Been doing some rudimentary sims of current feeedback + speaker.

[Kaz Kylheku]

Hi guys!

Having found this electronic model of speaker impedance, I set about doing some simulation.

Here is a screenshot comparing two current feedback situations. Both achieve almost exactly the same voltage transfer curve against identical speaker models. But one is some 7 decibels louder, due to using a smaller current sensing resistor. Using a smaller resistor in the CF line and larger DC blocking cap makes the curves nearly the same.

Indisputably, less is not more: more is more! Except maybe in current sensing resistors.

LTSpice .asc file is incluced; strip the .txt suffix. (Maybe the BBS should allow .asc files).

The speaker impedance page I linked to is guitar oriented: the first set of parameters given are for a "typical guitar speaker", so I used exactly those values.

Keeping things bone simple, I didn't model amp details like the output inductor, and Zobel network.

The CF topology and choices of feedback resistors, as well as the way input is is patterned after a power amp I use; adjust to taste.

[Kaz Kylheku]

BTW looks like I registered here on April 20, 2011. Just passed my 15th birSSday.　🎂🎉🥂

When I joined, Posh Spice and Sporty Spice were still 30 something, right?

The one for me that never grows old: LTSpice!

[Kaz Kylheku]

So I actually tried this exact thing illustrated in my above example: reduce the current sensing resistor to 0.1 Ω and make the other compensating adjustments. I was not looking to explore this exact aspect, but simulation has a way of taking you places.

I have to say, wow! While it has about the same amount of "scoop" as predicted by the bode plot, the sound quality is noticeably better. The top end is smoother and can take more presence out of the preamp. But not in a way that presence is lacking, such that you have to add it. If you know what I mean.
Probably the excessive negative feedback from the larger resistor was causing a problem for the amp, leading to distortion that trashed the high end? Wild-assed guess. Plus there is the fact that the input had to be 7 dB louder to achieve the same level as now.

There is a lesson here: for current feedback, use a small value current sensing resistor; you don't need more than around 0.1 Ω.

That also keeps the wattage down. For current sensing resistors, the dissipated heat is proportional to R. If you cut R by four, you cut the wattage by four.

[Kaz Kylheku]

Case in point: that Crate GX-20M amp, in a recent thread:

https://www.ssguitar.com/index.php?topic=5743.0

0.1 current sensing resistor, only 3W rating. That would still be enough even with a 4 ohm speaker and more powerful amp. You need 5.5A through 0.1Ω to get 3W. That current is well over 100W through 4Ω.

Anyhoo, that Crate schematic has some interesting features (in this regard, but also others).

The voltage feedback has fairly hefty treble bypass resistor of over 500 pF.  So that by itself would give the amp a rounded high end. Then in the current feedback though, there is a low pass filter. That will reduce the high frequency feedback, boosting highs.

I'm going to simulate that exact network to see what that thing is doing in terms of frequency response against the typical speaker model.

Almost certainly, they tuned the parameters to that thing's 10" speaker and cabinet.

Prompted by the thread, I went to Youtube to hear this thing in action and it sounds good!

In the top of the 1980s, Crate got a boost from an Yngwie Malmsteen endorsement, haha. He wouldn't have done that for something that is complete *s!!t*; it would have to be at most slightly *s!!t*. Around that time, they had lots of them in guitar stores all over so I remember playing around with them.

The schematic is impressively clean and professional. Everything is crisply readable, and there are useful notes to remove guesswork: like that unless otherwise noted, all resistors are 5% and 1/4W. They give a table of all test points, showing the expected approximate waveform shape, and voltage range. Amazing! Pretty much all you need to troubleshoot that thing and return to factory spec. And though the draft job is from 1992, it's exported to crisply scalable vector graphics shoved into a PDF. Kudos!