[Syzygies]

I raised this question first in PPA v2 Project Announcement, in reply to PPL's post The PPA 2.0 Buffer tech info. I did not intend my question to be a suggestion for revising the PPA, whose design I intensely admire, and whose size would not accommodate my idea. So I'm moving my question here.

Given enough board space to accommodate the experiment, would it make sense to incorporate three diamond buffers into the power supply stage of a headphone amp, in addition to using three diamond buffers for the output stage?

We are familiar with op amp stages without buffers (CMoy), and op amp stages that incorporate buffers (Jung multiloop topology). We are familiar with power supply stages without buffers. Stop reading for the moment, and decide how you would fill in this remaining square in the chart: design for yourself a power supply stage that incorporates buffers.

I was thinking of starting with wide rails, e.g. an 18 volts or higher battery of NiMh cells, and using voltage references to somehow derive +2, +9, +16 as the three equally spaced voltages of a V-, G, V+ virtual ground scheme. Now use these references as the input signals to three diamond buffers, powered by the original 18 volt rails. Use the output of these three buffers as my power supply for the later stages of the amp, i.e. the next three diamond buffers have L, G, R as inputs, and the V-, V+ from the power supply buffer outputs as their rails.

I understand that op amps and buffers have pretty good power supply rejection ratios, which is why sagging rails under buffer load don't affect the sound that much. On the other hand, there's got to be a reason why people want to pay good money for Black Gate power caps, or build themselves massive, "stiff" power supplies. I can only conclude that the quality of the power supply does somewhat matter. It would seem to me that a well-designed active power supply could outperform a passive power supply.

Quote:

Originally Posted by ppl Syzygies> the concepts you talk about while conceptually intriguing are not applicable for the small form factor of the PPA. Please remember that this is a portable Amp and space is limited. I cannot imagine where you’re going to put 3 more buffers at on the ppa board. I believe that with the power supply rejection ratio provided by the ground channel and the Jfet rail isolation network. I cannot see how you consider that we addressed this issue with massive capacitance since the PPA actually use substantially smaller capacitance that other premium Amps. However thanks for taking the time to look into this and comment upon it for us hear.

Sorry, I was replying to the general potential of your new diamond buffer design, I did not intend my suggestion to be directed to the PPA v2, which I see as a showcase for this buffer, but by no means its only application.

Do you think my idea would work?

 

[jogor]

i've seen that the meta42 has buffer in its power supply. for simplicity why not use IC buffer instead of discrete diamond buffer? this is a good bet if space is a premium. but i'm not sure how to implement this in 3 channel amps. i think your idea is really good, if i have the time i would like to check on this and see how psu with buffer respond to dynamic loading.

 

[Syzygies]

Thanks, I hadn't noticed that, I'd read that the MINT was a scaled META42 but I never studied what was dropped.

The META42 uses an EL2001 buffer after the rail splitter for its virtual ground, as shown in this schematic. Tangent didn't mention this in Virtual Ground Circuits; he instead describes the "ground channel" that evolved for the PIMETA and PPA. While people tend to match components in ground channels with the LR audio output channels, I had been thinking of ground channels as being instead part of the power supply, as his article implies. This is what lead me to wonder why L,R weren't getting equal treatment. Perhaps my idea only requires five buffers, not six.

In the META42 or any ground channel design, the +,- rails are coming straight from the batteries, unbuffered. I'm curious about shrinking their voltage using voltage references (so I'm not trying to swing my power supply buffers rail-to-rail) and buffering them. I haven't seen a design that does this. As you say, IC buffers are smaller. However, it would defeat the purpose of this idea if they couldn't keep up, if their slew rate was holding up the whole class. Start stacking five or six sets of them, the price adds up. Part of the appeal of using diamond buffers in PPAv2 is that the price is coming down.

PPL's diamond buffer has some amazing specs, by comparison, and it's cheap to build, more so if one buys extra parts for matching. If one already has the parts for six on the kitchen table, why not use them?

If an active, buffered power supply worked well enough, it could be the final filtering stage for a boost regulator, allowing the use of AA NiMh cells, 2x to 3x more juice by volume than what people tend to use in their portables, and it could allow the use of smaller caps. Suck it in and learn to solder surface mount like everyone is telling me to do, and the size shrinks again. Now slide two boards into a Hammond case (I'd rather have this than a fast charger like the NiMh Battery Board, trickle charging is fine for me) and this starts to look like a viable, affordable portable option to me.

 

[tangent]

Quote:

Tangent didn't mention this in Virtual Ground Circuits

Certain of that, are you? Look at the third schematic in the "Simple Buffered Virtual Ground Circuits" section. And from just below that, I quote, "...you can replace the divider with a TLE2426....That's how we did the virtual ground in the META42 amp."

And the reason we dropped it is because the output impedance of a bare open-loop buffer is up in the tens of ohms. One benefit of the "ground channel" concept is that the buffer is wrapped by an op-amp, reducing output impedance dramatically. Since ground is supposed to be an infinitely low impedance sink, this is highly desirable.

 

[Syzygies]

Quote:

Originally Posted by tangent One benefit of the "ground channel" concept is that the buffer is wrapped by an op-amp, reducing output impedance dramatically.

Thanks. I had been thinking of copying the ground channel intact for the other power supply channels, not that I understood the reason you gave, until I read PPL's post The PPA 2.0 Buffer tech info asserting that the diamond buffer could run naked:

Quote:

Originally Posted by ppl This lends itself to use as a stand-alone Buffer that will allow you to build the PPA without any op amps and related support circuits. By simply jumpering the op amp non-inverting input to the output and not installing the op amp casscode constant current sources your PPA can be a unity gain feedback free headphone amp with an all discrete component signal path.

If time, space, and components were free, do you think that copying the ground channel concept for power supply V+, V- could make a modest improvement in sound? Your choice, wrap an op amp in a buffer, or use PPL's diamond buffer alone.

 

[tangent]

ppl's buffer is still open-loop, and therefore still has high output impedance. TANSTAAFL.

 

[DaKi][er]

Syzygies- have a look at Gilmores psu for they dynahi's, thats pretty close to what your suggesting