Sound Quality Vs. Measurements

Status
Not open for further replies.
dvv,

After all that, we are in pretty much agreement on the crossover question. Now to you helpful suggestions on my learning about amps.

Your questions are all quite valid as usual. Again, sorry about the RED. Printkey 2000 started acting up.

Easy answers first:
"Budget" as it is just a learning experience.
36 V rails as that is what my available transformer will provide. Existing chassis means I am only buying the actual amplifier circuit, so it is a cheap project.

*Something around 50W. Not at all critical. It is what the transformer will supply. I don't have its ratings, but it came in a 60W integrated amp that was intended to be high end. (It isn't.) Torroid that is reasonably large physically, so I am somewhat confident it will suffice. Backwards way to work, but again, this is just for my education. I picked 27dB gain just to be compatible with my other amps if I were ever to use it in a bi-amp setup. Picked 1.54 V as the input. It will actually have a little preamp stuck on it, so I have control over that.

*It needs to drive "reasonable" speakers. Basically my conventional 2 ways, so not difficult. No exotic drivers. "Nominal" 6 to 8 Ohm. As you would guess, no extra effort put into Zobels in my designs as I have not been shown they are needed to help an amplifier, but may actually be a tiny detriment. No funny ribbons or anything.

*Yes, the physical construction would have the rails for the first stages back to the supply, not tapping the rails on the board. Stars for ground and power off stubs. The physical aspect I am pretty confident on. I am not doing separate supplies or winding's as I might if I were designing from scratch. As this is BJT, I don't need the higher driver stage voltages as is a benefit in a MOSFET.

* Back and forth on EF or CFP. Originally I had one pair of outputs, and I may go back to that. Even in the artificial world of Spice, it was a bear to stabilize. (current version posted below after that effort) I am concerned about bias thermal stability in the CFP. Just setting a fixed value in the model seemed that very small changes made very big effects. But heck, how more touchy can it be than the Exicon's I put in the Hafler? You are probably right in multible CFP not being the place to start

*I went double outputs as it showed somewhat less distortion at higher power. Kind of a hint from Self. I don't need them for power handling.

*The big feature for low distortion and noise was buffering the VAS. From what I gather, CFP is fine for single outputs, but with twin I should consider EF.

*I spent a huge amount of time modeling different CCS's and found the mirror to be a reasonable good enough answer. A more perfect input current mirror is possible, Wilson et-al, but it is totally swamped by output distortion so I left it as good enough. Back to it being "reasonable" not state of the art plagiarizing the expertise of many in these very threads!

Device selection is a very week point for me. I find a datasheet I like and is is not available, SMT, or I can't find a model for it anyway. I am starting with the junk and probably counterfeit parts that came with an e-bay Chinese ripoff kit. The board is close enough to hack up so it was an easy start. I do have some known quality of the small signal trans. I measured a bucket of them for the Hafler, so I should have some reasonably well paired inputs. Learning to write a SPICE model is a long term goal.

Over in the SolidState forum I got some hints on SOA protection and Baker clamps. Lots to learn. Most efforts I try there have very large negative impact on distortion. I found that playing with the DH-120 rebuild. Some is a fact of life, most is I don't understand that yet.
 

Attachments

  • FirstAmp.jpg
    FirstAmp.jpg
    950.2 KB · Views: 114
Got the picture now. I printed it and will, over the weekend, simulate it in my sim program, just to see what I get.

Overall, I was right, this is next to exactly the same as Sansui 9900 and 11000 were, almost down to the latest detail.

So, here's what I propose - I sim the schematic and report on what I got. Then, after I've had some experience with it, I tell you what I've found and what I think you should change to make it better.

I can also jot down the complete procedure of calculating the overvoltage and overcurrent protection, or you can settle for fuses. However, the way it can be done electronically is such that unless you're really going to town with that amp, it is most unlikely the protection circuits will ever activate, but in case of need, remember, the fastest fuse around is a transistor. I've seen my share of burnt out amps in which the "protection" fuse was as new. Also, experience teaches me that the bad name these ciruits have is in effect an urban legend; they can indeed choke an amp, but only if they were designed to be aggressive and/or to cover up for inadequately dimensioned output stage and/or sloppy PSU. Anyway, I'll give you the tools and you decide.

As a side note, for my drivers, I'd always use Motorola/ON Semi MJE 15030/15031 trannies; they are more than powerfull enough at 50W/140V/8A, and if you look at their data sheets, you will see that they hit 60 MHz quite easily.

Also, I would seriously consider making the VAS as a cascode stage. This practically eliminates many problems as is, so you compensation caps are usually much smaller (roughly half their present value), which means twice the voltage slew rate, among other things.

And I definitely would use a triple output stage, along Bart Locanthi lines; it usually sounds better and offloads your VAS for current requirements.

You say you've looked at quite a few schematics; I would still like you to look at a few I can supply, from Hikko and Harman/Kardon mostly, which I feel would be beneficial to you in this particular project. Please do send me your mailing address to dvv@bitsyu.net . You have nothing to lose, only something to gain.
 
tvrgeek,

Power supply is +/- 36V before or after the full wave rectification?

At first glance, if it was a 60 WPC amp, it should be +/- 25V before and +/- 36V after the bridge rectifier, but you never really know. My H/K 6550, nominally 50/70W into 8/4 Ohms, has +/- 51V PSU lines ...
 
Frank,

A wise German taught me something I find to be very important - before you get down to it, you need to make up your mind as to what you are designing it for, commercial Hi-Fi or straight music..
dvv, thanks for your good thoughts! My aim has always been the all-out assault on getting good sound, if that can be commercialised all well and good - I would find little satisfaction in producing a me-too product, there are plenty of others who would be far more competent in the bling and marketing factors.

My overall approach these days would be not to re-invent the wheel; rather, take a good standard of normal, value for money product and 'turbocharge' it - add the 'go faster' bits of real worth, to get a 10 fold or better improvement of the qualities that count, not silly spec. factors ...

Cheers,
 
The very pages I was looking at - yes, a transformer can deliver more than the nominal amount, but do it continuously and things might start looking sick. I don't like the distortion measurements - this is part of the trendy, 'nice' distortion movement, and that means the sound lacks real bite and composure when it's needed for the complex stuff - I must say some of my worst listening sessions have come courtesy of Krell ...
 
One can't just zip across to the technical stuff when it suits, to "prove the point". The chip amps "work", because they're opamps, with added grunt - give them a power supply with a big heart, and they'll do the job. As a counter example, I have a cheapy, purchased over the counter, untouched audio amp, 100W, that uses, surprise, surprise, chip amps that "do" 100W. Well, this is a gutless wonder, falls apart as soon as a decent wallop of sound comes along - and looking at the power supply it's obvious why, the absolute minimum which allows them to test that it 'works' ...

Well powered chip amps use feedback in the correct fashion, which means the sound stays consistent up to the point of clipping - as it should do ...
 
Last edited:
Status
Not open for further replies.