I have found the following design on the 'net:
http://xipix.de/elektro/gimos50.html
It looks interesting, although very unfamiliar to me: it's a soft clipping power amp with, by what I read, low global feedback, current feedback and (therefore) high output impedance. It looks very well suited to guitar duties, although I'm not an expert (I can find my way around preamps and I can also design them, but I'm not familiar with power amps).
I've tried deciphering the schematic, but as I said, a lot is unfamiliar to me: the input stage isn't the common differential pair and kinda looks like an output stage? Also, it employs MOSFETs while most of the things I've read up on employ BJT transistors (there shouldn't be much difference though, MOSFETs are biased similarly to bipolar transistors, with the difference that the gate is basically electrically isolated).
Could you please walk me through it and explain to me if and where it can be improved upon?
http://xipix.de/elektro/gimos50.html
It looks interesting, although very unfamiliar to me: it's a soft clipping power amp with, by what I read, low global feedback, current feedback and (therefore) high output impedance. It looks very well suited to guitar duties, although I'm not an expert (I can find my way around preamps and I can also design them, but I'm not familiar with power amps).
I've tried deciphering the schematic, but as I said, a lot is unfamiliar to me: the input stage isn't the common differential pair and kinda looks like an output stage? Also, it employs MOSFETs while most of the things I've read up on employ BJT transistors (there shouldn't be much difference though, MOSFETs are biased similarly to bipolar transistors, with the difference that the gate is basically electrically isolated).
Could you please walk me through it and explain to me if and where it can be improved upon?
The feedback appears to be the same as the Maplin MOSFET - 33k into 1k.
So this looks perfectly ordinary to me.
If you want to release the output devices from the chains of GNFB, to give IMO a nicer more open sound, I would suggest you make the VAS stage with FETs to get a nice tubey feel, with it's own (fast) NFB loop, and use darlington BJTs as the followers on the output device.
Then the local feedback of the darlingtons provides the gain for the low impedance output.
Then arrange the FETs to gently clip so the BJTs never do.
What about using mixed-mode feedback? It makes it possible to dial in exactly how large you want the output impedance to be. Then use some global feedback with diodes along the path to increase the amount of feedback over a certain point (and therefore introduce soft clipping). You would never achieve unity gain anyway, since you still need the input signal to go from, say, 4Vpp to 40Vpp (with some safety limiting kicking in over a certain point). As long as the power amp gain is kept over a certain amount stability shouldn't be a problem, right?
It's an interesting idea, but any global feedback, even at rare times, means you have to tie the speaker earth to the signal earth, and I'm not convinced that's a good idea, if avoidable.
Also diodes / zeners always leak a bit before the knee, as it were, so it may pollute the circuit.
Perhaps there's scope to have some zeners around the driver stage, which perhaps would softly limit the voltage excursion? Then they would interfere less? Just thoughts, I could be wrong
Bob Cordell's Power Amplifier book includes the "Klever Klipper" which introduces soft clipping upstream of the power amplifier itself; soft clipping is completely outside any feedback loops, global or local. Bob's KK has been discussed here on diyAudio, including its use in Bob's "Super Gainclone" chipamp PCB. On the schematic below, the Klever Klipper is dual opamp "U5" with its surrounding ancillary bits. U8 (a DC servo), U4B, and the burly chipamp U7 implement the downstream power amplifier.
_
_
That's my invention
I'm currently taking a different approach. The amp worked well, but had far too much power for a living room amp. Remember, a guitar speaker has around 95dB efficiency! I also changed the negative feedback slightly to make the speaker 'sound' like a tube amplifier. Finally, very important for the sound, DC blocking with a capacitor at the output. This softens the distortions in the low frequencies.
Can't mixed mode feedback be added to your previous design? The 50W into 8 ohm amp is actually enticing and might come useful for a higher-powered amp.
That said, DC blocking with a capacitor is only really feasible at relatively low power and a DC servo would probably be a better solution overall?
That said, DC blocking with a capacitor is only really feasible at relatively low power and a DC servo would probably be a better solution overall?
As I understand it, they are part of a current mirror, the current flowing through the positive input of each of the two LTP is mirrored into the VAS (Q9 and Q10). What I don't understand is where the very strong local feedback is. And the website states that there is very little global feedback, but I can see some feedback (lower than many designs, but not that low) flowing back into the negative inputs. The funny (but obvious) thing is that there technically is more feedback when it soft clips (feedback resistors go from 33k into 1k to 33k // 10k into 1k).
Correct,Q5,6, are part of the current mirror. R8,9,11,12 (current feedback) sets the local voltage gain of the differential stage. The higher these resistors, the lower the voltage gain of this stage. R16,19 sets the local voltage gain of the VAS.
Both together form the total voltage gain,Both together form the total voltage gain, and in this case it is small. That's why the negative feedback is weak, it has nothing to do with the ratio of R14,15.
Yes, it might by possible.
Why? Maybe you know more, but as far as I know, larger powers should also be transmitted via a capacitor. Two advantages: 1. If the power amplifier breaks, it doesn't take the speaker with it. 2. The signal edges of a low-frequency overload are smoothed in the direction of a triangle, with the correct capacity. That sounds less sharp.
Last edited:
Fine, but if the open loop gain is relatively low (if EdGr's calculations are correct, 46dB works out at a gain of 200, which is far too low for effective negative feedback) then the output impedance is already quite high (I'd guess around 8 ohms?), and the frequency response will be dictated by the speaker's impedance somewhat. Therefore I don't understand that comment referencing Rod Elliot.
This only affected the mixed feedback he used. This also gives the speaker more freedom (impedance).
As far as I can understand, mixed mode feedback is something you can add on to voltage drive amps (ie. high negative voltage feedback, very low output impedance). Your amplifier as is shouldn't have a very low output impedance and should already be, in a way, "mixed mode", or am I getting it wrong? If I am, could your design be modified in order to vary the amplifier's output impedance? On the usual voltage drive amplifiers, it is a matter of adding a 0R1 resistor to ground at the speaker's return, and then connect this resistor at its "high" node (connection to speaker return) to the negative input through a low value potentiometer or resistor.