Any real advantage to mixed mode feedback?

[Mick Bailey]

Has anyone done an A-B on implementing mixed-mode feedback in an SS amp? I'm wondering if it makes much of a difference or whether it's just subtle but useful. The low component count and ease of implementation makes me wonder why so many amps rely just on voltage feedback if mixed-mode is all it's (sometimes) claimed to be in giving warmer, valve-like response.

[Steve Conner]

I found that raising the output impedance of a solid-state amp (irrespective of how it's done) gives better results with guitar speakers.

The bass is boosted due to less damping of the speaker's mechanical resonance, and the presence band is boosted because more current gets driven through the voice coil inductance. The subjective result is a warmer, brighter, more lively sound.

[loudthud]

I agree with what Steve said and would add that the higher the output impedance, the better. This means crank up the voltage gain to 100 or so. Checkout the Randall RG100. No Voltage GNF except at DC.

[Mick Bailey]

Just taken a look at that schematic. Interesting, though it looks like it could be pretty unstable without a speaker load. I was thinking something more moderate - say 60% voltage and 40% current mix, maybe adjustable to keep within the limits of what the power amp can deliver. The Randall looks pretty extreme, but maybe the overall design keeps a lid on things. Modifying an existing amp could give problems with almost 100% current feedback.

Looks like a good reason to do some practical experimentation.

[J M Fahey]

FWIW I've used mixed mode feedback since forever (think 1972) when a tube shortage in Argentina made me stop making Fender clones and go SS.

And I found the latter dull and flat (in a bad way) with exactly the same speakers, until I noticed that the output voltage in my Tube power amps went rom, say, 3 V RMS to 6V RMS if I lifted the load, so internal impadance was 4 ohms so damping was = 1 .
Then I started reading for ways to raise voltage by 100% when load is pulled I would accomplish the same.

As in: "the speaker does not know what is driving it, just how it behaves" .

So as NFB signal I added an 1/40 output voltage sample (10K/220r) in series with another 1/40 output voltage sample derived from output current (4 ohms speaker/ 0r1 4W resistor) and I ended with an SS amp which had roughly 20X gain with a 4r resistive load , gain did rise at resonance and higher frequencies and would become 40X no load ... a Black Box that copied what my Tube amps did.

Been using that since, even in Bass amps.

Try it, you can use an spdt switch to either short the 0r1 resistor and open an extra 220r in series with the first one or do the opposite.

Basic gain does not change, timbre will.

[teemuk]

Also, the CFB path doesn't just have to be plain network of resistive dividers, you can also use it for overall shaping of response: Assuming you remain within limits of stability you can add different kinds of RC filters to the current feedback path to enhance or attenuate frequencies of interest. I've seen many amp designs that tame down the excessive treble response caused by rising (speaker) load inductance. Attenuating excessive peaks in response at "speaker's fr" seems to be equally common technique.

Also, a switch between VFB / VFB+CFB, or various levels of CFb, is a nice feature. IMO, rather use a switch because the effect is so subtle that continuous control over it is not required.

[nevetslab]

That's very intriguing, mixing the both voltage and current feedback to raise the output impedance.

I never did get to playing with current feedback on SS Power Amps, but some years ago, having bought a Bruel & Kjaer 2712 180VA Power Amp (normally used to drive shaker tables), I found I really liked the sound of this (as a bass amp) in High Output Impedance mode, driving a pair of 12" Gauss speakers in my small bass guitar rig. Max output voltage is only like 12V RMS, but good for 15A RMS. Output Impedance (Hi Z mode) is >50 ohms below 40Hz, >80 ohms 40-100Hz, back to >50 ohms 100Hz-300Hz, & >20 ohms 300Hz-1kHz. Never gets below 10 ohms. I have the two 12" 4 ohm Gauss drivers in parallel, for 2 ohm load. Real efficient speakers, so the relatively low output voltage isn't an issue.....at least in small practice rooms. I never did gig with it to find I could drive it into protection, where it gates the input. Handy feature when using it for its' intended purpose. Big shaker tables are like huge loudspeaker motors, intended to put mass into motion, test specimen mounted/strapped down to the driven platform. Made a very cool sounding bass amp with a good preamp front end.

Bruel-Kjaer-2712.pdf

I'll have to go back and look at this all over again, as well as put some bench time in on adding CFB on some of the power amps here collecting dust. Interesting Power Amp circuit in that Randal RG100

[Mick Bailey]

I've just had a good session experimenting with mixed-mode feedback and initial results are promising, though eye strain has caused me to wrap up the project for the evening. The main impression is the increase in bass, and an extended treble. With the ratio more in favour of current feedback the test amp (Sessionette 75 with a G12H-100 speaker) was clipping on the bottom string with dropped D tuning and had a speaker-flapping quality. The speaker reproduces the low-end well, but I never found the bass to be particularly strong with the Sessionette. Current feedback can increase the bass to excessive levels with this amp. Easing off the current feedback made the change less noticeable with mid-position playing, but gave a subtle extension to the bottom-end and slight increase in brightness. It wouldn't case me to write home to Mother, but definitely an improvement.

The other aspect I noticed is that with the current feedback set to a pleasing level at home-playing volume caused premature clipping on the clean channel at gig volume.

So, some more experimentation needed but I'm encouraged by the results so far.

[J M Fahey]

Well, you are applying low and high EQ to your power amp.

And unfortunately SS amps do not clip gracefully on their own.

But partial Bass cutting before it can improve things.

Download and read Teemuk's book, there's a chapter dedicated to enhancing SS distortion by careful EQ, both pre and post distortion.

In a power amp you have not control of post EQ,because it's basically the response curve of the speaker which is fixed, but pre EQ still helps.

In SS Guitar, user phatt has experimented a lot on it, and among other things found that cutting certain high frequencies before distortion helps tame harshness .... as counterintuitive as it may seem.

[Mick Bailey]

I revisited this today and got it pretty much sorted for my particular amp. Wonderful low end without booming or thumping and much more treble detail. Like taking a blanket off the amp and plugging it into a larger cab. Only better. To A-B it I made the current feedback switchable and there's no going back, so it will get hard-wired tomorrow. I did need to add a series cap to AC couple the feedback signal - without it the DC offset increases.

The only thing I want to work on is a background low-frequency hum that's only present with current feedback. Grounding the power amp input is totally silent with voltage-only feedback. Maybe due to the extended bottom-end?

My voltage-feedback setup is now 47k/560 Ohm divider with 47uf cap. Current sense resistor is 2x0.33 ohm resistors in series coupled to the feedback node via a 0.22uf cap. I tried various series resistors but it works better just like it is.

[pdf64]

Regarding the hum, at the cab's bass resonance there will be a high Q gain boost, so if that happens to around 100 Hz then the power supply ripple will get caught up in it.
Perhaps increase the degree of voltage feedback, as that may act to set a limit to the boost?

Has it passed the 'gig volume' test?

[J M Fahey]

My voltage-feedback setup is now 47k/560 Ohm divider with 47uf cap. Current sense resistor is 2x0.33 ohm resistors in series coupled to the feedback node via a 0.22uf cap.

That .22uF is way too small and you don't need it anyway.
Too lazy to draw something but your NFB network, with your values must be:

1) current sensing voltage divider: amp speaker out driving > speaker (8 ohms?) in series with > a 1 tenth to 1 20th its value so for 8 ohms it should be from 0.8 ohms to 0.4 ohms ... and in fact I'd use just a single 0.33 ohms resistor > ground.

2) voltage sensing NFB: amp speaker out > 47k ohms , the free end goes to the - input so it provides both DC and AC feedback to have low offset > 560 ohms > 47uF cap > joint/union/net where the speaker and the 0.33 ohms resistor meet.

You don't need the .22 uF because the 47uF already blocks DC .

And you don't need to send the "current voltage" straight to the - input, the way I suggested both voltages get added because they are in series.

[Mick Bailey]

Oops - cap value is 2.2uf. But I get your point that it shouldn't be needed, though for some reason my DC offset goes from 14mv to 180mv without the cap, so I need to take a detailed look at what's happening elsewhere. I'll reconfigure the amp as series-connected feedback and make some comparisons to how it is right now. The sound is how I want it, but the implementation I've got isn't right.

pdf64's reasoning for the hum looks like the answer - it's 100hz and there's a large peak in output across the speaker terminals at this frequency if I feed a signal gen right into the power amp. It passes the gig volume test - though it's a lot louder than before.