BA-3 Amplifier illustrated build guide

6L6

Moderator
Joined 2010
Paid Member
What is the BA3 gain?

About 30. This can be adjusted by choice of Mosfets and the ratio of R10 to R11

Does the P3 being set off mid resistance change the gain?

If it does, it's not enough to notice. The Jfet stage is a buffer. That said, it might be worth looking into to see if there is any change in gain. But as it will be dealing with a few percentage points off of unity, it's going to be inconsequential compared to the gain of the 2nd stage.

Can the gain be reduced or increased to suit other requirements?

Of course. And depending on your requirements you may want to build an entirely different circuit. But if you want to build a BA-3, a really good place to start is the stock values.
 
Thank you.
Really clear answers.

except I think you meant R13 to R11 ?

Yes I'm thinking of building a BA3 clone on some protoboard.
And wanted to know if it was suitable as a headphone driver.
But with gains way above 2times (+6dB) that would not be sensible.

I might even use latFETs instead of the 2sk2133/j313 to compare first build to second build.

I might even look at some of Juma's circuits to see what global feedback could do to a BA3. But I might need help with stability compensation If I get that far.
 
Last edited:

6L6

Moderator
Joined 2010
Paid Member
Yes, I said the wrong resistor pair. Oops.

For headphones you could use it, but a follower stage makes more sense to me.

But do you (the user, not Andrew specifically) need any gain from a HPA? (Which is really a question about your specific headphones.) For example, I have wickedly efficient, very easy to drive headphones 106dB/mW, 25ohm, so I need a buffer and an attenuator from even the wimpiest sources... Whilst somebody with planar headphones will need a different circuit.
 
Official Court Jester
Joined 2003
Paid Member
That's appears more complex than I had hoped. Sounds like a project for the advanced experimenter on a different thread. Thanks.

gain of input part is pretty much A1= ((((R6/(R3+(2 x R5))+(R7/(R4+(2 x R5))))
R5 figuring in twice , being common for both Jfets

gain of output part is pretty much A2=R13/R10 or R13/R11 ;
or (without adding) is because mosfets are working unison ....... (A class stage , their source voltages are in phase , drain voltages are in phase ........ so no voltage adding)

different mosfets ( meaning different xconductane ) will not influence gain that much , due to pretty heavy source degeneration

overall gain is sum of these two

ximage.php,qwidth=600,aimage=archive,_fig4.jpg,aauth=d69b4e5769f27dcb6ce8ede9fff00c18.pagespeed.ic.ulFJb0YcD5.jpg
 
Last edited:
That's clear enough for even me to understand. But as you describe, the minor increase in performance probably doesn't justify moving from the proven standard mosfet. I will most likely mess with the resistor values at some point in the future - as has been noted in various posts around the forum.

Thanks for the details. :up:
 
Official Court Jester
Joined 2003
Paid Member
if you can get Toshibas without harm to leg and kidney , go for them

if not , plain vanilla IRF will do the job .......... and P3 will ensure magic

to clarify :

....

gain of output part is pretty much A2=R13/R10 or R13/R11 ;
or (without adding) is because mosfets are working unison ....... (A class stage , their source voltages are in phase , drain voltages are in phase ........ so no voltage adding)

...


written little different : A2 = (((R13/2)/R10) + ((R13/2)/R11)) ...... A class output stage - each half (mosfet) is seeing half the load (R13)

or , if you wish , A2= R13/(R10+R11)
 
Last edited:
That's clear enough for even me to understand. But as you describe, the minor increase in performance probably doesn't justify moving from the proven standard mosfet. I will most likely mess with the resistor values at some point in the future - as has been noted in various posts around the forum.

Thanks for the details. :up:
except that Zen is wrong.

Firstly, the first stage gain is stated as unity gain and there is a formula. The formula gives a different result from unity gain.

Secondly, the second stage gives a formula for gain and then says double it. That could mean add the two halve gains or multiply, fortunately both these will give the same result. Then follows by saying that the gain is actually less than the formula predicts. What? How much?

Thirdly, the first stage gain and the second stage gain are multiplied.

Finally gm comes into this somewhere, but it's not stated to what extent gm varies the gain.

It was these anomalies that got me confused.
I asked and was told to read more.
So I did.
The anomalies still exist.
The answers given since I first queried this have not cleared this doubt up for me.
All we have is that the gain is not anywhere from unity gain to 60times, but is actually 30times.

It seems that the only way to find out is to build and measure. Much as you have said
I will most likely mess with the resistor values at some point in the future
 
Last edited:
as usual

:D

edit :

find How to gain gain_A Reference Book on Triodes in Audio Pre-Amps from Burkhard Vogel , read it few times from cover to cover

see ya in few years

then you'll know more and I'll be even dumber

entropy , that is

:p
Where is your justification for all the alleged errors in your post that I have pointed out to Bcmbob?

If you want to define your knowledge in the terms you have posted then I'll not argue.
 
Official Court Jester
Joined 2003
Paid Member
I'm lazy* to do that - split circ in upper (or lower) half , make current and voltage analysis , write down everything including xconductance , explain why one scaling factor for value of common degeneration impedance in first and why another scaling factor for second stage common loading impedance .... put few necessary sketches and post everything here

gm (S,µ,whatever) is in datasheets , deriving Ri from it is easy , Rl is known from schematic , proper value of degeneration impedance is clearly seen and scaled from schematic ..... so that's trivial ....... but also irrelevant , counting on few % difference to simplest formulas based on resistor values only

as I wrote before - put few parts in LTSpice , or fill few sheets with numbers and sketches and you'll know answers

I know both and I really don't care am I wrong or not ....

so , take it or leave it

*edit: for proper numeric analysis , one need to have exact value of gm for both JFets and Mosfets ...... and there is no other way than to measure it for chosen static working point , that being of most importance for mosfets - datasheet graphs aren't so detailed and precise as for small sinal JFets ........ I'm especially lazy to do that .....
 
Last edited:
Member
Joined 2009
Paid Member
So If I read again and more carefully
we have:
first stage gain = (R6+R7)/R5
second stage gain <R13/R10 and then double that giving < R13/(R10||R11)
The total gain would be first stage gain times second stage gain:
BA3 gain < (R6+R7)/R5*R13/(R10||R11) < (100+100)/100 * 332/(22||22) < 2*30.2 <60.4

So back to my questions:
What is the BA3 gain?
Does the P3 being set off mid resistance change the gain?
Can the gain be reduced or increased to suit other requirements?

or is unity gain (+0dB) to somewhere around 60.4Times (+35.6dB) close enough for every one else?

Hi Andrew.

There were many tries to clear Nelson's gain prediction of 36dB for BA-3.
Indeed Botte and me once wrote that the single BA-3 shows only 8x gain in vivo!

This link shows the combined knowledge of the best experts, only the results differ.......

http://www.diyaudio.com/forums/pass-labs/195303-burning-amp-ba-3-a-60.html#post2866494

But I have to shut my mouth because I understand nothing....

:)
 
Thanks for that link to a discussion that I had forgotten had even happened.

I will have to assemble and see what I get.
Then decide whether it will do what I need, probably not, so I may end up discarding it.
Shame that so much time will be wasted rather than modeling the circuit.
NFB would make this all so easy. Gain depends on resistor ratios until the open loop gain becomes too low.