Discrete input stage for Mic Pre

[sbeach]

Hi,

I'm new to posting here but I've been lurking for quite awhile. I have been designing & building mic preamps for a few years - starting with basic opamp gain blocks - but lately have been trying to learn about & incorporate more discrete component stages. Here's a mic input stage design I'm trying to get a better understanding of ...

http://groupdiy.twin-x.com/displayimage.php?pid=1523&fullsize=1

I lifted the JFET differential input from a discrete opamp input posted here. Is this a legitimate way to use this differential-to-single-ended stage in place of an input transformer? If so, is there a way to determine and/or adjust the gain for this stage? I would like to be able to choose between 5 to 25 db of gain here if possible. Would the gain be adjustable by increasing or decreasing the 1k resistor on the JFET drain? What would be the problems in using such an approach? Crappy inteference rejection? High noise? I imagine the actual JFETs used would greatly influence the gain available.

In a complete preamp design the following stages could add from 30 to 50 db of gain but for clarity & simplicity I have left that part out.

Any ideas, comments or helpful suggestions are very welcome. I really want to learn this stuff. Thanks,

Skip Beach

 

[PRR]

> Is this a legitimate way to use this differential-to-single-ended stage in place of an input transformer?

Who would accuse you of being illegitimate?

The general objection to FET inputs for low-level mike inputs is that you can get much lower noise with a transformer or BJTs. But JFETs are getting better. And typical mikes (and acts!) are a lot louder than The Olde Days, so noise may be a non-issue.

> Would the gain be adjustable by increasing or decreasing the 1k resistor on the JFET drain?

Affects gain AND maximum output level.

Cathode resistors (whadya call em... source resistors) reduce gain without reducing maximum output level, however they increase noise.

> I imagine the actual JFETs used would greatly influence the gain available.

With older FETs... not a lot. Gm was all over the place, and Idss was all over the place, but when you bias them all to the same current, Gm didn't vary much, within a type and not much more for assorted types.

> In a complete preamp design the following stages could add from 30 to 50 db of gain but for clarity & simplicity I have left that part out.

But the Classic way to get best overall performance from an input like this is to integrate it inside the overall feedback loop. Then you can reduce gain without (so much) increase in noise due to large cathode resistor values.

That input network was stolen from a BJT design. BJTs need large base coupling caps to swamp current noise. A JFET has no input current noise. The 47uFd could be 0.01uFd, the 2.2K could be 4.7Meg.

In hi-RF environments, you want more than 200pFd across the inputs. The mike should stand thousands of pFd, so go 470pFd or 1,000pFd. Also 470pFd from each input to ground. And 47 ohms or a small (uH) coil in series with each input, to give the caps some leverage. Yes, in my rooms I don't need any of that, but friends in the city can't work without massive RFI suppression.

You show the same ground symbol everywhere. In actual wiring, the XLR pin 1 must NOT run anywhere near the amplifier stages, because it is full of crap. It goes to the chassis, the chassis flows to the rack or ground-bus, and the amplifier stages have their own separate ground chain.

 

[Rossi]

[quote author="PRR"]
That input network was stolen from a BJT design. BJTs need large base coupling caps to swamp current noise. A JFET has no input current noise. The 47uFd could be 0.01uFd, the 2.2K could be 4.7Meg.
[/quote]

But what about the phantom power resistors? In a transformerless design, they limit the input impedance to a max of 13.6K. For a hipass cutoff frequency of, say, 10 Hz the input caps would have to be at least about 1uF.

 

[clintrubber]

[quote author="Rossi"]
But what about the phantom power resistors? In a transformerless design, they limit the input impedance to a max of 13.6K. For a hipass cutoff frequency of, say, 10 Hz the input caps would have to be at least about 1uF.[/quote]
HPF as formed by the 6.81k & the input caps ? The phantom-power-resistors are before the caps, so no HPF (if I interpret what you wrote correctly).

Regards,

Peter

 

[SSLtech]

[quote author="Rossi"]But what about the phantom power resistors? In a transformerless design, they limit the input impedance to a max of 13.6K. For a hipass cutoff frequency of, say, 10 Hz the input caps would have to be at least about 1uF.[/quote]
They would... if the 13.6k was the OTHER side of the capacitors! -Fortunately for us, the phantom has to be fed to the SOURCE side of the cap in order for the mic to drink the DC. The cap has the option of discharging through the 4.7 meg resistor, or the gate of the FET. -Either way is going to take a L-O-N-G time, so you're still good down well outside the audio band.

Keith

 

[clintrubber]

Hey Keith, shouldn't we better be listening to Serge i.s.o. nitpicking on Rossi's HPF-calculation ?! :thumb: :wink:

 

[SSLtech]

..I AM listening to Serge!
(Ford Mustang... BANG!")

Keith

 

[clintrubber]

[quote author="SSLtech"]..I AM listening to Serge!
(Ford Mustang... BANG!")

Keith[/quote]
:thumb: :grin:

Would you believe me if I told you I was listening to this one ?

Monsieur Gainsbourg Revisited by Various Artists (Polydor)
Packed with sultry interpretations from celeb fans such as Franz Ferdinand, Jarvis Cocker and Marianne Faithfull, this fabulous tribute to the sexy French lothario is a labour of love. Serge died 15 years ago, but he must be licking his lips up there in the sky over I Love You (Me Either), a lesbian translation of his Je T'aime? with Cat Power getting smoochy with model Karen Elson. Luxurious.

:thumb:

Peter

 

[Rossi]

D'oh!

Yeah you're right. The phantom power resistors do in fact limit the impedance the mic sees but they don't alter the HPF frequency set by the input caps. Late night math, I guess. :?

 

[sbeach]

Thank you for the valuable information PRR, although still ignorant, I am a good bit better informed thanks to you :grin: I have revised the circuit a bit & am posting another version for general perusal & also wish to pose a few additional questions.

http://groupdiy.twin-x.com/displayimage.php?pid=1532&fullsize=1

We've gone to a BJT for the input pair, maintaining the JFET current source, and have incorporated a negative feedback loop - just for this stage - with a 2x gain resistor in the loop. The subsequent gain stages (not shown) are both simple JFET amplifiers not unlike the well-known Hamptone circuit so I did not wish to use over-all global feedback for this project.

This differential circuit is most like a basic "opamp" 1st stage, but without the 2nd voltage gain stage & the current output stage. A stripped, bare-bones differential input amp. I felt I didn't need much gain here, plus the next stage has a 10MegOhm input impedance so very tiny current requirements precluding the need for a current boosting stage.

I originally wanted to see if I could come up with a way to unbalance the mic signal, add a little frontend gain and pass the signal along VERY cleanly. The challenge was to do this electronically without using an input transformer, or an instrumentation amp (SSM2019), or a commercial opamp in differential amp configuration.

Does this revised version of the circuit make good sense theoretically, given the aims of this project? Would it be just as well (or better) to go ahead and use regular opamps - either as an instrumentation amp or in a simple differential amp configuration?

I've often wondered why the basic opamp differential amplifier configuration doesn't get used for mic inputs very often? Other than DaviSound, Rane and Paia are the only companies I've seen use this approach for microphone inputs. I'd be curious to know anyone's thoughts on these matters.

Thanks,
Skip Beach

 

[SSLtech]

That's starting to remind me of some (Neotek?) mic pre that I saw a few years ago... Interesting, but you've GOT to have DC trims that don't drift with temperature. -For things like location recording, the temperature will go up and down more often than the hind leg of an incontinent dog, and if DC offset gets into the subsequent gain adjustment, it's a real irritation...

Just a thought... the pres that I worked on that looked like that had read DC-drift issues, though I never took the time to dig into why they were so irksome...

Keith

 

[clintrubber]

[quote author="SSLtech"]http://www.dustygroove.com/images/products/g/gainsb_serg_initialss_101b.jpg

Keith[/quote]

:razz: :thumb:

 

[recnsci]

Hi Skip and others,

first thing, second ckt you have posted will either be lousy oscilator
or lousy hysteretic comparator, cus you have positive feedback going
on, so back to drawing board.
Second thing, if we presume that magicaly feedback somehow
works, that balancing scheme (with 2.4k||47p in base of left transistor)
would not work. Thing is, common emiter stage thus diff amp has
guite large output impedance (1.5k in your case to be precise) that has
to be included in impedance calculations, and feedback change impedances
in circuit considerably. Take analogy with opamp based diff amp. If you
have R1 from input to negative input of opamp, and R2 from negative
input of opamp to opamp out, impedance you see is not R1+R2, its
just R1. And any debalancing scheme like this will depend strongly on
perfect symetry of circuit.
Plus debalancing toplology like this one will
need verrrry clean power supply (like, at least 80 dB cleaner than
lowest signal at collector).

cheerz
uros

 

[Boswell]

sbeach wrote:

We've gone to a BJT for the input pair, maintaining the JFET current source, and have incorporated a negative feedback loop - just for this stage - with a 2x gain resistor in the loop.

Where did you get that circuit?
The 2.4K is connected as positive feedback. Even if it were connected to the other base, it would destroy any d.c. balance conditions in the input pair. You don't really have enough gain in a differential pair like this to employ feedback round this stage.

 

[clintrubber]

Hey, when you post, then please don't forget to spoil this thread by linking to a Gainsbourg LP-cover ! :wink: :thumb:

 

[burdij]

I think I need to differ with a statement that PRR made about noise in FETs.

The general objection to FET inputs for low-level mike inputs is that you can get much lower noise with a transformer or BJTs. But JFETs are getting better. And typical mikes (and acts!) are a lot louder than The Olde Days, so noise may be a non-issue.

While it may be true that you can achieve some low noise gain advantage using a transformer, the case with jFETs and bi-polars is different. Noise performance in a junction FET will be better than the noise performance of a bi-polar transistor in a microphone preamp circuit (and in other circuits, like your satelite receiver antenna LNA) if presented with a relatively high impedance source. For low impedance sources, you may be able to come up with a circuit using bi-polars that will have somewhat better noise performance than with a comparable circuit using jFETs but is not sufficiently better to justify replacing the FETs. In your circuit, you might want to consider using a better FET model, one that is specifically designed and selected for low noise performance.

The semiconductor theory behind this phenomenon, as I understand it, is that the jFET has only one type of charge carrier as opposed to the bi-polars two charge carriers (the electrons and the "holes").

 

[rafafredd]

2sk30´s are good for low level low noise audio, but nothing compres to the famous 2sk170. That´s the only one! Or maybe two sections of 2sk389 in paralell, just like Forssell does in his JFET/opamp circuit.

 

[burdij]

The 2sk170 is a good choice. I was thinking of the Vishay/Siliconix low noise jfets like J202. The advantage of these, as someone pointed out here once, I think, is that the cutoff voltage range is higher than the 2sk170 (originally developed for smoke alarms) so the input can handle higher levels.

2sk170 cutoff range -.2v > -1.5v equiv. noise 1nV/Hz**.5
Siliconix J202 cutoff range -.8v > -4V equiv. noise <5nV/Hz**.5

For a low level app. like a condenser microphone preamp, where the transducer output is small, the 2sk170 would be an excellent choice.

Newark Electronics http://www.newark.com has the J202 in TO92 for $.44 in onesies.

 

[PRR]

> A stripped, bare-bones differential input amp.

The standard microphone's noise resistance is about 200 ohms.

With this plan, the feedback adds 800 ohms each side, or 1,600 ohms.

Noise voltage will be twice as high.

You'll need to block DC to the feedback network (ignoring the fact it is drawn positive). As-is, the feedback wants to force Q1's collector to zero volts, while the base is also zero volts. A transistor can work that way, but speed is low and output swing is low.

> why the basic opamp differential amplifier configuration doesn't get used for mic inputs very often?

The 1-opamp 4-resistor scheme is always noisy. The input resistors add to the source noise. If you use the usual rule of loading with 10 times the source impedance, noise voltage rises by a factor of 3 or 10dB. If you "match", noise is 6dB worse than a perfect preamp. "Pro" gear routinely gets noise figures within 2dB of ideal.

The 3-opamp "instrumentation amp" does not have this flaw. Most non-transformer mike inputs are based on this general plan. The shunt feedback resistor between the feedback inputs must be lower than mike impedance or it adds noise; at low gains, this can imply serious loading on the outputs.

> I felt I didn't need much gain

Low gain and low noise (comparable to low-Z mike noise) is a VERY tough problem.

You might do well to build two of the Hamptone modules, and a diff-amp combining their outputs. You sure get a hi-Z input. You won't get 60dB CMRR, but often 20dB CMRR is all you need.

If you continue with the diff-pair: forget feedback to the bases. Split the emitter current source into two resistors or FETs, one per amplifier device. Put an adjustable resistor between the emitters. For output resistor of 1.5K, a 740 ohm resistor will give gain of 2, a 140 ohm resistor will give gain of 10.

> I need to differ with a statement

Above 1K, either BJT or FET will be fine.

But 200 ohms is tough for any practical device.

For most of my life, JFETs were nowhere near competitive with transformers or rich BJT designs.

Yes, a few JFETs now challenge BJTs for 200 ohm audio noise figures. FETs have many other advantages, and low noise is not always the important goal. However, our selection of JETs is small and getting smaller, and the old-type RF JFETs will not give low-low-noise against 200 ohms.

> The semiconductor theory behind this phenomenon, as I understand it, is that the jFET has only one type of charge carrier as opposed to the bi-polars two charge carriers (the electrons and the "holes").

Yes, but for the same area and similar current, a BJT will always have higher Gm and lower noise than a field-effect device such as a tube or a JFET. In vacuum tube terms, you want the grid as close to the cathode as possible. Tubes and JFETs use a "field" with dimensions large compared to an atom. The BJT's "grid" is a one-atom wide junction between P and N. If you let the JFET's gate area grow (or parallel a lot of devices) the Gm and Nv can be low, but input capacitance and leakage will be higher than a BJT.