Help designing double differential symmetrical amp

[hpt]

Hi everybody,

I am a EE undergraduate, relatively new to audio amp design. I am working on a design involving an LTP input stage, LTP voltage amp stage, and two opposite symmetrical class-AB push-pull output stages driving a load in a bridged fashion. Essentially there are two signal paths, one of each opposite phase. Of course the input would be driven by a balanced signal.

This will be something kind of similar to the early 1980's hitachi designs. I may later consider implementing a JFET input stage, to keep the DC offset in check. I haven't thought too much about OPS dveices yet, but maybe MOSFETS maybe BJTs.

Here is where my design is currently and the time domain simulation (please note the output signal trace is divided by 1000):





Please, I would highly appreciate any advice, criticism, suggestions on how to move ahead, etc. Yes, I am aware that this circuit is probably far from optimised. I don't doubt that it may be entirely ill-concieved (I hope not). I have yet to do the maths properly, have just done some rough napkin calcs. Probably I will solidify the emitter degen values, and the current mirrors, design some proper ccs, once I figure out how to apply feedback to this thing? And DC feedback also. DC servo maybe?

I am working with Self, Slone and Cordells fantastic books, but the design I wish to do seems somewhat unconventional with respect to what is covered in these books. I believe for a newbie like me, it would have been much easier to work thorugh a simple 3-stage design, but unfortunately I must work with what I have shown you so far.

I have checked out Linsley-Hoods book, and this gave some good reference to the Hitachi application manual which I have now managed to download and investigate. Also I believe Duncan's book has some good examples of these 80's Hitachi amps, I haven't checked this out yet but I am in the rocess of getting it.

If anybody has copies of the 1981 Hitachi manual I am tlaking about, please let me know, I'd love to see more pages of these!

Regards,

hpt

[DF96]

You can't take a differential output from a normal current mirror, as the input and output sides have completely different impedances and DC voltages.

Personally I have never seen the point of fully differential designs, as the complications outweigh any small advantages, but that is a discussion for another day.

[tiefbassuebertr]

avoid true complementary output designs like the hitachi applications with 2SK135/2SJ49 and that one listed about
Audio (Lateral) MOSFET's for Class AB - Overview, P-Spice URL's

a good approach offer this thread:
NVMOS amplifier
especially post 297
more ideas I have listed there:
Only N-Channel MOSFETs (NMOS); better Audio from non complements by Audio Power?

[CBS240]

Have you tried a bridge topology, iow complementary cascade differential circuit? Having both phase VAS outputs operating in complementary push pull has some advantages. As far as the above schematic, 600mA could not be flowing for 2Ivas source, 300mA X 1K is 300V...not enough room. 2Ivas current is actually zero. Current could not flow to the bases of 2Q1/2Q2 because 2Q10/2Q11 has no base current DC bias and do not conduct, thus are merely open circuits. Jumping into linear audio amp design and starting with balanced topology may not be the best place to start. Even if this was to be typical LIN, there are several fundamental issues to be worked out in the above diagram.

Full complementary balanced topology works very well, but it has to be implemented properly.

I added a drawing of the VAS stage in this amp. The complementary cascade diff is outlined in yellow, along with all the wonderful cascode connections. The feedback type CCS or ring CCS is for the bias currents, in this case active, but they can be static by replacing the CCS circuits with resistors and removing the common mode loops.

[hpt]

Thanks CBS240. I realise I am jumping in the deep end with this kind of design, but unfortunately I am required to produce in some way a balanced or bridged design, and preferably with a differential VAS with two opposite phase signal paths. I also realise that many people believe there is no point of fully differential designs.

I was very suspicious of the normal current mirror on the LTP taking both phases, as it would try to force equal currents in both legs, but I tried it and it actually seemed to work okay in SPICE. Obviously I can't be taking things at face value and this was a silly idea.

So I will analyse what you have told me regarding the problems in my above circuit, and hopefully this will teach me some more things about such a design.

The problem is, I am referencing schematics in the Hitachi application notes, as well as a few wodd balls in Self and Slone, and then trying to apply the design methodology from Self, Slone and Cordell regarding a 3-stage Lin.

" Having both phase VAS outputs operating in complementary push pull has some advantages." Please elaborate on this point.

Please if you could suggest any references where I might find information regarding the design of thiskind of amplifier, this would be great. I originally wanted to steer clear of complementary differential as this looks even more complicated to me. But I will be looking through the amp you showed me.

Cheers

EDIT: Something like the front-end of the Kaneda amplifier http://www.bonavolta.ch/hobby/en/audio/kaneda_p.htm coupled to two output stages one for each phase, and driving the load bridged. Compared to a non-bridged scenario, should have better slew rate, less demand on power rails, and less load on amplifier output, no?

[jcx]

doesn't Bob's Book shows his "balanced" current mirror driving 2nd diff pair from his AES Mosfet Amp article?

[hpt]

Thanks jcx. I have read that part of the book; however after buying and digesting the AES paper, the amp makes a lot more sense. This is a great reference.

I now further realise how misguided my own attempt was at the circuit shown, lacking proper biasing, the mirror, the vas collector load resistors / ccs current and lack of rail headroom, and all. I guess I can't forget basic circuit theory when I am plodding through optimisations on example circuits and what not.

Now I am wondering how this amp might be implemented with a) balanced input, and b) differential output / bridge output.

[jcx]

you could load each side of the 2nd "VAS" diff pair with CCS for phase split drive but balancing the CCS current sum against the diff pair tail current may require a servo

or you could duplicate the whole 2nd diff pair/mirror VAS and connect it anti-phase to Bob's Q8,9

differential out amps need something setting/controling the common mode of the output - see the lit on recent diff in/out op amps - they have a common mode pin and internal servo subcircuit

[CBS240]

" Having both phase VAS outputs operating in complementary push pull has some advantages." Please elaborate on this point.

Better symmetrical slewing. Since your output is the difference between the two phases, then having a distortion in the negative phase will not be canceled by the same distortion in the positive phase, so you end up with the distortion in both halves of the signal, as far as the load sees. A symmetrical common mode distortion will not show up at the load with BAL outputs.

Your double differential is a good approach, try building the same circuit with opposite polarity devices and put them together. Symmetry between phases and symmetry between polarity of devices. Think of building two complementary amplifiers, mirror images of each other reflected over the common mode biasing points, one side of the differentials' for pos phase and the other for neg phase, together with the same common mode bias currents.

Originally Posted by jcx

differential out amps need something setting/controling the common mode of the output - see the lit on recent diff in/out op amps - they have a common mode pin and internal servo subcircuit

This is a good point. There must be a DC reference for the inputs that bias the DC output of each phase to zero volts. You should have no offset between each phase and gnd, this ensures no offset between the two outputs and that they have equal dynamic range.

[PMA]

Originally Posted by hpt

Here is where my design is currently and the time domain simulation (please note the output signal trace is divided by 1000):

You might like to rearrange it, to get the working balanced circuit:

But 1k resistors and 600mA CCS is impossible. You would get it saturated.