Transient Attack and Amp Design

There are innumerable design variables that could relate to the good transient response you have noted. And we wouldn't have adequate visibility into most of them without at least having schematics and other detailed design documentation on the amps being compared.

If you haven't seen it, you'll be interested in this review of the DNA-1 that appeared in Stereophile in 1992. Robert Harley had observations about its transient response that were very similar to yours:

http://www.stereophile.com/solidpoweramps/520/

Regards,
-- Al

Transients require large amounts of near instantaneous current without the voltage in the power supply dropping. Essentially the bigger a power reservoir you have the better an amplifier can cope with transients. It matters less how the power is stored (a capacitor bank or a couple of large capacitors compared to the total amount of stored energy). In theory, the lower the internal impedance from the power supply to the output transistors and teh lower the output impedance of the amplifier the less voltage drop will occur when drawing large amounts of current and hence a better transient response (less compression).

It helps to have a speaker with higher impedance. Speakers with dips of quite low impedance will be the cause of most observed problems (rather than the amp).

Shadorne ;
You stated ...
Speakers with dips of quite low impedance will be the cause of most observed problems (rather than the amp).

Can you supply a definition/value of this impedance dip ?

Thank you .

Saki70, the difference between 8 and 4 ohms can be all that it takes; nearly every amplifier made will sound faster (as well as less bright, more natural) on a speaker of higher impedance.

02-16-09: Shadorne
Transients require large amounts of near instantaneous current without the voltage in the power supply dropping. Essentially the bigger a power reservoir you have the better an amplifier can cope with transients. It matters less how the power is stored (a capacitor bank or a couple of large capacitors compared to the total amount of stored energy).

Your comments point to what I suspected. Now does the transistor proximity to the storage cap(s) make a difference? McCormack puts the caps close to output(?) transistors rather than a considerable circuit distance as I saw in my sunfire and marrantz amps. Or are we talking nanoseconds and thus it is not a significant consideration?

BTW, thanks for everyone's comments thus far!

Now does the transistor proximity to the storage cap(s) make a difference?

Atmasphere can speak to that more knowledgeably than I can, but I don't think that would make a significant difference. The concern would not be propagation delay (which would be a few nanoseconds at most, as you speculated), it would be that the inductance of the wiring or printed circuit board traces between the capacitors and the output devices would slow the transfer of energy.

But I don't think that would be a significant effect either. As a very rough ballpark the inductance of one foot of straight wire is around 15 nanoHenries, which would have an impedance of about 2 milliohms even at 20kHz, and much less at lower frequencies. The numbers would be somewhat different if printed circuit board traces were involved, rather than discrete wiring, but I would think they would still be totally insignificant.

What would be of potentially greater significance would be noise pickup on those runs, but that is presumably filtered out by much smaller decoupling capacitors (that have much better high frequency performance than electrolytic storage capacitors), which should be and presumably are located very close to the point of use in the output stage area.

Regards,
-- Al

Jragsda, glad to hear you are enjoying your DNA-1.

I came up with the DNA (Distributed Node Architecture) design specifically because I do believe that it is audibly beneficial to eliminate (as much as possible) the connectors and wiring between the output devices and the source of their "instantaneous" current. At the same time, the capacitors used may be smaller individually and still add-up to a large effective capacitance. The capacitors still have to be carefully chosen for their audio characteristics, but it has been my experience that these physically smaller caps often have better audio performance (particularly at high frequencies) than their larger brethren. While it is difficult (or impossible) to defend this position from a strictly engineering / measurement standpoint, I note that a fair number of other amplifier designers have adopted this technique since the DNA-1 came out.

Best regards,

Steve McCormack
SMc Audio

Hi Steve, I think there is plenty of evidence out there to support your position regarding smaller caps!

Steve, at the risk of gushing, I am honored that you have responded to my thread. Kudos on an outstanding, musical amplifier in the DNA-1.

In the old Stereophile review (noted above in this thread), your DNA-1 design is noted for its treatment of distortion spectra. Here are JA's comments:

The Test CD also features signals with distortion spectra representing "tube"-type THD, "solid-state"-type THD, and a mixture of harmonic and subharmonic distortion typical of a planar speaker driven at high levels. My experience with generating the signals for the disc using the Audio Precision System One Dual Domain suggests that the situation is more complex than RH describes. While the DNA-1 does have some upper partials present, these are not isolated but are accompanied by the lower partials in an almost regular descending series. This is typical of tube amplifier performance, tending to sound smooth and "fat" rather than grainy.—John Atkinson

I suspect, but have not read, that you have departed from the "tube" distortion spectra approach in your later designs, notably the DNA-500. Is this so? If so, would you please comment as to your reasoning?

Alamrg

Inductance of 1 foot of straight wire is not 15nH. According to this calculator:
www.consultrsr.com/resources/eis/induct5.htm
it is between 330nH for 0.1" Dia to almost 500nH for thin wires.

Large capacitors have large inductance. Bypassing them with film caps creates parallel resonant circuit (inductance + lossless cap) that tends to ring. It is always better to use bunch of smaller caps in parallel to further reduce already smaller inductance and to lower resistance (ESR).

There are better quality caps like slit foil type (used in Hypex kits - very expensive) where foil has cuts to prevent eddy currents and to lower inductance (losses).

Power supplies require large capacitance because they are unregulated. Regulated SMPS running at 100kHz is fast enough to response to any current demand at 20kHz in spite of small capacitors.

Kijanki -- Gee, that's the same calculator that I used! Playing around with it a little more, it looks like I must have set the units to mm instead of inches, to get the 15nH or so, although I thought I checked myself pretty carefully on that. In any event, thanks for the correction. Even at 500nH, though, the impedance would still be essentially negligible at audio frequencies.

Good points about the possibility of parallel resonant circuits, ringing, etc. Thanks!

Regards,
-- Al