LTspice Ncore simulation

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Then turned into a giant advertisement for nc1200 OEM, this whole forum was in force with it.

It better have delivered what it advertised for what it cost, with the many years of UcD behind them all things considered. I think everyone can agree Soren originally oversold the dam1021, however I think the latest models have very much lived up to the expectations and he doesn't overcharge for OEM parts and sells to everyone even with the top end board. Nobody would have an issue if Hypex did the same.

Not sure I follow how they were using the nc400 on this DIY forum as an advertisement for the nc1200, which was not and is not available to the DIY crowd.

In any case, Hypex and the Soekris company aren't even close to being in the same league when it comes to professionalism- delivering tested, debugged, ready to use products and standing behind their advertised performance. The nc400 was sold as a DIY product. When a problem arose, they said "just send it back and we'll repair or replace". When issues surfaced with the Soekris dam, he said "fix it yourself, this is DIY". It's a cruel joke.

"I think everyone can agree Soren originally oversold the dam1021"-- you think? Little good it does people such as myself that bought 4 original boards based on his "oversold" advertising that the latest and greatest, rev. 5 or whatever he is up to now, finally deliver what was initially promised. The latest shoe to drop concerning potential synch issues is just another kick in the jewels.
 
"I think everyone can agree Soren originally oversold the dam1021"-- you think? Little good it does people such as myself that bought 4 original boards based on his "oversold" advertising that the latest and greatest, rev. 5 or whatever he is up to now, finally deliver what was initially promised. The latest shoe to drop concerning potential synch issues is just another kick in the jewels.
What now ? please, what's going on with the Soekris DAC ? I've been planing to use it in the future on my diy r2r dac. I'll admit, I've not dug deep into this product, just found it a few months ago and thought, well, that could be interesting as diy dac.
 
Yeah, its like the original nc400 was flawless or something.

So what issues are you aware of that Hypex has ignored or told you to fix yourself?

I'm not aware of any inherent flaws that compromised the module and prevented it from achieving it's advertised performance. What's more, knowing Hypex, they have always been quick to resolve any customer issues in a completely professional manner. In my own case, when I needed a new set of connecting cables, due to my own mistake, they sent me a completely new set, for free, not even charging me postage. They know how to put the customer first.
 
Have been sim'ing and one note on the Ncore design (sim presented here) is that the gain is only around 4 times, which means you need a pre gain stage, which is of course not som much of a problem, but worth to mention.
Having compared to a pre/post filter fb implementation, and the Ncore does for sure get better results, but at a cost of a bit higher complexity. One must also wonder how much extra gain is sinsible to put in the loop to gain the better HD figures, and also how low in HD do you need to go ... with speakers producing at least in the 1/10% HD, id it then possible to hear the difference of say 0.01 and 0.05%?
I for sure don't pretend to know the answer, and for sure the Ncore design raises the bar and makes otheres wonder if they can do better on their own designs ... which is a very healthy thing. Thumbs up for Ncore.

(It is especially towards higher freq Ncore really makes a bug diff ... having much more fb at say 10 kHz)
 
Last edited:
Hello everyone,

I am currently revisiting the UcD and Ncore designs as a vehicle to compare LTspice to PSIM (to which I have been given a six month evaluation license). PSIM (like SIMPLIS and some other expensive simulators) allows simulating loop-gain measurements in the transient domain using DFT techniques like a loop-gain analyzer in the lab.

LTspice can do this measurement too, but it is much slower than PSIM (which uses piecewise linear models for the switches and other nonlinear elements). I am still learning the abilities and limitations of PSIM so I can't promise a meaningful outcome, but I will report back if I get something interesting (might take awhile).

On another note, I would like to touch upon the subject of Bruno's new Ncore style amplifier with the digital front end interface (CDs and other digitally encoded sources are not converted to analog, rather the digital music stream is fed directly into the amplifier). If Bruno is smart (and he is), I am guessing his amplifier reads the digital stream then delays it very slightly before feeding to the analog class-d amplifier.

A small delay would give the design a "look ahead" in time so that Bruno's amp could monitor voltage supply sag (and perhaps output current) so that the amp could intelligently pre-clip the input music stream just enough such that the amplifier is never driven into saturation at the rails. That would eliminate a lot of distortion from railing the amplifier. Since it's digital, the pre-clipping could track the possibly sagging wave shape of the supplies or be strictly flat topped (at a slightly lower value) or be rounded like a tube amp. Do any amplifiers already do this sort of thing?
 
Hello everyone,

I am currently revisiting the UcD and Ncore designs as a vehicle to compare LTspice to PSIM (to which I have been given a six month evaluation license). PSIM (like SIMPLIS and some other expensive simulators) allows simulating loop-gain measurements in the transient domain using DFT techniques like a loop-gain analyzer in the lab.

LTspice can do this measurement too, but it is much slower than PSIM (which uses piecewise linear models for the switches and other nonlinear elements). I am still learning the abilities and limitations of PSIM so I can't promise a meaningful outcome, but I will report back if I get something interesting (might take awhile).

On another note, I would like to touch upon the subject of Bruno's new Ncore style amplifier with the digital front end interface (CDs and other digitally encoded sources are not converted to analog, rather the digital music stream is fed directly into the amplifier). If Bruno is smart (and he is), I am guessing his amplifier reads the digital stream then delays it very slightly before feeding to the analog class-d amplifier.

A small delay would give the design a "look ahead" in time so that Bruno's amp could monitor voltage supply sag (and perhaps output current) so that the amp could intelligently pre-clip the input music stream just enough such that the amplifier is never driven into saturation at the rails. That would eliminate a lot of distortion from railing the amplifier. Since it's digital, the pre-clipping could track the possibly sagging wave shape of the supplies or be strictly flat topped (at a slightly lower value) or be rounded like a tube amp. Do any amplifiers already do this sort of thing?

I don't know of look-ahead limiting circuitry on board amplifier modulator(s), but some vendors are doing it using a SHARC DSP and probably today using an m4 ARM microcontroller or better.

LD Systems DDQ PA sub-woofers does look-head limiting.
 
Attached is the LTspice simulation schematic and loop-gain plot for the original UcD amplifier. This is taken from a transient simulation that is post processed with measurement statements to provide loop-gain in the time domain like a laboratory frequency response analyzer.

On my computer the simulation takes twenty minutes to process four curves of 101 points each (404 points total at about three seconds per point). Without the stepped dc bias a single run would take about five minutes.

What this shows is that audio band loop-gain decreases as the output approaches the rails. It also shows that the audio input should be rolled off at 30kHz to 40kHz to avoid artifact.
 

Attachments

  • UcD Loop Gain.jpg
    UcD Loop Gain.jpg
    301.1 KB · Views: 324
  • UcD xfer.asc
    3 KB · Views: 111
Last edited:
Albert Dunford of PSIM has made me a beta tester for their simulator. Betas often have bugs. In the version I have, the stepped parameters function is broken. I may be able to do stepping via scripts, but I haven't succeeded yet. The attached plots were taken with zero dc bias to the amplifier input.

The UcD amplifier is a self-oscillating design so the switching frequency is not clocked and varies a bit with operating point. This seems to be giving PSIM problems in determining the exact switching time so the loopgain is off compared to LTspice. Maybe this is a PSIM beta bug, but the difference is disturbing.

LTspice took 270 seconds (4 minutes 30 seconds) to run and post process the measurement commands. PSIM took 9 to 10 seconds. Scaling the output plots to be the same and creating an overlaid JPG took a half hour or more (much longer than run time).

I am using a beta release of PSIM so some bugs are expected.
 

Attachments

  • UcD Loop LTspice vs PSIM.jpg
    UcD Loop LTspice vs PSIM.jpg
    179.4 KB · Views: 248
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.