Idea of an simple mic pre.

[k brown]

Excellent points. It is sadly true that today we live in a rfi cesspool what with wifi, GSM, and bluetooth by the truckload. Back in the 70s the worst you had to contend with was a dodgy taxi parked outside the studio.

Cheers

Ian

Check out the copious RF filtering in this little Yamaha mixer (MG10/2):

 

[Matt Syson]

There is of course the suposed 'advantage' of wideband audio recording (sy up to 100KHz or more which neatly allows people to record and then mic multiple 'out of audio' band signals along with the wanted material below say 25kHz. Switchmode supplies are an (unfortunately necessary) curse as many operate in the 60 - 100KHz region and can act as simple AM radio transmitters (69KHz with 100Hz modulation) meaning the cheap switchers NEED a fair level of filtering (and enclosure) just to make the same ripple and noise figures that Linear supplies can manage relatively easily.

 

[Matt Syson]

The 100pF caps between the inverting and non inverting inputs on all the op amps 'trick' something I added to the Audix MXT1200 in around 1982 when the mixer had to 'reject' the RF of a 'walkie talkie' when operated that was lying on the control surface. as the inputs of that mixer had transformers it eased some of the problems of RF arriving by input cables.

 

[inkster]

TEAC TASCAM M15 had a discrete mic pre with transformer, kind of an interesting FET front end I always wanted to experiment with myself...never found the time

 

[user 37518]

TEAC TASCAM M15 had a discrete mic pre with transformer, kind of an interesting FET front end I always wanted to experiment with myself...never found the time

Looks to me like a JFET differential amplifier with an active load, followed by a VAS and a push-pull stage with a tap at the output to feed back the signal to the input of the diff amp. It is basically a standard 6 transistor discrete op-amp. It would benefit from either cascoding the VAS or re-arranging it as a Darlington.

Some quick estimations:

The 15 pF cap provides Miller compensation. With the voltages being displayed on the schematic, the tail current of the JFETs is around 1 mA, with that 15 pF compensation cap, the preamp has a very hefty 66 V/usec Slew Rate, or thereabouts.

The VAS runs roughly at 0.5 mA of collector current, which gives it a gain of around 450 V/V. Due to Miller effect, this would make the 15 pF cap look like 6.8 nF to the diff amp., depending on the output resistance of the diff amp will form a dominant pole with this cap.

JFETs have tiny transconductance so that is most likely why they decided not to degenerate the differential amplifier, a transconductance of 5 mA/V is probably as high as it gets for these transistors (without even knowing the part number it is hard to tell). I am guessing the complete amp doesn't have an open-loop gain larger than 80dB.

 

[user 37518]

Check out the copious RF filtering in this little Yamaha mixer (MG10/2):

The diff amp of the preamp is not inside a global feedback loop, I don't like that. It is so easy to just add an extra inverter to have balanced feedback. But I guess an extra op-amp per channel means more cost. Anyway....

 

[thor.zmt]

TEAC TASCAM M15 had a discrete mic pre with transformer, kind of an interesting FET front end I always wanted to experiment with myself...never found the time

It's a standard J-Fet input discrete Op-Amp. Better designs of that exist.

An OPA1641 would be a good modern choice to use a monolith op-amp instead.

Observe how the feedback DC blocker is implemented.

Thor

 

[thor.zmt]

The VAS runs roughly at 0.5 mA of collector current, which gives it a gain of around 450 V/V.

Incorrect. There is a 1.5k emitter resistor and ~ 22k load.

The VAS thus only offers 15V/V gain.

The capacitor at the emitter is shown as non-polar, I assume the "p" was dropped. Alternatively the symbol may be wrong and the capacitor is 330uF polarised, then the VAS Gain is indeed around 440.

Due to Miller effect, this would make the 15 pF cap look like 6.8 nF to the diff amp., depending on the output resistance of the diff amp will form a dominant pole with this cap.

Actually like 225pF. So slew rate with miller loop closed looks like 4.5V/uS and the low VAS Gain means modest input skewing may not saturate the the VAS.

JFETs have tiny transconductance so that is most likely why they decided not to degenerate the differential amplifier, a transconductance of 5 mA/V is probably as high as it gets for these transistors (without even knowing the part number it is hard to tell).

These are likely 2SK170 or 2SK369 or equivalent duals. So the Transconductance of the differtial Amp is probably around 15mA/V.

The current mirror is heavily degenerated, so we can consider it ideal, for a simple analysis.

am guessing the complete amp doesn't have an open-loop gain larger than 80dB.

Assuming 2SA970 for the VAS (another standard "Japan audio part"), the load for the input stage is likely around 400 * 1.5k or 600k, leading incidentally to a similar 80dB open loop gain estimation with around 18kHz -3dB Point.

If the questionable Cap on the VAS emitter is 330uF we instead get 100dB OLG (as the input impedance of the VAS is much lower) with 1.1kHz open loop bandwidth and similar gain at 20kHz as having 330pF.

Not a terrible design. The output stage runs class A and thus needs little feedback to linearise it.

The highly nonlinear BJT VAS is linearised with degeneration (local feedback) and seems actually compensated to give a slightly rising response before closing the miller loop (330p//1.5k on the emitter).

Almost all gain comes from the J-Fet and current mirror load, which in turn is heavily degenerated to make a very "ideal" mirror.

Thus NFB is closed around almost solely the pentode like trandlsfer curve of the J-Fet, producing very "tube Like" harmonics.

I would not be surprised if the circuit's sound would be described as "Tube like", which may have been the design goal.

Even in the "20dB more NFB"

Thor

PS, I tend towards 330pF as Teac/Tascam used to have a quite aggressive low TIM, low NFB Design Philosophy, which this would fit better, but both options are possible.

 

[rock soderstrom]

To the OP (and everybody elese) I would recommend Micro CAP for circuit simulation/analysis. It's been available 100% free for some years now (unsupported thought) but it's a professional tool and not so hard to start running for the basic stuff.

That was a really good tip, thanks for that!
I'm currently trying out Micro CAP and I got my first results without too many issues, although the programme really offers a lot of possibilities.

In contrast to LTSpice, the GUI and the shortcuts are much more common. A bit old-fashioned, but it does exactly what you expect, much better than LTSpice in my opinion.

Thanks again!

Edit: ...and they have tubes

 

[thor.zmt]

I'm currently trying out Micro CAP and I got my first results without too many issues,

I use Tina-TI and find it excellent.

In contrast to LTSpice

LTSpice is userunfriendly (to anyone who used a more mainstream tool before) and overall fairly poorly designed, because it started as "Switchercad" for designing DC-DC converters and had stuff bolted on.

For some reasons it is favoured by many hobbyists, perhaps a mix of Masochism in using something with very poor useability and groupthink reinforcement.

Edit: ...and they have tubes

Tina lacks these, but symbols are in the extended libraries and it is trivial to add new components as long as spice files exist.

Thor

 

[Newmarket]

I use Tina-TI and find it excellent.



LTSpice is userunfriendly (to anyone who used a more mainstream tool before) and overall fairly poorly designed, because it started as "Switchercad" for designing DC-DC converters and had stuff bolted on.

For some reasons it is favoured by many hobbyists, perhaps a mix of Masochism in using something with very poor useability and groupthink reinforcement.

LTSpice has a different operational approach in general compared to "standard" Windows type programs. Whether ECAD, ESIM or general. I didn't 'get it' for ages. Then I had some spare time to better understand it. Then it becomes very useful and schematics can be drawn quite quickly. Admittedly the graphics in terms of result plots are a bit old school and don't offer multiple windows etc afaik. Working in a professional environment people use a mix of LT Spice and Simetrix (£££) depending on personal preference and experience

 

[ruffrecords]

LTspice also runs very well on Linux. I don't know if any of the others do.

Cheers

Ian

 

[ct.waveform]

With all this 'to and fro' I have not seen much hrere about 'outside world proofing' of a design. things like the arrival of hf noise into the OUTPUT and input of any amplifier that needs proper RF filtering. The speaker cables of an amifier can and do bring received RF into a poorly filteres power amp circuit which then demodulates it and because of the inability of nFB etc to handle and correctly apply feedback at RF then demodulates it. The application notes for any chips are the starting point for a circuit that will function in a small environment of a test bench NOT in the real world outside. Stuffing a stray DC or a few volts of AC at powerline or RF frequencies into or out of a piece of gear is to be expected. Star earthing of a studio is great for protection against electric shock hazards but poor at RF noise 'control' The metal casing of any mains powered gear has to be grounded by law but RF and hum can't read so will sneak into gear by any path possible. Look at the shielding necessary to contain RF radiation from computer motherboards.

Yup. I do wish more people would realise that the speaker cables are just one resistor away from the input stage. And if it has got a capacitor across it, not even that!

 

[thor.zmt]

LTSpice has a different operational approach in general compared to "standard" Windows type programs. Whether ECAD, ESIM or general. I didn't 'get it' for ages. Then I had some spare time to better understand it. Then it becomes very useful and schematics can be drawn quite quickly.

PSpice which I learned Sim's with is by far more user-friendly than LTSpice. Tina (-TI) is very similar in user-friendlyness and obeys standard windows conventions.

As a result is is MUCH FASTER to create schematics there, for anyone used to standard windows conventions, without excessive relearning.

Going from PSpice to TINA had a near zero learning curve. I do have LTSpice and it only gets used in extreme self defense cases.

Admittedly the graphics in terms of result plots are a bit old school and don't offer multiple windows etc afaik. Working in a professional environment people use a mix of LT Spice and Simetrix (£££) depending on personal preference and experience

I am not what professionals use even TINA or PSPice, never mind others mentioned.

Modern EDA (Altium, Mentor Graphics etc.) integrate simulation, schematic capture, PCB Design and BOM / Supply Chain Management on levels few educational toys even manage to approach.

LTspice also runs very well on Linux. I don't know if any of the others do.

That may be why it is so utter tosh on Windows.

Thor

 

[ct.waveform]

Looks to me like a JFET differential amplifier with an active load, followed by a VAS and a push-pull stage with a tap at the output to feed back the signal to the input of the diff amp. It is basically a standard 6 transistor discrete op-amp. It would benefit from either cascoding the VAS or re-arranging it as a Darlington.

Some quick estimations:

The 15 pF cap provides Miller compensation. With the voltages being displayed on the schematic, the tail current of the JFETs is around 1 mA, with that 15 pF compensation cap, the preamp has a very hefty 66 V/usec Slew Rate, or thereabouts.

The VAS runs roughly at 0.5 mA of collector current, which gives it a gain of around 450 V/V. Due to Miller effect, this would make the 15 pF cap look like 6.8 nF to the diff amp., depending on the output resistance of the diff amp will form a dominant pole with this cap.

JFETs have tiny transconductance so that is most likely why they decided not to degenerate the differential amplifier, a transconductance of 5 mA/V is probably as high as it gets for these transistors (without even knowing the part number it is hard to tell). I am guessing the complete amp doesn't have an open-loop gain larger than 80dB.

I'd guess some long way below 80dB. The amp part doesn't seem to have anything; no cascodes, an unbuffered VAS (let alone buffered with a darlington), no driver(s) and doesn't look like it's a Class A output stage. And I'm not even sure we know what it's going to be driving, so some of that gain could be lost. This is puzzling as I have a friend who worked for TEAC in the '80s and I'd got the impression they took a lot more care than this. Maybe the budget got used up bythe transformer.

 

[ct.waveform]

Incorrect. There is a 1.5k emitter resistor and ~ 22k load.

The VAS thus only offers 15V/V gain.

The capacitor at the emitter is shown as non-polar, I assume the "p" was dropped. Alternatively the symbol may be wrong and the capacitor is 330uF polarised, then the VAS Gain is indeed around 440.



Actually like 225pF. So slew rate with miller loop closed looks like 4.5V/uS and the low VAS Gain means modest input skewing may not saturate the the VAS.



These are likely 2SK170 or 2SK369 or equivalent duals. So the Transconductance of the differtial Amp is probably around 15mA/V.

The current mirror is heavily degenerated, so we can consider it ideal, for a simple analysis.



Assuming 2SA970 for the VAS (another standard "Japan audio part"), the load for the input stage is likely around 400 * 1.5k or 600k, leading incidentally to a similar 80dB open loop gain estimation with around 18kHz -3dB Point.

If the questionable Cap on the VAS emitter is 330uF we instead get 100dB OLG (as the input impedance of the VAS is much lower) with 1.1kHz open loop bandwidth and similar gain at 20kHz as having 330pF.

Not a terrible design. The output stage runs class A and thus needs little feedback to linearise it.

The highly nonlinear BJT VAS is linearised with degeneration (local feedback) and seems actually compensated to give a slightly rising response before closing the miller loop (330p//1.5k on the emitter).

Almost all gain comes from the J-Fet and current mirror load, which in turn is heavily degenerated to make a very "ideal" mirror.

Thus NFB is closed around almost solely the pentode like trandlsfer curve of the J-Fet, producing very "tube Like" harmonics.

I would not be surprised if the circuit's sound would be described as "Tube like", which may have been the design goal.

Even in the "20dB more NFB"

Thor

PS, I tend towards 330pF as Teac/Tascam used to have a quite aggressive low TIM, low NFB Design Philosophy, which this would fit better, but both options are possible.

I was going to say they are bound to be 2SK170. We forget they used to be throwaway cheap and could be found in even the cheapest PA amps back then. I have a few TOA amps somewhere which have a few (and some original Toshiba power transistors too - though sadly not complementary pairs.)

 

[ct.waveform]

I use Tina-TI and find it excellent.



LTSpice is userunfriendly (to anyone who used a more mainstream tool before) and overall fairly poorly designed, because it started as "Switchercad" for designing DC-DC converters and had stuff bolted on.

For some reasons it is favoured by many hobbyists, perhaps a mix of Masochism in using something with very poor useability and groupthink reinforcement.

Edit: ...and they have tubes

Tina lacks these, but symbols are in the extended libraries and it is trivial to add new components as long as spice files exist.

Thor

I'm truly at a loss to understand how anyone can find LTspice unfriendly. I think it's utterly brilliant. I've used it for ~ 15 years and don't often have to go to the User group until things gets very specialised indeed. Funnily enough I finally downloaded Tina-free just last week to see if their models of some parts worked better natively, and I totally hated it. I couldn't even draw a point to point line without a spiral of squiggles at the closing end, let alone plop a component over a track and have it realise that it wasn't intended to be in parallel with a perfect short. Moreover, from the number of complaints I read, Altium appears to have been ruined and is now a pricey cash cow. Just as LTspice is now the de facto professional standard, KiCAD (and its integration with FreeCAD) appears to be becoming the same for layout. I haven't tried it yet, but I'm told the schematic editor in KiCAD has become very good indeed.

 

[Newmarket]

I am not what professionals use even TINA or PSPice, never mind others mentioned.

??? I think that sentence may have got a bit mangled along the way ?

But apart from that I should have said I also use TINA-TI and also like that. Do you have the full commercial TINA or the free TI version ?
Also Altium integrated SIM from some years ago. That's good but I find it only really copes with segments of a moderately complex circuit. Current versions likely more powerful ?

 

[Newmarket]

I'm truly at a loss to understand how anyone can find LTspice unfriendly. I think it's utterly brilliant. I've used it for ~ 15 years and don't often have to go to the User group until things gets very specialised indeed. Funnily enough I finally downloaded Tina-free just last week to see if their models of some parts worked better natively, and I totally hated it. I couldn't even draw a point to point line without a spiral of squiggles at the closing end, let alone plop a component over a track and have it realise that it wasn't intended to be in parallel with a perfect short. Moreover, from the number of complaints I read, Altium appears to have been ruined and is now a pricey cash cow. Just as LTspice is now the de facto professional standard, KiCAD (and its integration with FreeCAD) appears to be becoming the same for layout. I haven't tried it yet, but I'm told the schematic editor in KiCAD has become very good indeed.

tbf LT Spice can seem unfriendly if you're used to working in a "Windows" manner. And the graphical results format looks dated and limited in some aspects.
TINA-TI is easy to use. You are just doing something incorrectly wrt the wiring connections.
wrt professional standards - I guess it depends on industry sector, budget and location. But I'm not aware of a company if any size going with KiCad.

 

[ct.waveform]

tbf LT Spice can seem unfriendly if you're used to working in a "Windows" manner. And the graphical results format looks dated and limited in some aspects.
TINA-TI is easy to use. You are just doing something incorrectly wrt the wiring connections.
wrt professional standards - I guess it depends on industry sector, budget and location. But I'm not aware of a company if any size going with KiCad.

You're more likely to be right on that than I am. I'm mostly going on what I'm seeing on forums - mostly LTspice ones. And lots of nice things are being said about KiCAD now, though I hated it when I first tried it, which is now more than 10 years ago.

 

[rock soderstrom]

LTSpice is userunfriendly (to anyone who used a more mainstream tool before) and overall fairly poorly designed, because it started as "Switchercad" for designing DC-DC converters and had stuff bolted on.

For some reasons it is favoured by many hobbyists, perhaps a mix of Masochism in using something with very poor useability and groupthink reinforcement.

+1

Masochism is the right term - I don't even want to get used to this crude user interface. Horrible.

I use Tina-TI and find it excellent.

That's on my list, I'll have a look at it this evening. Micro Cap stopped being developed three years ago, which makes you feel a bit uneasy about learning it now.

 

[Newmarket]

You're more likely to be right on that than I am. I'm mostly going on what I'm seeing on forums - mostly LTspice ones. And lots of nice things are being said about KiCAD now, though I hated it when I first tried it, which is now more than 10 years ago.

Well it's horses for courses (and budgets of course). I can see KiCad and other stuff eg the free version of DesignSpark PCB is popular in the hobby / maker / startup sectors. No point in paying £££s for stuff you never want to get into eg Buried Vias; BGA breakout etc.
LT Spice is well used on all sectors ime. I just wouldn't say it was the default. imo there isn't really a default for simulation in any sector. In reality what you get out if simulation depends a lot on how far you want to get into the technicalities eg wrt timesteps / convergence tolerance etc. Cheers.