mid is probably a bit too prominent of course ... but that's just a feature of the K67 Chinese capsules we're all used to by now! 

 

[rogs]

I've been looking at optimising some of the component values when using alternative capsules to the one presently specified for the project  - which is a 'standard'  Chinese K67 style capsule.

I have a couple of 3 micron  edge terminated 34mm capsules, which have a much larger  capacitive value than the 65pF of the K67.
At around 90pF,  this capsule really needs several component changes to work optimally with the existing PCB...

The use of an 8MHz oscilllator in place of the present 10MHz - for example - helps increase performance by improving the 'Q' of the inductors, as they are operated closer to their specified  frequency..

I have also taken the opportunity to experiment with reducing the current drain from the phantom power by around 40%.

The performance using these 90pF capsules has improved by quite  a measurable  amount..... a signal to noise ratio of over 80dB, and a sensitivity of around -12dB.... (probably now a little too 'hot' for some applications). 

But reducing the sensitivity can also reduce the noise floor even further of course ... which is useful!

I've made some extra notes - and a new schematic - a copy of which you can find here:

www.jp137.com/lts/RF.AMX10.new.notes.pdf

 

[rogs]

Further to my post above - and to my notes - I have to confess that I cannot figure out the maths involved in working out the total loading presented to T1 primary by the inductor assembly as a whole ? -- ( As you can tell I'm no mathematician!)

I've attached a diagram of the relevant circuitry below..

It's easy enough to calculate the value of the series total of C4 and the capsule at around 47.4pF ..
But how does the introduction of T2  across the 'bridge' actually  affect the loading on T1?....

 

[abbey road d enfer]

Further to my post above - and to my notes - I have to confess that I cannot figure out the maths involved in working out the total loading presented to T1 primary by the inductor assembly as a whole ? -- ( As you can tell I'm no mathematician!)

I've attached a diagram of the relevant circuitry below..

It's easy enough to calculate the value of the series total of C4 and the capsule at around 47.4pF ..
But how does the introduction of T2  across the 'bridge' actually  affect the loading on T1?....

It is very complex; interaction between tuned circuits is counter-intuitive and maths are tidy and often not very informative. Calc methods are better applied. That's what simulation does.
The most significant factor IMO is that combining these circuits yields in addition to a low-Q primary resonance a secondary resonance at about 8MHz that is very narrow so the slope of the voltage vs. frequency curve is very sharp, giving high sensitivity.
The lack of accurate parasitic data makes the results quite error prone.
See attachment; Red is tank input without T2, green same with T2 and purple is output at T2's secondary.
Resuts are probably +100/-50% from reality because I neglected parasitic capacitances, resistances and losses, also made an assumption about coupling coefficient.

 

[rogs]

Thanks for that information Abbey .. and it seems to match up pretty well with a real world 'mock up' of the assembly I tried earlier, and helps makes sense of what I discovered.

It was using my original 10MHz design for the K67 capsule, so the 'Q' of the IF cans is likely to be slightly less than with an 8MHz  version (the coil data is specified for 7MHz operation).

I also need to make some allowance for some small scope lead loading error ( 10M and 15pF).

Even so, I can measure a gain of some 27dB between the input to T2 primary at the bridge junction, and the output of T2 secondary.

I think it's this kind of 'noise free' - 'Hi-Q' gain  that has made my results so much more impressive than I first expected from this simple circuit.

My thanks to the contributors on the original Yahoo Group Micbuilders thread who got me thinking about using IF cans and the benefits of  resonant 'Q' gain. (That would be Umashanker and Ricardo - thanks guys!)

 

[rogs]

Interesting new 'RF bias' mic from Rode -- the NTG5:  https://www.rode.com/microphones/ntg5

Looks very nice -  and seems to get good reviews!  The price is good too -- especially when you consider what you get with the whole package

Impressive spec -- Thought I'd compare some of the parameters to the latest version of the RF.AMX10  - (the one with the 90pF 'C12' style capsule)..

The sensitivity of this latest version of the AMX10 is probably a bit too 'hot' at c. -11dB , so I added a pad to match it to the sensitivity of the NTG5  at  c. -23dB.
I added a single 47R resistor across pins 2 and 3 of the XLR to achieve this, rather than fit RP1 onto the PCB. That way you 'pad' the noise from the output stage as well....
That mod brings the equivalent noise figure of the mic down to 4dBA  ( this was confirmed by checking against my Rode NT1, which also has a noise figure of 4dBA).

So ...

The good:

• The AMX10 now  has a signal to noise ratio of 90dB - against the NTG5 figure of 84dB
• The AMX10 is  around one tenth of the cost of the NTG5

The bad:

• The NTG5 has a weight of 76g against the 280g of the AMX10
• The AMX10 draws 4mA of  current against the NTG5 2mA

The ugly:

• The NTG5 is a very elegant short shotgun mic, with a length of 203mm  and a diameter of only 19mm
• The AMX10 still looks like a BM800 ! 


Still, at least the AMX10 signal to noise ratio is better!  ...

 

[abbey road d enfer]

An SDC cannot be as quiet as an LDC. SDC seems to be a requisite for shotgun mics; I must admit a 17x1.5" mic would be embarassing...

 

[rogs]

According to the Neumann notes here:  https://www.neumann.com/homestudio/en/difference-between-large-and-small-diaphragm-microphones  one of  the advantages of an SDC  is their excellent transient response and the 'reality' of the recorded audio. ...At the expense of slightly higher noise levels.

I think that most (all?) commercial RF bias mics are SDC shotguns ?...
The AMX10  is probably the world's first LDC  RF bias mic !

Not that there is any real demand for this type of mic  of course - but at least it is quiet!   

 

[abbey road d enfer]

According to the Neumann notes here:  https://www.neumann.com/homestudio/en/difference-between-large-and-small-diaphragm-microphones  one of  the advantages of an SDC  is their excellent transient response and the 'reality' of the recorded audio.

That is true. However, I see many people using LDC's for acoustic string instruments, where the difference is noticeable. I guess they love the presence peak in the upper midrange

I think that most (all?) commercial RF bias mics are SDC shotguns ?...

Since shotguns are often used in a difficult environment, RF bias makes sense.

The AMX10  is probably the world's first LDC  RF bias mic !

Indeed!

Not that there is any real demand for this type of mic  of course - but at least it is quiet! 

There could be a demand for a sub-zero noise shotgun...?

 

[rogs]

There could be a demand for a sub-zero noise shotgun...?

Well the AMX10 has a noise floor 10dB lower than the much vaunted EM172 Primo electret capsule,  so it might find an application for use outdoors - in nature recording  for example?....

It's not a shotgun mic of course, but the cardioid pattern is pretty impressive in the end address configuration....

 

[gyraf]

thanks

 

[rogs]

The main project schematic and parts list website pages have both been updated to version 2 (v.2)
This includes the  current  reduction mods originally made to the 8MHz (90pF capsule) version, as described in the posts above.
Links to the previous schematic and parts list are included on the new pages for anyone already working with the original version.

The project website URL remains the same: www.amx.jp137.com

 

[rogs]

EDIT: Please see post #378 below -- the inclusion of  R13 47R as an attenuator between pins 2 and 3 of the XLR can cause distortion with some pre-amps, and has now been removed.
C5 and C10 are now 220nF.
Please see the revised schematic attached to post #379
The main project notes have been revised.

 

[Khron]

Increasing R5/6 above 100k is not an option? Or what (detrimental) effects might that have?

 

[rogs]

I don't think  changing those for - say - 150k  resistors would make that much difference to the input impedance.
It's only going to be around 20k with a 100R  output load, using those transistors - so 100k resistors  are going to make that around 16.6k  150k resistors  would make it around  17.6k, so probably not really that  significant.

I do think I may need to investigate further whether it is a good idea to fit  R13, the 47R resistor across the output.
It works well enough as an attenuator, but I'm finding an increase in distortion figures at higher input levels using a separate test circuit
It looks as if  some pre-amps may not like having the differntial inputs - and the 2 x 6.8k phantom power resistors - effectively bridged with a 47R resistor...

It may be necessary to leave that out  and go back to adding the 'pad' resistors RP1 and RP2 in required.
Not quite such low noise floor of course..

It's not easy to test this whole circuit by simply substituting the capsule, as it is with 'standard' Schoeps type circuits -- I may need to try and knock up a test rig using a varicap diode.....?...

Still playing - and still finding out new things about this concept.. 

 

[rogs]

Further to my 2 posts above,  I have now realised that the inclusion of R13 (47R) across pins 2 and 3 - whilst working fine as an attenuator - can cause distortion with some preamps. I have thus removed it from the schematic.
The sensitivity is returned to around  -12dB.
C5 and C10 are now  fitted as 220nF to lower the LF -3dB point to 20Hz with Q2 and Q3  BC557B transistors, working into a 300R load.
The main project schematic has been revised.
Sorry for any confusion... 

 

[rogs]

I mentioned the new Rode NTG5 mic: https://ntg5.rode.com in my earlier post #368 above. 
Looks like a very nice mic - and the cheapest I've seen so far of the various commercial RF bias mics on the market.
(Not that many of course!)

I was impressed with their phantom power current drain figure of only 2mA, and wondered how well I could get the RF.AMX10 to perform only drawing that much current? -- The answer is, not too badly....

I had previously mentioned that the latest version of the AMX10 is a pretty 'hot' mic, with a sensitivity of  around -12dB. 
I wondered how the overall noise figure would be affected, if I reduced the sensitivity by around 10dB by reducing oscillator current.
A few quick experiments showed that the noise figure would only be degraded by around 5dB.

I've attached a schematic of a  version that  pretty much copies the NTG5, spec wise -- although the AMX10 does have an more extended LF response.
It's also - sadly - still much heavier and uglier. 
...But it is LOT cheaper 

By changing the values of  R3, R11 and R12 you can get to the required  spec.. a sensitivity of around -22dB, a total current drain of 2mA - and a S/N ratio of c.85dB.

You'll notice D1 has been re-introduced as a 36v zener. That's to protect Q4, which has a maximum Vgd of 35V. 
I found it necessary when using my dbx286s channel strip pre-amp, which has a phantom power voltage of  52V  (right on the upper limit of the phantom power spec)

I know we have discussed current drain at various times before in this project thread ... so I'm quite pleased to have been able to get it down to 2mA 

( Higher resolution PDF of the schematic here: www.jp137.com/lts/AMX10v.2.2mA.pdf  )