The man's just sharing some sensible information for someone to test his amp, which he's perfectly willing to have done. I think we could give him a break.
Except in the datasheets for the ncore you can clearly see the sharp rise in distortion vs frequency that is curtailed rapidly beyond 6.7k due to the measurement bandwidth limitation impacting on its ability to resolve the third harmonic.
Some amplifiers are obviously affected by this far more than others. The ncore is specifically because its midband distortion is so amazingly low. If it wasn't then the rise would be pretty much invisible as it would be swamped out by other forms of non linearity elsewhere within the amplifier.
This is Distortion at 100W/8 at 5KHz = .0024% THD of my 200W CFA mosfet amp. The GNFB is about 80 dB from 0 to 20 kHz and it sounds nice according to Richard Marsh http://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-80.html#post4355552
Damir
Damir
Attachments
I fully agree.. It depends on the application... The result with class D can be good but it takes careful considerations
Go Yell. I do to when reading this thread....
Well .. it is because noise starts to dominate when using 80KHz Bandwidth.
But class D is working Hell.
- Inductors in the output stage produces a lot EMI, also the inductors are coupling induviduelly
- Free running class D designs will produce EMI (This is also the case for unshielded SMPS) that will be mixed between the channels and will be seen in the audible range as moving artifacts
- To solve the problem with mixing the switch frequency they offset the switch frequency. For fixed frequency types it is okay. For Free running it does not as the switch frequency change with signal level and load.
- If they are not shielded they will inter-fer with other circuits producing artifacts that we can hear in the speakers.
- Many class D design go for high power... But most likely the S/N ratio is worse than a good designed class A/AB
Filters for measuring class D is to prevent the switch frequency beating the ADC in the measurement equipment. 300KHz on a 182KSamples DAC is not good for the nyquest theory....
It is not just something i have read about....
Last edited:
What does sound best to my ears?
Single gain stage, no multi feedback gain stages....
If we talk voltage feedback it is the inverted configuration with virtuel ground on the noninverted input. For many designs the Jfet input stage is an keeper because of the very low leakage and non existing current noise making it a good choice in high impedance circuits... >5K source impedance.
But the JFET has a highly Vgd depended Crss (Miller capacity). By using it in inverting configuration the voltage across Vgd becomes more or less static.. And you can hear it, regardless of feedback value.
If it is non inverting i will bet on Current feedback as long as we do not talk filters.
The design phase is 1 - 5% simulation, the rest is PCB design, prototyping and testing.
Single gain stage, no multi feedback gain stages....
If we talk voltage feedback it is the inverted configuration with virtuel ground on the noninverted input. For many designs the Jfet input stage is an keeper because of the very low leakage and non existing current noise making it a good choice in high impedance circuits... >5K source impedance.
But the JFET has a highly Vgd depended Crss (Miller capacity). By using it in inverting configuration the voltage across Vgd becomes more or less static.. And you can hear it, regardless of feedback value.
If it is non inverting i will bet on Current feedback as long as we do not talk filters.
The design phase is 1 - 5% simulation, the rest is PCB design, prototyping and testing.
I have some uCD180 modules, they are not shielded, and there is NO audible noise. Not even with horn tweeters, not even with my ear right at the horn's mouth.
How do inductors in the output produce EMI? They are filters, not generators.
high OS ratio 64-128x is standard today in delta-sigma audio ADC - do knock it down with a low pass but it doesn't need to be sub lsb
but if you really want to measure class D you should be using MHz sample rate ADC
Digilent Analog Discovery student demo/lab board has a 14 bit 100 Ms/s ADC
Inductors produce magnetic field lines around them.. Depending on the inductor construction it can be less but it is there even with "shielded" types.
RFID technology uses inductors for coupling the LF and HF region.
If you place an sensitive (high impedance of more than 10KOhm) circuit nearby they will pick up noise from the inductors. Try make an 1 turn loop antenna with scope probe and hold it near the inductors, you will be able to see the switch frequence...
RFID technology uses inductors for coupling the LF and HF region.
If you place an sensitive (high impedance of more than 10KOhm) circuit nearby they will pick up noise from the inductors. Try make an 1 turn loop antenna with scope probe and hold it near the inductors, you will be able to see the switch frequence...
Agree. I think there's some questionable ideas there, but the circuit is interesting and the data will speak for themselves.
One remark, don't ever put the inductors in parallel!! In this position the coupling between the inductors are at its best.
I love the certainty here. Just in case any of us unwashed unbelievers are not sure of the ears of gold at play. YOU CAN HEAR IT.
I am clearly old and deaf. 🙂
Take the inductors out and then look at emi. Gets much higher at distances beyond inches. The inductor body is too small to be an effective antenna at the relevant frequencies, the close coupling effect is all near field, not radiating.
ha ha ha ha ha....
It is not like cables. Try compare with source impedances of ~100k and make an AB comparison of an inverted design with 2x gain and an noninverted with 2x gain as well.
It should noted that:-
1. The ncore distortion plots have been done with module powered up from a bench power supply. I am willing to wager that it was with a regulated PSU (linear perhaps?)
2. All of Doug Self's plots in his book ditto - module powered off separate DC supply regulated
It is extraordinarily difficult to get below -110 dBV on mains noise where all the electronics is housed in the same chassis including the PSU.
1. The ncore distortion plots have been done with module powered up from a bench power supply. I am willing to wager that it was with a regulated PSU (linear perhaps?)
2. All of Doug Self's plots in his book ditto - module powered off separate DC supply regulated
It is extraordinarily difficult to get below -110 dBV on mains noise where all the electronics is housed in the same chassis including the PSU.
Last edited: