All OpAmp mic design (no FET at first stage)

[soliloqueen]

Which is why interpreting measurement results is probably even more important than the measurement alone. Caveat of what you are describing is the fact these effects most of the time get abused by marketing, claiming there's something more to the sound than the measurements can show. And don't get me wrong, there are many things measurements can't show, but i rarely see them discussed anywhere.

If you go this far, you have to take into the account all the non linearities, artefacts caused by the music instruments, and every component starting from diaphragm sputtering thickness to the dust on your speaker and air humidity and elevation. This inevitably leads to audio foolery and doesn't leave much room for creative process. Somehow almost as a rule this leads to seeking exotic, unobtainable, extremely expensive parts, instruments, wires, bla, bla, bla...

I can make a mic, measure it, publish all relevant data, but once you put it in your environment and start using it, it will perform, measure and most importantly sound different because all the variables involved. However these variables are beyond anyones control. They are also way above percentages expressed in zero point whatever.

by far the most interesting effect that couldn't be "traditionally measured" i've ever seen in product development was that one time, in an edge case with very unhappy parts being used out of spec, i saw a circuit that reacted differently to transients than steady state signal. that is, it reacted differently to playing an impulse than it did on sweeping and deconvolving. it measured flat on sweep and with noise, but distorted more and more the shorter the input signal was. absolutely fascinating. we found the problem and fixed it, but it definitely added several new tests to my suite when evaluating circuits. it was surreal to hear an obvious ~10dB boost somewhere that didn't show up on any normal test.

 

[rogs]

A unity gain OP applies maximum feedback, many OP amps can't handle that.
Try it open loop, like the tube config.
I'd be curios how an active loaded 6DJ8 would work out. Up to 60dB gain without feedback possible.
Of course an OP can work, and they are getting well past the 741 stage, but any better?

The OPA164* series of op-amps specifically includes unity gain stability as a feature of the device.
The data sheet lists a pretty impressive specification for audio use: https://www.ti.com/lit/ds/symlink/opa1644.pdf

I think it safe to say that op-amps have moved on quite a bit since the days of the 741 !

For this use -- a simple low distortion linear impedance converter - I think it's hard to beat?
I think a discrete FET front end should probably be able to beat it fractionally for self noise, with careful design?
The problem I have with using discrete JFETs as impedance converters is deciding whether to bias for minimum distortion - or symmetrical clipping ?
Neither of those are a problem with an op-amp like the 164* series!

 

[MicUlli]

The OPA164* series of op-amps specifically includes unity gain stability as a feature of the device.
The data sheet lists a pretty impressive specification for audio use: https://www.ti.com/lit/ds/symlink/opa1644.pdf

I think it safe to say that op-amps have moved on quite a bit since the days of the 741 !

For this use -- a simple low distortion linear impedance converter - I think it's hard to beat?
I think a discrete FET front end should probably be able to beat it fractionally for self noise, with careful design?
The problem I have with using discrete JFETs as impedance converters is deciding whether to bias for minimum distortion - or symmetrical clipping ?
Neither of those are a problem with an op-amp like the 164* series!

Yes that's true. OPA164x is a very good choice for audio applications. If you can live with noise it is okay. I am thinking about a hybrid design by combining a very low noise JFET with an OpAmp driver stage. It would solve the noise issues AND provide very low THD...

 

[rogs]

Yes that's true. OPA164x is a very good choice for audio applications. If you can live with noise it is okay. I am thinking about a hybrid design by combining a very low noise JFET with an OpAmp driver stage. It would solve the noise issues AND provide very low THD...

I compared the noise generated from my OPIC experiment with the noise from my Rode NT1. The latter has a published self noise figure of 4.5dB(A) and the single OPA1641 in the OPIC design only generated about 3dB more noise than that...... I decided I could live with that!

That noise was still essentally swamped by the ambient noise, in the sort of 'real world' recording situations I normally encounter.

I also no longer had to fiddlle trying to bias the FET front end, for the best compromise between low distortion and symmetrical clipping.
I've never been able to get the optimum settings for those two parameters to coincide.... Seems you have to choose one or the other ?

 

[Matt Syson]

I've never been able to get the optimum settings for those two parameters to coincide.... Seems you have to choose one or the other ?

The internet is a great way to introduce dissatisfaction to your life. The 'ideal' Op amp (and indeed any other) is ALWAYS constrained by real life issues.
Changing the ratio of oxygen and nitrogen (amongst the other gasses in 'normal air' will alter the sound you record with a microphone, as well as changing pressure and temperature, all of which are a distraction from a 'perfect' performance by the artist. Getting air conditioning noise level down is also a serious issue if you want to get picky as sub 10 Hz fluctuations due to the fan/impeller will add distortion (modulation) to any microphone signal and probably to playback from a loudspeaker if you really looked for it.

 

[Khron]

The 'ideal' Op amp (and indeed any other) is ALWAYS constrained by real life issues.

... But where does one draw the line? What *IS* "good enough"?

sub 10 Hz fluctuations due to the fan/impeller will add distortion (modulation) to any microphone signal and probably to playback from a loudspeaker if you really looked for it.

Sure, if you deliberately avoid high-passing anything, at least in the mix, if not even earlier. Not saying AC noise helps, of course, but those infrasounds might not be the end of the world.

 

[homero.leal]

Yes that's true. OPA164x is a very good choice for audio applications. If you can live with noise it is okay. I am thinking about a hybrid design by combining a very low noise JFET with an OpAmp driver stage. It would solve the noise issues AND provide very low THD...

Look for the Ocamm-3b design on the MicBuilders forum. May be you could adapt it to use LSK-189 and OPA1611.

On the other side, if what you want is low noise and low THD, why don´t you just use the Pimped Alice design with the LSK-189, biased for low THD?

The quietest mic I have built, as I remember, used that config with a 5G resistor instead of the 1G.

I don't know what your use case is, but I think for regular studio recording, OPA Alice is really more than enough.

 

[homero.leal]

What do you think? Will it work? Will be quiet?

 

[Khron]

1G in the feedback network? Won't that be akin to running the opamp basically open-loop? And it's inverting?

Will it work?

Well, apart from inverting the signal... Maybe? Ltspice could possibly show things...

(Later edit) Well, it looked weird, but at least in simulation, appears to work..? Still inverts the signal though, circuit gain about 2.5x (~8dB-ish), using an LSK189A model.

10v zener's sucking down a whopping 6.6mA(!!!). Bumping that up to 18v, reduces that to ~4mA; 1% thd with a hair under 4.5Vpp input (waaaaaay sooner than an opamp on its own). 24v zener only sinks ~2mA, pushes 1% thd point up to about 6.3Vpp input.

LSK189B and -C clip even sooner.

Of course, take all this with a grain of salt, it's just a simulation...

 

[homero.leal]

1G in the feedback network? Won't that be akin to running the opamp basically open-loop? And it's inverting?



Well, apart from inverting the signal... Maybe? Ltspice could possibly show things...

(Later edit) Well, it looked weird, but at least in simulation, appears to work..? Still inverts the signal though, circuit gain about 2.5x (~8dB-ish), using an LSK189A model.

10v zener's sucking down a whopping 6.6mA(!!!). Bumping that up to 18v, reduces that to ~4mA; 1% thd with a hair under 4.5Vpp input (waaaaaay sooner than an opamp on its own). 24v zener only sinks ~2mA, pushes 1% thd point up to about 6.3Vpp input.

LSK189B and -C clip even sooner.

Of course, take all this with a grain of salt, it's just a simulation...

Thank you for the info about sim, Khron! There is a comment on the original design, stating that C6 cap has an effect on the gain for the circuit.

Here is the note:
C1 (C6 for this design) <= Cx.
Voltage gain = Cx/C1
Cx is the capsule capacitance

So, for a unity gain on your sim, you could set C6 = to capsule capacitance.

And you are right, original design sends inverted signal to XLR3. WIll correct that.

Note that OPA1611 has a 3.6mA of quiescent current.

Regards!

 

[homero.leal]

1G in the feedback network? Won't that be akin to running the opamp basically open-loop? And it's inverting?



Well, apart from inverting the signal... Maybe? Ltspice could possibly show things...

(Later edit) Well, it looked weird, but at least in simulation, appears to work..? Still inverts the signal though, circuit gain about 2.5x (~8dB-ish), using an LSK189A model.

10v zener's sucking down a whopping 6.6mA(!!!). Bumping that up to 18v, reduces that to ~4mA; 1% thd with a hair under 4.5Vpp input (waaaaaay sooner than an opamp on its own). 24v zener only sinks ~2mA, pushes 1% thd point up to about 6.3Vpp input.

LSK189B and -C clip even sooner.

Of course, take all this with a grain of salt, it's just a simulation...

Also did a quick sim with OPA1641 and 2n3819 using 2V vpp. Here is the THD result:
Total Harmonic Distortion: 0.000865%(0.001011%)

Will need to look for LSK189 and OPA1611 models, but it doesn't looks bad

Kind regards!

HL

 

[Khron]

Also did a quick sim with OPA1641 and 2n3819 using 2V vpp. Here is the THD result:
Total Harmonic Distortion: 0.000865%(0.001011%)

Since we're dealing with opamps, i chose to "push the limits", especially in the light of @kingkorg 's recent capsule output measurements, hence the hefty signal levels I chose to try. Point being, this circuit clips way before an opamp on its own, if would seem.

24v zener only sinks ~2mA, pushes 1% thd point up to about 6.3Vpp input.

With the 100p "capsule capacitor" in my sim (where the ~8dB gain comes from, perhaps? identical values give 0.85x / -1.4dB "gain"), that still "only" translates to ~15Vpp at the output of the opamp, before thd going over 1%. That's into a 1.2kohm load, mind you, but the same opamp (model) only clips 1-2v away from the rails (so supplied with 24v it can drive 20Vpp clean as a whistle).

Sure, might have plenty of input impedance, but a restricted input range before breaking up. Can't say how JFET parameter spread (particularly Vgsoff) might affect the performance of this circuit, though.

 

[ccaudle]

What do you think?

Does that provide any advantage over just using an OPA164x device? Without bootstrapping the drain there is still gate-to-drain capacitance of the JFET, which is close to the input capacitance of an OPA1641.
The LSK189 is lower noise density than the OPA1641, but since the JFET is providing no gain I don't see how that helps, you still have the voltage noise density of the op-amp added to the voltage noise of the JFET. I certainly could have missed something in the analysis, but at first glance I do not see the advantage.

 

[Cqwet Dbdfte]

If a single JFET (amplifier/charge converter) with a 10:1 step down transformer has enough gain, (and galvanic isolation),
I do not get why an OP amp would add anything meaningful other than possibly getting rid of the transformer.
The gain of the JFET is set by capacitive feedback, as it is a charge converter more than a classic voltage amplifier. Is this also the case for the OP?
I see a few designs floating around without Neumann's uH+nF XLR filter.
Is the limited UHF bandwidth of the OP amp making this not needed?

 

[kingkorg]

I do not get why an OP amp would add anything meaningful other than possibly getting rid of the transformer.

Opamp has potential of 36v headroom, having low noise performance (even tho not quite as low as fet) it has way higher dynamic range. Also if right capsule is selected with posibility of taking higher polarization voltage the whole opamp based mic can smoke the fet one noise wise as well.

Noise is not just dependent on fet (or opamp spec) it is dependent on polarization voltage and capsule performance. Polarize a (right) capsule with 200vdc and 164x noise spec becomes irrelevant. Use fet, and it will crap out becaule it can't take level.

Fet into a transformer (km84) is pretty mediocre solution. I understand if it's used for nostalgia reasons or specific vibe, but performance wise it has both headroom, THD and noise issues. The design comes from an era where the whole point was to replace the tube with a solid state device, not much thought was put into the design.

 

[Khron]

an era where the whole point was to replace the tube with solid state device, not much thought is put into the design.

... And when the parts selection was pretty dire, compared to what's available nowadays.

 

[homero.leal]

Opamp has potential of 36v headroom, having low noise performance (even tho not quite as low as fet)

OPA1611 has 1.1nV/√Hz noise spec, also it's THD + N is rated to 0.000015%

But as it has bipolar input, it can't be used for the input stage, and that's where LSK189 comes into play. LSK189 has a noise spec of 1.8nV/√Hz.

Original Occam-3b circuit (by Zapnspark, on the MicBuilders forum) is based on a design published by Scott Wurcer (Charge Amplifier)

I understand that maybe the design doesn't makes sense, but is fun to revisit old stuff and try it with new and more modern parts.

Regards!

HL

 

[Khron]

OPA1611 has 1.1nV/√Hz noise spec, also it's THD + N is rated to 0.000015%

But as it has bipolar input, it can't be used for the input stage, and that's where LSK189 comes into play. LSK189 has a noise spec of 1.8nV/√Hz.

Sure, that's all fine and well, but an OPA1611 is a whopping 4-5 euros A PIECE And on top of that, add another 6(!!!!!) euros for an LSK189 - Mouser pricing, at least.

Now let's see, 10-11 euros just for the active parts for one microphone (whose gain depends on the capsule capacitance, AND inverts the signal), or a(nother) 2eu opamp that'll do the job at least just as well? The phrase "exercise in futility" comes to mind, unfortunately...

 

[ccaudle]

I understand that maybe the design doesn't makes sense

It does make sense after you pointed out the specifics, like using a bipolar op-amp specifically for the lower noise, but then needing to buffer with a JFET to get high enough input impedance at low frequencies for a capsule.

I think it is often the case as Khron points out that a 6dB noise improvement ends up costing 12dB in money, but as that is only less than the price of a restaurant meal in additional component costs it is a reasonable experiment for DIY.

I have an off-and-on interest in seeing if the AKG C460 style circuit can be designed with improved modern components. The C460 did not seem well-liked, and the later C480 design went back to discrete transistors, but it is possible that the components available 42+ years ago limited the true potential of the design. AES preprint 1876 from the Spring 1982 convention describe the design goals for the circuit.
Page 6 of the preprint describes essentially the circuit of post #128 but also including bootstrapping of the drain voltage (to eliminate the effect of gate-to-drain capacitance) as "current state of the art" (for 1982), with the note "The disadvantage of this circuit: the noise voltage generated in the 2nd stage is added unreduced to the output voltage."
The resulting C460 circuit used gain in the JFET buffer included in the feedback loop to reduce the noise from the op-amp (specific model not noted, but I am sure it was much noiser than a modern OPA16xx device).

I am not really sure how to analyze the C460 circuit, which includes a constant-current diode (using J500 JFET) in the feedback loop, presumably as a way to increase the gain of the JFET buffer stage.

Just mentioning this information in case anyone looking at this thread might be interested in that circuit but not familiar with it already.

 

[MicUlli]

Yes, it is always easy to apply some criticism, but this forum should be understant as an idea pool
Let me place some comments:
1. Maybe that under most circumstances ambient noise is higher than mic noise. But mic noise has a different quality. A good sound engineer can always distinguish between these different sources.
2. A lot of discusson has taken place regarding THD. Some guys claim 0,001% as barely acceptable. If we consider THD+N mic noise matters because it may be far higher.
3. The audio interface used matters too. If it produces a noise floor of -120dBV(A) it is not worth to spend effort to the mic noise itself. But the best audio interfaces are able to perform better than -134dBV(A).
4. Consider a very good audio interface in conjunction with a well designed low noise head amp. This would allow to decrease the capsule polarisation voltage considerably, a DCDC converter for 60V or so could be omitted. No mess with beat frequencies any more.
5. I admit that these points make only sense for high quality mic capsules, probably not worth the money when using electret china crap...

BR MicUlli