Here is my last creation - DOCAFET class A amp!

[slepro]

Hi everybody,

I would like to introduce you my last creation, the DOCAFET amp (for doubly cascoded JFET). Please don't laugh at his name, I'm not having a huge artistic mind!

The sound seems to be very good, I still did not build it into a casing, but I'm using it as a prototype for now.

I would like to have your opinion on the design because for both channels I built, they are having thermal noise that can be heard through the speakers. By looking at the schematics, does anyone of you have an idea on what may cause this noise?

The amp never get into oscillation, so I can say it is stable. The quiescent current is also very stable. Please note that the 50V dc supply comes from a regulated PSU, and the 35V dc supply is only filtered by huge capacitors.

I thought that the source of the noise may be the zener diodes of the cascoded differential amp, but even with capacitors across them, the noise is still there. Maybe the source is the JFET's? I don't know...

I'm open to all your comments, so don't hesitate to post your opinions!

Thanks all

Sébastien

 

[lineup]

noise HF filter RC-filter RC input cap

hi
nice work! Potential very interesting schematic!

NOISE.
What could produce noise so big we can hear it
is around input or the input stage itself.


Correctly setup, most input JFET would not create such noise.
Not so we could hear it.


My guess is HF noise. Very high above the audio.
This noise is present everywhere
and any UN-SHIELDED wire will act as an antenna for it.


To possibly cure this, add one 220pF - 470pF - 1nF, parallel across input resistor R13(47k)
and to make this RC filter for HF
you add this Resistor, 470 - 1k Ohm
in series with input Capacitor C5(10uF).
This 470 Ohm resistor could be After or Before C5.

Most usually resistor is placed between C5 and R13,
from end of C5 to 'top of' R13 = in series with input cap.

I would try 470 Ohm + 470pF first.
If needed, I would use 1 kOhm + 1nF, for high frequency roll-off.


If you have some potentiometer or other resistance
before the input of power amp you may skip the resistor.
Then just adding one 220pF or 470pF film cap in parallel Across R13
could be best solution.


This RC filter is a good idea, even if in your amp
the noise have another cause.
But try this first.
Then we may find something else, that possibly adds noise in your amp.


Good Luck!
lineup

 

[darkfenriz]

Seems like a grounding issue for me.
C4 value is drwing mistake, isn't it?

 

[slepro]

lineup : Thanks for your help, I'll try that as soon as I can. I am right to say that it would be better to put a capacitor of the same value across R14 as the one across R13?


darkfenriz: I don't believe that is could be a ground issue, since the pcb is having ground plane all around on both sides. For C4, you are absolutely right, the value should be 100pF. The prototypes are mounted with this value, the 4.7nF is an error on the schematic only.

Does anyone think that it may be the zener D2 and D3 or the transistors type for Q7, Q8, Q17 and Q18?

 

[Leolabs]

Seems like a grounding issue for me.

It's ok for a schematic session.Just need to pay attention during layout session.

C4 value is drwing mistake, isn't it?

Maybe not.But it's kind unusual value.

 

[lineup]

lineup :
I am right to say that it would be better
to put a capacitor of the same value across R14 as the one across R13?

No, slepro
you are VERY wrong, if say so

Because would make a lowpass RxC, RC-filter
with R11 x Cap

you remember I told you could put a resistor, 470 Ohm in series with input cap.
R11, 560 kohm is a little bit higher, in fact >1000 times higher
so the lowpass roll off would be like 1000 times lower!!!!!!!!!!


-3dB roll off frequency would be:
( PI ~3.14 )
f = 1/ ( 2 x 3.14 x R x C)
and
using 560000 Ohm and 0.000000000470 Farad
we get f ~ 604 hertz
.... would mean your amplifier would have certain problems, for example with 1 kHz sounds or higher......

..... instead if using 470 Ohm and 470pF
we get f ~ 720484 hertz = 720 kHz
which can't effect audio at all = 20-20.000 Hz

---------------------------------------------

besides, the input transistor
is having contact with wires, signal cables, that come from outside the amplifier circuit
maybe even from outside amplifier chassis, via input jacks.

This is why you do not need such a cap at the other JFET input.
The input RC filter limits what enters your amplifier.

I guess you are only interested in Audio signals, in this case,
and you do not want to use your amplifier as a radio receiver
catching those antenna produced RF signals from signal input wires.

The simplest radio antenna is an un-shielded wire, of suitable length,
as most people probably know is working fairly good.



lineup
a former DX-ing radio amateur - now limited to the audio frequency band

 

[fab]

noise problem...

Hi Sébastien,

Nice schematics.

my guess on your noise problem if we reject the grounding scheme issue and you have followed lineup recommendations:

1) without having an appreciation of the level of noise you have, the 2N5458/5461 are not the best jfet for low noise application. You may try 2SK170/2SJ74 for comparison purpose. Ensure to have the appropriate IDss current for your circuit.

2) you can bypass all zener with about 100nF cap.

3) R11 560K for feedback resistor is an unsual high value (hiher resistance value gives higher thermal noise. Use 47K like R13 and it will have a better DC offset optimization. Of course ensure to reduce R14 accordingly.

Notes:
I can not see the Zobel at your output.
C1 and C10 are unusual for compensation cap location.
Have you simulated your schematics to have an idea of the stability of your amp (gain and phase margin) in resistive and capacitive loads?

Good luck

 

[slepro]

Re: noise problem...

No, slepro
you are VERY wrong, if say so

Because would make a lowpass RxC, RC-filter
with R11 x Cap

you remember I told you could put a resistor, 470 Ohm in series with input cap.
R11, 560 kohm is a little bit higher, in fact >1000 times higher
so the lowpass roll off would be like 1000 times lower!!!!!!!!!!


-3dB roll off frequency would be:
( PI ~3.14 )
f = 1/ ( 2 x 3.14 x R x C)
and
using 560000 Ohm and 0.000000000470 Farad
we get f ~ 604 hertz
.... would mean your amplifier would have certain problems, for example with 1 kHz sounds or higher......

..... instead if using 470 Ohm and 470pF
we get f ~ 720484 hertz = 720 kHz
which can't effect audio at all = 20-20.000 Hz

---------------------------------------------

besides, the input transistor
is having contact with wires, signal cables, that come from outside the amplifier circuit
maybe even from outside amplifier chassis, via input jacks.

This is why you do not need such a cap at the other JFET input.
The input RC filter limits what enters your amplifier.

I guess you are only interested in Audio signals, in this case,
and you do not want to use your amplifier as a radio receiver
catching those antenna produced RF signals from signal input wires.

The simplest radio antenna is an un-shielded wire, of suitable length,
as most people probably know is working fairly good.



lineup
a former DX-ing radio amateur - now limited to the audio frequency band

It seems like you misunderstood me I'm not talking about putting a cap in // with R11 (feedback resistor) but in // with R14 which is the gate grounding resistor, as R13 for Q14 and Q9. I though that if some noise is generated across R13, there's should have some across R14, no?

I didn't had time today to test all that, but I'm impatient to see the results.

Hi Sébastien,

Nice schematics.

my guess on your noise problem if we reject the grounding scheme issue and you have followed lineup recommendations:

1) without having an appreciation of the level of noise you have, the 2N5458/5461 are not the best jfet for low noise application. You may try 2SK170/2SJ74 for comparison purpose. Ensure to have the appropriate IDss current for your circuit.

2) you can bypass all zener with about 100nF cap.

3) R11 560K for feedback resistor is an unsual high value (hiher resistance value gives higher thermal noise. Use 47K like R13 and it will have a better DC offset optimization. Of course ensure to reduce R14 accordingly.

Notes:
I can not see the Zobel at your output.
C1 and C10 are unusual for compensation cap location.
Have you simulated your schematics to have an idea of the stability of your amp (gain and phase margin) in resistive and capacitive loads?

Good luck

Thanks for your suggestion of JEFT's I will have a look at that!

The Zobel circuit may be added outside the pcb if necessary, but I would like to avoid it.

The reason I put C1 and C10 where they are, is to gain some output span of the voltage gain amplifier section. And I don't see any problem with these capacitors.

It is a good idea the scale down R11 a bit. But will it make a real difference since I'm driving JFET gates wich don't need any current to be driven (I know that at higher frequency's it not the same thing because of Cgs, but for audio frequency the effect is negligeable).

So, I will try (as soon as I can) to put a 330pF resistor across R13, I will add a RC network in series with C5, between C5 and R13, and I will scale down R11 to 47k and R14 to 3.9k.

 

[lineup]

no
I did not misunderstand you
"I am right to say that it would be better
to put a capacitor of the same value
across R14 as the one across R13?"

And the resulting RC filter
would be the value of R11 x Cap = 560k x 470pF (or 220pF)
even if Cap is across R14, attached to ground.

So my posted comment is still correct.
It would destroy the bandwidth of your amp.

You should wait with adding little cap in the feedback side
because you might not need it.
----------------


i agree with previous poster here,
Reduce the resistor divider
R11/R14, (560k/47k) setting the gain of amp =11.9+1 ~ Gain 12.9

for example using 100kohm for R11
and 6k8 for R14
would give Gain:
(100/6.8) + 1 = input x 15.7

-------------------------------------------------------------------
for a voltage supply of 35 V DC. ( 50V DC in input/VAS stage)
you may set the gain to 15-20, something

Gain=15.7, would make 1.0 Volt RMS input
have peak voltage swing output: +/- 22.2 Volt

Gain=20, would make 1.0V RMS input
have peak voltage swing output: +/- 28.3,
which still leaves a few volt margin before output transistor start clipping
35.0 - 28.3 = 6.7 Volt, across output transistors including emitter resistors 1R = 1 Ohm = R20,R23,R25

-----------------------------------------------

Start calculating, R11 and R14 and what gain you want.
But Keep R11 - the feedback resistor <= 100 kohm.
Most amplifiers do not have this resistor higher than 47 kohm.
And there are good reasons for this!


Regards
lineup

 

[wojtek5001]

I think, hat You shuld put small (~7 Ohm) resistors in bases of output transistors. The noises You hear may be results of HF oscillations.

I also think that C1 and C10 shuld be throw off. I din't done any calculations, but I think that open loop gain will be high enough without them. well, a bit subjectivism - I simply don't like caps in such way javascript:smilie('')

C4 seems to be too high in value. C4 and C8 shuld be equal (look at Leach amplifier, You can learn a lot from that, and i'ts very similar to Your). Again, I didn't done any calculations, but C4 is just to big for me javascript:smilie('')

I also agree with all above about input lowpass filter and grounding. You shuld carrefully check grounding topology;

 

[nikwal]

Did you try capacitive load? hmm.. I did a mini-version with almost the same topology.. http://hem.bredband.net/nikwal1/kifs.gif
and also a more advanced with current generators almost the same as yours and spice says 0.00002% thd on that one..
, kifs works ok , i built a simple prototype(and spice says 0.01%thd) but i did'nt listen to noise yet.. you might want to put more capacitors in the feedback somewhere??..

 

[AndrewT]

Hi,
you tell the future as well

I would like to introduce you my last creation,

or is a suicide attempt on the way?

 

[slepro]

Hi,
you tell the future as well
or is a suicide attempt on the way?

Maybe you'll prefer in french (my first language)

Je vous présente ma dernière création (dans le sens de la plus récente et non pas dans le sens de ma dernière à vie).

You may use http://babelfish.altavista.com/ for your convenience.

Thank you!

 

[slepro]

I did some research about noise and JFET and I saw (again) that a way to have less noise generated by the first stage transistors, is to put a resistor in series with the gates. So each JFET will have it's own gate resistor. The value should be something between 1k and 10k.

What do you think of that?

 

[lineup]

I did some research about noise and JFET and I saw (again) that a way to have less noise generated by the first stage transistors, is to put a resistor in series with the gates. So each JFET will have it's own gate resistor. The value should be something between 1k and 10k.

What do you think of that?
.

good Idea
very good idea!

not only JFET
but also Power FET, Power BJT
and even small signal bipolar, like BC550, 2N3904, 2SC2240 ...
And OpAmp both input PINS the same.
have good use of gate / base series resistor many times!


For Power MOS, we called them GATE stoppers

Note!
These resistors should ALWAYS
... be a short leg (cut off unecessary lengths at the Semiconductor pin side)
... and mounted as close to Semiconductor PIN
as possible ( without damage the resistor / semiconductor )

For BC550 and such small TO92 bipolar, this resistor can be:
like 470 - 2.200 Ohm
For small JFET I think the same values goes.

For Power devices, MOSFET or BIPOLAR,
suitable resistor can be lower.
Not often more than 470 Ohm,
but I think I have seen as low as 15 Ohm, in some Hifi amplifiers.


Best Ohm resistor value, depends on
1). What Semiconductor transistor
2). Your actual Circuit
3). Optimal value often requires real life testing. Using Signal Generator + Oscilloscope


lineup

 

[lineup]

hi
again
slepro

lineup here

You can see me use exactly such base resistors for input pair
in
My Ultra Performance
Discrete BC550C BC560C Op-Amp operational amplifier
Here you have
Link to simplified schematic of my circuit:
Mister Lineup AD797 Ultra Clone
Really extreme data for that circuit!

To you I say ( don't tell everybody all my secrets, please )
I have used 1.000 Ohm
in BOTH Op-Amp input stage transistor pins.
It did not hurt my test data when I did my Spice Simulation.


note:
For my bipolar BC550C this value worked good.
But I think small JFET can use, 1.000 - 2.200 Ohm, too
maybe even a bit higher, 4k7 ...... ?


Regardas
lineupas

 

[Roar Malmin]

You need to remove C1 and C10. Also you need to put a resistor in parallel with C6, to define the current flow in the driver transistor. As it is now, you are only pumping current in and out of the bases. This is dangerous, and most possible the source of your problems. A reasonable value would be in the range of 22R to 47R. C6 should be at least 220nF. R10 and R18 should be in the range of 15R to 22R. It depends on how much overcompensation you prefer, but 33R is too high. R11 can be from 22K to 33K, with R14 from 1K to 1K5, for a total of 22x gain factor. 1R as emitter resistor on the output stage is also to high. Try R22 to R33/5W, noninductive. To shape up your squarewaves, you can put from 4p7 to 10p across R11. Styroflex would be best. R3 and R22 is way to low. Why are you running the zeners at 20mA? Try 8K to 10K here. D1 and D4 should be paralleled with some capacitance. 10uF to 100uF. Implement these changes and report back.

 

[forr]

Hi Slepro

---The Zobel circuit may be added outside the pcb if necessary, but I would like to avoid it.---

There is no good reason to avoid a Zobel network at the ouput and many bad reasons not to have it. The noise problem can be due to a very HF local oscillation of the output stage not visible on a scope, a Zobel network usually tames it.

A warm resistor in an output Zobel network is useful as a good indicator of an unsuspected HF oscillation. It should be located near the power transistors.

Add a Zobel now and remove it only when your amp will be finalised if you still are convinced it has a deleterious effect on the sound.

 

[lineup]

.

slepro, I do not know if somebody has suggested this already.

See DOCAFET attachment
In your circuit:
Those drivers Q19,Q5 (MJE15032-MJE15033)
Normally would have a resistor between their emitters.
You have one Cap C6, (2nF2).
I think this is good, better than having nothing from Q19-E to Q5-E.

My choice, would to allow a DC current to flow here.
To set THE LOWEST working current level of Q19/Q5.
My choice would make resistor value such, that 50mA-100mA flows in this resistor.
Because I like my stages mostly have a good high Class A level of working point.
I know many other, that select this resistor to get only 5-10 mA.
But maybe they have Class AB low bias in output stage??

I run 99% of all my Lineup Audio Lab orginal output stages in
Absolutely True Class A. For power Amplifiers.
Same goes for any input or driver stage.

I have however done some clones, of op-amps that uses low Class AB output bias.
Explore AD797 Zero Distortion Clone - by Lineup Audio Lab


You could make some tests, if you should keep C6 in parallel with this resistor.
I am not sure C6 would improve anything, at least as low value as 2.2nF,
if you use a resistor allowing DC current.


Regards
lineup

 

[slepro]

You need to remove C1 and C10.

Why should I remove these capacitors? I don't see any reason why I should remove them. The amplifier was carefully simulated with Pspice before ANY prototypes, and the results were better with there capacitors than without.

Also you need to put a resistor in parallel with C6, to define the current flow in the driver transistor. As it is now, you are only pumping current in and out of the bases. This is dangerous, and most possible the source of your problems. A reasonable value would be in the range of 22R to 47R. C6 should be at least 220nF.

I know that lineup is having the same remark as you for C6 and a resistor in // with it. I put that capacitor there to have less gain at high frequency, but again, I don't see any reason of putting a resistor to allow a DC current flow. The DC current flow is made through the output transistors bases and emitters (don't forget that I drive this amp output stage in class A, so I have at least 5mA through the emitters of Q19 and Q5.

R10 and R18 should be in the range of 15R to 22R. It depends on how much overcompensation you prefer, but 33R is too high.

I think 33R is close enough to your 22R you required. For now, I will let these resistors as is.

R11 can be from 22K to 33K, with R14 from 1K to 1K5, for a total of 22x gain factor. 1R as emitter resistor on the output stage is also to high. Try R22 to R33/5W, noninductive. To shape up your squarewaves, you can put from 4p7 to 10p across R11. Styroflex would be best.

You are right, R11 and R14 are too high. I reduced them to 47k and 2.2k respectively. I don't believe 1 ohm emitter resistor for output stage is too much as this permit to have a very good sharing of current through all of the output transistor. And in AC, all these resistor are in //, so a total of 0.33 ohm. Less than 10 times the load, so it is fine.

R3 and R22 is way to low. Why are you running the zeners at 20mA? Try 8K to 10K here. D1 and D4 should be paralleled with some capacitance. 10uF to 100uF. Implement these changes and report back.

the reason why I'm driving these zener at 20 mA is because I'm using 1W zener diode. To allow them to be in their linear region, they need to have that much current flow. Go see the datasheet!
You are right about the 10uF to 100uF capacitors, I will add them.

Don't think I'm just denying your recommendation, but when I designed the amp, I was not "going" only by feeling, but everything was calculated and optimised. I you have good arguments about your suggestion, I will be happy the read them