The Composite / Hybrid- Linear + Switching Power Amplifier

[Craig405]

Hi all,

This is a subject that has essentially never been discussed at DIYaudio.com, maybe for good reason, maybe just because it’s not widely known..

Here is a previous thread I have found:
SMALA Amplifier concept
There was one other I think but I can’t find it.

I have recently done some work on these amplifiers, spent about 10 weeks researching and simulating. No building though. I had some reasonable results with a self oscillating parallel composite amplifier, here are some rough figures:

-100W @ 89% efficiency up to around 10KHZ <0.1% THD
-100W @ 82% efficiency at 20KHZ <0.02% THD

Load = 8R//100nF

If there is enough interest with DIYers I will post my results and write up, it will require me finding a decent way to upload it first (3MB PDF).. i’ve got a crud load of Christmas shopping to do last minute beforehand!

Cheers
Craig

[sakis]

ok ... i am interested i dlike to see more
thanks

[Bigun]

Read this: http://quad405.com/jaes.pdf

and then note this: Devialet High End Audio integrated amplifier

This thread has my interest too.

[destroyer X]

YouTube - Csöves single ended VS Quad405

regards,

Carlos

[ontoaba]

Your simulation show the digital self oscillating with 60 times higher frequency than linear (sine) signal. I am using less than 10 times higher frequency (says 100Khz) and still wishes lower than 100KHz to be used, because of sensing and controller limitation. Please built it and we may sharing in problems. You will need very fast equipment for your project, and it is hard job.
Similar but used for supply:
http://www.diyaudio.com/forums/solid...ml#post2223627

[Craig405]

Bigun, yep exactly, the idea is very similar to the Quad 405 current dumping / error correcting principal, a few additional considerations are necessary for dealing with a switching dumper amplifier. Thanks for the interest.

Ontoaba, I think I allowed frequencies up to 500KHz without seeing masses of loss in SPICE.
Really the switching frequency is best kept lower, I agree. This is easily tweakable in my implementation.
I used IRFB4020 N-type FETS, so my results do depend on how good the model is. Driver was IXDD404, this was not a great choice but I was having issues getting driver SPICE models to work.

Cheers
Craig

[Bigun]

The issue highlighted by vanderkooy was the time delay at the output of the ClassD power stage - resulting in a higher current draw from the ClassA stage. Do you see anything of concern in the simulations ? - or is your Class A stage simply beefy enough not to worry about it (I think this is what Vialet did, they used a beefy Class A stage).

[Craig405]

Vanderkooy does mention this, he is correct, the delay in the control loop of the SMDS (switched mode dumping stage) is directly linked to the rise in correction current in the LCA (linear correction amplifier).
This is the phase lag he talks about and is a function of the hystersis thresholds, type of SMDS, bridging network and the active control loop component delay.

This can be controlled, the simplest method is by adjusting (lowering and shrinking) the hysteresis bands in the self oscillating controller (in my implementation anyway).

But in my sims the LCA was certainly beefy enough to cope, I didnt have time to work on a custom LCA so I used two good fast fairly powerful designs from the late (sadly) Professor Marshall Leach (Low TIM) and Giovani Stochino (Ultra fast power amp, thanks to Alex M's kindness).

Cheers
Craig

[Bigun]

well it looks interesting - I've been reading the vanderkooy paper after seeing a review of the vialet amp and thinking about this approach for the past month - with no intention of ever doing anything about it but maybe I can follow your work...

[Bigun]

Originally Posted by destroyer X

YouTube - Csöves single ended VS Quad405

regards,

Carlos

I didn't seriously expect to be able to hear anything properly via my computer speakers. But I did hear a difference, just don't know if it was real or my brain foolin' with me. I focussed on the first part i.e. Johnny Cash. To my ears, via an iMac, the tube sounds different from the Quad. I preferred the tube for Johnny Cash. When it switches to the Quad the high's don't sound right and the mids have lost something. The next track, Queen, sounds to me as if the bass is tighter on the Quad but I didn't like the sound as much. Weird, it can't be possible to really hear these differences via a microphone, youtube and an iMac, so I put it down to biassed expectations.