NMOS for final stage UcD

UcD?

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation

Ah, UcD = Universal Class D, not University College Dublin or University of California Davis etc..

--sp

I thought of adding "universal class D" between parenthesis, I don't know w hy I didn't. Some explanations can be found here --

/ucd.html -- and the basic schematic I'm currently testing in LTspice is he re -- -- but, in fact, the fi nal stage is common in all class D amplifiers.

why I didn't. Some explanations can be found here --

gy/ucd.html -- and the basic schematic I'm currently testing in LTspice is here -- -- but, in fact, the final stage is common in all class D amplifiers.

Something wrong with the FETs used there? PHP18NQ11T ~$1 at DK.

Hey can I ask a stupid question?

Where is the oscillator (sets the PWM frequency) in that circuit?

George H.

No, they seem right, but what about their performance? That's why I asked f or suggestions, hoping that people with more practical experience would giv e better(?) alternatives, maybe some lower gate charge/Rds ones with a bett er price, maybe something more commonly found that are similar, etc. I mean , usually, test schematics like these are made with a rather more general p urpose in mind. As far as I see, they may as well advertise dusty transisto rs forgotten on their shelves.

You're looking at it :-) The explanation is in the first link I gave and he re is a more palpable result for LTspice (I used non-inverting input, it ca n be changed):

*** start Version 4 SHEET 1 1220 680 WIRE 368 16 288 16 WIRE 512 16 464 16 WIRE 560 16 512 16 WIRE 672 16 640 16 WIRE 688 16 672 16 WIRE 720 16 688 16 WIRE 784 16 720 16 WIRE 288 32 288 16 WIRE 512 32 512 16 WIRE 688 32 688 16 WIRE 784 32 784 16 WIRE 240 48 16 48 WIRE 176 96 160 96 WIRE 240 96 176 96 WIRE 16 112 16 48 WIRE 176 128 176 96 WIRE 16 224 16 192 WIRE 176 224 176 192 WIRE 672 224 672 16 WIRE 672 224 176 224 FLAG 784 112 0 FLAG 688 96 0 FLAG 512 112 0 FLAG 464 48 0 FLAG 368 48 0 FLAG 288 112 0 FLAG 720 16 out FLAG 16 224 0 FLAG 80 96 0 SYMBOL voltage 16 96 R0 WINDOW 3 9 109 Left 2 SYMATTR Value sin 0 1 20k rser=100 ac 1 SYMATTR InstName V1 SYMBOL cap 160 128 R0 WINDOW 3 28 49 Left 2 SYMATTR Value 470p Rser=1k Rpar=10k SYMATTR InstName C1 SYMBOL g 288 16 R0 WINDOW 3 11 122 Left 2 SYMATTR Value table(-1.1m,-0.2k,-1m,0,1m,0,1.1m,0.2k) SYMATTR InstName G1 SYMBOL res 176 80 R90 WINDOW 0 5 56 VBottom 2 WINDOW 3 27 56 VTop 2 SYMATTR InstName R1 SYMATTR Value 1k SYMBOL tline 416 32 R0 WINDOW 3 0 28 Top 2 SYMATTR Value Td=200n Z0=50m SYMATTR InstName T1 SYMBOL res 496 16 R0 SYMATTR InstName R2 SYMATTR Value 50m SYMBOL ind 544 32 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L1

SYMBOL cap 672 32 R0 SYMATTR InstName C2

SYMBOL res 768 16 R0 SYMATTR InstName R3 SYMATTR Value 8 TEXT 120 -56 Left 2 !.tran 1m TEXT 120 -32 Left 2 ;.ac dec 1k 2 2000k

*** end

It's a self-oscillating (variable frequency) thingie..

Den tirsdag den 6. maj 2014 18.56.41 UTC+2 skrev Spehro Pefhany:

now why I didn't. Some explanations can be found here --

ology/ucd.html -- and the basic schematic I'm currently testing in LTspice is here -- -- but, in fact, t he final stage is common in all class D amplifiers.

it's it really just a variation of a hysteretic buck converter?

-Lasse

As far as I can see in their description (and several other pdfs found on their website), it's no hysteresis and the feedback is after the filter, not before.

I could be wrong but I think L1 in the output is what causes the oscillation. The MOSFETS switch fully on and L1 delays it long enough to allow a full on and the feed back gets a phase error which causes a nice oscillation.

Just my guess of course..

Jamie

Den tirsdag den 6. maj 2014 21.37.38 UTC+2 skrev Vlad:

still this:

looks pretty much exactly like a hyst-buck with RL/CL as a bit of feed forward on the feedback

-Lasse

Or Up Counter Down (that's infix notation).

On a sunny day (Tue, 6 May 2014 12:24:28 -0700 (PDT)) it happened Lasse Langwadt Christensen wrote in :

Right, the patent should not have been allowed,

People have used variable power supllies for audio before.... Even Elector.

On a sunny day (Tue, 6 May 2014 12:37:38 -0700 (PDT)) it happened Vlad wrote in :

No hysteresis is normal in such a supply (see my lab supply design), and the voltage feedback is always from the OUTPUT.

I know it comes from the output, but that is conventional fixed frequency d esign (like your lab supply), no self-oscillating configuration, no? Doesn' t the HBC rely on the ESR of the output capacitor? The schematic I posted d oesn't have hysteresis (I'm not talking about physical devices that may hav e some due to imperfections). According to their explanations, the UcD seem s to rely on the phase shift the LC filter induces, plus the feedback loop. The patent for the HBC also seems to be dated 2011 (if I got the right one ), which is years after the UcD. Don't get me wrong, I'm not making a case for the UcD, even if I like it's simplicity.

On a sunny day (Tue, 6 May 2014 23:33:37 -0700 (PDT)) it happened Vlad wrote in :

Yes I have read that site, and the pdf. It oscillates at the frequency where the phase shift is 180 degrees, amplifier delay plus the 22 uH output coil.

The only thing a bit different would be the lead RC filter. Its a strange patent world, the specs look good, but if you have to pay Philips a lot of money to use it, what good is it?

IIRC (could be wrong) they (Philips) made a chip for this? I tried looking for it but could not find it, They do make a lot of other stereo class D in chip form, like the TDA8920B, but its app note shows a different design, with a normal PWM generator.

If I come across a class D chip from them with that system I will let you know.

PS last night hen thinking about this I concluded that I would not want an other 400 kHz oscillator in the room. Too much RF interference already messing up short and medium wave reception, not to mention long wave..

I scrapped a switching power supply from ebay for just that reason a while ago, and now use the case for my SDcard logging GPS based radiation counter:

So this project box was free... cheaper than a new one anyways..

This is what is inside:

Maybe that class D (Philips patent) could be useful in a closed bass speaker box, shielded as well.

I'm not paying one cent (except for devices, unless I can scavenge), the di yaudio forum has quite a few posts with people actually cuilding these thin gs, not to mention the inventor, Bruno Putzeys, actually responds to some a nd gives feedback/advices. So, no, no money from me in that regard. Of cour se, if I should somehow be a millionaire because of this, I'm willing to ch ange my mind...

ther > 400 kHz oscillator in the room.

Sadly, you make a good point, I'll have to consider some serious shielding/ pcb ways.

other > 400 kHz oscillator in the room.

No matter what I do I can't get the frequency above 300kHz (in LTspice). At full modulation it drops to as low as 86kHz, zero crossings are around 280 kHz. Even so the THD with plotwinsize=0 and numdgt=16 says 0.0924%@21W/

4R (efficiency is 93.37%). It's also true the sources are modelled with Rse r=0.1 Cpar=1m and parasitics near the end stage (on all pins, D, S, G) are 25nH Rser=25m. Output filter is a Chan core inductor and the capacito r has Lser=25nH and Rser=25m. Could these mean I can go for some cheape r (higher) Rds FETs? Last choice is a PSMN017-30BL and the dissipated power is ~250mW.

Changing the supply from 15V to 35V and the output FETs to some RJK0651DPB gives 92.96% eff. and ~0.13% TH @105W/4R. But these are harder to "glue". I was thinking of some DPAK ones, to keep it as little as possible, but will they be able to cope with that much dissipated power?

On a sunny day (Wed, 7 May 2014 06:04:09 -0700 (PDT)) it happened Vlad wrote in :

I do not know the answers to these questions, but what happens to teh frequency when you use a real loudspeaker spice model like from here?

I mean loudspeakers are not pure resistors in any way.

IMNSHO the FETs should not be critical, but the Cgs could change a few things. I use IRLZ34N for everything...logic level drive, but its only 60 V IIRC :-) Did not try use it for audio yet.