High Input Impedance BJT Stage Idea!

[Bill Mountain]

I have been looking at ways of creating a high input impedance BJT stage.  I have been studying biasing techniques and I realized out of all the ones I've tried I haven't seen one like this:



I hooked this up on the breadboard and I played it with and without a buffer and I couldn't hear any extreme differences in the high frequency content (but I was playing a bass).

I imagine that this can affect the current at the base but the transistor seemed to function fine except that the base was sitting at closer to .5 volts over the emitter.

Any thoughts from the resident geniuses on this as an input stage or a standalone high input impedance BJT boost?

Edt:  These weren't the exact values but it was something close to this.  Basically textbook biasing from one of my text books!

[GFR]

Gus has posted some designs here that use bootstrap to get a very high input impedance.

example:

http://www.diystompboxes.com/smfforum/index.php?topic=90020.0

[R.G.]

That gets a high input bias resistance, OK. If you put a capacitor between the junction of the 110K, the 10K and the 1M bias resistor to ground, you get the "noiseless biasing circuit". There is also a voltage drop across the 1M resistor from the bias voltage to the base, caused by the actual value of the input bias current. This may or may not be a problem.  It's really good for minimizing the noise contributed by the bias network, though.

However, that 1M bias resistor is in parallel with the input resistance of the transistor. This is approximately the current gain times the 1K emitter resistor. For a transistor with a gain of 100, that's 100K; for a gain of 400, it's 400K. So the input impedance is close to 1M parallel with 100K, or 400K, or whatever you have in the actual transistor.

A bipolar with hfe = 1000 gets you to 1M || 1M, or 500K.

The bootstrap versions take some of the signal voltage from the emitter through a capacitor to the junction of the DC bias network - where all three bias resistors meet. This voltage is only slightly smaller than the input signal, so it "cancels" the voltage across the bias resistor and makes the AC impedance much higher.  Bootstrapping can get you to quite high input impedances if done well.

[Gus]

some circuit I posted here in the past
Bootstrapped EF

Bootstrapped NPN instead of a JFET

Bootstrapping in a bottom bump


The univibe has bootstrapped transistors.  Bootstrapping is old.
The beginner project has bootstrapping GFR posted a link

[PRR]

> I haven't seen one like this

[gritz]

^Aah, thermal noise and d.c conditions dependent on hfe.^

I suppose it depends on how much input impedance you actually need.

[Bill Mountain]

> I haven't seen one like this

I played around with higher values and the highest I got up to was 5.6M and 1.5M but the math no longer looked good.  I was afraid there would be too little current.  But I guess if it works then it works.

R.G.:  So...whould a darlington with an HFE of 5-10k be usable like this?

Gus:  As soon as I wrap my head around bootstrapping I'm gonna give some of your circtuits a try.

Thanks guys!

[R.G.]

I played around with higher values and the highest I got up to was 5.6M and 1.5M but the math no longer looked good.  I was afraid there would be too little current.  But I guess if it works then it works.

Note that the two-resistor network is, as Paul says, the Thevenin equivalent of the three-resistor one, but it doesn't give the noise advantage of attenuation of thermal noise at the voltage junction, combined with the low current-noise of the bias resistor. Three resistors make this possible, but for DC the two-resistor and three resistor setups are equivalent.

R.G.:  So...whould a darlington with an HFE of 5-10k be usable like this?

It would, and it gives high input impedance, but loses some noise performance unless really well done and with the right devices.

As you can guess, I'm a little nutzo on noise issues. It can be tricky to get it all working right at the same time.

[Bill Mountain]

I played around with higher values and the highest I got up to was 5.6M and 1.5M but the math no longer looked good.  I was afraid there would be too little current.  But I guess if it works then it works.

Note that the two-resistor network is, as Paul says, the Thevenin equivalent of the three-resistor one, but it doesn't give the noise advantage of attenuation of thermal noise at the voltage junction, combined with the low current-noise of the bias resistor. Three resistors make this possible, but for DC the two-resistor and three resistor setups are equivalent.

R.G.:  So...whould a darlington with an HFE of 5-10k be usable like this?

It would, and it gives high input impedance, but loses some noise performance unless really well done and with the right devices.

As you can guess, I'm a little nutzo on noise issues. It can be tricky to get it all working right at the same time.

Ah yes...My ignorance of electronic theory strikes again!

I would say that most of my designs are simple bass preamps and dirt boxes so noise isn't a big issue for me as much as reliability and consistency of components.  I have built some nice distortions with FET's but I'm attracted to the ease of use and repeatability of BJT's.  Noise isn't huge for me but I honestly haven't been affected by any noise related issues so it's possible I'll change my tune eventually.

I'm taking a break from opamps so I've been exploring transistor circuits and I'm on a search for a good multipurpose input stage.  I'm using MPF102's and the ones I have seemto be fairly consistent but with the scarcity of (good) JFETs nowadays I'm looking at other options.

Thanks so much for your help!

[PRR]

> doesn't give the noise advantage of attenuation of thermal noise at the voltage junction

Thermal noise at the 10K-110K junction is lower, but to that we must add the ~~14uV thermal noise of 1Meg to get the noise at "input". The thermal noise is the _same_ either way.

The power-supply hash could be lower IF Bill hung a cap at the 10K-100K junction. But he didn't. And depending on his supply, it may be moot.

In real life, both thermal and supply noise/hash are further attenuated by whatever source impedance (plus couping-cap impedance). Given a large cap and a 100 ohm opamp source, effective input noise is very-very low. Given 5K of naked guitar in the lower audio band, very low; but the 50K-200K rise at pickup+cable resonance can give an annoying ringy-hiss.

The proper answer for this is even higher biasing resistors (so that source impedance sucks-off resistor hiss). (Or bootstrapping a smaller resistor to a higher effective impedance, but that is a very advanced topic and nowadays rarely a best approach.)

Fact is that 1Meg was good enough for Leo Fender so it is good enuff for us. (Altho Ampeg tube-amp tended to 3Meg-5Meg.)

> the math no longer looked good

What math?

The real drawback of super-high input bias resistors is the uncertainty of input grid/base/gate current. In a silicon BJT this is roughly Ic divided by hFE. You know what Ic you want. But hFE varies 3:1 from part to part and also with temperature. For a _specific_ transistor and temperature you can always hand-pick a 12.34Meg resistor which will bias-up right, but change the temp or transistor and you may need 13Meg or 7Meg. (How hot was it today and how cool is it tonight?)

The 3-resistor ploy with nothing extra at the T-point is just a waste of 12 cents. (As you can guess, I'm a little nutzo on excess parts.) Unless you need the _power_supply_ filtering of an added cap at this junction, forget it.

With the 2-resistor network, the bias-string DC current should be "much more" than nominal base current. If you expect "about 2uA" into base, flow 20uA through these resistors. But the definition of "much more" depends on circuit stability. If you have a large drop in an emitter resistor, you are probably OK. If you have no emitter resistor, you are sure to be mis-biased (except in SPICE, which lies).

For a nominal 0.75V at base, you don't have much drop across your emitter resistor.

A different way to look at it: the effective resistance seen by the base should be much less than the emitter resistor multiplied by hFE.

For a transistor type of nominal hFE=100-300, then the effective resistance seen by the base should be much less than 1K*100 or much less than 100K. So already we are "low" impedance by guitar standards.

Also the transistor base alone will be Re*hFE or 100K, again low for gitar.

A part of hFE=300-900 is available cheap. This moves us to <300K bias and 300V into the base, still low-ish for guitar.

The real issue is the fairly high collector-emitter current. 0.5mA is not unreasonable, _unless_ you need a HIGH input impedance.

A Darlington is not wrong but not magic. As R.G. says you now have two base-emitter junctions at the same voltage gain, "double the hiss". Well not quite, but even so. Also you have double the Vbe and Vbe-variation, so device and temperature issues are worse.

Interestingly one of Darlington's other plans, PNP-NPN, _cancels_ Vbe (almost) and makes biasing a trice easier. But that's two devices instead of a handy pre-pack device.

The real fun does come with 2-device topologies. One BJT alone is enough power gain to be useful but not enough power gain to be amazing. A 2-BJT scheme can be far better.

[Gus]

Another circuit to think about


One could add series input and output resistors for stability.

[R.G.]

> doesn't give the noise advantage of attenuation of thermal noise at the voltage junction
Thermal noise at the 10K-110K junction is lower, but to that we must add the ~~14uV thermal noise of 1Meg to get the noise at "input". The thermal noise is the _same_ either way.

Yep, the thermal noise is the same either way. The noiseless biasing approach gives you the ability to get lower excess noise from currents, plus a more solid Vbias by running a higher current through the bias dropping string - if that matters. For one transistor, it doesn't, much. If you can reuse the bias voltage, you break even at two transistors and save on average on resistor per other use of the bias string.

If that matters.

But yes, thermal noise is a function of temperature and resistance, always. It's really only the excess noises and amplification of those that can be sidestepped or avoided.

[PRR]

> Another circuit to think about

Yes, 2-BJT compounds are much more powerful (and interesting).

That one is unity-gain. Bill seems to be aiming for gain of 10 (seems high but not my project). How can you do that CE-CE compound with gain?

OTOH, sticking with your unity-gain and returning C3 to "4.76V" will give input impedance nearer 20Megs.

It is probably possible to get some of both (gain and hi-Z) but this gets into small differences of large ratios so I won't bet on it without more thought.

[PRR]

> get lower excess noise from currents

I was sneaking around Shot Noise. This is lower in the low resistance but higher with the higher current. IMHO true Shot Noise is rare in audio (gets killer at high RF or when you soak in liquid Helium to kill Johnson Noise). In fact, for values like Bill's, whatever random noise appears at the low-Z node is completely swamped by the 1Meg's self-noise, which in turn is more-or-less shunted by our source (with naked guitar a tough case due to ~~3KHz resonance).

Likewise any "bias stability" is thwarted by the 1Meg.

Yes, there are EXCELLENT other reasons to go to a low-Z node. Add a cap, cut power crap. Share it among several bias-users, save a few parts.

[Bill Mountain]

These are some excellent answers guys.

I don't have a specific amount of gain in mind but something that would be a good input stage.  I've seen and used a ton that work so I'm not hurting for one at the moment.  This is more of an exercise for future projects.

It looks like 2 tranny's or bootstrapping is the way to go.

I guess a buffer into a simple boost stage might work just fine.  I'm all for keeping parts counts low so I may have to revert back to opamps (for input stages at least).

Seriously, thanks again guys!

Oh...when I read that BJT input opamps have lower input inpedance, how low are we talking?  And if they are high enough for our purposes, could I just copy the opamp input stage (is it a simple differential pair???)?

[GFR]

> Another circuit to think about

Yes, 2-BJT compounds are much more powerful (and interesting).

That one is unity-gain. Bill seems to be aiming for gain of 10 (seems high but not my project). How can you do that CE-CE compound with gain?

Like this?



I've read it's not so interesting as the unity gain version, but AKG seems to like this topology a lot.

[PRR]

When working as intended (heavy negative feedback), the BJT opamp input impedance is more than high enough for any guitar purpose.

The naked (no NFB) may be as low as 10K, but the NFB jacks it up.

A further issue is how large a DC bias resistor you can get away with. If the output DC must be within milliVolts of a specified voltage, even 50K may be high-ish. This is where FET inputs shine. If you only have to get near 4V-5V, then much higher values can work.

[Gus]

Another X10 gain sim idea using a Sziklai pair instead of a Darlington transistor.  Green input,  blue output.  I have not built this.  Simple single coil guitar, cable sim C5 and to the left on the screenshot.
Bill Mountain have you tried the NPN boost ver 2?