Power Supply Resevoir Size

[fas42]

Originally Posted by OnAudio

Its interesting that the SYMEF gets loud without clipping. And is pleasurable too. People you need to experience this amplifier.

That amplifier sounds very interesting, OnAudio, you're making the right noises about its capabilities. In particular, that you can go to high volumes with no or little change in apparent tonality, an excellent marker. I shall investigate further ...!!

Frank

[dkemppai]

Originally Posted by gootee

The high currents should probably be confined to small-geometry local loops whenever possible. And I still think that transient response accuracy ("speed"?) is what separates the merely great amps from the exquisite ones. After all, it appears to be relatively easy to make an amp with vanishingly-small THD (i.e. almost-perfect steady-state response) that still doesn't sound good-enough. So therefore the key must be in the transient response, right?) So focusing on decoupling cap configurations seems like a key, to me, which has been neglected too much, by most (apparently). It's not always easy to get the decoupling good-enough, even at audio frequencies, especially when high currents are involved. It usually requires accuracy to hundreds of kHz and the physical geometry becomes a significant problem, with most DIY construction methods. The main thing there is to even think about it in the first place, and actually calculate the minimum required decoupling capacitance AND the maximum tolerable inductance in its connections. THEN we might see that it's sometimes VERY difficult to get a low-enough impedance to be seen by the pins where it's needed.

It's ridiculous to talk about a PSU's low output impedance without also talking about the inductance of the conductors from its output to the load, for example, and about what the load's actual impedance vs frequency requirements are, specifically, and at what current levels.

Tom,

I really like you train of thought. I've been trying to keep on this thread, but posts are coming faster than I can read them.

Your design methodology makes sense, and is basically used in high speed design both analog and digital. What you are proposing as important almost points towards a switching supply, where small amounts of very low impedance capacitors could suffice in ripple filtering (As high speed caps with low impedance are practically a requirement). With a small supply, those little low impedance caps could be right next to the amp, with minimal feed impedance by wiring.

I agree with your assessment of transient response. I am coming to believe that phase linearity in the amplifier (i.e. constant delay vs. frequency) is extremely important in good reproduction. However good phase linearity may be hindered by a poor supply. I believe you are right in your assessment that the rising edges and falling edges of high dv/dt waveforms is crucial in faithful reproduction. Power supply and power supply feed impedance is most likely very important.

Some attention needs to be paid to capacitor construction in this thread. Many technologies exist, some exhibit horrible microphonics, some change capacitance with applied voltage, and many have extremely high impedance values. Even the geometry for an existing technology should be assessed . Consider a 1000uf aluminum electrolytic that is tall in skinny vs. one that is short and fat in their construction.

I'll be trying to keep up with interest!

Dan

[dkemppai]

Originally Posted by Terry Given

...multilayer heavy-Cu PCBs for my designs, I always twist wires (plaiting works great for a center-tapped winding) and prefer to use planar transmission lines (our DC bus caps were bolted to a pair of 2mm Al (or Cu for the big ones) plates separated by 1mm lexan. its pretty easy to get stupidly low inductance like that.

...many in parallel - low ESR, low ESL, lots of surface area for cooling) caps, a planar transmission line to the amp with a bit of series R, and another parallel cap bank at the amp.

most of the audio amp layouts that I have seen have been single-sided and appallingly bad. actually, thats unfair - almost every PCB layout I've ever looked at has been bad. minimum PCB cost is not necessarily optimum wrt EMI.

I believe you are spot on here. Board layout is as important as circuit design. Type of capacitor, and number in parallel. It's amazing what a few uF of SMT capacitors can do.

I agree that a lot of boards are poorly designed. Spend some time looking at some RF layout for the 2-3Ghz range. Build a few circuits for that range, and you'll get to see how important layout is!

Will be following with interest...

[gootee]

The three worst things that I often see in diy power amplifier board and circuit designs:

1. Input signal and input signal ground are not kept as close together as possible, everywhere.

Ditto for AC mains pair, pairs from secondary to rectifier to caps, DC power/gnd pairs, output pairs. If they are in wires, twist them tightly together, all the way from end to end. If they are on a PCB, keep the traces right next to each other, or (better) right on top of each other on opposite sides of the board (or in adjacent planes), or put one of the pair on a plane, either all around the other or above or below it, everywhere.

If there is any open space between the input signal and its ground, ANYWHERE, then they have "enclosed loop area".

A time-varying magnetic field in the air will induce a corresponding time-varying current in a conductive loop, in proportion to the geometric area enclosed by the loop. (Part of Faraday's Law and Maxwell's Equations.)

But also: Any loop with a time-varying current will induce a time-varying field in the air, in proportion to the geometric area enclosed by the loop.

So sometimes they are mostly receiving antennas and other times they are mostly transmitting antennas.

Don't make antennas! Keep all pairs TOGETHER, very close together, everywhere and always!

2. Shared ground conductors:

All conductors have parasitic inductance and resistance. A time-varying current flowing in a conductor will induce a voltage, across the inductance, that is proportional to the time-rate-of-change of the current, i.e. v = L di/dt. (Note that the induced voltage can be large even if the current is small, as long as the current's amplitude is changing fast-enough. And the current will also induce a voltage proportional to the resistance, i.e. v - iR .

If the conductor is a ground-return conductor, then the voltage back at the non-ground of the conductor will not be "ground". It will be a time-varying voltage.

If the amplifier small-signal input ground reference point shares that conductor, or part of that conductor, on its way to the main ground, and that ground-return conductor carries the currents from, for example, the decoupling capacitors' ground and/or the output/zobel ground, then the input small-signal ground-reference point for the amplifier will be a time-varying voltage, which will effectively be ARITHMETICALLY SUMMED with the input signal. That would be "a BAD thing".

Use star grounding. Keep the input signal ground separate from the other grounds. Run a completely-separate ground conductor all the way from the input signal ground to the star ground point.

3. Improper Decoupling Capacitor Configuration:

In addition to a high-frequency bypass cap right across the pins, use enough decoupling caps, with low-enough inductance in their connections to the device (i.e. close-enough!), right at each power output devices power and grouund pins.

[Nico Ras]

Those are very true words spoken Tom.

[apassgear]

Tom,
20 years ago I was able to do some good enough etchings, but on my attempt last year was a complete disaster transferrin from different medias and I got discouraged In the meantime I did scribed PCB’s which are less than good enough. I will follow your link with an open mind and maybe I can do again. Regarding double sided mmm but can try…that would awesome.

My layouts (art work) are done on autocad which I’m very familiar with. At different times I have tested a number of dedicated software from sophisticated to more humble ones but I did not feel comfortable with them and I guess mainly because of poor libraries. Now I have a simple one loaded on the PC Design Spark which is free but haven’t tested it. I know some other guys also use acad for their work too but of course a dedicate program should be easier to use.
Since I’ll be going to F5T V1 I will try that DesignSpark prog and see what happens.

Thanks for your encouragement

Cheers, Tony

[mightydub]

Originally Posted by gootee

All,

Attached are the four files for a very simple power supply test-simulation. I have much-more-detailed ones that I will upload soon. But this one will get everyone started.

I'm late to the party and haven't made it through the entire thread, however here is my contribution from almost 3 years ago to simulation on this topic:

AC to DC Power Supply - diyAudio

This is just a very basic model and can easily be extended to include the effects of ESR, lead inductance, etc. The amount of rail droop with real music is significant - granted what I used was very bass heavy.

[Nico Ras]

Originally Posted by apassgear

Tom,
20 years ago .....I know some other guys also use acad for their work too but of course a dedicate program should be easier to use.
Cheers, Tony

Tony,

your are not alone. Problem with ACAD is that you need to memorize the net-list . In a dedicated PCB program the layout is referenced against the schematic which makes a big difference.
The design spark program is not bad at all. I use the commercial version called EasyPC Win. What may be nice but also restrictive of design spark is that the libraries are tied in with RS Components.

[jacco vermeulen]

Originally Posted by Terry Given

a pair of plates separated by 1mm lexan.

Just think how stupidly easy it is for a DIYer to make bus bars like that (instead of a bowl of spaghetti)

[OnAudio]

Originally Posted by fas42

That amplifier sounds very interesting, OnAudio, you're making the right noises about its capabilities. In particular, that you can go to high volumes with no or little change in apparent tonality, an excellent marker. I shall investigate further ...!!

Frank

Welcome to the club