OK, I looked at the Ap Note and it's hard to follow. On the layout they provided there is no Q1 and two Q3's. The board side view doesn't line up with the component side view. There are also places where you need to add jumper wires that are not shown. So I built a new version and I'll show my layout sketch and the mods I did to the schematic. One additional mod I did was to add a capacitor to the input.

Most things I build are on a proto board with holes on 0.1 inch centers and copper pads on one side. To this I add eyelets from Keystone in three sizes. In some places you need to remove the copper pads because the pads cause shorts between eyelets. There are multiple ground wires that come off the power amp board to a star on the power supply board that I haven't built yet. I built the power amp board and got it running on my lab power supplies. Still might make some changes there. Another mod I allowed for is mixed mode feedback. There is an eyelet on the power amp board labeled IFB. That hooks to a low value resistor that goes between the low side of the speaker and ground. If you just want Voltage feedback, ground the IFB pin.

The attached schematic shows the short circuit protection and zener diodes on the MOSFETs.

I use Vector Flea Clips.

Link: https://www.ebay.com/itm/Vector-Push-In-Flea-Clip-Terminal-Printed-Circuit-Board-/362593814146

Notice how they extend from the hole in only one direction. You use two rows of holes on the board, each with 3 holes. For the center pin (3), orient the extended portion of the clip towards you, away from the heatsink. To the right and left of that pin, orient the extended portion of the clip way from the center pin to the right (pin 5) and left (pin 1). On the row between that first row and the heatsink, you use the holes behind those two outside pins except you orient the extended portion of the clip towards the center so you can solder to pins 2 and 4 of the chip. Did that make any sense ? Read it again.

From the top it looks like this:

HEATSINK
2x4
135

x= hole not used

JFETs with low pinchoff are easier to bias like tubes. If the pinchoff is too high, the bias will drift around when you overdrive the JFET unless you great lengths to control the bias.

This is an old design that was in an IR data book in the 80's. It doesn't have short circuit protection, but that can be added. You also need gate protection zeners for the output MOSFETs. I built one and it sounded good. You can get to 50W at 4 Ohms with +/- 25V rails.

Link: http://www.cieri.net/Documenti/Altri%20marchi/International%20Rectifier%20-%20Linear%20Power%20Amplifier%20Using%20Complementary%20HEXFETs%20(AN948).pdf

Edit: Attached below is the PCB layout from the Ap note article edited to show the components X-ray style over the etch.

Not exactly a cascode circuit because the gates are connected together.

A quick experiment on the curve tracer shows that it knocks Idss down to about 60% of the single JFET value. Gain is also lower. The curves look slightly more linear.

Big filter caps are inductive above a certain frequency. It depends on the internal construction of the cap. The smaller caps (which have lower internal inductance) are added in parallel to insure that the impedance across the big filter caps remains low across the audio band up to some upper limit probably to a couple of hundred Kilohertz.

This is mostly an audiophile thing that may or may not be measurable.

If you ask 50 people your question, you will likely get 50 different answers. Now if you set up a test where you ask 50 people to play an amp and ask them if it's tube or solid state, it's one of those situations when you can fool some of the people most of the time, but you can't fool all of the people even once.

In my experience, you can come pretty close, but you can fool more people if you think outside the box (non-conventionally).  A tube power amp has a high output impedance (low damping factor), it has some gain compression and produces a non-symmetrical square wave when you overdrive it. If you can reproduce those characteristics,  you've come a long way.

Quote from: phatt on March 14, 2020, 09:17:52 AM
The Elephant in the room;
It took ~~ 40,000 Yrs to reach 200 Million people on this planet
At the time of Christ it's estimated that it was a shade under 200Million.
By the 1830's we cracked the 1 Billion mark.
Less that 200 Years later we have hit the 7.8 Billion.
Now if that does not tell you the real reason for the worlds problem I don't know what will. 8|
Unlike the FAKE hockey stick curve presented by Al Gore this one is REAL.
https://www.worldometers.info/world-population/
Scroll down a bit for the graph,, scary

There are people that suggest that the human population on Earth be limited to 500 million. The first place I ran across this was here:
https://en.wikipedia.org/wiki/Georgia_Guidestones
It's interesting that they can't seem to be able to track down the people that paid to put these things up.

Justin Bieber had a big tour booked in the US but ticket sales were slow so they had to change venues to smaller places. This was before the virus really got serious press. Now they may have to cancel the whole thing or postpone it.

The question I have is: If you get it once, are you immune or can you get it again ? Or is it like the flu and it mutates every year ?

CMOS circuits can fizz, especially when you lower the Voltage (this increases the gain) and you won't see any ringing on the waveform. Increase the Voltage and the gain goes down and they start clipping softer. Fizz is mostly from too much feedback. Solid state circuits have hundreds of times more feedback of tube circuits. A tube power amp may have 20dB of feedback. A typical opamp has 40 to 60dB or more feedback. Same goes for chip power amps.

A low damping factor just sounds better. Those attenuators that are basically just a Voltage divider pot across the output don't sound very good until you add series resistance to the speaker. A solid state power amp doesn't require a load, and the argument can be made that it sounds better without it. My latest Proof of Concept experiment compared a switching power supply against a soft analog power supply. The switching supply didn't sound that bad. Different, but not bad.

Those TIP parts don't have a very good SOA (Safe Operating Area) and the ON Semi parts are a little better than ST. So a safe limit for the rail is 33V or what you get with a 24VAC transformer with a bridge-cap type filter. But you can use a 4 Ohm load and slam the rails all day long if your heatsink is big enough.

A 40 VA wall wart transformer should get the job done and be safe for amateur builders. If possible, get one with a safety ground, it will help keep the hum low. In the USA you can use MG Electronics MGT2450. Be aware that these units have an internal fuse, if you blow it, you have to replace the transformer   That should get you close to 15W at 8 Ohms or 25W at 4 Ohms.

I'll experiment with short circuit protection and update the schematic. It'll take 6 or 8 diodes.

Playing at bedroom levels is always going to be a compromise. You never get really good tone. A simulated speaker load and headphones might get you close.

Phatt,

The MC12DAR was posted about half way down the first page in this thread.

Link: https://www.ssguitar.com/index.php?topic=3747.0

This is just a simplification of the circuit I actually use, it's intended for beginners. If you can't obtain the TIP darlingtons, you can try any other big darlington, but some changes to the circuit may be needed. The two 2.2K resistors may need to be adjusted to get full output. Everyone thinks an emitter follower should have a low output impedance. In actuality, the output impedance is the impedance at the input, divided by beta. On my prototype, the output impedance IIRC is about 4 ohms (measured by the delta E over delta I method using 4 and 8 Ohms). So adjust the 2.2K resistors up or down so you can obtain rail to rail output, but don't go lower.

If there is any interest, I could do a new layout, maybe even order some PCBs.

I think you lose too much when you are not driving a power amp to the rails. It kills the feel of the amp.

For comparison below is a THD plot of a TDA2030 chip amp and the MC12DAR that I linked above.

Note the Chipamp has low THD for much of the graph, then rises steeply at the right where clipping starts to occur.

The MC12DAR has no global AC feedback to it's THD is higher but it has a lot more gain. The THD rises gently and reaches several percent before clipping occurs.

I built a JLH last week and here's a THD graph I plotted. The power supply is only about 30V so power is about 10W at 8 Ohm. Notice how THD rises slowly and then takes a turn about 10W at .5% THD. This is where hard clipping starts to occur. The heatsink is 1 degree per Watt runs 55 degrees C. Too hot to keep your hand on.

Contrast the above to the Nelson Pass design called the ACA. This is a little amp that does about 10W at 4 Ohms. Two channels are bridged to give about 20W at 8 Ohms. Notice how THD is rising slowly but doesn't take the turn until you get near 20W at 5% THD. Much more tube like. The ACA kits you can find at diyAudio run on a 24V switching supply so you don't have to mess with mains wiring.

Kevin O'Connor devotes a chapter to the JLH in his book, Tonnes of Tone. He runs his version on slightly over 50V which should give about 25W into 8 Ohms. He calls it the "Hood" amp. The heatsink and power transformer run very hot is his design and he doesn't give you a way to adjust the idle current. It really needs twice the heatsink and a bigger transformer.

Another short coming is he doesn't include mixed mode feedback. It would only take one resistor and would go a long way to make the amp sound more tube like.

If you want something solid state that sounds like tubes, look into the designs of Nelson Pass. He's on diyAudio and has a couple of websites. You'll still need a big heatsink.