Current-mode Feedback - the Right Way?

[elwood]

Hey everyone, I've been working on an 1x12" amp design and learning a LOT in the process. My "reference" amp is a Fender Hotrod Deluxe which I do like. I'm not trying to emulate it with solid state circuits, but it has been a good reference point so far.

One thing that's baffling me is the "presence" control. On the Hotrod I always have it cranked... I find things sound terrible otherwise. I've plotted the curve of the Fender power amp with the presence control cranked and have been trying to get something similar in my power amp. I know that the presence is part of the current feedback circuit, reducing the feedback at the higher frequencies. I'm having trouble though when it comes to knowing how much feedback I should be using in my own circuit.

With respect to current-mode feedback, I have implemented it around my amp using a 0.25R resistance in series with the 8ohm speaker. I have buffers and such set up to amplify the recovered signal from the shunt and mix it back into the signal feeding my amp. I use a dummy load that is switchable to 8, 12 or 24 ohms and I've been using this to see how the load affects the output signal from the amp. When I change from 8 to 24 ohms the amp output changes by about 3-4dB based on my configuration. But I have no idea what is the "normal" amount of feedback to apply, and how much my load should affect the output voltage of the amplifier.

Can anyone explain what is the normal amount of feedback and how much it should affect the output of the circuit? So far I haven't come across any hard and fast numbers based on practical implementations. Thanks!

[teemuk]

I think that it will vary from design to design and the best you can do is to tweak the circuit to your own liking. You can find plenty of practical examples by studying schematics but I'm pretty convinced there is no "golden rule" to this stuff.

The way I see it: If the change in response is only few decibels it is almost inaudible and therefore pretty pointless. On the other hand, if the change is several decibels it severely reduces the amount of output power the amplifier can feed to its nominal load impedance – at least in practical applications. This is because such a set-up will also require that the amplifier's output can swing to a moderately high voltage, otherwise the increased gain causes the amp to clip as soon as the loading impedance rises sufficiently high. Most current feedback amps are prone to this sort of clipping. So, too much current feedback becomes detrimental in many respects. I'm pretty sure that most designers want a compromise and make the response have about 6 – 10 dB difference between the impedance "peaks" and the "notch" that represents the nominal load. This way the change in response can be perceived rather easily while the output power difference between the peak and notch levels is still somewhat "manageable". The response will of course vary with different speakers.

Just for kicks I simulated the basic feedback circuit of two Fender power amplifiers that use current feedback and stepped the load from 8 to 24 ohms. Both showed a gain difference of about 7 dB. When I plotted the response with a model of an 8-ohm speaker the gain difference between the resonant peak and the nominal load "notch" was about 10 dB. Due to those practical reasons the difference is not usually any higher than this but I do remember it being somewhat less (say, 4 – 6 decibels) in some other amplifiers. Some amplifiers also reduce the amount of current feedback at lower frequencies.

I think you should experiment with few different configurations and just let your ears be the judge.

The classic feedback controls, presence and resonance, work better when you place them to the conventional feedback loop.

[elwood]

Hey Teemuk,

Thanks for the detailed response. I certainly realized the "not enough is not useful" side of things. Since many of configurations didn't yield much difference in sound at all. But I wasn't sure how far most amps go with it. Most of my design projects for work and play are industrial/utilitarian stuff and a lot of things with micro-controllers. I'm really finding the need to listen and decide what I prefer to be a gratifying and also frustrating part of the design process. 

I think I might have (from some trial and error) come up with values that I like. I just have to validate that they're reasonable.

Just out of curiosity, what software do you use to simulate circuits? How about for schematic and PCB layout? I've been learning Proteus at work, but it's pretty expensive for hobby use.

[teemuk]

For simulation I use SwitcherCAD III/LTspice by Linear Technology. I tried several freeware softwares few years ago and that one was definitely the best one of them due to many reasons. The most notable ones being unlimited amount of imported models and circuit components (many freeware softwares limit these) and a feature to assign ".wav" files as voltage/current source signals and extract ".wav" files from voltage/current plot. Schematic capture worked greatly too and The GUI was intuitive and easy to learn. That software is absolutely amazing considering that it is freeware.

For PCB design I use a combination of PCBExpress and... MS Paint.  :lmao:

I make the layouts and trace route design with PCBExpress, as it is quite fluent for that. However, that software doesn't allow printing or extracting the layouts to some other software so I take a screen capture of the board to MS Paint, which I use for the final editing and printing. It's not an ideal technique but works well since I still make most of the boards the "old school" -way by drawing the traces to the copper planes with a permanent marker and using the computer-made layouts only as drilling templates. I can't recommend that technique for any excessively complex circuit though. I would rather use a CAD but my Windows version is so old that the typical CAD softwares won't work in it. I have heard good things about "Eagle" so perhaps I should try that some day.