Any Maplin MosFet Amp Guru's on here?

If you are taking the amp on the road or otherwise subjecting it to heavy vibration you might want to stick the polypropylene caps to the pcb using a small blob of silicone or glue gun.

Good Idea, although they won't be doing any traveling, I'm interested to see how they stack up against my Quad II Mono Blocks after I've built the Rod Elliot Pre Amp and Phono (RIAA) Pre Amp, I think the Quad 22 control unit lets things down, but that's a story for a different thread 🙂
 
C4 is a mystery...why the rather specific value of 6.8nF?
R12?

Another spin on it 🙂

Think of R12 as providing the base bias current to turn on the current source (TR5). Almost 10ma flows which is a bit excessive. Lets make R12 a 30k and ensure TR5 is a high gain device. That cuts the dissipation.

Do we need TR3 ? Lets remove it and link it out.

Is R10 helping or not ? I would say not. Its reducing the transconductance of that stage. Lets link it out.

Switch on. What happens ?

It works. Outwardly and not a lot changes. DC offset is a bit higher but that could be addressed. All is stable.
 

Attachments

  • Capture.JPG
    Capture.JPG
    239.7 KB · Views: 277
Is R12 still needed, in the 90's Maplin article they even mention VR1 being connected between the collectors of driver transistors as if R12 didn't exist, and with no further mentioning at all of its role.
Does R12 really minimize start-up thump as earlier mentioned in this thread?
 
There is a lot of talk about maximum power and SOAs.

In reality a big amplifier is not normally ever subjected to max power RMS.

Very loud can be as little as about 10W with plenty of headroom for brief crescendos.

I actually measured the power going into my speakers at very loud from a 100W amplifier and it was only 6W RMS.
The listening averages are indeed very low.
The peaks of the transients on the other hand can be very high.
That's why many of us allow for peaks that are 20dB above our average listening levels, in the hope that the transients don't get clipped. Clipping is quite common. That's where the amplifier has been asked to output MORE than it's rated maximum.

That 20dB ratio between average and peak only allows for a ten fold increase in signal voltage. I believe some music can go beyond 30times on voltage ratio (~+30dB).

I typically use <1W for adequate or normal listening averages. But I build ~100W amplifiers. There's my 20dB for the transients.
 
Has anyone commented on the symylarity between this "Maplin" and our Forum's Symasym and Roender amplifiers?

Note in both of these that R12 has been replaced by a grounded common base transistor (Q10 in the Roender).
 

Attachments

  • rmi-fc100v1.1.jpg
    rmi-fc100v1.1.jpg
    84.3 KB · Views: 281
  • Symasym5_2sch.GIF
    Symasym5_2sch.GIF
    16 KB · Views: 264
Has anyone commented on the symylarity between this "Maplin" and our Forum's Symasym and Roender amplifiers?

Note in both of these that R12 has been replaced by a grounded common base transistor (Q10 in the Roender).
I think you mean R9 = 12k. The Roender looks like something designed in a simulator! The double-differentiator, if you will, arrangement is common. But although it may seem a clever way to cancel distortion it actually brings with it extra problems that are rarely addressed. The Maplin circuit has some nice properties to it. I'm not sure this was by design or accident, tho.
 
Another spin on it 🙂
You've turned it in to the more common, "RCA"/Naim type architecture. Single differentiator and simple Miller VAS. That will work. You introduce some distortion because the dynamic current to the upper FET goes through the variable resistor (RV1) but the lower FET is fed directly. In the Maplin circuit the current is fed equally(ish) to both FETs and the current through the variable resistor is constant(ish). In your circuit you would need to bypass RV1.

R10 in the original also reduces the transconductance, as you say. This reduces the loop gain and should help stability. Raising the loop gain is a mixed blessing because you have to be a lot more careful about stability. I think the Maplin design deliberately sacrifices feedback for stability. It's a cautious design. The pcb layout doesn't help it, though.
 
Last edited:
Use LTSpice and slightly overdirve the amplifier at 1kHz. You will see steps in the voltage on TR3 as C4 charges. Make it too large and the design latches after clipping for a while.
R9 (12k) limits current during clipping and improves recovery.
R4 (47R) is harder to explain. It might have been to modify clipping symmetry. The amplifier seems to work with it shorted
 
Here is the PSSR for the original. I used 470uF reservoir caps to generate significant ripple. Not the greatest sim in the world but it gives an idea.
 

Attachments

  • R3.JPG
    R3.JPG
    210.7 KB · Views: 230
  • R2.JPG
    R2.JPG
    259 KB · Views: 227
  • R1.JPG
    R1.JPG
    255 KB · Views: 252
Realistic STA-2200 circuit similarities

I apologize it this doesn't fit here, but I'd like to show a very similar circuit. (Mooly, please delete this post if it is deemed as thread jacking).

The Realistic STA-2200 receiver is from ~1980 and is rated at 60 W/channel using one pair of the Hitachi 2SK134/2SJ49 (running at ~+/- 48V). Although not the most powerful Realistic receiver produced, several have noted that they think this was the best sounding one they made. Stereo Review found the 1K onset of clipping at 65W (16 ohms), 74W (8 ohms), 97W (4 ohms) and 80W (2 ohms). THD was measured at .003% at 1W and 1K Hz, rising to .008% at 60W.

The amplifier section (shown below) is very similar to the Maplin one being discussed here.

Steve.
 

Attachments