2-Stage N-Channel Class A

[Rupopulles]

Hi dearest Fellows,

Hope you rolled in the new year smoothly.

The christmas break left me with some time on my hands and after dabbling with a lot of designs of ever increasing complexity it was time to build something simple. I had been thinking of a single JFET gain stage and a class A buffer with single supply. The goal is to get a dominant second harmonic of negative phase around normal listening levels. Distortion may be audible if its the right distortion.

What I conjured up turns out to sound rather pleasing but I feel it can be optimized much further. Some pico-C here and there and it is already well behaved on breadboard and nicely quiet. Now running it at 48V and 1.35 Amps and getting about 10W into 4 ohm.

it actually is similar to the Pass' zen amps only with a separate buffer. The buffer also being Pass inspired, enjoys an active current source that creates some push pull action. A JFET driver was added in between as not to disturb the gain stage too much. The gain stage is cascoded and bootstrapped together with the driver JFET. The whole is placed within a light negative feedback loop to yield practical damping factor.
I have for you some LTspice shots. Better pictures and perhaps Arta measurements will follow.

Im leaving this here to hopefully gather your thoughts and opinions.
Much cheers and thanks to all of you.
-Ruben

1W 20k

20W 20k

 

[EdGr]

Kudos for building what you simulated. 🙂

I think the design is good as a learning exercise, but offers few real opportunities for improvement. Improving it means going to a completely different design with split supplies and direct coupling.
Ed

 

[abstract]

I actually like single supplies because it's a simple way of avoiding a very tricky problem: common-mode mains hum caused by leakage capacitance between primary and secondary transformer coils. Because it's common mode, hum can still come through on ground even if +/- are kept perfectly steady.

On top of that, if there's no 'ground' available as such, I like to create virtual grounds locally where needed.

 

[Rupopulles]

Kudos for building what you simulated. 🙂

I think the design is good as a learning exercise, but offers few real opportunities for improvement. Improving it means going to a completely different design with split supplies and direct coupling.
Ed

thanks, Indeed. I for one have done too much simulating too little building due to how time consuming it can be.

I could give it a traditional front end with diffpair and common emitter follower. (Ignore this OPS as it will kill itself during startup)

Or keep the design but move the coupling capacitor to the input where impedance is high and use a DC servo.

The question remains whether eliminating the output capacitor weighs up against needing an extra secondary, regulator, reservoir cap and the servo. Like you I certainly prefer direct coupling. But here single supply seems to make sense. The negative effects of the electrolytic will be negated somewhat by placing it within the NFB loop. Furthermore there is barely any low frequency attenuation. I might try this DC servo implementation anyhow.

Cheers!

 

[juma]

...The question remains whether eliminating the output capacitor weighs up against needing an extra secondary, regulator, reservoir cap and the servo....

Hi Ruben,
my experience shows that output cap included in NFB loop (as you did in post #1) is transparent and does no harm. What I would do to improve the design (without changing it into something too different) is adding some OLG and profiling the harmonic structure by :

• using CCS instead of R4, R1, C2 (to suppress H3 and higher harmonics) ;
• changing J2 into CCS loaded BJT emitter follower with Ic=10 mA or so (makes mumbling note go away) ;
• adding about 10mF from M3's Source to GND (Amplifier needs energy storage along with MOSFET CapMultiplier ) .
• I would also try to get rid of C4 (or reduce it to 1-2pF) - you'll hear the difference in strong, fast transients and ambiental information in higher register.
  
  By doing this you'll get cleaner, more natural and refined but very pleasant sounding amplifier.

 

[abstract]

What about a "full bridge" but with a simplified RHS providing an idle voltage instead of an inverted signal? I did this for a headphone amplifier but I see no reason why it couldn't be scaled up.

The power rails would tend to see a rectified version of a half bridge load, which may or may not be an improvement, but it would allow you to keep the LHS performance.

 

[Rupopulles]

Hi Ruben,
my experience shows that output cap included in NFB loop (as you did in post #1) is transparent and does no harm. What I would do to improve the design (without changing it into something too different) is adding some OLG and profiling the harmonic structure by :

• using CCS instead of R4, R1, C2 (to suppress H3 and higher harmonics) ;
• changing J2 into CCS loaded BJT emitter follower with Ic=10 mA or so (makes mumbling note go away) ;
• adding about 10mF from M3's Source to GND (Amplifier needs energy storage along with MOSFET CapMultiplier ) .
• I would also try to get rid of C4 (or reduce it to 1-2pF) - you'll hear the difference in strong, fast transients and ambiental information in higher register.
  
  By doing this you'll get cleaner, more natural and refined but very pleasant sounding amplifier.

Much appreciated Juma. What I would like to improve is exactly the harmonic structure. Im yet to find out how to use the load lines to purposefully shape these. OLG= openloop gain? Ill play with your suggestions. lets see what they do.

However:
-When R4, R2, C2 Are replaced with CCS, how do we now define the voltage at the collector of Q1?
-C4 gets rid of squarewave overshoot. How else to fix overshoot? Lets see if it can be smaller though.

Big cheers!

 

[Rupopulles]

Some scope shots:
Clipping at just under 20W into 4 ohm:

20k square-wave (C4 = 5pF):

 

[Rupopulles]

Given this is an inverting amplifier, the voltage at the input JFET gate should show us the approximate error signal. So looking at this voltage should also give us an idea about the harmonic structure?
The green trace shows the J1 gate voltage when driving 4 ohm at just about 20W now running about 1.75A at 48V:

Here is the voltage across M1 source resistor to show us what the active current source is doing again in green:

And here the current source current for a square wave (note the overshoot, problem?);

 

[EdGr]

The question remains whether eliminating the output capacitor weighs up against needing an extra secondary, regulator, reservoir cap and the servo. Like you I certainly prefer direct coupling.

All commercial hi-fidelity amplifiers are direct-coupled. Output capacitors are only for low-powered amplifiers.
Ed

 

[juma]

OLG= openloop gain?

Yes

However:
-When R4, R2, C2 Are replaced with CCS, how do we now define the voltage at the collector of Q1?

For example, like this (the point is active loading of input/gain stage) :
https://www.diyaudio.com/community/attachments/gf3amp-jpg.662222/

... Lets see if it can be smaller though.

👍

 

[Rupopulles]

At last some Arta measurements taken with a (compromised) Focusrite Scarlett.

I bumped up to current again to 1.95A still at 48V. Much more and the case temperature will surpass 70 degrees ℃. I bypassed all source resistance at the input JFET (it needed about 30ohms IRL) and played with the active current source gain. Now this crude little amp actually performs somewhat respectably.

1W into 4 ohm:

19W into 4 ohm:

And frequency response which looks a little rough though Ive seen the Scartell do this before. Note that this is not absolute closed loop gain. Actual gain that im running is just under 10X or 19.something dB.

In the future id like to test at higher frequencies, IMD and openloop gain but for today I'm happy.

Kind Regards,

Ruben

 

[Rupopulles]

What about a "full bridge" but with a simplified RHS providing an idle voltage instead of an inverted signal? I did this for a headphone amplifier but I see no reason why it couldn't be scaled up.

The power rails would tend to see a rectified version of a half bridge load, which may or may not be an improvement, but it would allow you to keep the LHS performance.

Thanks Abstract!

Excuse me for asking but what are LHS and RHS? I find myself unfamiliar with these abbreviations. With full bridge do you mean balanced outputs? Or are we talking rectifier bridge?

Cheers!

 

[abstract]

Sorry, "left hand side", "right hand side". On the schematic the speaker is placed in the middle between 2 power stages, usually operating symmetrically, but not as I described above.

 

[Rupopulles]

Using a little more pre-amp gain on the Focusrite Scarlett got me this 1W into 4ohm 1kHz measurement. Scheme is unchanged other than being trimmed for a tad more push pull. Pretty cute but it could do with less third.

About temperature:

Currently I'm having the single pair of IRFP240's dissipate just under 50W per device. Biased at 1.95A under 48V. An NTCALUG01A104G is mounted directly on one of the Mosfets with a little bit of thermal paste in between. Now running for 2 hours it actually drops to 14k8 ohm indicating 75 degrees Celsius. However when placing the UT139A multi-meter temperature probe in the notch at the side of the to-247 package it reads 90 Degrees (!). Where exactly should case temperature be probed? Or is this the time to turn down the bias.

There are more scheme tweaks to be tested but constructing a second channel and powering it from a toroid for proper stereo evaluation comes first.

Cheers,

Ruben

 

[WhiteDragon]

making the math simple and assume 1 C/W case to junction

if your dissipating 50 watts then the actual junction/die is 50 C higher than the case.

so if the case is 90 C then the actual Die is at 140 C

assume the typical plastic case is rated @ 150 C absolute max

and depending on voltage being used.
the gate would pretty much likely fail instantly
with higher voltages at 140 C

 

[kodabmx]

FWIW, I've run MOSFETS with a heaksink temp of 110°C and they still work.

 

[cumbb]

Positive feedback per differentialamp does not sound as good, clean, clear, naturally as a simple negative feedback. The other active component is clearly audible.
A capacitor at the output offers a very good and simple and controllable way to influence the sound. These things sound sooo different like wines or beers taste;-)
If ONE set of components is used for only one power supply (one voltage) or two power supplies (two voltages)... a difference is to be expected;-)

 

[Rupopulles]

making the math simple and assume 1 C/W case to junction

if your dissipating 50 watts then the actual junction/die is 50 C higher than the case.

so if the case is 90 C then the actual Die is at 140 C

assume the typical plastic case is rated @ 150 C absolute max

Indeed. Although using the real conditions and Rth case to junction from the datasheet we get 24{V}*1.95{A}*0.83{C/W}= 39 C., Actual die temp will be 129 C. 20 C margin is still too little commercially but for the brave DIYer just on the edge of acceptable. Im planning to turn down the current to yield 15W. Im never actually using 20W anyway.

Big cheers,
Ruben

 

[Rupopulles]

Positive feedback per differentialamp does not sound as good, clean, clear, naturally as a simple negative feedback. The other active component is clearly audible.
A capacitor at the output offers a very good and simple and controllable way to influence the sound. These things sound sooo different like wines or beers taste;-)
If ONE set of components is used for only one power supply (one voltage) or two power supplies (two voltages)... a difference is to be expected;-)

Hi Crumbb,

What scheme are we referring to here? When using a differential pair input like proposed in post #4 it still is using negative feedback, only now arranged to be non-inverting. Or am I misunderstanding?

Kind regards,
Ruben