Still avoid the 100u lytic. Very slow charge up with 1.5-2mA from the Jfet. 56K//2u2 R5 1W, C3 MKP/MKT seems OK. Prefer MKP.
Godfrey, while the diode will work, it will take a while for the reg to get to full rail, in which time the amp may be clipping constantly.
Salas, I didn't see your edit. When we finally get the schematic finished, would you please delete all the bad schematics to avoid confusion?
- keantoken
Salas, I didn't see your edit. When we finally get the schematic finished, would you please delete all the bad schematics to avoid confusion?
- keantoken
Attachments
1: Ground plane for decoupling. ???
Double sided is good for the boardhouse but sucks for us DIY'ers. All through hole discretes here , no need for a groundplane .... just star grounding and good layout (shortest paths - separate amp into building blocks ... Input/VAS/OPS - PS).
BTW - I have both Abacom CAD packages (below - alex's software , no demo's)
OS
Double sided is good for the boardhouse but sucks for us DIY'ers. All through hole discretes here , no need for a groundplane .... just star grounding and good layout (shortest paths - separate amp into building blocks ... Input/VAS/OPS - PS).
BTW - I have both Abacom CAD packages (below - alex's software , no demo's)
OS
Attachments
I am totally jealous and disenfranchised, me and my poor little KiCad.
IF you are going to do this now I will stop half way though. I do not mind being trumped again.
Thanks
Okay, no ground plane PCB then. As long as we are careful, I think the miller compensation should be enough.
- keantoken
- keantoken
The very cool thing about the program that Alex uses, is that you can pour a ground plane on the component side with out any input from the operator. SO, IF some of us WANT a ground plane we can use it and those that don't can just skip that part. This amp has too much bandwidth for me not to build without a ground plane....
Thanks
Thanks
I have to agree with Krisfr, good HF design requires a ground plane. These days board houses don't charge much more for a double sided board and they have become the norm. If OS complains I will donate a set of boards to him. 😛
The only problem I see is you try to etch the board yourself............but that is another can of worms. 😱
Jam
..............or we could run two sets of boards.
The only problem I see is you try to etch the board yourself............but that is another can of worms. 😱
Jam
..............or we could run two sets of boards.
Last edited:
Just do not make a ground plane on your DIY board, make a single sided board...
Just hold on you will see what I am talking about as this goes forward...
Thanks
Just hold on you will see what I am talking about as this goes forward...
Thanks
If we used ground fill on just one side of the board, it might work but would be sub-optimal.
I think two board versions would be best. It seems like OS has already volunteered to do the non-GP version. 😎
- keantoken
I think two board versions would be best. It seems like OS has already volunteered to do the non-GP version. 😎
- keantoken
When I get the same software I can do the the other, for myself or anyone else that is interested.
Thanks
Thanks
No , you can do... If you want the goods , go DEEP into my "ZIP1" directory(E-software/Abacom XX .zip). I won't directly link it for obvious reasons. I also will redo my whole "mongrel" collection at the "alex" level for learning curve reasons.😀
Compare Kicad to the Abacom software first (feature vs. learning curve vs. integration) , you might find Kicad is the best for through hole discrete work. after one could even export kicad to sprint for further "play".
edit: $$7K genesis stealth amp DID NOT have a bloody ground plane...
OS
Last edited:
I dont want to mess with you
but please consider to do simple clean boards lines
without too much shapes changing, dead end, traps, etc
I know it may be crazy and silly, but I think of a signal path like leading a horse to the water
the strickt meaning of star ground doesnt seem very practical
maybe better think of it as a tree with branches
looks a bit different
but probably works exactly the same way, or better
personally I think starground looks like a misunderstood fashion, and maybe even utterly wrong
but you are the experts
but please consider to do simple clean boards lines
without too much shapes changing, dead end, traps, etc
I know it may be crazy and silly, but I think of a signal path like leading a horse to the water
the strickt meaning of star ground doesnt seem very practical
maybe better think of it as a tree with branches
looks a bit different
but probably works exactly the same way, or better
personally I think starground looks like a misunderstood fashion, and maybe even utterly wrong
but you are the experts
The starground does not have to look like a "star" , it can be a "U" ,a "X" , or even 2 "u's" ... as long as positive return is symmetric with negative return. Same with Speaker and NFB takeoff , right in the middle. The speaker return does not have to be close to the output terminal , neither.
Keep miller loops real close to the associated devices and as short as possible. Segment/partition the different stages of the amp (LTP/VAS/OPS) ... AC and power supplies on a separate board. COMPLETE and utter silence shall Ye have , all alone with what is left (-120DB noise from input pair junctions 😀 )
OS
I guess it only matters with a 100% symmetric design
I would expect it to be more important having a short ground connection, and as direct as possible
Sillas/Kean, what is the new shunt regulator supposed to do? It is neither a stable output as it changes with input voltage and its load impulse response is awful. Maybe I have it wrong again and it should sound great. halving the load 1K to 500 ohms drops the output by half and it takes more than 500 mS to recover.
Attachments
Don't know man, Kean just put a Norton Vref there that we discussed about only. The rest he will explain, he must have had it fully simulated most probably. The should sound great no matter faulty remark is not nice on Kean IMHO. He investigates enthusiastically & purely technically, making no claims, and he could be given some slack. Churning out simulated CCTs can include presentation errors. If we catch some is help.