LTspice Ncore simulation

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello,

I haven't posted here in a while, but I do a lot of simulation with LTspice and have written a number of pages on the LTwiki including most of the Undocumented LTspice page.

Perhaps some may remember the threads on my class d Leap-Frog feedback design methods (including full LTspice models, of course). I have modeled the original UCD amplifier and run some interesting analyses of the distortion (coloration) it produces as it approaches clipping. I can't remember if I posted those models here or on the LTspice Yahoo group.

I thought it would be a fun challenge to try to model the Hypex Ncore design. My resources are the latest white paper and the USA patent. Can anyone point me to other potentially useful material? Is an Ncore schematic available? Do any simulation files already exist?

I have created a 5-pole, 4-zero (I think - haven't derived the transfer function yet) network that reproduces the magnitude loop-gain curve in the white paper. The phase response curve is not published (at least I haven't seen it), but from the simulations (both ac and transient) it is critical to always keep the zeros from drifting into the right-half plane. This is not easy to do without extreme care - kudos to Bruno.

The Ncore design is conditionally (un)stable* and can oscillate both at the intended frequency of several hundred kilohertz or at about ten times lower frequency. By shorting out the second pole pair in the feedback network one may unconditionally kick it back up to its intended operating frequency. This works well in simulation (not surprising as it is key to the operation of the real life Ncore).

The point of going all this trouble is to achieve a significant increase in (flat) loopgain across the entire audio band. When this is combined with an open-loop transfer that has been optimally linearized, the result is extremely low audio distortion without TIM or other coloration (clipping artifacts are still a weakness, at least in the simulation).

Before I upload a simulation file that is off the mark, I was hoping to collect any further detailed information available.

By the way, I hadn't really looked into the Ncore before now - Bruno is a genius.
___________________

* Not sure of the correct terminology for the fascinating feedback topology of the Ncore.
 
Welcome back Analogspiceman, :)

I do remember your leapfrog method and its inherent overcurrent protection feature. Your contribution here has helped in learning so much about the different set of approaches possible in a switching amplifier and ofcourse extracting the different set of waveforms in Ltspice.

I reckon you still main your hefty library of ltspice models, mind sharing it with us.

Here is your typical
NCORE

There is one more exact reference but I can't post it here.

Thanks for this informative thread.


Cheers
KAS
 
Last edited:
Here is a related thread I just found on the EEVblog forum:

An experimental 4-th order linear audio power amplifier - Page 1

The subject is not a switching amplifier, but a linear amp that uses a conditionally stable fourth order feedback network with an oscillation detection circuit to quench instabilities by reducing the order to one.

The last post was in 2016. Simulation schematics are presented at the beginning of the thread.
 
Just for your reference ;)
 

Attachments

  • ncore.png
    ncore.png
    54.7 KB · Views: 1,283
I looked at Ncore section in the NAD M22 service manual. It seems to use a twin tee active filter for the low frequency pole pair and the output L-C for the 3x higher frequency pole pair. Everything is fully differential so it is more difficult to follow. Also much of the circuitry is just a discrete fast comparator that drives the output MOSFETs.

For the purpose of analysis, understanding and system simulation the output can be replaced with a simple behavioral power comparator with settable delay and the feedback and analog front end may converted to a single ended equivalent. The saturation protection can be behaviorally simplified as well. If correctly done, the result should simulate correctly both in the frequency and time domains.
 
Okay, I have simplified and reduced the NAD M22 Ncore fully differential design to its single-ended equivalent. The output stage (which is just a fast power comparator) has been replaced with a behavioral subcircuit. The front-end opamp (also a behavioral model) is driven single ended but it has complimentary outputs in order to achieve proper phasing in the feedback loop.

The behavioral model of the output includes a stimulus source to enable measurement of loop-gain (which then is measured at the comparator input fb node).

Self-oscillation at the undesired lower frequency is suppressed with a bidirectional behavioral diode (its clamping parameters are adjustable).

Because it is cut down to the bare essentials, the simulation runs very fast. It probably could run even a little faster by removing the linear time delay element required for the ac analysis (it would be replaced by a simple a-device delay).

The feedback network around the opamp is not really a twin-tee. In the original circuit it is a dual differential balanced low pass tee (two tees, but not twins in the usual sense).
 

Attachments

  • Ncore.gif
    Ncore.gif
    57 KB · Views: 825
  • UcD Ncore.asc
    5.5 KB · Views: 204
I haven't analyzed the influence of R1-R3, C1. It must tweak the closed loop transfer function a bit. I wonder why they are there. Maybe they help to linearize the open loop transfer function.

I made a small mistake in the conversion to a single ended topology. The value for C1 should be 4n7 rather than 1n0. Please correct this in your simulation schematic.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.