DIY BJT buffer design

Hi

This is my version of the class A BJT transistor buffer from Building a buffered Gainclone chip amp. ("Simple class A discrete buffer circuit"). I've omitted the DC blocking capacitor as my power-amp circuits (LM3886 units) already have them on their input.

Will four 1.5uF MKT capacitors be sufficient for decoupling? It will be powered by a regulated +/-15V supply.

Other than that, just running this design past people more expert than myself (there are a lot of you, trust me) to see if my version is sane and workable before building it.

Regards...
 

Attachments

  • lm3886 bjt buffer pcb.gif
    lm3886 bjt buffer pcb.gif
    17.8 KB · Views: 719
  • lm3886 bjt buffer circuit.gif
    lm3886 bjt buffer circuit.gif
    25.1 KB · Views: 750

Ian Finch

Member
Paid Member
2010-04-11 4:22 am
Coffs Harbour, NSW
Hi
I assume you simulated this circuit or at least adapted from something successful,
but why the 1 amp BC639? Sure, this a class A output stage but what do you have
in mind as a load and your planned quiescent current? If you need high output current
capacity and hence will have high quiescent current and dissipation, the T0126 version
BD139 is a lot easier to arrange cooling for. OTOH, if this is a marginal rating issue,
BC327 or 8 are no slouch at this type of duty.

A simple buffer will not have many layout issues provided your supplies are well filtered
for the current required. Bypass requirements are pretty much related to that too.
Not sure that placing the power input terminals central and running the 15V rails between
the in and out signal terminals is a good idea, though. Generally, keep them well apart.

Good luck and watch pinouts if trying substitutes!
 
Hi
I assume you simulated this circuit or at least adapted from something successful,
but why the 1 amp BC639? Sure, this a class A output stage but what do you have
in mind as a load and your planned quiescent current? If you need high output current
capacity and hence will have high quiescent current and dissipation, the T0126 version
BD139 is a lot easier to arrange cooling for. OTOH, if this is a marginal rating issue,
BC327 or 8 are no slouch at this type of duty.

A simple buffer will not have many layout issues provided your supplies are well filtered
for the current required. Bypass requirements are pretty much related to that too.
Not sure that placing the power input terminals central and running the 15V rails between
the in and out signal terminals is a good idea, though. Generally, keep them well apart.

Good luck and watch pinouts if trying substitutes!

Thanks for comments. As it's driving a chip-amp, the load is small, way way below, as you suspect, the 1 amp 639s specification. The choice of BC639 was dictated by what is locally available in retail stores. There are many other choices but that unit was the only one the original article at Decibel Dungeon mentioned that I can source locally, so at least I know it works without any major problems. I may get into trying different BJTs in the future (in case it turns out I don't have a life after all - you never know). Your point about the supply rails going close to the signal lines is an excellent one - thanks for pointing it out.

Cheers
 
Hi,
from my experience the 1.5µF caps should be way sufficient for local onboard buffering of the supplies.
Especially when there is only an intermediate load to drive like 10k.
Wouldn't run the buffer on much lower load impedances than that though without increasing the standing current, as the max out voltage before (current limit induced) clipping is 0.6V/R5*R_load.

Also take a look at your voltage-regulators in combinination with the buffercircuit, and what else may be running off them.
Voltage-regulators are amps too, so they have a chance of oscillating too if not treated with care.
Agreed, its not very likely - haven't whitnessed an oscillating voltage-regulator-IC myself so far.
But I also don't know what your voltage ragulation looks like. Hence the hint

Regards
 

Bonsai

Member
Paid Member
2003-07-25 10:44 pm
Europe
www.hifisonix.com
R3 and R9 seem like very high value for a bipolar input. Are you feeding this circuit from a DC coupled source?

The current source emitter load is around 5mA. This means, assuming an hfe of 300 (my guess), the base current of Q1 and Q4 will be circa 17uA. Across a 100k bias resistor this is a voltage drop of 1.6V - which translates directly into loss of headroom. Ugly.

Most source equipment (CD player, iPhone/iPod)now days will easily drive 10k, and 5k is usually no problem.
 

Ian Finch

Member
Paid Member
2010-04-11 4:22 am
Coffs Harbour, NSW
Hi
I had a look at the Simple Discrete Buffer Circuit you referred to on the the Decibel Dungeon forum. Now I understand how the original poster came up with BC637 actually but it seemed he was throwing in absolutely anything that came to hand and could only have assessed the results by ear.

The BC 547 would have been the appropriate choice. Really, that was one sure way to the bin for the project when you would still still have sub-optimal results after countless guesses at what combinations of parts might sound better before the smoke escapes.

Bonsai gives you a good starter to designing a properly working circuit there. I would follow that up with the aim of learning to design your own electronic circuits and make part selections properly. Then you will be streets ahead of that guy's guessometry method and a lot happier with the sound. :cool:
 
Last edited:
R3 and R9 seem like very high value for a bipolar input. Are you feeding this circuit from a DC coupled source?

The current source emitter load is around 5mA. This means, assuming an hfe of 300 (my guess), the base current of Q1 and Q4 will be circa 17uA. Across a 100k bias resistor this is a voltage drop of 1.6V - which translates directly into loss of headroom. Ugly.

Most source equipment (CD player, iPhone/iPod)now days will easily drive 10k, and 5k is usually no problem.

10K is more reasonable for R3/R9?

I expect the input will be a CD player or the output of my PC's Xonar Essence sound card, so, no, not DC coupled.

Thanks for all the info!

Hi,
from my experience the 1.5µF caps should be way sufficient for local onboard buffering of the supplies.
Especially when there is only an intermediate load to drive like 10k.
Wouldn't run the buffer on much lower load impedances than that though without increasing the standing current, as the max out voltage before (current limit induced) clipping is 0.6V/R5*R_load.

Also take a look at your voltage-regulators in combinination with the buffercircuit, and what else may be running off them.
Voltage-regulators are amps too, so they have a chance of oscillating too if not treated with care.
Agreed, its not very likely - haven't whitnessed an oscillating voltage-regulator-IC myself so far.
But I also don't know what your voltage ragulation looks like. Hence the hint

Regards
Thanks. I'll post my regulator circuit today and provide a link - it's already been under scrutiny in the Power Supply forum but I haven't posted my last revision there yet.

Do you recommend changing R5?

Hi
I had a look at the Simple Discrete Buffer Circuit you referred to on the the Decibel Dungeon forum. Now I understand how the original poster came up with BC637 actually but it seemed he was throwing in absolutely anything that came to hand and could only have assessed the results by ear.

The BC 547 would have been the appropriate choice. Really, that was one sure way to the bin for the project when you would still still have sub-optimal results after countless guesses at what combinations of parts might sound better before the smoke escapes.

Bonsai gives you a good starter to designing a properly working circuit there. I would follow that up with the aim of learning to design your own electronic circuits and make part selections properly. Then you will be streets ahead of that guy's guessometry method and a lot happier with the sound. :cool:

Thanks. I disliked the pick and mix randomness of transistor selection too (as my experience/education levels increase, I keep finding things on that site I don't like the look of). The simplicity of this circuit is its attraction, especially as more popular JFET circuits often use difficult to find components.

I just looked up the 547 and can get it locally through AZtronics. Looks like one of those parts available everywhere.

Thanks for the help!
 

Ian Finch

Member
Paid Member
2010-04-11 4:22 am
Coffs Harbour, NSW
I keep finding things on that site I don't like the look of. The simplicity of this circuit is its attraction, especially as more popular JFET circuits often use difficult to find components.
If you haven't already visited this popular site, do - often!
Elliott Sound Products - The Audio Pages (Main Index)
Project 37 (The DOZ preamp) might look a little familiar in its topology and good to learn from.
 
10K is more reasonable for R3/R9?
it would reduce the offset, as bonsai pointed out.

Of course if you had a source that can't drive 10k then the circuit is pointless.

What is the input impedance of your gain clone. if it is somewhere around the 10k your source should have no problem driving it, again as bonsai pointed out.

So if that's the case (your sources beeing able to drive the GC directly, and the ones you mention very likely are able) then there is no technical reason to add this stage.
Do you recommend changing R5?
Only if your going to drive loads with significantly less impedance than 10k like 500 Ohm, because then your buffer would clip at levels of around 2.5Vpeak (see my first post for the math) which is roughly the level of CD-Player outputs (correct me if I'm wrong here).
 
Try get the BC547B (or C). They have better Hfe (i.e. current gain) than the BC547A. See the datasheet here for how the grading works.
Good point. They're actually quite hard to find. :(

If you haven't already visited this popular site, do - often!
Elliott Sound Products - The Audio Pages (Main Index)
Project 37 (The DOZ preamp) might look a little familiar in its topology and good to learn from.
I am, indeed, familiar with the site and have used many of the designs (verbatim or as inspiration) in other circuits. The higher complexity of the Elliot circuit makes me worry this Dungeon design is going to have issues. I'll build the Dungeon buffer anyway and do some tests - should be interesting.

Does one omit the feedback resistor in the ESP design for 0db gain? I'm familiar with NFB in op-amp ICs but have much to learn about fundamental transistor and amplifier theory. Am looking into doing an EE course at the TAFE (apologies to foreign readers: TAFE = partly government funded vocationally oriented polytechnic-like tertiary education organisation).

it would reduce the offset, as bonsai pointed out.

Of course if you had a source that can't drive 10k then the circuit is pointless.

What is the input impedance of your gain clone. if it is somewhere around the 10k your source should have no problem driving it, again as bonsai pointed out.

So if that's the case (your sources beeing able to drive the GC directly, and the ones you mention very likely are able) then there is no technical reason to add this stage.

Only if your going to drive loads with significantly less impedance than 10k like 500 Ohm, because then your buffer would clip at levels of around 2.5Vpeak (see my first post for the math) which is roughly the level of CD-Player outputs (correct me if I'm wrong here).
Thanks. Sorry for the stupid questions: I should know more about this subject than I do! :)
 

Ian Finch

Member
Paid Member
2010-04-11 4:22 am
Coffs Harbour, NSW
The higher complexity of the Elliot circuit makes me worry this Dungeon design is going to have issues. I'll build the Dungeon buffer anyway and do some tests - should be interesting.
Does one omit the feedback resistor in the ESP design for 0db gain? I'm familiar with NFB in op-amp ICs but have much to learn about fundamental transistor and amplifier theory.
No need to apologise for noobiness. We are all noobs at some time and are just fortunate to now have DIYAudio to help speed up the learning process.

The only complexity of the ESP preamp circuit is due to inclusions like quiescent current adjustment and some linearising circuitry, including a little feedback. In practice, you'll need to do some things to your buffer to establish acceptable operating conditions and, since they will be a critical part, the power supply(s) too.
Note that just removing the feedback resistor R5 will disturb the DC conditions of the input transistor and lead to some problems you won't need.
Feedback also determines the gain of the amplifier by the ratio of the voltage divider formed by R5+R4, so the appropriate change is to increase R4 which, due to the cap., only affects AC gain. You won't be able to remove all gain and the reduction of feedback will result in increased distortion. However, this is a good testbed for checking the amp theory out and show you the problems ahead for such simple buffer designs.

Having a scope, even a 16bit, 11kHz bandwidth crappy soundcard one will make things very interesting with the right free software. At the output levels here, you should have no overload issues but use some protection such as 2 series pairs of IN4148 diodes wired in opposite direction across the analog input terminals anyway. A blown input is a blown card so don't try it with on-board cards if you can help it! Make sure that you can derive a 1kHz test tone that works in real time while the device is being measured or you will need to buy or build a very low distortion oscillator.

The ESP tutorial on amplifier operation will bring the theoretical issues to your attention but a good simple text on amps, even opamps, (Check Amazon, Fishpond etc.) is the go. By all means play with the simple buffer circuit, but use the opportunity to learn something that will give you much wider understanding.
Personally, I think the net is absolutely great for cherry-picking facts, papers and docs. here and there and the 'wow factor' of stunning graphics etc.
For formal learning, there is nothing like a good book and a more comfortable place than squinting at poorly scanned documents on your computer's VDU.
 
Last edited:
No need to apologise for noobiness. We are all noobs at some time and are just fortunate to now have DIYAudio to help speed up the learning process.

The only complexity of the ESP preamp circuit is due to inclusions like quiescent current adjustment and some linearising circuitry, including a little feedback. In practice, you'll need to do some things to your buffer to establish acceptable operating conditions and, since they will be a critical part, the power supply(s) too.
Note that just removing the feedback resistor R5 will disturb the DC conditions of the input transistor and lead to some problems you won't need.
Feedback also determines the gain of the amplifier by the ratio of the voltage divider formed by R5+R4, so the appropriate change is to increase R4 which, due to the cap., only affects AC gain. You won't be able to remove all gain and the reduction of feedback will result in increased distortion. However, this is a good testbed for checking the amp theory out and show you the problems ahead for such simple buffer designs.

Having a scope, even a 16bit, 11kHz bandwidth crappy soundcard one will make things very interesting with the right free software. At the output levels here, you should have no overload issues but use some protection such as 2 series pairs of IN4148 diodes wired in opposite direction across the analog input terminals anyway. A blown input is a blown card so don't try it with on-board cards if you can help it! Make sure that you can derive a 1kHz test tone that works in real time while the device is being measured or you will need to buy or build a very low distortion oscillator.

The ESP tutorial on amplifier operation will bring the theoretical issues to your attention but a good simple text on amps, even opamps, (Check Amazon, Fishpond etc.) is the go. By all means play with the simple buffer circuit, but use the opportunity to learn something that will give you much wider understanding.
Personally, I think the net is absolutely great for cherry-picking facts, papers and docs. here and there and the 'wow factor' of stunning graphics etc.
For formal learning, there is nothing like a good book and a more comfortable place than squinting at poorly scanned documents on your computer's VDU.
Thanks for all that and thanks so much for your (and everyone's) help.

(Incidentally, I have a CRO, rather old and definitely uncalibrated but it works - rescued off the top of someone's curbside hard rubbish dump. :) )