Another Annoying Beginner Question: OPAs vs BJT/JFET Mics

[MicUlli]

FET's are STILL better. I use them myself... e.g. 2SK117's and 170's (Real Toshiba ones!)

Completely true!

FET based mic circuits have very little gain. The typical Schoeps circuit has the FET in no gain phase splitter followed by PNP emitter followers. So no gain at at all.

The Schoeps circuit has +5,1 dB gain, that is because of the phase inverting property.
IMHO her is no need for voltage amplification at all for mic headamps. They shall provide current gain.

When using a JFET the following aspects are important:

1. Use the JFET as source follower,
2. Bootstrap the drain pin (to eliminate Cdg-nonlinearities) i.e. by a pnp-output transistor,
3. Provide a very high impedance as load at the source (use a current source or a simulated inductance).

Obeying these rules lead to discrete designs that are capable to deliver 2V rms at a 3kOhm load with THD < 0,1%. I have measured 0,008% with 200mV rms, which corresponds to typical SPL values under normal recording conditions. No need to use an OpAmp when you can catch it with only 3 discrete transistors..

MicUlli

 

[Mr. Moscode]

I'm not a fan of OPAs as followers. I won't start an argument as to why.

JR has a great point - the wheel is round. If it ain't broke, don't fix it.

If 3 to a dozen parts get the job well done what are we looking for? Isn't the technology running close to the theoretical limits for noise? Which to me makes the case for sound quality as the goal.

(Not do discourage R&D, have at it!)

BTW - At what point of measured THD % do you guys and gals, hear THD? .00001, .0001, .001, .01, .1, 1? Maybe that's for another thread.

 

[Voyager10]

discrete fets often have lower gains and require less feedback, so you can get nice-sounding audible lower order harmonics in the signal. this contributes to the "vintage sound" by some amount that i'd love to see someone measure.

There are some plots of various FETs fed with a simple sine wave here: https://groupdiy.com/threads/comparison-of-jfets-for-mic-applications.86559/post-1132551

These were measured at 100mV in, which is 110-120dB SPL for a typical capsule, and there's not a huge amount of harmonic distortion going on.

Obviously there are lots of things (drain currents, Vds) we could play with here, but this seems a long way short of an 'overdriven tube' sound.

 

[MicUlli]

BTW - At what point of measured THD % do you guys and gals, hear THD? .00001, .0001, .001, .01, .1, 1? Maybe that's for another thread.

That's the point! Our ears introduce a heavy amout of distortion from 90dB SPL upwards, it seems that nobody cares about it

 

[soliloqueen]

I'm not a fan of OPAs as followers. I won't start an argument as to why.

JR has a great point - the wheel is round. If it ain't broke, don't fix it.

If 3 to a dozen parts get the job well done what are we looking for? Isn't the technology running close to the theoretical limits for noise? Which to me makes the case for sound quality as the goal.

(Not do discourage R&D, have at it!)

BTW - At what point of measured THD % do you guys and gals, hear THD? .00001, .0001, .001, .01, .1, 1? Maybe that's for another thread.

Lower audible limit for me is about 0.06% at normal listening levels. Honestly 0.1% more reasonably. That's in addition to the THD of playback. That's above most mic amps at normal SPL isn't it? So unless the input is really loud I'm probably not going to be able to hear the distortion on a mic circuit

Modern opas don't really have that much more noise than discrete circuits, and in another 10 years they'll probably have less. They also don't require individual biasing or as careful parts selection for essentially the same performance. There are lots of advantages to an OPA. Especially considering now they have true jfet opas with no phase reversal that are totally unity gain stable with under 5nV noise that can easily run on phantom power. They also have opas designed specifically for high source impedance applications that have noise less than resistor noise now.

 

[kingkorg]

Completely true!

The Schoeps circuit has +5,1 dB gain, that is because of the phase inverting property.
IMHO her is no need for voltage amplification at all for mic headamps. They shall provide current gain.

When using a JFET the following aspects are important:

1. Use the JFET as source follower,
2. Bootstrap the drain pin (to eliminate Cdg-nonlinearities) i.e. by a pnp-output transistor,
3. Provide a very high impedance as load at the source (use a current source or a simulated inductance).

Obeying these rules lead to discrete designs that are capable to deliver 2V rms at a 3kOhm load with THD < 0,1%. I have measured 0,008% with 200mV rms, which corresponds to typical SPL values under normal recording conditions. No need to use an OpAmp when you can catch it with only 3 discrete transistors..

MicUlli

Idk if you caught some of my latest posts regarding the reasons i went for opa based mics. I record lots of sources, drums can easily go beyond 2v rms. Plosives as well. Hint - sm7b handheld metal vocal. Very popular style of vocal recording because of reasons... Noise both in case of drum and close up vocal recording takes the back seat...

Plosive into Neumann k105 capsule, unity gain modified Rogs' opamp circuit with external psu. The capsule can easily take 150db. I believe i measured THD to be 0.009% at this level.

Question: Aesthetics aside, most OPAs need 32 v rail to rail. At that voltage the most current available from phantom pwr would be 16/v drop across 3400 ohms (2x 6.8k phantom pwr Rs in parallel) =4.7 ma.

Is that enough to drive a 600 ohm mic input w/o clipping?

Exactly the reason why i built external PSU for my latest opamp mic.

 

[Cqwet Dbdfte]

There are some plots of various FETs fed with a simple sine wave here: https://groupdiy.com/threads/comparison-of-jfets-for-mic-applications.86559/post-1132551

These were measured at 100mV in, which is 110-120dB SPL for a typical capsule, and there's not a huge amount of harmonic distortion going on.

Obviously there are lots of things (drain currents, Vds) we could play with here, but this seems a long way short of an 'overdriven tube' sound.

I did not consider the added gain of 5.1dB in the Schoeps topology.
This corresponds to about 1.9x gain, so the 100mV would be 200mV for the above example.
As OP amp gains goes, this would be in the buffer category.
The human ears added harmonic distortion is even order in nature so 0.1% odd order products will not sound the same as 0.1% of even order products.

 

[rogs]

One thing to add is that many engineers are operating under a time-crunch in the studio (due to artist budgets). In those situations, “fixing it in the mix” often isn’t feasible. There are many situations where I have grabbed a mic that I knew would soft-clip or band pass the source I was recording so that I would get what I needed to begin with. I don’t always have the time to go though and high-pass or transient-shape all my tracks, and in those cases, microphones that behave in a known (but not linear) way are invaluable.....

I think my 'channel strip' comment in this post might also apply in this situation ?....

 

[rogs]

24 bit A/D and D/A is still unachievable due to real world physics when you consider NOISE over a 20 to 20K bandwidth because the signal is either too quiet or needing excessive voltage rails at the 'other' end of the range, so although 32 bit is undoubtedly good for PROCESSING a signal with unrestricted heqdroom, ultimar=tely you are stuck at the analogue interface. Yes 24 bit is possible if you restrict the bandwidth but by the time tou have put all the tricks together to make it work it becomes impractical. Arguing for a bandwidth limit above 20KHz has some merit but it largely makes things DIFFERENT but not necessarily technically correct AND can introduce problems like the recording of spurious HF interference (mobile phones and switchmode supplies that we are surrounded by) and the need to filter these signqls in the 'correct' place. Having to work in a Faraday caqge and have practically no gear using power or signql frequencies above the audio band (20KHz) so where do you put the restrictions? Marketing and the internet have a lot to answer for.

Some quite interesting 24 bit v 32 bit float comments and audio samples in this page from Sound Devices. The low level comparison audio samples were quite revealing, I thought?... Probably not often encountered in 'real world' recordings though!

They don't seem to mention that 24 bit A/D and D/A 'is still unachievable' ... Whether that's because things have moved on technically - as they seem to do at an amazing rate in the audio electronics world - or whether those notes are merely just 'marketing hype'?
That's not really typical of Sound Devices technical notes though, in my experience.

 

[Khron]

They don't seem to mention that 24 bit A/D and D/A 'is still unachievable'

Why would they? That article's there to extol the virtues of their own products, not to point out their limitations Not necessarily "hype" as much as "glossing over details that the marketing team considered detrimental / unessential".

 

[rogs]

Why would they? That article's there to extol the virtues of their own products, not to point out their limitations Not necessarily "hype" as much as "glossing over details that the marketing team considered detrimental / unessential".

Actually, reading again I think the 'unachievable' part may refer to the fact that the theoretical dynamic range of a 24 bit signal is some 15dB greater than the minimum achievable noise level, ref to 0dB.

32 bit float recording is a completely different animal, with a theoretical dynamic range in excess of 1500dB !
Importantly, the upper limit is not determined by the usual 0dB level.

I think it's going to be the analogue electronics that are always likely to be the limiting factor with that format...

( Some interesting notes on 32 bit float in another Sound Devices article HERE )

 

[JohnRoberts]

Lower audible limit for me is about 0.06% at normal listening levels. Honestly 0.1% more reasonably. That's in addition to the THD of playback. That's above most mic amps at normal SPL isn't it? So unless the input is really loud I'm probably not going to be able to hear the distortion on a mic circuit

I try to avoid arguing with people on the WWW about what they say they can hear, BUT from my personal experience the audibility of distortion depends very much on what "kind" of distortion. Some kinds are much more audible than others.

Modern opas don't really have that much more noise than discrete circuits, and in another 10 years they'll probably have less. They also don't require individual biasing or as careful parts selection for essentially the same performance. There are lots of advantages to an OPA. Especially considering now they have true jfet opas with no phase reversal that are totally unity gain stable with under 5nV noise that can easily run on phantom power. They also have opas designed specifically for high source impedance applications that have noise less than resistor noise now.

There will usually be lower noise discrete devices than op amps but the ICs continue to get better. At some point if/when ICs get good enough the market for some discrete devices may go away. The very low noise bipolar transistors that I used to use in mic preamps, are no longer available (in production quantities).

JR

PS: FWIW the LM394 a once popular, very low noise, matched dual NPN transistor is actually an integrated circuit using multiple NPN transistors in parallel internally.

 

[MicUlli]

Idk if you caught some of my latest posts regarding the reasons i went for opa based mics. I record lots of sources, drums can easily go beyond 2v rms. Plosives as well.

The question is: How much clipping free output can you expect by a phantom powered condenser mic ? It is simply the peak to peak output current capability of the driver stage multiplied with the mic preamp input impedance. With a differential class A transistor driver stage you may get 7mA pp and a voltage clipping limit of 48V pp (output unloaded). This of course would require an operating current of 7mA..

 

[rogs]

The question is: How much clipping free output can you expect by a phantom powered condenser mic ?......

The output current of the OPA1641 op-amp I use in my simple OPIC preamp is limited to an absolute maximum value of 30mA.
A more 'realistic' 20mA limit is used in the performance graphs. Where a 47uF primary decoupling capacitor is used, we mght expect that to store a maximum charge of 1.5 mC. That would translate as 0.0004mAH -- or more usefully, about 1.44mASecs..... Not a lot!

But if we can assume that the kind of voltage peaks we are trying to 'capture' from a very loud sound source are transient (like a snare drum) rather than continuous (like a jet engine) then the stored energy in the decopuling capcitor might provide some useful 'back up' current for the very short time needed?

It might be a good idea to increase the value of the capacitor fitted?
Say that is now made 100uF..... Then with 30v DC rail applied, the capacitor could store (in theory) enough energy to supply 30mA for 100mS.
Real world losses will make the actual energy available quite a lot less I suspect, but it might be enough to improve the performance - transient repsonse wise - by a measurable amount?

The other alternatives mentioned in this thread -- either reducing the polarisation voltage dramatically, or using an idependent DC supply for the preamp will of course produce more predictable results.........
But the energy stored in a primary decoupling capacitor might assist with those occasional transients?......

 

[MicUlli]

That is certainly true
Some simulation shows: With continuous RMS signal your circuit delivers an output voltage of 4V (rms) into a load of 2kOhm with 0,02% THD. That is 11,3V (pp). For my oppinion completely sufficient. A transient wave (100ms length) produces a distortion free (<0,1% THD) output of 18V (pp). Not bad at all

 

[Andy Froncioni]

What you state is true, for example, for a measurement microphone. There we want an absence of "character".

For most recording tasks a truly characterless microphone is not what recording engineers want. Which is why rarely see "measurement" type microphones used for recording.

As 'ideal' microphone we best use a large array of completely digital MEMS micro sized microphones (High SPL version) to get a flat response, no (read very low) distortion and a complete absence of analogue electronics, then we add desired directivity, frequency response shaping and distortion in DSP.

That and not Op-Amps is the right tool for microphones in the year 2567 (Buddhist Calendar).

Thor

And does an OPA mic really have a lack of character? You still have all the spatial and frequency shaping of the condenser capsule working for you, don't you?

 

[thor.zmt]

And does an OPA mic really have a lack of character?

Compared to the classic options, at higher spls, yes.

The characteristic tone of close miked female voice or brass you get from a Fet or Tube Microphone is not possible using an op-amp mic.

You still have all the spatial and frequency shaping of the condenser capsule working for you, don't you?

And that is all there is to tone?

Again, a lot depends what you actually want from the mic.

Thor

 

[baaccs]

On the other hand sound sources like snare drum, have initial very sharp and high in level transient. It is for the most part useless. It can randomly trigger compression where you don't want it, drum buss master buss...

Many engineers count on specific mic, or more often mic preamp to soft clip part of that transient and act as a limiter (neve 1073 is often used for this), so it does part of the job for them. Many don't even know what it does but they like it. Op-amp clipping sounds very bad for this kind of job, so you don't want any clipping coming from an op-amp in this case.

Having that said, i wanted a condenser with it's capsule properties but with high headroom op-amp circuit, so I can get that transient into my DAW without any clipping. So i built one. I like the flexibility of dealing with these transients in post.

Which capsule do you prefer using for the snare microphone. Also, which version of the opamp circuit are you using for the snare. Is it the one based off of the circuit Jules did?

 

[baaccs]

With the (gradual) move towards 32 bit float recording, digital clipping will slowly become much less of a problem.
The line input of my Zoom F3 recorder can accept a +24dBu input before clipping ( That's 34.7v p-p !)

Now, whether there will now be a whole new generation of 'compression' VSTs to try and copy the famous 'soft limiting' effects of analogue devices - like the Neve 1073 - we shall see.
The dreaded hard digital clipping should become a thing of the past, once 32 bit float becomes the dominant format...
How long that will be?.....

It sure has taken the industry long enough to get there. Samplitude was doing 32 bit float in the 90s

 

[kingkorg]

Which capsule do you prefer using for the snare microphone. Also, which version of the opamp circuit are you using for the snare. Is it the one based off of the circuit Jules did?

K105. The one from Neumann kms105. 104 would be great too. They are built for stage, insane rear rejection (hi-hat), takes 150db of spl, has low end roll-off which helps with close miking, just ideal by all means. No, it's Rogs' opic48 based + external PSU for continuous, stable 36v.