2-Stage N-Channel Class A

20230126_123058_HDR.jpg

Hello There,

Finally got to building the second channel. Heres a picture of when I first achieved breadboard stereo. Im now powering it from the wall through an old and noisy telecom 1000VA toroid, which started a whole journey of combatting ground loops.

This single ended topology shouldnt have much PSSR but with the bootstrap load combined with the mosfet Cap multiplier it gets under -100dB of PSSR at 50hz.
Depending on what other devices in my room were doing I was still getting hum and buzz. I introduced a ground isolator as proposed by Eliotsound. Seemed to do the charm but then hum again. I proceded giving each channel its own isolator. Humm still. Only a couple days later I realized the negatives of the osciloscope probes are earthed (of course). This created the occasional ground loop. Disconnected those, quiet at first but then once more: hum. Lastly, 10 ohm ground lifters. Finally silence, but now I got conditional oscilation(!) So I bypassed the groundlifters with 2n2F caps. Now finally unconditional silence (and stability) even when the refrigerator goes full beans or with dimmers turned on. The toriod itself does hum though, sometimes more, sometimes less. Not surprised as it was cheap and third hand. Could power factor correction help there? Its now (much) quieter then my Cyruss Misson One and earthed safely. Ear on the woofer and I hear nothing but my blood pumping round. Ear on the tweeter and I hear the faintest white noise. The speakers are MB Quart Q600. 90dB 4ohms. Im using 2A5 fast fuses at the moment. I can short the output, replace the fuse and all is well.

Another issue I faced was a very inconsistant soft low frequency crackling or shuffling noise with the occasional gentle sqeek. Luckily this turned out to be inconsistant contact resistances in the source to drain high current path. The wires and the breadboards im using are not really up for the task. Sometimes the contact resistance would drop .5V and generate enough heat to melt the wire or contact points. This can make breadboarding a power amp quite annoying, especially since the contacts get less grippy over time with the consequent heat exposure. Any tips and tricks for breadboarding power amps? Perhaps you know breadboards that are more rugged? For now I just soldered some connections and the crackliness disapeared.

In all this the performance remained largely untouched and im honestly quite charmed by the sound. Its warm and solid. Pinpoint between and behind the speakers and sometimes it stretches a little beyond left and right. Vocals and saxophones excell. I do feel like my old Misson One can reach a little taller but id have to do more AB testing to confirm.
Im getting about 0.05%THD at 20W, 0.004 at 1W. 1k, 2k, 3khz. At 6kHz 20W rises to .06% which matches spice simulation. At 20khz its expected to be 0.08 but lets say its 0.1%. Id say thats pretty neat for a single gain stage.

I feel its time to move on to pcb, give it casing and name the design. Green Comet feels suiting.

More Arta FFT's and the proposed PCB design may follow shortly.

I know no one is asking for it but thank you fellows for letting me unpack here.

Much cheers,
Ruben
 
Last edited:
  • Like
Reactions: decramer and Jcris
Hi Mr Borko,
Thanks for your reply and input. You're probably right. Its just that my breadboards are getting tired.
Yeah you recon the FR shows oscillation? I attributed the 0.05dB wiggles to my signal source. Ill have to do a direct loopback to confirm. If there is oscilation its above the 100MHz bandwidth of my oscilloscope, which is feasible since we're using mosfets here. Im carefully looking for oscillation everytime I test.
I shot these FFT's last week at 1.8A bias without averaging. These shots really show the limitations of this design. it performed better last time I checked. Hope to share new shots soon so dont take my word for it yet.. Still im liking how its second dominant and how the harmonics taper off. I do wonder how to combat intermodulation in a low feedback two stage amplifier. How does Nelson Pass do this?

Inter-modulation at 10W per fundamental:
intermodulation at 20W summed power. 1.8A bias.png


40Hz at 20W:
40Hz 20W 1.8A bias. .png


6khz at 20W:
6kHz at 20W 1.8A bias.png

Do note bias here is lower then before by 150mA. My FFT screenshots are not the most consistent so please excuse me for being messy.

Much cheers and thanks!

Ruben
 
Last edited:
  • Like
Reactions: decramer
Dear Fellows.

As promised I'd like to share with you the proposed PCB layout. I took the liberty of using two copper layers. Capacitive decoupling groundplane and powerlines on the back, signal and reference groundplane on the front. I'm adding the rectifier and reservoir cap to the board so power supply impedance can be minimized. M3 screw termination for external reservoir capacitance or lab supply is also included. The secondaries connect to the board directly.

Front:
Green Comet PCB Front .png

Green Comet PCB Front 3D.png


Back:
Green Comet PCB Back.png

Green Comet PCB Back 3D.png


Playing at ear-protection levels:

After playing different kinds of music, I did discover more limitations, I noticed the cascode reference LED's cutting out at clipping. (nice clipping indicator?) This was pared with occasional oscillations but only when playing my fav minimal techno record: Plastikman/Richie Hawtin's Consumed, which seems to be punishing source material. Playing this record at full beans easily clips the amplifier. I gladly conclude that its clipping is not sonically intrusive, I gave the LED's more current and bumped up the driver current and have not spotted these oscillations since. Noteworthy is that the oscillations occurred in both channels at the same time.. still not quite sure how this happened. It appeared on the bottom of heavy bass notes in high current moments around clipping, Anyway they're gone now. Furthermore I made sure the regulated voltage comes up more slowly to avoid startup chonk noises and now soldered ALL high current paths point to point. Man she sounds gorgeous, spacious, reach out and touch-like. (to my rather inexperienced ears) I'm surprised how much point to point seemed to help.
The mosfets are now mounted directly to the heatsinks without mica insulators as only now I noticed the non-conductive coating. Still reading several mega-ohms between the drains.

KiCAD schematic:
Green Comet Schematic 4-2-2023.png

Please find the pdf if unreadable.

Lastly I added the kiCAD project in a zipfile. Feel free to have a go or point out possible improvements/mistakes.
Much thanks and cheers!

-Ruben
 

Attachments

Last edited:
  • Thank You
Reactions: huggygood
A second board has been constructed successfully.

Not quite smooth sailing.
When attempting to test with signal generator while supplying the amplifier from the toriodal transformer the negative input line exploded is spectacular fashion. Turns out I had only connected the amplifier to earth through the negative line of the signal generator output. The negative input of the amplifier is ground-lifted through a 10 ohm resistor. Since I left the power supply floating a large current flowed from its ground through the 10R to signal generator earth. The wire that exploded was rated at 1A so the input must've been pulled down to at least -10V. This sadly killed the input JFET and the driver BJT. The 10R had also failed and became open circuit. The input zenerdiode clamps were left intact as was the rest of the amplifier.

How to protect against such a connection error?
My initial idea is to clamp the 10R with high current diodes so it cannot develop more than +-.7V. I'm not 100% sure if my assessment and reasoning are correct here. If you have any ideas please let me know.

I replaced the failed parts and MADE SURE to earth the amplifiers from their power supplies. PCB stereo is living happily ever after.
PCB stereo.jpeg


Good but not perfect.
The heat-sink mounting is not ideal either. The (from the picture) leftmost power transistors are mounted closer to the edge than in the breadboard phase, causing them to get hotter so I am forced to turn down the bias to about 1.75A. I also had to bend the legs to the side a little as my measurements of the mounting holes turned out to be offset by a couple of millimeters. Still I'm rather happy with the result. Sound seems more detailed than I remember (constructors bias or not). The layout did not create any hum or noise. Without music playing a slight hiss is perceived only with ears pushed onto the tweeters, other than that my 90dB 4 ohm speakers are dead-silent. Furthermore it is most liberating not having to worry about all the little contact resistances on a breadboard.

Much cheers,

Ruben
 
  • Like
Reactions: decramer
After a good number of days listening I though I hook up the audio interface again to take some Arta shots.

Performance is largely the same as it was on breadboard. Unfortunately more PSU components appear in the FFT's, however by ear the channels seem dead silent. Like before (I think) the audio interface is referenced to signal ground via the amplifier input/interface output. Correct approach?

Do you think this is pcb layout related or has it more to do with the closer proximity to the transformer?

I will experiment more the position of the (Elliot Sound Projects) earth isolators. I noticed when the earth wires from amp to isolator are in close proximity with the transformer hum was introduced to the speakers. Perhaps it will help to place these right by the amp channels instead of making a 50cm run around the heatsinks.

Nevertheless the PSU components are at almost -100dB referred to a 1W signal which ain't half bad🙂

I'd love to hear your thoughts.

Much Cheers,
Ruben

20W into 4ohm:
Schermopname (112).png
1W into 4ohm:
Schermopname (111).png
 
  • Like
Reactions: bogdan_borko
So interestingly placing the earth isolators right by the amplifier channels actually introduced audible hum/buzz and made gave a louder turn off thump. Now ground makes a 50cm run around the heatsinks before going through the earth isolators to earth and no hum is percieved.

Next ill try some kind of transformer shielding. The quest to get FFT's as clean as on breadboard has begun...
 
Help Needed! Sparks and Smoke!

Oh Well,
I destroyed one PCB by shorting ground to the secondary with my voltage probe... (never using alligator clips to probe ever again) Secondary tracks burned through and the rectifier got fried. After repairing the PCB and replacing the rectifier I switched on to check it the PSU section was OK now with the fuse removed so the amp does not get power.
Supply read 60V like intended but a smell of burned resistor was perceived.
I checked the all resistors including the ground lifter which read 10R. Reconnected and checked the earth/ground-lines. All OK.

So I popped in the fuse and switched on anyway. Now measured the Offset voltage before the output capacitor. 18V and the cascode leds were brighter then before. The Input Jfet and cascode transistor had died (again). Base was shorted to emitter. JFET did not pass signal.

Turns out when checking the ground-lifter I was only measuring the one in the other channel and really I should've measured 5 ohms with both inputs connected and channel grounds tied together. When disconnecting the broken channel PCB altogether the ground-lifter turned out to measure 35k ohm. So before grounds were tied together. Maybe not when checking PSU voltage alone....

I'm still trying to wrap my head around whats happening. Some large current sometimes run through and burn out the ground lifter that im sure of. Maybe it develops enough voltage to exceed the Vgs and Vbe ratings of said transistors before burning out. Or perhaps when the ground lifter fails open this kills said transistors. Perhaps the ground line from the channel in question was iffy and temporarily disconnected, leaving current nowhere else to go but through the ground lifter(?). These ground /earth connections are made much sturdier now. After replacing the cascode and input transistors all is well and now I'm listening to music again.

This is getting rather expensive as I've killed quite some JFET's in similar manners. I'd love to hear your opinion on my reconstruction of destructive events (and on how stupid I am). I'm hoping to prevent these disasters in the future but your help is needed!

Much Cheers,
Ruben

Here is the latest Schematic which inhabits the PCB:
Schermafbeelding 2023-04-05 221222.png
 
  • Like
Reactions: decramer
Meanwhile, I've had a tube amp actually catch fire and continue to play music. Transistors are so damned fragile compared to tubes LOL
I don't see any protection devices on the fets - Strapping a Bidirectional 12V TVS like P6KE12CA from sounce to gate should help. Or whatever smaller device they have for J1
I turned a few power MOSFETs into wires before I started to strap them 🙁
Ugly but functional 🙂

1680726685406.jpeg
 
  • Like
Reactions: Rupopulles
Hi Kodabmx,

Thanks! You reminded me to check the clamping zeners.
There are zeners across Gate source of the input jfet. (D9 to D12) However I just measured them and notice D10 has failed and D12 Read only 2V. This does indicate they (over)conducted and the and Vgs was exceeded once they failed. These should obviously be higher wattage types. I'd love to dabble in some tubiness, This topology would actually work with tubes. Tubes are just so very pricey even more so than those LSK170's.

Much Cheers!