Ive repaired class D amplifiers for a few years now, but I decided to attempt at building one. I eventually want to build one thats 1000w+ but for now, I have to start somewhere.
I found a PDF on the web from a university/student that used delta sigma modulation.
I tried building his modulator, and I couldnt get it to work at all. Ill post a gif I ripped out of the PDF. I used LF356N and LM311s instead of what he has listed, Also using SN74AS74 as thats what I have on hand.
any ideas as of why? the 100nf caps on the + pins of the op-amp I know arnt right, I tried biasing these to 2.5V. and it sorta works, but it really doesnt.
any ideas? or does someone have a better schematic for a crystal based modulator.
As far as what I am building, I am building a single supply rail bridge amplifier. using a pair of IRS21844 driver ICs, driven from the sigma-delta modulator. But the modulator isnt working. lol.
I dont want a self-oscillating design, I want an external crystal/oscillator based design so I can sync it up with the power supply, and MCU I will be using to control the protection, etc... But ill worry about that later. Thats why I chose his method at first as it uses a crystal oscillator.
Thanks.
I found a PDF on the web from a university/student that used delta sigma modulation.
I tried building his modulator, and I couldnt get it to work at all. Ill post a gif I ripped out of the PDF. I used LF356N and LM311s instead of what he has listed, Also using SN74AS74 as thats what I have on hand.
any ideas as of why? the 100nf caps on the + pins of the op-amp I know arnt right, I tried biasing these to 2.5V. and it sorta works, but it really doesnt.
any ideas? or does someone have a better schematic for a crystal based modulator.
As far as what I am building, I am building a single supply rail bridge amplifier. using a pair of IRS21844 driver ICs, driven from the sigma-delta modulator. But the modulator isnt working. lol.
I dont want a self-oscillating design, I want an external crystal/oscillator based design so I can sync it up with the power supply, and MCU I will be using to control the protection, etc... But ill worry about that later. Thats why I chose his method at first as it uses a crystal oscillator.
Thanks.
Attachments
Last edited:
I didn't look at the circuit design but I noticed that there are no decoupling capacitors shown. If you want it to have a chance of working, you would need to add decoupling caps from each chip power supply pin to ground (preferably directly to the chip's load's ground). if you don't want to calculate the capacitances needed, you should be able to try a 10 uF electrolytic in parallel with a 0.1 uF X7R ceramic, from each power pin to ground.
Briefly looked at the 100nF caps from opamp +input pins to ground, per your comment. It is "POSSIBLE" that they are meant to be that way. I haven't analyzed the circuit, even in my mind, really, but, those two (U1A and U1B) opamps remind me of differential integrator opamp circuits, except for the unequal capacitances.
Briefly looked at the 100nF caps from opamp +input pins to ground, per your comment. It is "POSSIBLE" that they are meant to be that way. I haven't analyzed the circuit, even in my mind, really, but, those two (U1A and U1B) opamps remind me of differential integrator opamp circuits, except for the unequal capacitances.
I built that uni delta sigma amp once, it was the noisiest pos ever.
Then i discovered UcD and i was hooked, no other class d can compare to UcD in sound quality.
Pre filter feedback class d is designed to run into one load impedance only, ie you have to change thew output filter capacitor to get a flat frequency response for a given load impedance.
Post filter feedback self oscillating class d(UcD) does not have this problem.
Then i discovered UcD and i was hooked, no other class d can compare to UcD in sound quality.
Pre filter feedback class d is designed to run into one load impedance only, ie you have to change thew output filter capacitor to get a flat frequency response for a given load impedance.
Post filter feedback self oscillating class d(UcD) does not have this problem.
Not being a class d expert, the questions arise:
What would it buy you, to be able to synchronize the oscillation with the power supply and mcu? And if there would be some significant benefits from doing that, then why couldn't you still synchronize them, without a crystal-driven oscillation? Or, assuming that you could, what is nature of the trade-off space?
What would it buy you, to be able to synchronize the oscillation with the power supply and mcu? And if there would be some significant benefits from doing that, then why couldn't you still synchronize them, without a crystal-driven oscillation? Or, assuming that you could, what is nature of the trade-off space?
I recomment starting simple, take a single channel gate driver like the TC4422, add a LM311 comparator and the LC output filter, then add the feedback loop around that, either pre or post filter feedback. This will make a simple low power class d amplifier but demonstrate the basics of a self oscillating class d amplifier.
Here is a circuit i am working on, it is based on the Philips UcD concept along with snippets of technology by Anaview in Helsingborg Sweden.
Note that this circuit is not for beginners.
Here is a circuit i am working on, it is based on the Philips UcD concept along with snippets of technology by Anaview in Helsingborg Sweden.
Note that this circuit is not for beginners.
Attachments
See thats just it, I dont want to do a self oscillating one. Maybe later, but I dont want to for now. I come from a repair background, not an engineering background and its possible that you cant cross breed those, but from my experience the self oscillating ones are the WORST ones to repair when things go awry.
the fixed frequency ones tend to work much better and dont give issues when repaired.
But I assume the self-oscillating one is required under multiple load conditions? and a fixed-frequency amplifier only works under a single load?
Lets say I did do self-oscillating, How can I make this work in a full bridge setup? I built my output board already setup for full bridge using the IRS21844, driving a pair IRFB4127s on each side, and its setup for single rail. So i would need a blocking cap for a single channel?
the fixed frequency ones tend to work much better and dont give issues when repaired.
But I assume the self-oscillating one is required under multiple load conditions? and a fixed-frequency amplifier only works under a single load?
Lets say I did do self-oscillating, How can I make this work in a full bridge setup? I built my output board already setup for full bridge using the IRS21844, driving a pair IRFB4127s on each side, and its setup for single rail. So i would need a blocking cap for a single channel?
Last edited:
For fixed switching freq (clocked) and for Pre filter feedback self oscillating, you need to change the output filter capacitor if you change load from 4 ohms to 8 ohms to keep the frq response flat.
Post filter feedback self oscillating remedies this, you can apply 2-16 ohms and have the frq response remain as flat as on a class AB analog amp, if anything the switching frequency might change alittle.
I've built a clocked class d, ran it at about 250kHz and it sounded ok, clocked class d has its drawbacks that self oscillating doesent have, something with distortions and stuff.
A video: Improved feedbackless Class D amplifier - YouTube
Post filter feedback self oscillating remedies this, you can apply 2-16 ohms and have the frq response remain as flat as on a class AB analog amp, if anything the switching frequency might change alittle.
I've built a clocked class d, ran it at about 250kHz and it sounded ok, clocked class d has its drawbacks that self oscillating doesent have, something with distortions and stuff.
A video: Improved feedbackless Class D amplifier - YouTube
Attachments
Ok, I built a quicky circuit just goofing off.
I built a R/C triangle oscillator, and used another op-amp to feed that into along with the audio signal. I then take this output into a buffer, then into the IRS21844, using 15V as the voltage for this IC.
I take the low-side output streight into a filter and into my 32 ohm headphones, both headphones are in parallel-mono, so thats 16 ohms.
It is working, sounds ok, but I am getting a crapload of carrier in my filter.
So what is the filter calculation for Fc of 150khz, load of 16 ohms? I can actually feel the headphones heating up on my head because of the carrier that I obviously cant hear. hehe.
Thanks.
I built a R/C triangle oscillator, and used another op-amp to feed that into along with the audio signal. I then take this output into a buffer, then into the IRS21844, using 15V as the voltage for this IC.
I take the low-side output streight into a filter and into my 32 ohm headphones, both headphones are in parallel-mono, so thats 16 ohms.
It is working, sounds ok, but I am getting a crapload of carrier in my filter.
So what is the filter calculation for Fc of 150khz, load of 16 ohms? I can actually feel the headphones heating up on my head because of the carrier that I obviously cant hear. hehe.
Thanks.
At that low switching i'd say 50-100µH inductor and a around 1µF cap.
However since you are also using a single 15V psu, you need a dc blocking capacitor in series with the headphones, otherwise the 50% duty cycle makes it so that your 16 ohm load always see 7.5V DC, hence the heating you experience.
However since you are also using a single 15V psu, you need a dc blocking capacitor in series with the headphones, otherwise the 50% duty cycle makes it so that your 16 ohm load always see 7.5V DC, hence the heating you experience.
Yea, I am using a blocking capacitor, I have them on the return side of the load, 1 between the load and +15v, and the other one between load and ground, basically to bias the headphones at 7.5V. the other end of the load is going to my inductor.
I am using +/-15v for the op-amps, +5V for the buffer in between the op-amps and the IRS21844. Then I am running the IRS21844 at 15V. Using a 1K ohm resistor between the PWM output op-amp, and the buffer so i dont overdrive the buffer and blow the logic IC. I am using the buffer inverter so I can chain into another driver at inverted signal for full bridge setup. Otherwise i would just route the PWM direct to the driver IC.
I am not targetting a full range setup, just a subwoofer amplifier so full range isnt a necessity, hence why I am using a lower switching frequency.
I am using +/-15v for the op-amps, +5V for the buffer in between the op-amps and the IRS21844. Then I am running the IRS21844 at 15V. Using a 1K ohm resistor between the PWM output op-amp, and the buffer so i dont overdrive the buffer and blow the logic IC. I am using the buffer inverter so I can chain into another driver at inverted signal for full bridge setup. Otherwise i would just route the PWM direct to the driver IC.
I am not targetting a full range setup, just a subwoofer amplifier so full range isnt a necessity, hence why I am using a lower switching frequency.
I guess the inductor is not making an LC low-pass filter with the cap then. Is it possible to put a cap to ground downstream from the L? What impedance is in series just upstream from the 0.1uF?
A series-shunt passive filter, like a series R followed by a C to ground, is just a frequency-dependent voltage divider. So it's easy to derive the equation for one, and the cutoff frequency. For the RC one, the cutoff (-3dB) frequency is
f = 1 / (2 • 3.14 • R • C )
The usual resistive voltage divider, a series R1 then R2 to ground gives
Vout / Vin = R2 / ( R1 + R2 )
When a capacitor or inductor is used in place of R1 or R2, or both, then if phase angle is ignored you can just use the frequency-dependent "resistance" of the C and L, which are 1 / ( 2•3.14•f•C ) and 2•3.14•f•L .
So if you want to know when (at what frequency) Vout has dropped to 0.5 as much, for the same Vin, set Vout/Vin to 0.5 and solve for f. Then plugging in L and C values will give the frequency. Or, if you know what cutoff frequency you want, solve for L or C. If you still have C or L on the right side (if you have both L and C in the divider), you get to pick a value for it, and for f, and the other L or C value that is needed is the result.
REMEMBER to express f in Hertz, C in Farads, and L in Henries.
A series-shunt passive filter, like a series R followed by a C to ground, is just a frequency-dependent voltage divider. So it's easy to derive the equation for one, and the cutoff frequency. For the RC one, the cutoff (-3dB) frequency is
f = 1 / (2 • 3.14 • R • C )
The usual resistive voltage divider, a series R1 then R2 to ground gives
Vout / Vin = R2 / ( R1 + R2 )
When a capacitor or inductor is used in place of R1 or R2, or both, then if phase angle is ignored you can just use the frequency-dependent "resistance" of the C and L, which are 1 / ( 2•3.14•f•C ) and 2•3.14•f•L .
So if you want to know when (at what frequency) Vout has dropped to 0.5 as much, for the same Vin, set Vout/Vin to 0.5 and solve for f. Then plugging in L and C values will give the frequency. Or, if you know what cutoff frequency you want, solve for L or C. If you still have C or L on the right side (if you have both L and C in the divider), you get to pick a value for it, and for f, and the other L or C value that is needed is the result.
REMEMBER to express f in Hertz, C in Farads, and L in Henries.
Last edited:
smallest inductor I have on hand is a 1mh, they are round blue ones that came in a mixed bag of inductors at the local electronics shop. Most are 10mh, 22mh, 33mh, all the way up to 100mh.
So i been tossing those in parallel to get it divided down. But with a 1UF cap at the end near my load it just isnt making a whole lot of difference, sounded crackly and distored with quite a bit of carrier getting through.
But placing a 0.1uf cap in series with a 100ohm resistor on the drive side makes all the difference in the world.
So i been tossing those in parallel to get it divided down. But with a 1UF cap at the end near my load it just isnt making a whole lot of difference, sounded crackly and distored with quite a bit of carrier getting through.
But placing a 0.1uf cap in series with a 100ohm resistor on the drive side makes all the difference in the world.
was looking at the micrometals utility, trying to figure out what inductor i am going to need for 2500W into 1 ohm max.
Not entirely sure, but I keep getting T225-2B as my core in the program, with 6AWG wire. thats huge. ive seen manufactured 2500w amps with way smaller cores than these, unless its not really 2500w. just says it is.
Not entirely sure, but I keep getting T225-2B as my core in the program, with 6AWG wire. thats huge. ive seen manufactured 2500w amps with way smaller cores than these, unless its not really 2500w. just says it is.
How is the power rated? You should use "RMS Watts, continuous" ratings, not peak ratings.
The RMS Watts Continuous rating implies using the output power for a single sine tone at the maximum attainable output amplitude without clipping.
The maximum possible output power ratings can easily be determined from just the maximum possible peak output voltage and the load resistance:
If the output sine's peak voltage is able to be Vpk volts, continuously, then the sine's RMS voltage is Vrms = Vpk/sqrt(2) and the RMS power is
Prms = (Vrms)²/Rload, which is
Prms = Vpk²/(2 x Rload)
which also gives
Vpk = sqrt(2 x Prms x Rload) for a given Prms and Rload.
So, to get a maximum of 2500 Watts RMS continuously into 1 Ohm, you would need to be able to drive a sine wave output with a maximum peak output voltage of
Vpk = sqrt(2 x 2500 x 1)
Vpk = 70.71 Volts
-----
The momentary peak output power would occur (ignoring phase) when the output voltage was Vpk, and would be
Ppk = Vpk²/Rload
and solving for Vpk gives
Vpk = sqrt (Ppk x Rload) for a given peak power and Rload.
In order to get 2500 Watts Maximum Peak Output Power, you would need a maximum peak output voltage of:
Vpk = sqrt (2500 x 1)
Vpk = 50 Volts
The RMS Watts Continuous rating implies using the output power for a single sine tone at the maximum attainable output amplitude without clipping.
The maximum possible output power ratings can easily be determined from just the maximum possible peak output voltage and the load resistance:
If the output sine's peak voltage is able to be Vpk volts, continuously, then the sine's RMS voltage is Vrms = Vpk/sqrt(2) and the RMS power is
Prms = (Vrms)²/Rload, which is
Prms = Vpk²/(2 x Rload)
which also gives
Vpk = sqrt(2 x Prms x Rload) for a given Prms and Rload.
So, to get a maximum of 2500 Watts RMS continuously into 1 Ohm, you would need to be able to drive a sine wave output with a maximum peak output voltage of
Vpk = sqrt(2 x 2500 x 1)
Vpk = 70.71 Volts
-----
The momentary peak output power would occur (ignoring phase) when the output voltage was Vpk, and would be
Ppk = Vpk²/Rload
and solving for Vpk gives
Vpk = sqrt (Ppk x Rload) for a given peak power and Rload.
In order to get 2500 Watts Maximum Peak Output Power, you would need a maximum peak output voltage of:
Vpk = sqrt (2500 x 1)
Vpk = 50 Volts
Last edited:
Just a question, Why the level shifting? the IR2110 has an isolated logic ground/supply so you can run them off of a true 0v ground reference and 5V on the input side. But the output side uses COM and VCC for its own output stages, which can be at +/-RAIL
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.