MF500 | My 500W/4Ω Hi-Fi All-FET Amp Design
Time has come that I've completed my amplifier design into a schematic that can reliably be built! All of the concepts turn out really well in the final design. The design went through multiple phases of which you can read all about in the thread "State-Of-MOS": 200W Ultra Low Distortion Pure FET Amplifier, along with various posts describing its workings.
This amp uses a radically new "VAS". The reason I'm quoting that word VAS is because of the fact that the stage is now actually two separate stages that are highly tied together. However, this yields enormous benefits in so many areas that this new "VAS" stage I developed, significantly improves performance over both a conventional single ended VAS and symmetrical VAS. I'm rather excited about it and to put it to a real test. But that will be one of the following steps to take; to design the PCB, build it and then measure and listen it.
This thread will also become the build log as this is the final design as I'm going to build it. Just shoot if you have any questions or have anything positive to add :)
But for now, I will introduce to you MF500, an amp with (for now in the simulator) exceptional distortion and TIM figures; thank you for your time..
Well, I have some problems viewing the picture. Can't you make it higher resolution?
This is really good news Sir MagicBox, I will be waiting on the PCB design of this project so in the future I can also built one of this totally work of art amp design.
I'm hoping it won't take too long to get started on PCBs. Right now I'm still doing some last minute changes, most notably having added the required opamp input stage due to the rather low input impedance of the amp itself. In the simulation I picked an NE5543AP which has offset inputs that can be used to trim the amp output. Overall distortion numbers don't change as this opamp has a huge bandwith still at unity gain.
The other change is to put an image together of a magnified view of the schematic but I'll have to make several screens and glue them together for that to obtain a hi-res image. It'll be up in the main post soon :)
It seems that Mr. MagicBox is not only in the Need for Speed, but the Need for Mr. MagicPCB: Alex!!:wiz:
What big mirrors you have grandma! (from Little Red Riding Hood story)
You going right for the big power version or going to build a proof of performance at lower power?? :D
No! They are deflectors! Photon deflectors! To protect the Warp Core in the back don't you see? Oh wait...
Yes, I'm going for this big power version right away; the sim already took me to the proof of concept comparing its relative performance to all kinds of (conventional) topologies so I am very confident about the outcome, this amp certainly won't blow up in your face and trample your music like it is a ripe grape :)
The mirrors are prominent to the design yes, it's what makes the concept work. The cascoded mirrors perform a number of tasks:
-Voltage domain conversion (fixed 30/30 at the IPS, 100/100 VCS for this design)
-Decoupling control voltage from the output voltage; output voltage is the signal swing, the input voltages can be anywhere between 30/30 theoretically.
-Sets a low gain input/output current ratio of 7 which results in:
-Strong outputvoltage rail current of 45mA
-Proportional input currents decoupled from any absolute voltage
-Power balancing between input/output current, most of the power goes into driving the output rather than wasting half of the power in the input rail with a 1:1 mirror otherwise
-Input currents drop in favourable range for the 4 gain devices that modulate the input currents.
-The 4-transistor gain unit can now be referenced against any voltage, in the design it is voltage centered in the middle of the IPS output voltage range.
There are more details, but these are the prime functions that the mirros perform on both sides. The whole stage operates in a 'current mode' of thinking :)
merely curious, I didn't follow the development thread (I think that is what it is) and am not an expert on my next question. But, with the current mirrors providing "gain" are you depending on their linearity, or feedback to correct their linearity?
Hi Bear, thanks for asking. No extreme linearity is required, the gain by parallelling the output devices in the mirror is merely to control the relation between the input/output current. This relation by nature is more lineair than a voltage controlled Vgs/Id slope. The task of the mirrors are not to provide the actual gain, that's left to the 4 devices in each mirror input rail. Hence the mirrors are easy to drive with a low current and still have a powerful voltage output rail, with a rather lineair relation between them two.
P.S. I looked at your website, you are a good man :)
Well, I managed to put a good quality (I think) schematic image together now, you may have to scroll some. I also took some time to log some of the measurements from this amp. They are:
-THD1K Full Power
-THD20K Full Power
-TIM spectrum analysis 19K/20K with 50Hz resolution
-TIM spectrum analysis 4K/5K with 20Hz resolution
These numbers oughtta back up the schematic a bit ;) When I find some time I think I will write a detailed description to the workings of the whole circuit, there's quite some thought behind it along with the compromizes (such as why 7 mirror transistors) to make things work at this power scale.
For now I'm qurious whether folks realize only half of what what this topology accomplishes other than being a trunk of trannies together :p
Edit: I did not connect the opamp null offset; this particular model did not match the datasheet on pin out so I just left it. You can basically pick any decent opamp frontend, but this NE5534 works well, amongst the best I've tested in the sim. It can cope with the required 10mA full signal drive and you could even apply Douglas Self's opamp parallelling here because they are configured as unity gain buffers whose outputs are then simply summed into the J-FET virtual ground node. All you need to do is to divide the input impedance resistor by the number of opamps you'll use. In the schematic, if you'd parallel two 5532's you'd use an output resistor of 200 Ohms per opamp. It makes the life of the opamps twice as easy. In this case this would be applied to share the relatively high drive current of the amp input.
|All times are GMT. The time now is 12:41 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
Copyright ©1999-2016 diyAudio