The MONGREL (supersym II)

[AndrewT]

looks like I need to explain further.
A shield protects what is inside from outside influence. There are a couple of conditions (I think it was Conrad Hoffman that posted these).
cond1.) the shield must be very close to circular.
cond2.) the core must be very near the geometric centre of the shield.

In this situation any interference passing near the shielded core will be attenuated by the shield so the the core carries it's signal with little contamination.

Next the shield must be connected to a low impedance reference to conduct away the interference we are trying to avoid. Admittedly most situations use signal ground as the shield reference and often use the shield as the return route for the flow in the core.
But the output is also a low impedance reference. It will work just as well in shunting the interference that the shield is intercepting.

Why would a long wire not be good for taking an NFB signal from the output to the -IN terminal? Because it may pick up interference. How does it do that if it is connected to a low impedance reference? Because the inductance of the long wire raises the wire's impedance to HF and as the wire gets longer the inductance gets higher and the -IN end of the connection becomes more susceptible to picking up interference.

If the core is at the same potential as the shield then it cannot pick up interference from the shield, nor can it have parasitic capacitance to conduct a potential from shield to core, they are at the same voltage. If the shield is attenuating external interference then the core is carrying a less contaminated signal to the -IN end.

[ostripper]

Next on the "chopping block" is a real classic. Called the "APT1" , it was designed in Cambridge ,Mass. by a Tomilson Holman. I had a chance to chat with the designer APT 1 power amp – undeservedly forgotten when I was "green" (still am). He said " but I don't think it can be reproduced today since some of transistor types are no longer made."

HA HA , eat your heart out.... (BELOW) I am VERY pleased with this one as it spec'ed out to the original's performance BEFORE the refinements. On "first click"
thd was .015% and even dropped to .01% at higher freq's - 20Khz/100w (strange). I then added a cascode under the level shifting current mirror to allow for the use of the KSA992, which is a much better device than the originals or the mpsa92 that I used before to handle the 150v rail to rail voltage requirement.These changes dropped the THD to .003% .. all the way to 10khz... and just slowly increasing to .0075% all the way up to over 30KHZ!
Same story at 200w , it "held" that same .003 to .007% all the way into the ultrasonics while also keeping the same H2 dominant harmonic distortion "signature". H5/7 were real low , as good as the AX (blameless).

Below is the GX. I am so impressed by it's harmonics that I might build this one for myself FIRST.

Another thing unique to this topology is that it seems to "force" a balance between LTP currents.. (last pix) at all power levels and frequencies, I attribute this to the use of the level shifting CM.

OS

[keantoken]

Andrew, That's exactly what I was thinking, though for some reason I didn't say it. There is an application note by Linear that describes a probe using an amp to buffer the shield at the same voltage as the conductor to cancel out parasitics effects...

- keantoken

[ostripper]

Originally Posted by AndrewT

looks like I need to explain further.
A shield protects what is inside from outside influence. There are a couple of conditions (I think it was Conrad Hoffman that posted these).
cond1.) the shield must be very close to circular.
cond2.) the core must be very near the geometric centre of the shield.

In this situation any interference passing near the shielded core will be attenuated by the shield so the the core carries it's signal with little contamination.

Next the shield must be connected to a low impedance reference to conduct away the interference we are trying to avoid. Admittedly most situations use signal ground as the shield reference and often use the shield as the return route for the flow in the core.
But the output is also a low impedance reference. It will work just as well in shunting the interference that the shield is intercepting.

Why would a long wire not be good for taking an NFB signal from the output to the -IN terminal? Because it may pick up interference. How does it do that if it is connected to a low impedance reference? Because the inductance of the long wire raises the wire's impedance to HF and as the wire gets longer the inductance gets higher and the -IN end of the connection becomes more susceptible to picking up interference.

If the core is at the same potential as the shield then it cannot pick up interference from the shield, nor can it have parasitic capacitance to conduct a potential from shield to core, they are at the same voltage. If the shield is attenuating external interference then the core is carrying a less contaminated signal to the -IN end.

I understand , It will be easy to implement as well.
OS

[AndrewT]

It's taken me a while to remember the name, but this is similar to "guard ring". This is a standard technique for sensitive inputs to remove or attenuate the effects of parasitics affecting the amplifier input.

It is that idea I'm trying to explain here.

I wonder if this guarding technique would be useful for all NFB traces. A pair of output connected traces laid one to each side, close to and parallel to the route all the way back to the -IN point would attenuate on PCB interference from affecting the NFB signal?

Discussion???

If there is a useful discussion to be had, then a new thread would be appropriate.

[ostripper]

Originally Posted by AndrewT

It's taken me a while to remember the name, but this is similar to "guard ring". This is a standard technique for sensitive inputs to remove or attenuate the effects of parasitics affecting the amplifier input.

It is that idea I'm trying to explain here.

I wonder if this guarding technique would be useful for all NFB traces. A pair of output connected traces laid one to each side, close to and parallel to the route all the way back to the -IN point would attenuate on PCB interference from affecting the NFB signal?

Discussion???

If there is a useful discussion to be had, then a new thread would be appropriate.

There is an excellant article about that here ...
Grounding Strategies for Printed Circuit Boards

as you can see below , I already do the "guard ring" technique on my VB's(input section). BTW the "APT" GX1.1VB is DONE! Look for the errors.. Ha HA. Compare schema to layout , for this was done mentally , It COULD happen !!

OS

[csy]

Hi OS

I'm keen to build the PB60 with CX1, and have a few queries.

1) KSC3503 appears to come in different hFE values depending on the suffix letter, for example KSC3503D (hFE:60~120). Could you please confirm which variant to use? (or not matter)

2) I can't source Qnjw0302g/Qnjw0281g, but can source njw0302g/njw0281g. Likewise, I can't source Q2SA1837/Q2SC4793, but can source 2SA1837/2SC4793. Could you please confirm these are the same devices?

3) Could you please explain how to build the 3.3uH inductor (gauge/turns/diameter)?

4) I'm not familiar with getting PCB's professionally printed. Is it sufficient to simply submit your PCB jpeg and parts mirror jpeg?
I take it that the parts mirror is intended to be top-side?
and is it normal to submit the screen-print in a mirrored format?

5) I take it that the output and driver transistors protrude out beyond the PCB for mounting onto the heatsink (as per photo examples of your previous builds), but this creates a form-factor issue for people like me retro-fitting this into slimline cases/heatsinks (which I imagine would be typical for the PB60). My case is 80mm high with 75mm high heatsink, and your PCB is 76mm plus the height of the transistors, so it doesn't fit. I have a work-around but is a bit messy. Your original powerboard prototype would have been perfect. Anyway just a bit of "food for thought".

Thanks for your assistance, and I really appreciate you making this public domain for DIY'ers like me!

Craig

[Rob_S]

Hi Craig,

For mounting the output devices you could use L brackets on the heatinks so as not to mount the devices out from the board. Think it would still be very tight but may make it work

Rob

[ostripper]

Originally Posted by csy

Hi OS

I'm keen to build the PB60 with CX1, and have a few queries.

1) KSC3503 appears to come in different hFE values depending on the suffix letter, for example KSC3503D (hFE:60~120). Could you please confirm which variant to use? (or not matter)

2) I can't source Qnjw0302g/Qnjw0281g, but can source njw0302g/njw0281g. Likewise, I can't source Q2SA1837/Q2SC4793, but can source 2SA1837/2SC4793. Could you please confirm these are the same devices?

3) Could you please explain how to build the 3.3uH inductor (gauge/turns/diameter)?

4) I'm not familiar with getting PCB's professionally printed. Is it sufficient to simply submit your PCB jpeg and parts mirror jpeg?
I take it that the parts mirror is intended to be top-side?
and is it normal to submit the screen-print in a mirrored format?

5) I take it that the output and driver transistors protrude out beyond the PCB for mounting onto the heatsink (as per photo examples of your previous builds), but this creates a form-factor issue for people like me retro-fitting this into slimline cases/heatsinks (which I imagine would be typical for the PB60). My case is 80mm high with 75mm high heatsink, and your PCB is 76mm plus the height of the transistors, so it doesn't fit. I have a work-around but is a bit messy. Your original powerboard prototype would have been perfect. Anyway just a bit of "food for thought".

Thanks for your assistance, and I really appreciate you making this public domain for DIY'ers like me!

Craig

1 - ksc3503/ksa1381 Hfe... classification E or F will be sufficient. I use the "E"

2 - The "Q" is a remnant from my simulator (I will correct that on schema) ,your assumptions are correct.

3 - Inductor - use 1/2" dowel or pencil , wrap 18 ga. magnet wire 15 turns one layer or 8 turns 2 layers. Mount 3W resistor (R98) on board first , then bend the wire on the outside of your new inductor into the inside of the 1/2' circle. Inductor mounts vertical on PB60 V1.1.

4 - About the PCB's , print them out on (staple's) photo paper , iron them on to your pcb material at high heat , soak in warm water , etch in peroxide/HCL (muriatic acid). they look as good as what a pro will do. I have never not made my own stuff , so I can't tell you. Mirrored image is top screen print.

5 - If you could hold on a few days I will make a PB60 "low profile" , I made these (PB60-250) for the standard 125mm (5") aftermarket heatsinks.

As far as boards ... I am going crazy with "Updates" 1/3 of amp performance is layout. The PCB/schema programs do not have all the "tricks" integrated into them.

The PB250 (Below- 1) - TONIGHT ..I will have an advanced version with Dr. jagodic's Cap multiplier , larger voltage board caps, LED'S better star grounding , AND the proper implementation of Doug self's magic Vbe resistor. NFB return will go shielded direct from inductor to euroblock.

PB60 - I did not change that. ALL is good.!!

NEW!! BX1.1 (below-2) the "BaBy" smooth sounding classic RCA bootstrapped. amp..


TO DO ... I will make the low profile board (will be the PB60LP) couple days.

FX - super pair amp
EX - "Leach amp" fully complimentary input stage with classic leach VAS.

OS

[AndrewT]

r16=4k7 for +-70Vdc????
R18=4k7?