Mic Preamp Schematic Collection

[Samuel Groner]

Hi

Over the last years I spent much time studying microphone preamplifier topologies. This resulted in more design ideas that I'll ever be able to build (or at least I have a need for). Imprisoned thoughts don't help much so I decided to draw some schematics in case anybody's interested in building something new next year. Here we go:

Design A: A_r1.pdf
Based on the famous topology firstly presented by Cohen, this is a pretty simple no-nonsense circuit. It uses PNP input transistors instead of the ubiquitous NPN for lower noise. In addition to this, a negative bias at the base of the input transistors improves headroom and provides correct bias for the input coupling capacitors even if phantom power is switched off. I've built this design with some minor changes with great success. Just recently had a classical recording in one of the largest concert hall of Switzerland using 16 channels of those.

Design B: B_r1.pdf
An attempt to skip electrolytic capacitors (and DC servos). Needs many expensive precision components, but could be pretty stunning performance wise--we're talking about 110 V/us slew rate and around -130 dBu EIN (150 ohm source, 20 kHz) here, right? Don't expect an output offset in the uV range, but about any line input should be able to eat one or the other mV.

Design C: C_r1.pdf
An extension to the first design--gain is now distributed amongst two stages. This will greatly reduce distortion at high gains. In addition to this, the frontend is modified for even lower noise. If I'd got asked for a mic preamp optimised for ribbon and moving coil microphones, this would be one possible answer.

Design D: D_r1.pdf
That's an elaboration of the "shared gain" topology of which I have already posted two (less detailed) schematics. I can give reference to Steve Dove for the basic idea. The gain setting with the linear pot is pretty neat (D_r1_gain_law.pdf). Overall a simple yet versatile and good performing circuit.

Design E: E_r1.pdf
A solution for all that stocked 2SK170's. Given proper care to the implementation, this design will provide a very low noise figure, essentially limited only by the transformer's DC resistance. Likely very pleasing for dynamic and ribbon microphones. Includes a gain trim option.

Design F: F_r2.pdf
A fusion of the transformerless and transformerbalanced topologies used here. Fully balanced and wide gain range yet simple to implement. Note very low quiescent current. Expect low distortion and noise over a very wide gain range. Revision 1 suffered from oscillation, revision 2 corrects this.

Design G: G_r1.pdf
Based on the high performance AD797 IC opamp, with a 1:2 input transformer. Fully balanced for very good input CMRR, doubled slew-rate and high maximum output level. Only few parts needed and hence a simple build.

I believe that all designs are well thougth through and conservatively rated regarding stability and component tolerances. Nobody's perfect though, so I recommend breadboarding/prototyping and appreciate any comments.

If someone builds one of these, please let me know. And if there's some serious interest in one design, you might talk me into designing a PCB for it. Not promised though.

That's it for today. Two to four other designs in preparation!

Samuel

 

[jdbakker]

Very interesting ! I'm halfway a two-stage design for a Green-like pre for my HDD recorder which looks a lot like your Design C.

I have a couple of remarks about your protection circuitry, though. First, if C4/C5 are low ESR, a short between IN+/IN- and ground can easily generate a current well in excess of the peak pulse rating I(FSM) of D1...D4, depending on the winding resistance of L1. While breadboarding my Cohen/Greenish pre (with, admittedly, a different CMC) I found I could zap 1N4148 protection diodes this way. Furthermore, this kind of protection with diodes to the rails only works well if the decoupling capacitors on the supply are >> the input coupling caps, otherwise the rails may well 'bump' by over 10V, possibly exceeding the supply maximum for the op-amps.

Is there any particular reason why you tie compensation capacitors C7 and C8 to the negative input of the op-amps rather than across R16/R17 as in the Green/Cohen designs ?

[quote author="Samuel Groner"]That's it for today. Four to five other designs (mostly transformer balanced) in preparation![/quote]
Great stuff, keep 'em coming !

Thanks,

JDB.

 

[SALSA]

Hi Samuel, what does it take to talk you into designing a pcb for design B ? :wink:

I am pretty intrigued by that design and would really like to know how it performs.

The components are expensive, but not more than a commercial DOA with good in- and out transformer ...

Cheers,
:sam: :sam: :sam:
Mike

 

[JohnRoberts]

[quote author="Samuel Groner"]Hi

Over the last years I spent much time studying microphone preamplifier topologies. This resulted in more design ideas that I'll ever be able to build (or at least I have a need for). Imprisoned thoughts don't help much so I decided to draw some schematics in case anybody's interested in building something new next year. Here we go:

Design A: A_r1.pdf
Based on the famous topology firstly presented by Cohen, this is a pretty simple no-nonsense circuit. It uses PNP input transistors instead of the ubiquitous NPN for lower noise. In addition to this, a negative bias at the base of the input transistors improves headroom and provides correct bias for the input coupling capacitors even if phantom power is switched off. I've built this design with some minor changes with great success. Just recently had a classical recording in one of the largest concert hall of Switzerland using 16 channels of those.

Design B: B_r1.pdf
An attempt to skip electrolytic capacitors (and DC servos). Needs many expensive precision components, but could be pretty stunning performance wise--we're talking about 110 V/us slew rate and around -130 dBu EIN (150 ohm source, 20 kHz) here, right? Don't expect an output offset in the uV range, but about any line input should be able to eat one or the other mV.

Design C: C_r1.pdf
An extension to the first design--gain is now distributed amongst two stages. This will greatly reduce distortion at high gains. In addition to this, the frontend is modified for even lower noise. If I'd got asked for a mic preamp optimised for ribbon and moving coil microphones, this would be one possible answer.

I believe that all designs are well thougth through and conservatively rated regarding stability and component tolerances. Nobody's perfect though, so I recommend breadboarding/prototyping and appreciate any comments.

If someone builds one of these, please let me know. And if there's some serious interest in one design, you might talk me into designing a PCB for it. Not promised though.

That's it for today. Four to five other designs (mostly transformer balanced) in preparation!

Samuel[/quote]


Sorry if this is old news but that "Cohen" circuit looks very much like the Transamp topology that I though Buff pioneered several years before Cohen's '84 paper. Did Cohen have some earlier priority?

One difference from memory is that Buff used TL07x opamps and cap coupled the collectors of the input devices to establish dc operating point in the gain amps (a large electrolytic cap between collector and - input with large value feedback resistor to provide DC feedback path). He used additional resistors from the supply to provide operating current for input devices.

The "Cohen" approach to self bias the input devices from the feedback resistors looks similar to a moving coil phono preamp topology I published in Popular Electronics magazine in Mar'81. This is a similar topology to mic preamps I used in consoles in late '70s early '80s but it all seemed like old news to me after Buff's work. I used very low noise Japanese transistors developed for use in head amps (2SB737-2SD786) .

BTW the 4403 is just a general purpose transistor and IMO not a great choice for low noise. Paul Buff used a general purpose medium power transistors based on the premise a higher current device would have lower Rbb and thus less noise.

I think I saw a discussion about selecting devices for Transamp use. AFAIK he selected mic pre input devices more to reject those with high process related (1/F) noise rather than for close Vbe matching as device manufacturing processes were not very pristine back in the '70s and noise was not a concern in med power devices. Even if he matched to a mV it would still be noisy switching at 60dB of gain. He did select and match devices for Vbe in his VCA products where mismatched Vbe did cause distortion. If anything he might have used drop outs from the VCA selection process in the mic pres. In the mic preamp the negative feedback of the opamp wrapped around the input device caused them to operate at roughly constant current.

JR

 

[bcarso]

The Motorola process for the 4403 was very much better than others (see the curves in Motchenbacher and Fitchen). I don't know whether On Semi has preserved the tradition. Amusing that they were merely trying to make a switch, and had no interest in low noise.

Not to sound like a Toshiba commercial, but the rbb' of the 2SA1316 and 2SC3329 is even lower than the Rohm parts mentioned. I have yet to have the time to play with them.

 

[JohnRoberts]

[quote author="bcarso"]The Motorola process for the 4403 was very much better than others (see the curves in Motchenbacher and Fitchen). I don't know whether On Semi has preserved the tradition. Amusing that they were merely trying to make a switch, and had no interest in low noise.

Not to sound like a Toshiba commercial, but the rbb' of the 2SA1316 and 2SC3329 is even lower than the Rohm parts mentioned. I have yet to have the time to play with them.[/quote]


Yes those PN sound familiar, I think they changed to those at my old gig when ROHM parts went obsolete (happened after I left).. The 737s that I still have a few of in the back room are also 2 ohm Rbb and on paper a red C hair less noise .55nV/rt Hz... but into 150 or 200 ohms no worries...both parts are much better than needed.

Yes I recall the 4403 in M&F ... but keep in mind when that book was written ('70s). The supposedly "low noise" transistors like 5088? or something like that were a joke into lower impedances... IMO Buff was wise to look at 500mA med power devices and screen for 1/F.

At least the Japanese were fanatical enough about their Hi-fi to develop devices for head amps. As I recall ROHM bought the small company that originally tooled the 737/786 parts.

I called and spoke with Fitchen who was a Prof. at University of Bridgeport as I lived a couple of towns away at the time. I wanted to pick his brain about low noise design beyond what was in his book and he ended up asking me if I wanted to contribute a chapter for his next book... I was flattered but wrong answer. I was more like an empty sponge than a font of wisdom. I'm self taught so writing for a college level engineering text is the last thing I needed to be doing, especially 25-30 years ago when I was but a mere pup... Maybe he was just trying to get rid of me.. It worked.

JR

 

[Guest]

So, what NPN beasts may be called today low noise devices for say 600 Ohm input?

 

[JohnRoberts]

[quote author="Wavebourn"]So, what NPN beasts may be called today low noise devices for say 600 Ohm input?[/quote]

I would use the Toshiba parts "bcarso" cited. Probably run them at lower current density than you would for 150/200 ohm nominal of low Z mics. If you look at the curves for constant NF vs. bias current in the data sheet that will help you dial it in for 600 ohms. Lowest NF is best case.

JR

 

[Samuel Groner]

BTW the 4403 is just a general purpose transistor and IMO not a great choice for low noise.

The Motorola process for the 4403 was very much better than others (see the curves in Motchenbacher and Fitchen). I don't know whether ON Semi has preserved the tradition.

Todays 2N4403 do have low noise (at least those from Fairchild). A frontend as in design C has an EIN of -137 dBu for a 0 ohm source (-131 dBu for 150 ohm), with dead flat spectrum. Any improvement over this is hardly anything more than academic, and will result in much harder parts sourcing (at least for DIY purposes). Not that I'd dislike acadamic improvements though...

Not to sound like a Toshiba commercial, but the rbb' of the 2SA1316 and 2SC3329 is even lower than the Rohm parts mentioned.

True, but they seem to be impossible to source in the higher beta range. I believe (without having actually measured things) that three and possibly even two 2SA970BL in parallel will make a better frontend than one 2SA1316 at the same overall current as it will provide much lower current noise (some dynamic microphones go well above the standard 150 ohm impedance) and very likely lower 1/f noise as well. Paralleling 2SA1316's (to get 1/f down) seems like a dangerous idea to me, considering the low fT and high capacity.

Paul Buff used a general purpose medium power transistors based on the premise a higher current device would have lower Rbb and thus less noise.

rbb is surely low, but beta and 1/f noise is likely worse than the figures we are used from small signal devices.

What does it take to talk you into designing a PCB for design B ?

Well, at least one or the other person who's interested as well. And likely three month time until all is done.

Is there any particular reason why you tie compensation capacitors C7 and C8 to the negative input of the op-amps rather than across R16/R17 as in the Green/Cohen designs?

I had stability issues with the capacitor in parallel with the feedback network--remember that this is actually a current feedback topology! Perhaps with a 2.2k feedback resistor you'll get it stable, but in the 1k case there's likely no way around adding some compensation. The original Cohen design uses the capacitors the way I do.

I have a couple of remarks about your protection circuitry, though.

Thanks, I'll think over it.

BTW, forgot to mention that bcarso contributed with helpful remarks to to the slightly different version of design A I've built, thanks!

Samuel

 

[JohnRoberts]

[quote author="Samuel Groner"]

BTW the 4403 is just a general purpose transistor and IMO not a great choice for low noise.

Today's 2N4403 do have low noise (at least those from Fairchild). A front end as in design C has an EIN of -137 dBu for a 0 ohm source (-131 dBu for 150 ohm), with dead flat spectrum. Any improvement over this is hardly anything more than academic, and will result in much harder parts sourcing (at least for DIY purposes). Not that I'd dislike acadamic improvements though...[/quote]

I stand corrected on the 4403... Yes I saw them in M & F's "Low noise design" years ago but the 4403s I remember from my bench time developing low noise design (yes I looked at them) did not have a beta of 500 and were not in the same league as proper low noise devices. Your cited -131 dBu @ 150 ohm is about a 1 dB NF (IIRC) and quite respectable. There is only one dB of possible improvement available if everything involved was noiseless and we know that will never happen. My only concern in a production environment is what kind of guaranteed performance you can expect. I have been bit before by vague or off sheet specs. For DIY they will be cheaper and more available, and if your results are typical they will equal or better a lot of audio paths in use.

Paul Buff used a general purpose medium power transistors based on the premise a higher current device would have lower Rbb and thus less noise.
-----
rbb is surely low, but beta and 1/f noise is likely worse than the figures we are used from small signal devices.

I never worked with the newer parts, I settled on the 2SB737s years ago and stopped looking (the "s" was a beta grade, something like 600x). At my last day job, the npn compliment (2SD786) was already in the system so I just used it. The 4403 is a 500mA device so I call it a medium power device especially when compared to a GP 3904 or such, but I may not be consistent with industry terminology. To handle 500 mA it will have lower base spreading resistance so if it's process noise is well controlled it should work well, and apparently it does.

Is there any particular reason why you tie compensation capacitors C7 and C8 to the negative input of the op-amps rather than across R16/R17 as in the Green/Cohen designs?
-------
I had stability issues with the capacitor in parallel with the feedback network--remember that this is actually a current feedback topology! Perhaps with a 2.2k feedback resistor you'll get it stable, but in the 1k case there's likely no way around adding some compensation. The original Cohen design uses the capacitors the way I do.
Samuel

Indeed putting capacitors across the 2k feedback resistors could introduce voltage gain in the feedback path (1.5k/2k with parallel C) and cause instability. The ratio of feedback (2k) and collector load (1.5k) must be factored into stability compensation. In this case it is less than unity (1.5/2). That works fine for the unity gain stable 5532 but to work with a decompensated 5534 you might need to tweak values.

From a quick glance you could save a few parts by locating the polarity switch at the very front end. I would be tempted to use a little less loss in the last differential stage (A) as the earlier gain stages will clip a few volts sooner in one direction than the other due to DC voltage drop across 2k feedback resistors (to supply operating current for transistors). The DC drop is more like 4V not .7V as noted on schematic. Alternately you could provide that operating current with separate resistors from the appropriate rail but IMO you will probably get slightly better CMRR as drawn.

In ver C. There is no practical benefit from the staggered gain stages. The noise will be dominated by the first gain stage and the beauty of that primary gain stage topology is that open loop gain actually increases as closed loop gain increases. Total open loop gain is opamp gain + collector load divided by gain leg between emitters. That's why this topology can deliver such good performance at high gains.

JR

 

[Samuel Groner]

Thanks for your notes.

From a quick glance you could save a few parts by locating the polarity switch at the very front end.

Can you elaborate--why would this save parts? It just needs a switch, and I could hardly think of a way to eliminate this... And I like it at the output, as it is much less likely to introduce contact problems here.

I would be tempted to use a little less loss in the last differential stage (A) as the earlier gain stages will clip a few volts sooner in one direction than the other due to DC voltage drop across 2k feedback resistors (to supply operating current for transistors). The DC drop is more like 4 V not 0.7 V as noted on schematic.

I guess you missed that I apply some DC offset at the input. This offset is chosen in order to maximise headroom at lowest gain. The uncorrected offset is more than 4 V because of Vbe and the considerable input bias current across the 100k resistors (R8/R9).

BTW, the input bias current is the reason why I lowered the according resistors in design C, as otherwise beta mismatch in the input transistors would have caused a large offset and considerable drift. This reduces input CM impedance a bit, but you can't win them all (at least if you don't like bootstrapping as I do)...

In ver C. There is no practical benefit from the staggered gain stages. The noise will be dominated by the first gain stage and the beauty of that primary gain stage topology is that open loop gain actually increases as closed loop gain increases. Total open loop gain is opamp gain + collector load divided by gain leg between emitters. That's why this topology can deliver such good performance at high gains.

Well, the increase of o/l gain is limited by the input impedance of the inverting input--it would need to be well below 1 ohm to come close to the theoretical result at 66 dB gain.

I don't have any practical measurements to offer but in simulation the bandwith of the first stage of design A decreases from about 15 MHz to 300 kHz as you increase gain. This is a strong indication that distortion will--more or less inversely proportional--increase. Up to about 40 dB things look OK, after that it quickly drops. That's why I distributed maximum gain by 48 dB + 24 dB amongst the two stages.

Samuel

 

[Guest]

Samuel;
why don't you use continuously variable gain control?

 

[JohnRoberts]

[quote author="Samuel Groner"]Thanks for your notes.

From a quick glance you could save a few parts by locating the polarity switch at the very front end.

Can you elaborate--why would this save parts? It just needs a switch, and I could hardly think of a way to eliminate this... And I like it at the output, as it is much less likely to introduce contact problems here.
[/quote]
If the polarity is managed at the microphone input there is no need to duplicate the output differential. I guess if this is intended as a stand alone box, the second output provides a differential balanced output, inside a console it's a redundant opamp and associated parts per input. Pad and polarity switching at mic level signals is common practice and not problematic.

I guess you missed that I apply some DC offset at the input. This offset is chosen in order to maximise headroom at lowest gain. The uncorrected offset is more than 4 V because of Vbe and the considerable input bias current across the 100k resistors (R8/R9).

BTW, the input bias current is the reason why I lowered the according resistors in design C, as otherwise beta mismatch in the input transistors would have caused a large offset and considerable drift. This reduces input CM impedance a bit, but you can't win them all (at least if you don't like bootstrapping as I do)...

You are quite correct, I didn't see the input DC offset. That 5V comes out of your input peak swing but 10v negative swing should still be quite adequate for a mic input signals.

Well, the increase of o/l gain is limited by the input impedance of the inverting input--it would need to be well below 1 ohm to come close to the theoretical result at 66 dB gain.

The 22 or 33pf feedback capacitor will reduce OL gain at HF (I think I used 5-10pf). I have never performed computer analysis on this but it seems in practice there were limits for very low emitter leg resistance, I assumed from non ideal characteristics of the capacitor routinely used in series to block DC effects. I don't know how much effort you have made to investigate how your 6800uFs will act at 2 ohms in the top audio octave (ESR, ESL, etc). Dealing with this may be a better justification for avoiding operating front end at reduced emmiter leg gain resistance.

I did one custom design for a studio that used no blocking capacitor in emmiter leg. If you cap couple between collector and opamp and provide a local dc feedback path you will have stable DC operating point. Switch will still click but no big nasty cap in audio path. Sounded quite good.

I don't have any practical measurements to offer but in simulation the bandwidth of the first stage of design A decreases from about 15 MHz to 300 kHz as you increase gain. This is a strong indication that distortion will--more or less inversely proportional--increase. Up to about 40 dB things look OK, after that it quickly drops. That's why I distributed maximum gain by 48 dB + 24 dB amongst the two stages.

Samuel

Back when I messed with these topologies simulation was not widely available, I will have to take your word for it that the model is comprehensive. My bench experience was quite good at 60dB of gain using slightly lower GBW opamps than 5532 (IIRC 3 MHz vs. 10 Mhz). If the GBW of the first stage is added to the GBW of the opamp (loop wraps around both) it seems to me you still have lots of loop gain margin for good accuracy. I trust the good bench results that I recall but that was all we had to go on back then. I concede I'm a bit of a luddite regarding simulations. I have routinely used these topologies at more than 40 dB gain with good performance.

JR

 

[Samuel Groner]

Why don't you use continuously variable gain control?

There are several reasons:
* it's hard to set the high gains precisely with potentiometers due to the very variable contact resistance
* you will not get down to 0 dB gain with potentiometers (unless you use a 100k pot, but that's clearly impossible because of the issue mentioned above)
* the gain-law is perfect with rotary switches--a simple 6 dB/30°
* tracking between different channels is much better with rotary switches
* the feel of a rotary switch is still unsurpassed IMO...

I don't know how much effort you have made to investigate how your 6800 uFs will act at 2 ohms in the top audio octave (ESR, ESL, etc). Dealing with this may be a better justification for avoiding operating front end at reduced emmiter leg gain resistance.

In fact this is another reason, just forgot to mention it. Typically electrolytics start to generate measurable distortion at 10-20x the -3 dB frequency if they act as a high-pass filter. We'd need ten 6800 uFs if we'd like to move low-frequency distortion below 20 Hz at the highest gain setting of design A--clearly an unhappy solution. With two stages, it's easy.

I have routinely used these topologies at more than 40 dB gain with good performance.

I hope I'll be able to run a few measurements on the AP System One next week when I'm back at work. I checked a few datasheets of preamplifiers using this topology but all carefully avoided the quotation of distortion figures at higher gains.

But we can look at the INA103 (which uses the same topology) datasheet: ina103.pdf. It shows (page 6) both falling bandwidth and rising distortion (they forgot to mention the measurement BW though so we don't know how much of the THD+N reading is due to noise).

But please don't get me wrong: I consider a single stage design to be very usable for most applications, just want to give reasons why I considered a dual stage architecture.

Furthermore, this kind of protection with diodes to the rails only works well if the decoupling capacitors on the supply are >> the input coupling caps, otherwise the rails may well 'bump' by over 10 V, possibly exceeding the supply maximum for the op-amps.

I updated the schematics and included a zener to clamp the rail bump. I left the 1N914B there for the moment beeing as they do have lower capacity than higher power devices, hoping that a short to ground is a rather rare event (unless you decide to patch mic lines with phantom power on). Any suggestion for a diode having low capacity but high peak current rating?

Samuel

 

[JohnRoberts]

[quote author="Samuel Groner"]

I don't know how much effort you have made to investigate how your 6800 uFs will act at 2 ohms in the top audio octave (ESR, ESL, etc). Dealing with this may be a better justification for avoiding operating front end at reduced emmiter leg gain resistance.

In fact this is another reason, just forgot to mention it. Typically electrolytics start to generate measurable distortion at 10-20x the -3 dB frequency if they act as a high-pass filter. We'd need ten 6800 uFs if we'd like to move low-frequency distortion below 20 Hz at the highest gain setting of design A--clearly an unhappy solution. With two stages, it's easy.

[/quote]


I'm not sure we're on the same page here.. I don't know about LF capacitor distortion as in a non-linearity, but the expected ESR and ESL (perhaps DF?) will be an issue just like using electrolytics in a passive speaker crossover. Into 2 ohms at "high" frequency it will not be a very ideal capacitor.

JR

 

[Guest]

[quote author="Samuel Groner"]

Why don't you use continuously variable gain control?

There are several reasons:
* it's hard to set the high gains precisely with potentiometers due to the very variable contact resistance
* you will not get down to 0 dB gain with potentiometers (unless you use a 100k pot, but that's clearly impossible because of the issue mentioned above)
* the gain-law is perfect with rotary switches--a simple 6 dB/30°
* tracking between different channels is much better with rotary switches
* the feel of a rotary switch is still unsurpassed IMO...

[/quote]

I meant in addition to. You know, sometimes it is convenient to have a gain trim pot to align faders in line visually.

[quote author="Samuel Groner"] Any suggestion for a diode having low capacity but high peak current rating?
[/quote]

b-e junctions.

You are lucky you have no vacuum tubes in your amp; try to measure 1N914 capacitance around 80 degrees C...

By the way, what kind of transistor sets on a single crystal are available currently?
I need some thermostabilized diodes for similar purposes...

 

[JohnRoberts]

[quote author="Wavebourn"][quote author="Samuel Groner"]

Why don't you use continuously variable gain control?

There are several reasons:
* it's hard to set the high gains precisely with potentiometers due to the very variable contact resistance
* you will not get down to 0 dB gain with potentiometers (unless you use a 100k pot, but that's clearly impossible because of the issue mentioned above)
* the gain-law is perfect with rotary switches--a simple 6 dB/30°
* tracking between different channels is much better with rotary switches
* the feel of a rotary switch is still unsurpassed IMO...

[/quote]

I meant in addition to. You know, sometimes it is convenient to have a gain trim pot to align faders in line visually.

[quote author="Samuel Groner"] Any suggestion for a diode having low capacity but high peak current rating?
[/quote]

b-e junctions.

You are lucky you have no vacuum tubes in your amp; try to measure 1N914 capacitance around 80 degrees C...

By the way, what kind of transistor sets on a single crystal are available currently?
I need some thermostabilized diodes for similar purposes...[/quote]

gain trim would not be a simple pot due to gain law of that topology unless pot varied both feedback resistors. Then tracking could be significant CMRR issue.

---
Input clamps were discussed recently. FWIW A jfet gate is a pretty low capacitance diode but perhaps not very robust or appropriate here. I don't know that C is huge issue wrt low input impedances locally.

JR

 

[jdbakker]

[quote author="Samuel Groner"][quote author="I"]Furthermore, this kind of protection with diodes to the rails only works well if the decoupling capacitors on the supply are >> the input coupling caps, otherwise the rails may well 'bump' by over 10 V, possibly exceeding the supply maximum for the op-amps.[/quote]
I updated the schematics and included a zener to clamp the rail bump. I left the 1N914B there for the moment beeing as they do have lower capacity than higher power devices, hoping that a short to ground is a rather rare event (unless you decide to patch mic lines with phantom power on).[/quote]
That is apparently the most common cause for ground shorts, yes.

[quote author="Samuel Groner"]Any suggestion for a diode having low capacity but high peak current rating?[/quote]
Not really. All other things being equal, a higher-current diode will need more junction area and thus have higher capacitance. A b-e junction will likely not withstand the 20+V reverse voltage. On the other hand, the inductance of your CMC may be enough to 'tame' the current surge. Other designs use 4R7..10R resistors in series with the input caps; this will naturally degrade your noise figure. With pulses this narrow you have to rely on dissipation in the semiconductor itself; using a larger pad for heat sinking will likely not be enough.

On a related note, in the DIY-990 thread John Hardy hinted that the 1N4448 might be a better candidate than the 1N4148/1N914 due to having tighter specs. According to the Fairchild Semi datasheet that I have for these diodes, the 1N4448/1N916B have a max total capacitance of 2pF versus 4pF for the 1N4148/1N914B. (In his post, JH says "I believe that the 1N4448 is identical to the 1N914B"; this is at odds with the Fairchild DS. Buyer beware).

Why are you worried about a few pF coupling to the supply rails ? Other mic pres are known to use much higher-capacitance diodes; the TI ref design for the PGA2500 has 80pf Schottky diodes! The distortion caused by the non-linear relation between reverse voltage and diode capacitance may be a bigger issue. Time to match protection diodes ?

[quote author="John Roberts"]Input clamps were discussed recently.[/quote]
I remember, but for the life of me I can't find it back. This thread comes closest, but it isn't quite what I was looking for.

JDB
[ISTR that PRR or someone mentioned a 1/Vreverse relationship for diode capacitance... Oh Search function, why do you fail me so ?]

 

[JohnRoberts]

[quote author="jdbakker"]

[quote author="John Roberts"]Input clamps were discussed recently.[/quote]
I remember, but for the life of me I can't find it back. This thread comes closest, but it isn't quite what I was looking for.

[/quote]

I posted about JFET gate diodes being low capacitance, I was thinking more low leakage, but doesn't matter not the right tool for the job.

I've seen clamps make with 2 transistors connected with bases floating on both and connected back to back (Cs to Es) with one end to ground and the other to the input line. This would clamp when the be junction zenered + the collector diode approx. +/- 7.2 V or so. Another approach is zener inside diode bridge connection. A few diodes in series will reduce capacitance somewhat.

I'm sure somebody somewhere wrote an overview of all the sundry approaches.

JR

PS: I think there may have been some discussion in the THAT mic preamp chip white paper... Perhaps in the thread about floating up the mic pre to 48v

 

[Samuel Groner]

In his post, JH says "I believe that the 1N4448 is identical to the 1N914B"; this is at odds with the Fairchild DS.

The datasheet suggests that the 1N4448 should have lower capacity than the 1N914B, but otherwise being a match. Looks like a typo, no? The Vishay DS confirms this assumption.

Samuel