D-Fend

[Dan30]

Awesome! I have 1u of empty rack to fill.

[Bruce Weldy]

Should be able to put a bunch of 'em in a rack space....any idea what the layout will be?

2-channel, 4-channel, 8 channel?

User programmable?

[LelandCrooks]

Not a clue. But I've been out of the loop for a while.

[Grant Bunter]

I've been quizzing josh.martin@eminence.com and he is happy for me to share this info :

"D-fend will be available in 3 forms for end-users. It will be available as the "raw" board that you see in the pictures with a small heat sink on the back with mounting holes. It will also come in a stand-alone unit, much like the size and configuration of a standard direct box. This units will have mounting holes or brackets as well. And thirdly, it will come in 4 and 8 channel rack mount units.

cheers,
Rich.

Here's your answer Bruce, it was on page 2

[Bruce Weldy]

I've been quizzing josh.martin@eminence.com and he is happy for me to share this info :

"D-fend will be available in 3 forms for end-users. It will be available as the "raw" board that you see in the pictures with a small heat sink on the back with mounting holes. It will also come in a stand-alone unit, much like the size and configuration of a standard direct box. This units will have mounting holes or brackets as well. And thirdly, it will come in 4 and 8 channel rack mount units.

cheers,
Rich.

Here's your answer Bruce, it was on page 2

Thank you, sir.....

[Grant Bunter]

^ You're welcome

[Tom Smit]

It steps impedance. The amplifier never delivers the voltage over the setting. It just never produces it, that's how they get away with very small heat sinks.

I thought by ramping up impedance the current from over voltage was reduced? Voltage being unaffected.

There would be a voltage drop across the resistance of the Dfend, thus reducing current through the circuit.

I couldn't figure this out simply because I was thinking in a traditional way. That is, impedance would mean resistance. Which would imply more and more resistance, which in turn would turn the power into heat, in my mind anyway. But last night at work, I got to thinking...impedance means just that. Something impedes the flow. So, how about this? Might the D-FEND have capacitors to store energy, and release that energy against the positve or negative speaker wire as is needed in it's specific time domain, to impede the flow. If the current power supply units can switch at, what , 100,000 times per second, why couldn't that be applied to this theory? IOW, if the limit is supposed to be 35v and the amp is capable of 70v, the device would apply up to 35v against the positive (and then the negative) amp terminal as is needed (think AC).

[SirNickity]

I'm not quite following your theory. Putting a cap in series with the driver would form a high-pass filter, so it would have to be a sizable cap to not interfere with subwoofers. Putting one in parallel would create a short at high frequencies (low-pass), also determined by the size of the cap. Since many amplifiers have an inductor on the output (to counter capacitive loads, which are hard on the transistors) I'm not sure, in practice, a paralleled cap would do much of anything at all.

The explanation from Eminence got broken somewhere between the engineers and the salesmen. The amp won't "just not produce the voltage" -- as we've been told countless times on this forum, the amp will deliver a certain voltage, PSU willing, despite the load on it. However, it won't pass quite as much current, which is what actually makes the speaker cone move and the voice coil heat up.

I would imagine the added resistance is simply in series with the load, and any voltage drop across the resistor would turn into waste heat. I don't know how they intend to get around that. Obviously, they don't entirely -- there are heatsinks after all. There's a bit of a curve where small additional resistances (say 1 ohm or less) don't drop much voltage (most of it is dropped by the load), and thus don't create much heat. Large resistances (higher than the load) will start to restrict current flow to the point where there isn't much heat because there isn't much current, but you start to reduce the power delivered to the speaker drastically -- like beyond protection and into "is this thing on?" The area between (1/2 to 2x load impedance) shoulder the load proportionally to the driver, so that means 10s or 100s of watts being dissipated in the D-Fend.

So I wonder what sorcery they've come up with, or if they're counting on low duty cycle to sink heat. Remember, the resistance is switched based on output. Under the threshold, there's no (appreciable) added resistance, so heat build-up is only a problem when overdriven. (Maybe after too long in the red, the system inserts impractically large impedances to protect itself, and the load. Call it negative reinforcement for the loose nut on the mixer.)

As always, I'm still learning, so if I made incorrect assumptions or I'm just plain wrong, please correct me -- nicely.

[Gregory East]

Blowed if I know how they do it, we need a headspin smilie.

Raising impedance ( resistance ?) momentarily would drop current through the floor so the amount of heat generated would be mimimal. Makes sense to me.

[Rune Bivrin]

It's most likely a switching circuit, operating at 100+ kHz. If the voltage produced by the amp is too high - as determined by the in-circuit DSP - it will simply start switching on and off rapidly, with a pulse width that averages to the limit. Capacitors and inductors will smooth that, although close enough to the driver that isn't very important. Pretty much the same as a Class D amplifier.

[Gregory East]

How does it switch off and on electronically?

[Rune Bivrin]

Same way a Class-D amp or a switch-mode power supply does: using fast MOSFET devices. The switching per se isn't rocket surgery these days, but linking it to a DSP and powering it from the speaker signal is clever and requires some attention to detail. The switching devices had better be closed when the power is to low to be able to power the electronics.

[SirNickity]

I was wondering if the "switched resistance" was actually a bank of resistors or a transistor being driven by the microprocessor. In the linear range, a transistor will look like a variable impedance. I assume it would work like this anyway. Only concern is that it can never truly by 0 ohms, even wide open. (But then, what can?)

Controlling the transistor via a micro doesn't seem that tough. I know how I would approach it (in theory at least.) The trick is having response times fast enough to stop peaks. With a fast enough ADC and a low-speed ARM, it should be pretty simple.

I've only ever used ATmega micros, which are good to about 20MHz, so maybe not quite fast enough for this kind of thing. I'm looking at building a clip detector, so a similar detection circuit, but I have the luxury of being able to slow down the sub-ms transients with a cap. Some of the audio guys in the micro / embedded community have actually used them for processing audio, but 44kHz or above is pushing it in terms of the chip's ability to get the samples in, do anything more than simple arithmetic, and stream it back out in a timely fashion. I digress..

Anyway, clever idea. Not rocket science, just an innovative application.

[Rune Bivrin]

No switched resistors. That would incur actual power loss, with corresponding heat generation.

These days the process is well within the reach of cheap circuitry. Heck, you can buy a simple digital recorder for something like $20, and that includes 16 bit CD quality (almost) D/A conversion and MP3 encoding.

[LelandCrooks]

It's most likely a switching circuit, operating at 100+ kHz. If the voltage produced by the amp is too high - as determined by the in-circuit DSP - it will simply start switching on and off rapidly, with a pulse width that averages to the limit. Capacitors and inductors will smooth that, although close enough to the driver that isn't very important. Pretty much the same as a Class D amplifier.

+1