Class D amp using LM319 comparator

In Fig. B circuit I increased the resistance from 1k to 15k and it improves a lot giving small vertical voltage difference of 19mv compared to circuit A with 2.68 volts! The waveform is shown below.

Fig B circuit
Overlayed triangular signal Fig B

Reducing the charging current further thru the resistor by doubling the resistance from 15k to 30k make it very near linear (yeah almost linear, lol). As you can see from the sim in Fig. C,
Fig C circuitFig. C, Overlaid Capacitor voltage and linear ramp generator are look alike now.

Overlayed triangular signalFig C

Circuit sim showing reduced charging current using 30K resistor. Then overlaid with linear signal. Now you cannot see the bended one because they are almost identical in linearity. Zooming-in is the only possible way to close examine the vertical voltage difference of the two

There is small amount of voltage difference plotted along vertical axis as small as 4.2 mv which is too small and might be considered benign. See Fig. C.
Peak to peak charging current decreased to 200ua.

Triangular current fig C

3 thoughts on “Class D amp using LM319 comparator”

  1. Hey, just read your post on how to build a class D amp using lm319 comparator. Im willing to build a class d amplifier and i´m gonna generate the pwm in the same way as you did cause it seems to be pretty wise. I got a couple questions for you: 1)After simulating your first circuit on LTSpice I couldn´t help noticing that the negative input of the second comparator has a DC value different from the DC value of the triangular signal that is at the positive input. I read several times the third page of you post in which you explain wih a simulation how the negative input should follow the average value of the triangular wave form. Therefore I can´t understand why there is a little offset difference in the simulation (this affects the comparison between the audio signal and the triangular) . And I don´t understand what you intend to achive by adding the refference to the original circuit : “I will add tuning mechanism at the output of U2 inserted between R3 and R7 because this is the culprit in making the duty factor to ~51%. R7 effectively add resistance to R3 during charging of C2 but discharges quicker via output transistor without R7 ugh!” Guess the question here is: what am I missing?/ why the offset of the signals at the input of the comparator don´t equate on my simulation?. 2) I was wondering if by any chance did you measure the THD for the amp?. I know it sounded good, but just wondering. As it regards the class D I´m designing i want to make a 100watt (RMS) over an 8 ohm load. I´m gonna use IRF2110 and mosfets in a full bridge topology (probably taking feedback from the load). Any suggestions?

  2. Hello,
    Thanks for your interest regarding this article. I will answer your question soon after I fixed my laptop that refuses to boot after windows 10 update. As of now, I am answering you via my smartphone. There are at least 7 visitors asking the same way (privately) as you did but I want to publish your question with my answer via comment section so that it will serve as my answer to those dudes as well. Regards.

  3. Hey, just realize what the mistake was. The two RC networks after the first comparator act as you explained in your post. The thing is that the RC network that clamps the negative input of the second comparator to the offset value of the triangular waveform (cross over point ) takes over 500ms to reach the rigth offset value. I should have looked at the .op rather the transient simulation . I realized when I saw that in an image you posted a transient simulation for 1000ms.

    I understand what your tunning mechanism is now. You add the trim pot along with the diodes to be able to change the period of the triangular, guess I was way too tired yesterday when I first read it.

    Very good job dude, your circuit is both cheap a reliable with good outcomes.
    I´ve been strugling to find easy class D amp circuit implementations.

Leave a Reply

Your email address will not be published.