LM386 buffer and gain+"bass boost"

[brett]

I'm building a little amp and have a couple of questions. It's an LM386 chip with an input buffer.
The schematic is below.

Instead of a JFET buffer, I thought a BJT emitter follower might work well.  My first question is will the buffer have enough base current to work as expected?  Background:  JFETs often bias low (e.g. J201 has Vgs around 0.5V), so there's not much headroom.  I wanted to bias the input/bass near 4.5V, and keep the input impedance high (300+kohms), so I chose 2 x 1 Mohm resistors and a high hFE transistor (input impedance = parallel of the two bias resistors and Re x hFE).  The result of that there's just 4.5 uA of base current.  Is that enough?  Will the hFE of the transistor be low with such a low Ib and Ic?  I need a hFE of 200+ to keep the input impedance high (see formula above). 

My second question is whether I can successfully combine the usual gain control (1 kB pot with 10uF DC blocking cap) with the treble cut "bass boost" circuit shown in the Datasheet.  Background:  I've used the "bass boost" circuit before, and like it, especially when playing at low volume.  As you can see in the schematic, output is fed back via a bass-cut RC filter to an early stage in the chip.  This treble cut (called bass boost on the datasheet) is about 6dB when the gain control is at a minimum (as shown, it's around 30).  However, the gain control alters (lowers) the resistance that the treble cut is bypassing. The gain control resistance varies from around 10k to almost zero, while the minimum resistance of the treble cut feedback is 4.7 kohms.  It seems to me that the treble cut will be minimal at high gain. 

I've made a similar build before, and decided that the interaction of the treble cut and gain controls was just an annoying fact of life (on one build I replaced the 4.7k resistor with a 10k pot, but got oscillations at low resistance/high treble cut. 
I wonder if one of you smart people can work out how to get these two feedback systems somewhat independent and happy?

Thanks for your answers and help.

[ElectricDruid]

How about using a buffer like the one on the TubeScreamer?

https://www.electrosmash.com/tube-screamer-analysis/pedals/distortion/tube-screamer-analysis.html

The calculated impedance for this is 446K, so that's fine for your needs. Plus it avoids using two large-value (hence noisy) resistors. Ok, you need two *other* 10K resistors and a 47u cap to make the 4.5V Vref supply, but that's not a problem.

I don't know much about the LM386, but I can read datasheets like the next guy! It looks like the 4K7/47n combination in your circuit are described as a "bass boost" in the datasheet, rather than a treble cut. It also says that you shouldn't use a value less than 2K, so that pot would be problematic at lower settings.
The gain control should work in combination with this, but again, I'd worry about going to 0 ohms with the pot turned right down - it might be better to have a small value resistor in series with the pot to provide a minimum resistance.

<edit>This is the datasheet I'm looking at:
https://www.ti.com/lit/ds/symlink/lm386.pdf

[Clint Eastwood]

If 446k is not high enough for you, you can use a bootstrap input:



This has ~ 1Mohm input impedance.

[antonis]

How about using a buffer like the one on the TubeScreamer?

How about NOT using all those %&(*^% lazy designer buffers..??

@brett: Implement circuit (B) in thread below:
https://www.diystompboxes.com/smfforum/index.php?topic=129250.msg1248337#msg1248337

P.S.
Tom, Electrosmash Input buffer calculations lack in both input impedance and voltage gain values..
(due to wrongly both h_FE and Emitter load estimation..)

[ElectricDruid]

How about NOT using all those %&(*^% lazy designer buffers..??

Aww, go on! Everyone *likes* those classic buffers, even if they're not that great! There are loads of famous pedals that use something very much like that, and no-one's complaining about them (except you, maybe?!? ).

Personally I prefer an op-amp buffer anyway - cleaner, better impedance, no messing about. Transistors are for tone or distortion in my book.

P.S.
Tom, Electrosmash Input buffer calculations lack in both input impedance and voltage gain values..
(due to wrongly both h_FE and Emitter load estimation..)

Gagh, that site always lets you down!

[PRR]

....large-value (hence noisy) resistors

Series resistors should be small. Shunt resistors should be large.

...the 4K7/47n combination in your circuit are described as a "bass boost" in the datasheet, rather than a treble cut.

Datasheets are written by committees of distracted staff.

The '386 is touted for gain of 20(-200), or 26dB. The booster/cutter gives midfrequency gain of 19dB, returning to 25dB at 100Hz. On women's choir there is no "bass boost" (all tones above 200Hz). Nor treble cut, because all tones are on the flat zone. Go by the graph, not the verbiage.

[antonis]

Everyone *likes* those classic buffers, even if they're not that great! There are loads of famous pedals that use something very much like that, and no-one's complaining about them (except you, maybe?!? ).

I've never complained either 'cause I never used them..

More seriously now, LM386 input impedance is 50k (p-n-p Darlington Base to GND resistor)..
That resistor is set in parallel with 10k Volume (input level) pot..
You DO agree that it isn't very wise to implement a CC amp of Emitter resistor bigger than the driven load, don't you..??

[PRR]

....4.5 uA of base current.  Is that enough?  Will the hFE of the transistor be low with such a low Ib and Ic?  I need a hFE of 200+ to keep the input impedance high ....

4.5uA in the un-loaded bias divider. At hFE=200 you need 9uA base current.

Or: 2.4k emitter load, times hFE=200, is 480k impedance into the base. This in shunt with 1Meg puts you far off of halfway, closer to 1/4.

And of course with nothing plugged in you will hear half the power supply crap, times 20. Battery may be tolerable. Only the best wall-supplies will be "quiet". 

Why did you pick these values? Especially 2.4k? That leads to over a mA of emitter current, which is mighty generous for a 50K chip. Yes you want a potentiometer but 10K may not be optimum.

Voltage divider bias with generous idle current is stable in mass production, but in one-off you can do better by cheating.

[brett]

Thanks PRR and others.
Wht 2.4k?  Because I use quick n dirty rules of thumb.
Lm386 =50k ohms.  For low losses, Zout = Zin / 5, so pot =10k, then for BJT output impedance Re = 2k (I had 2.4k on the shelf).  And hFE x Re is about 1M for hFE = 400.  Nothing more complicated than that.
But the BC549C datasheet only shows as low as 0.1mA IC, and hFE is looking quite 'sad' with that.  That's why I thought Ib of just a few uA and hFE of 50? 100? 200? might not work too well.  But now I see you use a 2.2M resistor for an even smaller Ib and Ic. 
I guess I should "dead insect" this and find a good bias point and measure the hFE.
The idea of "lossy" Re 47k and pot 50k is, I now realise, a potential benefit if the 0.4 V Max input of the LM386 is to be avoided (datasheet suggestion).  That's assuming that overloading the input sounds bad.  Without much evidence, I've worked off the observation that an LM386 powered by a battery gives great "crunch" because the battery resistance comes into play when the output pair are working hard.  It's only a couple of ohms, but that's probably enough, given it's a fair % of the output impedance.
Thanks again

[antonis]

for BJT output impedance Re = 2k

BJT output impedance is calculated by 0.026 / I_Emitter (V/A)..

2.4K (or any other R_E) is effectivelly set in parallel with 10k pot and LM386 input impedance when considering AC load..

If you insist to implement a single BJT, bootstrapping is one-way for high input impedance..   

P.S.
Let  h_FE alone unless, of course, you deal with a trasistor tester..

[brett]

Thank you Antonis
So 0.026 / Iemitter = 0.026 / (2V/2400 ohms) = 0.026 / 0.0008 A  = 30 ohms.  Tiny !
Heck.  I can easily afford to make Re = 47k.
While I'm at it, I'll do the PRR recommended trick of removing the "base to ground" input resistor, which lifts/improves the base bias point from about 2.5 V to 6V.
Thanks everyone.

[antonis]

I can easily afford to make Re = 47k.

IMHO, not so easily..

You see, a 47k Re couldn't swing down more than 50% of quiescent voltage for a load of 47k..

In case of 8k3 load (10k pot // 50k input impedance) and 47k Emitter resistor, you can easily calculate Emitter bias voltage for symmetrically undistorted signal...
(which, definatelly, should be much higher than mid-supply..)

[PRR]

I can easily afford to make Re = 47k.

IMHO, not so easily..

You see, a 47k Re couldn't swing down more than 50% of quiescent voltage for a load of 47k..

How much swing do we need at the input of an LM386? At stock gain=10, like 0.2Vrms, 0.3V peak. With the network on pins 1 and 8, gain can be 200, 30mV peak. (And that's what we'd need for easy e-guitar strumming.)

9V supply implies potentially 4V peak. So we have 10X-100X more than required. We don't HAVE to do the 5:1 impedance ratio.

The plan I posted, all 50K loads, will swing 2V peak clean, 2.2V peak with a tiny clip. Very very few guitars will peak this high for long enough to notice clipping.

[antonis]

The plan I posted, all 50K loads, will swing 2V peak clean, 2.2V peak with a tiny clip. Very very few guitars will peak this high for long enough to notice clipping.

Of course, for Emitter biased at level considerably higher than mid-supply..
(that's exactly what I was trying to tell OP..)

I think it makes no harm to design an Emitter follower for as high as possible undistorted output, despite input signal amplitude..

[brett]

Thanks all.
Although originally wanting to avoid JFETS, I took a trusty MPF102 out of the drawer and tied the Gate to ground and Source to ground via the 2.4k resistor that was on the board.  Voila.  0.65 volts on the gate (about 0.8mA Id).
That's too convenient to ignore, so that's what I'll use.
As PRR said - nobody is going to notice the little bit of clipping it does on guitar signal.
On the other hand, it'll clip asymmetrically when fed from a big Boost.  0.6V on the neg side, 8V on the pos side.

I'll post schematic and PCB after some testing.  Especially around the behaviour of the Treble Cut control with various levels of gain. 
This one is going in a 1 x 12" cab with a Celestion Rocket (quite efficient - from memory about 95 dB at 1m from 1 W).

Thanks everyone!

[antonis]

I took a trusty MPF102 out of the drawer and tied the Gate to ground and Source to ground via the 2.4k resistor that was on the board.  Voila.  0.65 volts on the gate (about 0.8mA Id).

Are you sure about that..??