CMoy with Jung multiloop & output buffer?

[dhaninugraha]

hi guys

having built 5 CMoy amps with good success (first 2 amps followed Tangentsoft's schematic, last 3 amps followed Chu Moy's), I've been wondering what improvements I can make to the CMoy, besides using better parts and buffering the virtual ground. balanced CMoy came to my mind a few times, but I had difficulties looking for resources to study about it, so I kinda scrapped the idea (they'll be some sort of "far in the future" project).

anyways. when I was skimming through the PPA pages at Tangentsoft, I noticed that Tangent mentioned "Jung multiloop" there. and the multiloop happened to appear at the PINT page too. I was wondering, what the hell is this multiloop stuff? so I looked at Walt Jung's explanation as well as looking at the PPA schematic showing the multiloop (honestly I couldn't really understand the stuff at Jung's page -- they're too technical for my critter brain), I came to understand that the Jung multiloop involves two loops: local loop and global loop. and as I examine the PPA schematics further, I also note the presence of a buffer there. a thought came to my mind: why not incorporate the Jung multiloop and buffer into my CMoy?

so this is the schematic I drew yesterday (the bottom schematic is an alternative to the top):

my questions are:

1) would the 551/552 be good at buffer job?
2) I read the 551/552 datasheet and it says that 551 is good for unity gain, so is it okay for me to connect the 551's output straight to the inverting input? and is it also okay to make the 552 buffer have a gain of 6, as AFAIK the 552 is stable at gains above 5?
3) I wanted the AD825's local loop to have a gain of 100 and the global loop to have a gain of 11 (hence the resistor values), are these values good?
4) I'm planning to bypass the AD825's power rails with 0.1uF Wima MKS2 in parallel w/ 0.01uF C0G ceramic. is this good? do I need to bypass the buffers' power rails as well?
5) on the bottom schematic, I put a question mark on the resistor symbol between the AD825's output and OPA551's input. do I need this resistor or do I not, and if I do, what is the best value to put in there? because I got the notion that both inputs must see similar resistance, am I right?
6) any alternative(s) besides the 551/552? IIRC, majkel said that the AD8397 is better than 2x BUF634 at this job... however it was also mentioned that the 8397 is somewhat cranky due to its bipolar input... what modifications do I need to do to my schematic to accomodate the 8397?


any inputs would be very very much appreciated

and big apologies for my bad English and bad drawing

 

[FallenAngel]

1) Not really, tried them, didn't like them as buffers (they're opamps actually). What about the good old BUF634 or LMH6321
2) OPA551 can run unity gain, but read #1
3) I'd keep the local loop at 500K, but read the PPA pages for parts choices
4) Can't hurt, but still make sure there is a few hundred uF of capacitance in the PSU (reasonably close to opamps/buffers)
5) 100 ohms sounds reasonable
6) AD8397 is both cranky and simply an opamps, not a buffer

Consider just building a Pimeta V2

 

[TzeYang]

i think the performance is a tons better by using the BUF634 or the National Buffer LMH6xxx.

OPA55x is a bit slow for an output stage so you must really make sure that the input stage is not too fast otherwise you'll run into problems.

Second FallenAngel, build a pimeta.

 

[cobaltmute]

You could use the Monofied Sijosae Buffer as your buffer:

Very happy with the results with this.

As the above said, you're trying to re-do the Pimeta.

 

[dhaninugraha]

true indeed, it's pretty much like I'm redoing the PIMETA except without an opamp on the ground channel

FallenAngel, many thanks for the insight

TzeYang, the AD825 is only 46Mhz, would it be "too fast" for the 55x to handle?

cobaltmute, thanks for the sijosae buffer!

I wonder if I can redo the layout so I can fit SIP sockets under the board (it'll look pretty much like an Audio-GD discrete OPA), so I can just plug them straight into a DIP8 socket on the board...

 

[TzeYang]

the OPA551 has about 3MHz and 552 at 12MHz IIRC (cant remember at which gain)

bandwidth drops as gain goes up, so even if the AD825 is slowed after factoring the external gain, it's still a bit too much for the 551 to handle.

 

[dhaninugraha]

I see...

therefore, is it wiser that I just go straight to BUF634 (or the Sijosae buffer above)?

 

[cobaltmute]

Quote:

Originally Posted by dhaninugraha /img/forum/go_quote.gif cobaltmute, thanks for the sijosae buffer! I wonder if I can redo the layout so I can fit SIP sockets under the board (it'll look pretty much like an Audio-GD discrete OPA), so I can just plug them straight into a DIP8 socket on the board...

You leave the leads long on the devices noted as V+, V-, In, Out and they will plug straight into the DIP8 socket to replace the BUF634.

From my Pimeta V1:

 

[tangent]

You can't have gain in the output buffer like in your initial schematic. It will look like an "error" to the op-amp, and it will cancel it out. If you want two stages of gain, you have to move the OPA55x outside the other op-amp's feedback loop.

 

[jcx]

actually gain in the output op amp can be a good idea, but there are stability concerns that need to be addressed

the overall gain is determined by the outer feedback loop, but with gain in the output amp the input amp doesn't have to work as hard - for instance a fast output amp like AD811 as shown by Jung or other CFA can "multiply" the input amp's slew rate by the gain around the output amp

another reason for local V gain in the output stage is where input common mode range doesn't extend to the supply rails - the AD8397 is an example where the input can only take ~1/2 the supply V but the outputs can swing nearly rail-to-rail so a local gain of 2 in a AD8397 output amp is a good idea in a Jung multiloop

 

[dhaninugraha]

many thanks for the answer guys, I'm still studying and restudying each and every answer you gave me ATM

so is it just wiser if I just use a BUF634 on the buffer position there?
because what I'm looking for isn't a dual-stage amplification but rather a buffer after the gain stage

EDIT: new schematic w/ the BUF634
(I guess this is actually redoing the PIMETA, after all... LOL

)

do I need to operate the BUF634 in its wide-bandwidth mode?

 

[TzeYang]

If you're comfortable with the use of SMD buffers, use the National equivalent of BUF634(refer to PimetaV2).

It sounds better and performs a bit better IIRC.

Gain of 11 sounds a bit too high IMO. Get something lower.

The rest of the circuit looks okay so far.

 

[dhaninugraha]

@ TzeYang:

the LMH6321 isn't it?

when I looked at the datasheet, there are pins EF and CL, what am I to do with these pins?
and I'm supposed to connect its GND pin to the circuit's virtual ground plane, am I correct?

 

[TzeYang]

CL is like bandwidth pin for the BUF634 <---refer to the datasheet and find a suitable value. Or you can just use the value of Pimeta V2

EF is the error flag pin used to thermal shutdown of the chip. leave this open you don't need it.

connect GND to your ground.

 

[dhaninugraha]

thanks a lot for the enlightment, TzeYang

I've redrawn the last schematic, from the perspective of the board layout...
the thick lines around and across the middle are the ground planes.

I don't show the power rails out of the splitters as well as the buffer's bandwidth control resistors so it'll look less obfuscated than it already is

the power source will be a 30VDC 400mA wallwart, and both splitters will be connected in parallel to it.

PS: the TLE2426 will be replaced by a discrete rail splitter which is pin-compatible, made by psychaudio of audiophile-id forum. he made two versions, one with 100mA output current and another one which is capable of 800A. I think I'll be getting the 100mA version for my amp. I'll post the picture and/or link soon.

what do you guys think? please do let me know