Transistor sounds

[lumanauw]

I'm aiming at making good sounding power amp. So, I decided that I have to know what kind of sound every different transistor make.
The idea is to make buffer (gain=1) with different kind of transistor. Because I'm using low rail (+/-15V), I use power transformer's primary (0-110V-220V) as load.
The configuration is differential preamp, but with different kind of transistor. With input R=4k7 and feedback R=4k7, I can get AC gain=1. I put A/B switch, so I can switch the signal feeding through the preamp or bypassed (not thru preamp), so I can hear what difference it makes compared with original signal.

A is bipolar (BD139). Compared to bypassed, it gives less detail, the sound is so-so.----having voltage gain.

B is mosfet (IRF510). Compared to bypassed, it gives more detail in sound, especially in trebles. Nice sound----having voltage gain.

C is CFP Jfet (2SK30+BD140+2k2 resistor). Nice sound. But something is strange. It don't have voltage gain. Even if I eliminate 4k7 feedback resistor, it already perform a buffer. If I put this 4k7 feedback resistor, it attenuates the sound.

Why is this CFP have no gain? How can I make it to have gain?

[MikeB]

hi !

i think, you shouldnt drive the jfets without any resistor to ground.
for example some 47k to ground ?
Maybe you should change this to all 3 circuits to get a more practical
testing ?
have you measured the current through the jfets ? it shouldn't
be more than ~1ma. (this might be the problem with your other amp,
maybe the current was to high, and the jfets "saturated")
according to calculations it is ~320uA (0.7v across the 2.2k)

You maybe forgot the 4th option, small bjt + cfp ?
My experience with these was very good sound (dynamics,details,mids,bass)
but something wrong with trebles. (sss and zzz with vocals)

Mike

[X.G.]

Originally posted by lumanauw
Why is this CFP have no gain? How can I make it to have gain?

hello,

the reason maybe the Idss of K30 is too small.Check the Idss,and set the idle current of K30 within the Idss( 1/2 of Idss maybe the best)

[wenye]

Hi lumanauw,

For circuit C, it's really a unity gain buffer with active NFB. The BD139/BD140 perform full feedback for input k30 pair.

If you want to have gain in this circuit, you should connect C pin of BD139/BD140 to negitive supply then it gain. But, in the other hand, the output distortion will increase a lot dut to None feedback was emploied.


Wenye

[lumanauw]

i think, you shouldnt drive the jfets without any resistor to ground.

Hi, Mike,
I think the basic cct is a differential with inverting output. So the gnd in the right transistor should do it for both of the transistor's reference.

have you measured the current through the jfets ? it shouldn't be more than ~1ma. (this might be the problem with your other amp, maybe the current was to high, and the jfets "saturated") according to calculations it is ~320uA (0.7v across the 2.2k)

Yup, the drop in the 2k2 resistor is 0.62V. I think its OK here.

You maybe forgot the 4th option, small bjt + cfp ?My experience with these was very good sound(dynamics,details,mids,bass)
but something wrong with trebles. (sss and zzz with vocals)

I got this idea of making C cct as Jfet CFP is from your design!. I wanted to compare the sound of BJT, MOSFET, JFET. For all BJT CFP, will it differs so much than cct A?
As for your experience with sss, zzz, you should try this JFET CFP. It has no sss or ZZZ at all, very nice sound.

the reason maybe the Idss of K30 is too small.Check the Idss,and set the idle current of K30 within the Idss( 1/2 of Idss maybe the best)

Haven't check the Idss yet. But in this CFP, it runs only less than 0.5mA. Idss of 2SK30 is less than 0.5mA?

For Jfets, we can match 2 things, Vgs and Idss. If we want to make differential pair with Jfets, which is more important to match?

For circuit C, it's really a unity gain buffer with active NFB. The BD139/BD140 perform full feedback for input k30 pair.

I dont quite follow it. In CFP output stage, it is really a unity gain voltage, since we take the output from the sources of FET.
But in this design, I took the output from emitors of the BJT (---drains of the FET). How come it is still a buffer at upper output?

If you are right, then Mike's suggestion on all BJT CFP should act the same too.
How about your SIM on your design, Mike? Does all CFP is a buffer with no gain, no matter where we take the output (at top or bottom?
Usually I can trace it is a buffer or not by connecting drops of Vbe. But in this CFP differential, I still haven't found why it should perform buffer if we take the output from the top.

[MikeB]

hi lumanauw !

You need some correct working/bias point, which MUST be identical
on both sides. The difference between these is amplified by the
diffamp.
I think you really should add the resistor, just before the 200ohm.
I jfets are not biased correctly, they can open the gate.

Ok, drop of 0.62v is 281uA, this is not much !
I don't know the k30, can't imagine that it saturates at this current.
Do you have acess to k170 ? I know this one works !
Or bf245 ? This one is very common.
I matched the jfets only for Vgs at ~300uA (difficult enough)
Did you measure the drop on both sides ?
Did you measure both actual Vgs ?

The problems with the sss,zzz are now without jfet. With jfets
i had none of these problems. But still need to check if its the symetrical
topology. I will simply replace the bjts in this amp by sk170/sj74,
then i know ! (Maybe the bc546b/556b is too cheap ?)
I think the cfp has a great influence on all devices.
It seems that the combination jfet+cfp is very promising.
My actual amp is symetrical bjt+cfp. (the 12transistormonster)
What do you recommend as small complementary bjt ?
I like the bc546b/556b as they are fast (ft=300mhz), have a real
hfe of 300, go up to 60v and cost only 4cent...

And the CFP explodes the gain of the diffamp. Without CFP its about
1:50 (with sk170). With CFP it was > 1:1000 !
(Without emitter-degenaration-resistors)
With emitterdegeneration the gain is about Rc/Re (~98%)
This means: Re = 22ohm, Rc = 2k Gain = ~1:88
With Re = 0ohm Gain = very very high (divide by zero )
You can say, the CFP flattens out the Vbe-curve.
In combination with Re you get some "ideal" transistor.

Do you remember what CFP stands for ? i only remember ComplementaryFeedbackP...)

Mike

[Eva]

That test circuit is useless. Any minimal difference in bias currents between both legs will cause severe transformer saturation and severe output distortion [the transformer starts to act as a pair of resistors and push pull effect disappears]

The results obtained from that circuit in each case are much more dependent on biasing imbalance and transformer saturation than the gain devices themselves

Note also that a 220-110V dual primary is extremely inductive [both leakage and coupling] and that the voltage swing required at one leg to get +-1V at the other leg at 20Khz may be more than +-15V

The test circuit by itself is already a nice distortion generator

[sam9]

My understanding, which may be faulty, is that a simple CFB arrangement has unity gain (i.e., no gain) because it is 100% feedback. I think you can create some gain by inserting resistance betwenn the source/emitter of the upstream device and the collector of the downstream device. This will reduce the feedback, of course. You see this ocassionally in output stage configurations - appearently it is thoght to subdue potential oscillation, but I'm not clear if that is the real reason.

I suggest you look around the web for audio test software that can use a soundcard. This lets you confirm frequency response, distotion, do some spectral analysis, etc. Subjective evaluations can be tricky. For instance, it's an old audio store trick to crank up the trebble a bit on units they are pushing as the impression is often heard as "more detail", "more revealing". There are even some signal processors that deliberately add distortion to high frequencies to make them sound "more alive". Since you are dealing with input stages, anything that happens here is going to be amplified approx. 30db down stream. What seems like a minor change, quirk or tweek here can end up being really big by the time it reaches the speakers.

[MikeB]

hmm, i've already spent some time in spice, the numbers i wrote
are from sims. In sims the cfp in this configuration creates a lot
of gain. I've already built 2 amps using cfp in the input, and they
behaved as they should.
Maybe you should look at the cfp-bjt as a helper. A slight currentchange
from the input-transistor changes the current in the cfp-transistor.
The cfp-transistor simply scales up the behaviour of the other
transistor.

I don't fully trust in spice, but spice is never that wrong...
I do fully trust realworldcircuits, and they do what spice predicted.

The CFP itself does not have gain, but it behaves like a transistor
capable of very much gain. (like a transistor with a very flat Vbe-curve)

I like CFPs, as you can use very high OL-gains, without getting an
unstable circuit. (like currentmirrors on top of diffamps tend to)
In spice the the cfp outperforms the currentmirrors. The combination
of both is getting difficult, that's too much gain. Then you get
openloopgains > 1:1.000.000
But it worked with jfets...
At 20khz full power harmonics were at -120db. (in sim)

Mike

[lumanauw]

What do you recommend as small complementary bjt ?

For small voltages I like BC550C and BC560C. Low noise and very high gain. For bigger voltage I use MPSA06/56.

With CFP it was > 1:1000 !

What gain? Current gain? Voltage gain? Strange, I got no voltage gain at all.

That test circuit is useless. Any minimal difference in bias currents between both legs will cause severe transformer saturation and severe output distortion [the transformer starts to act as a pair of resistors and push pull effect disappears

The results obtained from that circuit in each case are much more dependent on biasing imbalance and transformer saturation than the gain devices themselves

Note also that a 220-110V dual primary is extremely inductive [both leakage and coupling] and that the voltage swing required at one leg to get +-1V at the other leg at 20Khz may be more than +-15V

The test circuit by itself is already a nice distortion generator

What kind of distortion? 2nd harmonic?

I can use Resistor for output drop, but since the voltage is low, I choose using transformer primaries.
Thanks for the info. I will put 47k resistor in left transistor (like Mike's suggestion), but not to ground, but to wiper of 100K VR, which left and right end goes to +15V and -15V.
I also will put 10ohm resistor in 0 and 220V point to set the balanced current between each leg, adjusted by the VR.
If I can set the same current between those 2 legs, will it be that what you are saying is minimal?
EVA, you should try this preamp. Inspite of it "distortion generator"/flux imbalanced from your point of view, it sounds nicer than to put R for load. Maybe this distortions makes it sounds nicer.

My understanding, which may be faulty, is that a simple CFB arrangement has unity gain (i.e., no gain) because it is 100% feedback. I think you can create some gain by inserting resistance betwenn the source/emitter of the upstream device and the collector of the downstream device. This will reduce the feedback, of course.

Resistor between source/emitor of upper device and collector of down device? Isn't that is the 2k2 resistor? Or I get it wrong?Where to put the resistor to have gain in CFP? you have a drawing maybe?

Since you are dealing with input stages, anything that happens here is going to be amplified approx. 30db down stream. What seems like a minor change, quirk or tweek here can end up being really big by the time it reaches the speakers.

Thats exactly what I'm searching. Something good upstream will be very good downstream.

hmm, i've already spent some time in spice, the numbers i wrote are from sims. In sims the cfp in this configuration creates a lot of gain. I've already built 2 amps using cfp in the input, and they behaved as they should.
Maybe you should look at the cfp-bjt as a helper. A slight currentchange
from the input-transistor changes the current in the cfp-transistor.
The cfp-transistor simply scales up the behaviour of the other
transistor.

I don't fully trust in spice, but spice is never that wrong...
I do fully trust realworldcircuits, and they do what spice predicted.

The CFP itself does not have gain, but it behaves like a transistor
capable of very much gain. (like a transistor with a very flat Vbe-curve)

I like CFPs, as you can use very high OL-gains, without getting an
unstable circuit. (like currentmirrors on top of diffamps tend to)
In spice the the cfp outperforms the currentmirrors. The combination
of both is getting difficult, that's too much gain. Then you get
openloopgains > 1:1.000.000
But it worked with jfets...
At 20khz full power harmonics were at -120db. (in sim)

Mike

I think you are talking about current gain between base and emitors. Is it possible that CFP have enormous current gain, but no voltage gain?