JAT EZ amp - idea by John Audio Tech

Hello Christian. Thanks for showing us your amplifier explorations and sims. I have also been pursuing an amplifier with a single transistor VAS, and bumped into the same quandary as yourself regarding R21 on Q5.

After reading keantoken's posts on the "diyAB Amp - The Honey Badger" thread from # 2264 onward, I removed the VAS degeneration resistor to reduce distortion. ( surprising ! )

1737506309636.jpeg
WHAT !

The ratio of re' to RE doesn't matter because it is the Ic swing that sets the harmonic generation across Re', and the Ic swing is the same regardless of RE.

If our VAS was voltage driven and the feedback loop did not keep the IC swing the same in all conditions in order to meet the output demands, then RE would decrease the harmonic generation of re' by reducing Ic swing (at the same time as reducing gain). But that is not the circuit that is on our bench.

The resistor does however increase the conversion of OPS drive current to voltage at the VAS input which CAN increase distortion, but luckily again our VAS is current driven so the effect is not very strong.

On the other hand, RE is actually equivalent to 136ohm resistors in series with the driver bases, and you already showed this is undesirable when you bypassed R34/35 to improve stability.

That is one useful features of RE, it can act like base stoppers for the drivers without interfering with the bias loop. But in the circuit on our bench, RE is acting like base stoppers that are far too large.

I have not once, ever, found the idea of "local feedback linearizing the VAS due to Miller capacitance or RE" to be helpful, and in fact every time I've seen it brought up it leads in totally the wrong direction. It's like a computer virus which hides itself by causing the CPU to skip over countless important error checks.

Best, David
 
  • Like
Reactions: chermann
with my setup i got with 10kHz just 3,4V/µSec is that good?
with setting the bias of vas up to about 14mA i just got 4,1V/µsec
I test at 10kHz and higher to see how difficult the Darlington's are to drive.
And how slow they are.
Shouldn't be more than 8ma bias from 2nd gain stage.

If turn on time is poor difficult to drive at high frequency. likely why I mentioned I have abandoned BDW83/84 also Tip142/47
Of course differential slew is limited with such simple design as well.
Maybe I run some new tests, havent played with BDW83/84 in awhile. Slew is measured at 10% and 90% of peak.
So 10 volt peak to peak would show 8 volt peak to peak. 8 volt divided by Delta Y/X

I posted designs with faster and better Darlington's with better gain. Same thing the usual blah blah blah telling me all sorts of negative things.
Not you of course, haven't shown you any my higher voltage Darlington stuff.

Vas/ Second Gain stage never should need to be over 8 to 10ma low distortion and easy to drive power stages need no more than 3 to 6ma.
Guess your finding the fun of it all. Differential never more than 2ma needed. signals over 200mV at input will make Differential current jump. Not linear.
So Degen is essential for more linear current and more stable amplifier. Specially if current is set High to 2ma
 
  • Thank You
Reactions: chermann
I just try shortly to push the input differential pair with more CCS current..about 3,5mA...i just got 4,6V/µsec 🙄
You cant do much with such simplified topology. The main goal, not blow up transistors and have a stable amplifier.
I have seen many many simple amplifiers like this, they crank differential current or Vas. Refuse to use degen.
So they just get a oscillator trying to improve something that will never breach .05% distortion.
I have even seen 680p capacitors across Darlingtons to keep them stable. Because they cannot use degen and are not aware
of phase margin. It needs to have typical 45 degree to be stable. So Cdom and degen will always be high. No matter what to stabilize.

This is all you need to be stable and not blow up transistors with current. Expect .07% distortion at 1 watt 1kHz
Crank Bias up to 100ma maybe then .05% so pointless for high bias as well 30ma to 45ma per device is fine
wont even bother looking at 20 kHz. At higher power levels.....
I tested BDW transistors to re fresh my memory. Not to hard to drive actually, better than TIP types.

JAT_FixTEST1.JPG
10 kHz 10Vpp 5Vp 2.6V/us Slew Rate
1737520952384.png
Phase Margin 47
1737522643462.png
 
Hi David

Thanks for your info.
I have not once, ever, found the idea of "local feedback linearizing the VAS due to Miller capacitance or RE" to be helpful, and in fact every time I've seen it brought up it leads in totally the wrong direction. It's like a computer virus which hides itself by causing the CPU to skip over countless important error checks.

i will see in the lab at sqaure wave tests.....then i will jumper the RE of Q5
 
  • Like
Reactions: DavidT
I test at 10kHz and higher to see how difficult the Darlington's are to drive.
And how slow they are.
Shouldn't be more than 8ma bias from 2nd gain stage.

If turn on time is poor difficult to drive at high frequency. likely why I mentioned I have abandoned BDW83/84 also Tip142/47
Of course differential slew is limited with such simple design as well.
Maybe I run some new tests, havent played with BDW83/84 in awhile. Slew is measured at 10% and 90% of peak.
So 10 volt peak to peak would show 8 volt peak to peak. 8 volt divided by Delta Y/X

I posted designs with faster and better Darlington's with better gain. Same thing the usual blah blah blah telling me all sorts of negative things.
Not you of course, haven't shown you any my higher voltage Darlington stuff.

Vas/ Second Gain stage never should need to be over 8 to 10ma low distortion and easy to drive power stages need no more than 3 to 6ma.
Guess your finding the fun of it all. Differential never more than 2ma needed. signals over 200mV at input will make Differential current jump. Not linear.
So Degen is essential for more linear current and more stable amplifier. Specially if current is set High to 2ma
Hi whitedragon

i am sorry i do not want to blame you. because of keeping my ideas of BDW93/94 for this smple amp. it is just i bought it for AA9MD.
it is my first "big" amp to try to "design" and build.
it was partly fun....because if i want to finish something i go deep in the rabbit hole...as good as i can of course.

slew rate 10% -90%---thank you

signals over 200mV at input will make Differential current jump. Not linear.

that is what i guess if i push the amp hard...that the current of Q5 at squre wave was not a square...not?

thank you
chris
 
lol yeah fun stuff. Depends on the sim. I have real time oscilloscope in mine. If the amp is unstable then yes sometimes yes you get no waveform or sum check errors or just straight crash when you hit say a Squarewave at 22 kHz. Happened to me when first starting, doesnt happen anymore.
But is a comical way to find out a amp isnt stable. Actually pretty good opamp models in mine. Certain opamps very sensitive to DC offset.
So youll find out in strange ways.
 
signals over 200mV at input will make Differential current jump. Not linear.
that is what i guess if i push the amp hard...that the current of Q5 at squre wave was not a square...not?
line level standard is 1 volt RMS so I usually set my amp gains to reach full clean power with 1volt RMS and clipping 1.2 volts.

So typical normal listening levels will be 200 to 600mV at the input. And around 300 to 400mV is were current will really jump without degen on the input differential. So its good practice to see differential degen.

Good stability test , I do since I have real time scope. Will set my function generator to step up and down in 100mV input levels and see if the amp
rings or does anything strange at high frequency. My early designs without degen would have slight bursts of oscillation sometimes and then recover.
This is real world problem with many amps when stability is random. People that try to tell me degen is bad get ignored pretty quick.
 
  • Like
Reactions: chermann
You cant do much with such simplified topology. The main goal, not blow up transistors and have a stable amplifier.
I have seen many many simple amplifiers like this, they crank differential current or Vas. Refuse to use degen.
So they just get a oscillator trying to improve something that will never breach .05% distortion.
I have even seen 680p capacitors across Darlingtons to keep them stable. Because they cannot use degen and are not aware
of phase margin. It needs to have typical 45 degree to be stable. So Cdom and degen will always be high. No matter what to stabilize.

This is all you need to be stable and not blow up transistors with current. Expect .07% distortion at 1 watt 1kHz
Crank Bias up to 100ma maybe then .05% so pointless for high bias as well 30ma to 45ma per device is fine
wont even bother looking at 20 kHz. At higher power levels.....
I tested BDW transistors to re fresh my memory. Not to hard to drive actually, better than TIP types.

View attachment 1410996
10 kHz 10Vpp 5Vp 2.6V/us Slew Rate
View attachment 1410997
Phase Margin 47
View attachment 1411001

Again..thank you Whitedragon
i understand that to push current for nothing doesnt help. generally i build a bit bigger heatsinks becausee of better cooling. up to now i dont use mini housings for amp that can more then 20WATT at 4R.

but other things i do not really understand or missinterpret it.
Q1:why are base resistors needed. the darlingtons are slow so why?

Q2:you set the input filter downe 18k and 100p . ...instead of 220pF/22k...is that because of getting faster response from the "beginning" of the amp?

Q3: CCS is set is i to ~2mA...wow its the same as me 😉 . you did a other stabilisation with zener - it is the "better" version ...okay i have the Bob C. book but not read yet. it has about 770 pages but feels like 3000.😆

Q4: you re-config the degen of input LTP back to 237R -this fit to the degen 10R at Q5 and now we have degen factor of 10:1...right? less resistor value is not needed because current is just 2mA.

Q5: C7 FB cap and C8 miller cap of Q5 has to be changed. that was clear for me that in the real amp or a guy like you can better simulate it as me.

Q6: Bootstrap cap and resistors: the ratio is 1:1. R7 1k2 and R8 1k2?. i read somewhere it should be a ratio about 1:4-dont know why and where i read that...wrong? cap just 100µF...because of faster charge/disscharge?

Q7:C13 C15 are miller caps for the Darlingtons prevent oscillation - easy to implement for my for my prototype

sorry whitedragon because of my noob questions

kr
chris
 
Last edited:
i am sorry i do not want to blame you. because of keeping my ideas of BDW93/94 for this smple amp
Its fine with simple amps like this almost better to use darlington packages. And I was pretty much obsessed with darlington amps when first starting.
And even with some high power projects done recently I still use them. Had lots of fun with BDW93/94 my local electronics supplier had them in stock.
Usually got them for 50 cents to 1 dollar. Now when bored I sim with them a lot too.
Working on different simple amp to show you. Low parts count like this one hopefully .008% or better instead of .05% but very simple low parts count too.
Fun stuff
 
  • Like
Reactions: chermann
i am not realy a friend of TO220 packages (OPS) for an amp > 20W because as we know the power/heat dissapation is limit by the case.
i am very interested for a simple amp from you. you are the expert.

for more serious amp i want to use bigger transistors - that is physics.
 
Q1:why are base resistors needed. the darlingtons are slow so why?

Q2:you set the input filter downe 18k and 100p . ...instead of 220pF/22k...is that because of getting faster response from the "beginning" of the amp?

Q3: CCS is set is i to ~2mA...wow its the same as me 😉 . you did a other stabilisation with zener - it is the "better" version ...okayi ahve the Bob C. book but not read yet. it has about 770 pages but feels like 3000.😆
Q1 base stoppers used mainly for high frequency oscillation and stray inductance. Usually in board design if those pathways are short and trace is small usually many get away without them. poor practice either way ,board should include them. the value can be 10 to even 100 ohms.
VBE is sensitive to high frequency issues. and any surging from output section can cause oscillation. So even just a small impedance like thin short trace helps adding just 10 ohms is more a guarantee. If traces are long or lead wires to transistor can be up to 100 ohms. Needs to be on board regardless, yes in this case kept value low. Wont slow down transistor at all. Protects mainly VBE from big surges from the output section. Also why decoupling on power rail. 10 ohms and 470u traces can also be small at that point. Small traces have resistance already 10 ohms guarantees impedance not a easy path for surges. Good isolation for sensitive VBE with good layout, short small traces and added impedance from resistors guarantee decoupling from oscillation. Any further high frequency issues is tamed by small 100n capacitor tied across VBE..stable stable stable

Q2 I set gain with R9 at 18k value for R5 will always match this value. yes high for good impedance, matches R9 value so it wont cause DC offset at output.
much like servo circuit we need ref to ground and R9,R5 will always be matching. I keep 18k or higher for impedance. R4/C3 1k 100p is generic input filter to remove Radio Frequency interference/cell phone etc etc. People make wild magic guesses for this circuit. is easy 1k 100p good enough. if radio frequency interference still a problem raise 100p to 220p or 330p normal equation to determine frequency of simple RC filter.

Q3 yes good read to understand, short answer LED voltages can vary I just use zener, more stable over wide temperature range. LED assume 2 to 3 volts so current source impedance is higher if you use higher voltage reference for current source. with typical .7 to 1.4 volt ref expect current source impedance maybe 1meg if you raise voltage ref impedance is higher or better 2 meg. you can use 6 or 8 volts etc etc, but loose that voltage to the rail. So current source for say 2nd gain stage. We use feedback type so you only loose .7 volts to the rail. With differential dont matter. We want current source to be stable over wide temperture range so DC offset on output doesnt shift at higher temperatures. With low distortion circuits with mirror differential we can get better distortion performance with higher impedance current source. So higher voltage reference like 3 or 6 even 8 volts more useful for nice 2meg or higher current source impedance.
 
Q4: you re-config the degen of input LTP back to 237R -this fit to the degen 10R at Q5 and now we have degen factor of 10:1...right? less resistor value is not needed because current is just 2mA.

Q5: C7 FB cap and C8 miller cap of Q5 has to be changed. that was clear for me that in the real amp or a guy like you can better simulate it as me.

Q6: Bootstrap cap and resistors: the ratio is 1:1. R7 1k2 and R8 1k2?. i read somewhere it should be a ratio about 1:4-dont know why and where i read that...wrong? cap just 100µF...because of faster charge/disscharge?

Q7:C13 C15 are miller caps for the Darlingtons prevent oscillation - easy to implement for my for my prototype
Q4 yes 10:1 basic rule of thumb, yes can allow Cdom capacitor to be made lower to maintain good phase margin. This case Cdom remained high to get 45 degree margin. I could raise degen higher to get better phase margin, but yes current is high. Any higher value than 200 to 300 ohms at 2ma current it would cause distortion to go up. I cant lower Differential current lower than 2ma because will have issues driving simple VAS. if lowered to 1ma then yes degen value can actually go higher. Even with max degen and Cdom at 100p phase margin was still around 40 degree so I added small C7 cap across feedback. I just raised the value till I finally hit 45 degrees with generic standard cap values ended up being 47 degrees.

Q5 I wont go higher than 100p would rather be lower for better slew rate, but phase margin wasnt good enough still at 100p
Degen was maxed and finally more added to feedback path to finally hit 45 degrees. Technically little more advanced to calculate what cdom should be.
Simple amp, whatever, same old 68 to 100p. More a issue with beta enhancement transistor added, and cdom will get significantly smaller

Q6 value left higher to more likely maintain current and keep voltage drops the same, doesnt usually happen current is kept constant with resistor after the Capacitor. First resistor voltage/current will swing a little more
Since the cap in theory provides constant voltage. Like any current source, you wont have constant current without a constant voltage. capacitor should be large enough since it causes a phase shift at low frequency. Distortion at low frequency could increase if not large enough. Depends on resistor values.
1.2k and 100uf is about 1 Hz filter or likely where phase shift happens so no big deal. It could actually be 47u even 22u. I tend to keep it large. If the cap used for higher voltages. I guess it more about board space if you dont want a huge cap. Such low voltages like this a 100u wont be that big.
We can just do AC analysis in sim to see if there is a phase shift to high. 100uf will be more than large enough, specially at higher voltages when you resistors can be 2 to 3k or higher

Q7 power decoupling caps, high frequency is usually your biggest problem so the small 100n need to be as close as possible to the power transistors on the board.
larger caps around 100u to 470u should be as close as possible too. They are larger so harder to place on the board. The 100n are small and very much need to be as close as possible. If the wire length or trace length going to the big power supply caps is short you can often get away 100u to none. If your power supply is far away with leads its good to use 220u to 470u close to the power transistors. I didnt even show them in the schematic. Small amp like this with linear supply you can put rectifiers and main large caps right on the board if wanted.
 
Last edited:
There are still some issues. It is a very simple amplifier and the simplier they are the more careful you have to be in order to get the very best out of it.
Many problems are solved now but there are still some to improve.
1. The current generator with a zener diode is noisy as all zeners are. If there is a little imbalance between T1 and T2 that current (noisy) will add to the input of the amplifier. It is better to replace the diode with a transistor with the base to T4E collector to T4B and emitter to + supply rail. The R21 should the be R3/2.
It works identically good with BC557 as BC640. You can look at the super amps like the wolverine and they all have current generators with transistors.

2. You dont need the R1 and R2. They may reduce the risk of self oscillation but carefully values of C7 and C8 takes care of that.

2. R19 reduces the open loop gain of the amp. 0 ohm will give about half the distortion.

3. You probably dont need R17 and R13.

4. i would choose the R19 and R20 a lirttle bit higher, 47 ohm, to get less hum modulation when the amplifier is clipping.
If you never play that load it does not matter.

5. I test all my amplifiers in TINA ( can be downloaded from TI:s homepage). It is Spice with a simplified outer layer. And it works the same but dont have all the features.
First i connect the C2 to ground instead of the generator. Then I connect C1 to the generator instead of ground.
Now you have the same amplifier but inverting. You only need to add a measuring point at the base of T2.
Now You can see the open loop gain and phase as the difference between VF1 and VF2.
There i optimze the C8 value. You will see that a high value reduces the open loop gain at high frequencies so take the smallest value with a smooth phase curve.
Then I go back to the original noninverting configuration and optimise C7 to have a smooth high frequency roll off. Normally you dont have to adjust the C8 any more.

About the layout there is a lot to say.
Normally an amplifier layout is using a ground point near the input. From that separate leads go to speaker ground, The connection between C4 and C5, connection between C11 and C12 and C1 + input ground very short.
The lead between T5C to C4, C3 and T6C must be short as it is high current with high distortion going that way and it transfers magnetically to other leads nearby. It is easiest to mount the T5, T6 and T6 on the same cooler to get best possible thermal tracking.
I would replace R14 to a 220 ohm trimpot.

Good luck! It is possible to get a good performance of a simple amplifier if you just optimize it carefully.
 
Test it in TINA (it is free) and make your own decistions. You can make bode plots and noise test there and decide on your own.
Distortion test is very diificult so i would skip that. The amplifier did improve a lot with the suggestions from whitedragon. I didnt have the lust or power to make all the changes to the original scheme so respect to him.
Continue on the way to be your own expert.
 
  • Like
Reactions: chermann