Figure 1

You might notice that there are numerous possible circuit configurations that may be used in producing good quality class d audio amp using this chip. Like for example to use normal opamp instead of 2N3904 as the pre amp. Figure 2 shows how it is implemented.

Figure 2

This circuit basically functions the same way as in figure 1 but use opamp as the pre amp for driving pin 5. Direct coupling is necessary to fulfill the required bias voltage of ½ VCC on pin 5. Fine adjustment to 50% duty cycle is done via R4 and R9.

Some folks may want to try without master sync oscillator, so another possible approach is to configure all channels wired in “astable mode”. The frequency determining networks are pot R1, R2 and capacitors connected to their pin 6 and 2 respectively. The capacitor is being charged from VCC via R4 thru D1 and the pot then discharged via D2 from pot by pin 7 to obtain near square wave output as required.

Figure 3

The circuit shown above is a “preliminary audio sound test” circuit that will give some folks the taste of 555 timer as class d audio amp that can drive your stock 32 ohm MP3 headphones. If the sound excites your enthusiasm, then you may exploit it by using different approach that will fit your taste. Of course final circuit will require pre amp to match the required input voltage of the 555 timer and to facilitate the inclusion of negative FB. This is the simplest approach and is good in giving a sound quality “different” to figure 1.

This circuit utilized two pots for more precise offset null adjust at the output. Pin 5 have internal bias of 2/3 VCC equal to 8V, giving near 50% duty cycle when the pot is at the center. The thing is when pin 5 is terminated with an audio source having normal ½ VCC output, the duty cycle drops to 40%. These can be easily corrected by tweaking pot R1, R2 to get back to 50%. Please take note that direct capacitive coupling should not be use here because the output audio source itself will be the one to give pin 5 the correct bias of ½ VCC to pin 5. Of course if you insist to use coupling capacitor for audio testing purposes, then you need voltage dividing network like the one already shown in figure 3. The only caveat here is that the input sensitivity is +/- 5 volts to get full dynamic swing at the output.

Although no negative FB was implemented here, the sound is still surprisingly good. I noticed that even the audio source I used is my Acer aspire one netbook at mid volume setting, the audio sound output is still loud and good enough for my audio testing needs. The voltage dividing network R7, R8, R9, R10 along with C3 can be omitted if direct coupled audio source is implemented and that the source is ½ VCC. In any case you can adjust to 50% duty via pot R1 and R2.

The purpose of assigning more than 200 KHz channel spacing is that even their RC timing network are ideally matched, they still tend to drift in oscillation giving audible beat frequency to occur if they are oscillating at same clock frequency. Not only that, this suggested 200 KHz 400KHz channel spacing is essential for the reason I will discuss below.

Some folks may ask why 200 KHz 400 KHz channel spacing? Here is the answer;

Figure 4. Waveform without modulation

Figure 5. Upper channel circuit being modulated by +/- 1volt at 1 KHZ.

Figure 6. Waveform with 5 volts modulation at 1 KHz.

It is quite noticeable that by modulating pin 5 by +/- 1 volt at 1KHz, its oscillating frequency modulates to +/-26 KHz and when I increased the input voltage to +/- 5 volts, the frequency modulation becomes staggering. The frequency deviation stretches up to +/- 150 KHz!  That’s darn huge! This is the reason why at least 200 KHz 400 KHz channel spacing is required to avoid cross-talk on their respective adjacent channel. If they do over-lap, then it would be audible.

My first thought regarding the behavior of the circuit made me to think that this very wide stretching of bandwidth might jeopardize the sound quality of the music being played, but the truth is—it’s benign. In fact even no matter how large the frequency modulation is, when the output is being filtered the result is just always stays like the original. This is depicted in figure 7.

Figure 7. Modulation at 5 KHz with 5 volts amplitude shows how severe stretching of frequency from V max to V min of audio.

When I listen to the headphone with this kind of circuit configuration, I noticed there is a little mixture of crisp treble being reproduced compared to figure 1. Maybe this is the secret spice inherent to all self oscillating class d audio amp? Mixing of FM with PWM might produce useful harmonics that tend to make it more crisp in audio reproduction. Well, I have to investigate this a little bit further or maybe I am on the placebo effect only, LOL!

Unfortunately, I tested this circuit using only one channel due to no available 2 channel audio jack in my assorted box, (I’m still here in Norway) so full testing implementation with 2 channel is still on hold until I got that stuff. I think 200 KHz channel spacing is not enough (400 KHz would be adequate, (edited Aug. 8)) therefore, negative FB must be strictly implemented to reduce the severity of the FM modulation.

Thanks for looking,

Cheers

IQspice-4

]]> http://geekcircuits.com/2010/08/active-crossover-network-using-555-timer-as-class-d-amp/feed/ 4 http://geekcircuits.com/2010/07/class-d-amp-using-tl494-dc-to-dc-converter-chip-2/ http://geekcircuits.com/2010/07/class-d-amp-using-tl494-dc-to-dc-converter-chip-2/#comments Sun, 04 Jul 2010 22:33:59 +0000 IQspice4 http://geekcircuits.com/?p=433 Class D amp using TL494 DC to DC converter chip

It’s almost 5 years now since I publish a topic regarding how to use TL494 as class d amp in a popular DIY audio forum. I think it’s quite good to include this article here also. By the way, before I publish that topic I thoroughly googled the said subject to see if there are some geek guys who have already done such a playful exploit of this chip but I simply got zero hit. This could mean that I am an early bird on this regard? Or maybe I am the only one who have a bizarre taste in choosing the right chip for such an application.

I tried to check that thread three years ago and noted that the counter say 15K views as I can remember, but when I checked it out recently, it increased to more than 24K views even the thread was buried and not being resurrected. This could mean that some are still peeping on that thread and it made me to think that I am not lone odd guy after all.

The circuit shown below is an improvised version from the original. The circuit use a NFB held at pin 3 of TL494 then feed back to pin 2 via resistor Rfb as required. The error amp output at pin 3 of the tl494 requires about 2 volts to bring the output pin 11 and 8 to 50% duty cycle. This will need a bias voltage of the same magnitude on its non-inverting i/p pin 1 as provided by voltage divider pot. The inverting input pin 2 having same level of 2 volts via Rfb completes the balance of those 3 nodes just like biasing a conventional opamp. Take note that these three pins are linear nodes and that connecting pin 3 to pin 2 in series with Rfb is a normal thing but what if we connect Rfb to the switching output of power FETs?

Figure a.

Schematic of a typical opamp biasing scheme. Pot R1 is used as DC offset null adjust and also 50% duty cycles adjust. You can connect a headphone in series with a coupling capacitor from the output to check if everything is fine.

The basic circuit above has a voltage gain of 10 based on conventional opamp gain equation of Av= R3/R4. This will require an input sensitivity of +/- 150mv to give an output of +/- 1.5 volt on pin 3 centered at 2 volts. The output at pin 11 and 8 will then modulate to approximately 5 to 95% centered at 50% on quiescent state (no audio).

The circuit showed below implements a FB also known as “pre filter negative feedback” (NFB) and obviously far better to reduce distortion and to improve frequency response. Since this output node is a switching mode with a voltage higher than 12 volts then there must be a constant amplitude regulator before we feed the NFB to pin 2. TL494 requires single-ended supply; this makes the audio input always referenced to ground. It’s a good option to use strings of diodes to act as “amplitude chopper” in such a way of accepting only “constant amplitude PWM” signal but not amplitude modulated signal. This is an attractive option because this technique will accept only the PWM signal from the power FETs and does not require filtering before feeding back the signal to the i/p of the opamp. Another unique feature of this FB schemes is that this will make the circuit inherently immune to power supply variations without affecting the NFB voltage. This means if a fellow is using either single ended or split bus supply of 30 volts, and he decides to use 50 volts or more, as long as the drivers and tl494 supply is regulated to 12 V, and the R11 is matched with a correct wattage, then it will not affect the NFB. But if the bus supply is increased or decreased, the NFB will aggravate the output of the amplifier if it were done without those diode strings. Substituting a simple voltage divider in a form of two resistors is possible but the required voltage for FB will change if the power supply varies so diode strings is far better. For 2 volts required for feedback, the diodes strings should have 4 volts (.62V + 3.3V approx.) square wave at 50% duty cycle on idle as seen from scope. This will yield an average of 2 volts DC from your DMM. Note! Please pay attention to the “phase inversion within the loop”. This means starting from input pin 2 via Rin up to the nodes of power FETs, there should be phase inversion. Some newbie may lose their attention to this inversion when the signal is being split into two as required by FET driver. Watch closely dude.

Figure b.

There are two important things to focus your attention regarding the circuit above. These are; implementation of dead time, and make sure phase inversion is present within the loop. Disregarding these 2 things may create trouble for the newbie to achieve correct operation. Dead time network D6, D7, R6, R7 along C8 and C9 are “optional components” just in case power FETs input dead time circuit is not enough. If high idling current is present then cross conduction is at work so you know what I mean. Second thing is the phase inversion within the loop must be present. Neglecting proper pull-up and pull-down of the input pins of exclusive OR gates will jeopardize the beauty of negative feedback. EXor gate output with input “pull-down” must be on the Hin pin 10 of IR2010 and the other gate output with input “pull-up” must be on the Lin pin 12. If those gate input pins were carelessly swapped differently from the drawing, then the entire loop will be converted into positive FB and will ruin the entire sound department. Please try to peep on your EXOR truth table to see what I mean.

Figure c.

Another filtered feedback variation can be seen above. This circuit utilized RC filter with -3db roll-off at 32kHz. If you want to listen to the sound of the circuit in fig. c, you can connect directly R11 to output pin 11, 8 because phase inversion is also present within that loop (from pin 2 to pin 11, 8). Sorry for this mistake, I noticed that my 32 ohm headphone effectively “overload” the 680 ohm at pin 11 (with capacitor) giving only 0.93 volts output as measured by my DMM. I got a  distorted sound due to that and this loading also create wrong bias voltage to pin 2 as well. So, listening to this output with NFB and headphone together is not advisable.

Take note that removing RC filter R4 and C6 and connecting R11 directly to output pin 11,8 will produce good sound quality as well (yes, this is true if headphone is not connected there) possibly if additional buffer is available.  Take note that small amount of DC offset is still present from the output and this is a function of parameter spread variations such as diodes voltage drop per degree C, supply voltage changes and etc. Simply adjust the pot at pin 14 to reduce the offset to the lowest as possible. I always favor a pot to null the DC offset or adjusting to required 50% duty cycle.

Designing a class D from scratch is a tough proposition especially for a beginner. This might put Tl494 in the prime list of a newbie’s baby step towards such first attempt. The availability and low cost nature of this chip coupled with high degree of repeatability, simplicity and stable in operation will surely improve the skill of a beginner. If he is creative or little bit ambitious enough; with guts so to speak, he will be surprised to see how practical tl494 to be use as a class D audio source. The very first Tl494 that I experimented was just extracted from a junk ATX power supply but when I put it on the breadboard with some caps and resistors the sound is surprisingly good!

Case in point: Tl494 contains three independent i/p pins that can be tailored to suite your listening needs. In my prototype an active band pass filter is tuned to 20 Hz (I love 20 Hz bass), another one as full audio band and then pin 4 take care of the treble at 7kHz, (I like rich in treble also). This practical approach simply eliminates my needs to build additional active crossover network. It’s totally complete inside the chip.

Figure d.

Suggested circuit employing tuned band pass active filter using the lower opamp as “mega bass enhancer” that peaks at 20Hz with +/-7Hz bandwidth. Then upper opamp is a simple full audio band, while pin 4 takes care of the treble with -3db roll-off at 7 KHz.  Java script calculator for this active filter can be found here.

Did you notice the additional pot R9 from the circuit above? This is another adjustment for removing the offset voltage from the output because pin 4 requires 1.65V bias voltage as compared to pin 2 that needs 2V.

Let’s explore the innards of Tl494. Data sheet says; if the voltage at dead time pin 4 is varied from 0 to 3.3 volts, the output duty cycle will change from 5% to 100% (that’s non-inverting relative to output pin). Also, if the voltage is varied from .5 to 3.5 volts on pin 3, the duty cycle will vary from 97% to 0% (that’s inverting relative to output pin). But take note that the input signal is feed thru Rin of pin 2 which makes the loop also become “non-inverting” relative to output transistor. By the way, this pin 3 voltage should be centered at an average of 2 volts (.5+3.5/2) and also to pin 4 (0+3.3/2= 1.65V) to get 50% duty cycle from the output. Also take note that the author of the PDF did not mention if the output transistors are configured as emitter follower or just common emitter. These statements on PDF may confuse some folks (mind you, me too 6 years ago) but I just simply think that the author’s assumption referred it as emitter follower which is just fitted to the statement above. I also verified that on breadboard long time ago and it’s correct. But I preferred that the output transistors must be “common emitter” due to its lower saturation voltage than emitter follower. In this case, phase inversion will appear to pin 11, 8 output. Still confuse? Take a look at the schematic below for better enlightenment.

Synching of tl494 for multiple channel application.

Synchronizing multiple channels using tl494 is easy without any problem regarding audible bit frequency to occur compared to self oscillating class D approach.

Here is another suggested circuit employing the popular “bass and treble” tone control being used from linear circuits that most of you are familiar with.

Figure e.

The circuit above requires +/- 1.5 volt of input to get the required +/- 1.5 volt output from pin 3. This is due to the nature of this tone control network. When the 2 pots are both on their midpoint position then the voltage gain is 1 on pin 3.

Before you try any of the circuit shown above, try it first on the breadboard so that you can tweak the bass, the mids and high frequency by altering the R or C to your desired audio band. Or you can simply use only one opamp with one fb resistor to be simple one.

OK, here we go. Lets say a fellow wants to build a 2 channel class d amp using 555 timer. This is a bit tricky because using 2 of this chips will oscillate in different manner even if the frequency determining networks are ideally matched. So the need for syncing is a must. Quick thinkers immediately had in mind is that the circuit should have a master and a slave part for proper syncing of carrier signal to avoid audible bit frequency. Yes, this approach is correct but how to implement it? So here is a possible circuit that may pop-up in the mind of quick thinkers.

The circuit shown above was copied from my previous circuit I discussed recently. Here, pin 6 and 2 of slave A and B is being connected to the master pin 6 and 2 for syncing. The master controls the frequency of oscillation. This is the easiest and maybe the simplest syncing scheme that you may think (mind you, me too). But upon simulation, when the audio signal is introduced to either slave A or B, the output pulses to pin 3 does not behave as expected. Even if I carefully tuned pin 5 to any offset voltage that could bring the output more usable seems difficult.

The output waveform of either slave A or B looks promising if no signal is present on pin 5 (pin 5 is floating).
But when audio signal is present, and the collector of 2N3904 is connected to pin 5,my enthusiasm to the circuit suddenly collapsed.

Output waveform of pin 3 in both slave A and B. Here, the output pulses are intermittently put into a halt by the Vmax and Vmin of 1 kHz audio input. I think that there is no simple remedy available here for the moment.
This made me to conclude that this type of circuit is only applicable as single channel amplifier. Careful scrutiny within the entire circuit followed by several simulations attempt leads me to think that it is only necessary to link pin 6 and 2 to its very own oscillation or modulation and should not rely to external one. There must be another way to implement two or more channel and the simple answer was already discussed by me at the very first article I post.

The circuit shown below will work in multiple channel application because the master’s output pin 3 has negligible loading effect of the trigger pin 2 of any slaves connected to it. Also, this circuit is quite unique in its own right because the operating principle does not conform to a conventional setting where saw tooth is generated to sample the audio signal thru comparator’s two inputs to produce PWM. Here, the master generates continuous but fixed pulse width of approximately 98% for the purpose of two important things; To reset the internal flip flop of all slaves, and to discharge the timing capacitor via pin7 simultaneously. The most attractive feature of this circuit is that they are darn cheap. This circuit can produce wider modulation from 0% to 95%. But 100% is not allowed because the master is required to pull down pin 2 to ground of all slaves in short duration fast enough to accomplish its two task as already mentioned above. Opamp bias network R11 and R12 should be replaced with potentiometer for 50% duty cycle adjust.

The master produce an output of very short negative going pulse which is sufficient to change the output state of all slaves into logic 1 but the “width” of that “on” state is governed by modulation voltage on pin 5 of all slaves.
Please take note that even though the circuit looks good and simple, proper matching of the master and slave’s frequency is very important. Improper matching will result into unsuitable for class d application. This mismatched frequency scenario I’m referring to is; instead of producing class d amp, you will end up to get a frequency divider instead. Try to simulate it so that you know what I mean.
Let’s scrutinize the basic function of the circuit. The master is wired as an astable multivibrator having a fixed frequency F determined by;

and the duty cycle is;

The slave which is configured as monostable or “one-shot” having a variable pulse width (PW) which is a function of;

PW= 1.1 R1 x C1

Take note That there are three possible ways in modulating the width of the pulse of the slaves; Its either to vary the timing resistor R1 or to vary the capacitance of C1, but the most easiest way is to modulate pin 5 with a correct signal. Some newbie might wonder how does PWM is being generated by simply modulating pin 5? Yeah, It is more easier to grasp the theory behind PWM generation by altering either R1 or C1, but how about pin 5? The answer is simply based on the fact that if the 3 internal voltage divider of 555 timer varies its voltage via pin 5, the cut-off and cut-in of upper and lower comparator voltages will “set” and “reset” the internal flip flop of all slaves accordingly thereby producing PWM from their outputs.
In my breadboard, R2 doesn’t exist and I can produce 97% duty cycle for use as master. If you cannot produce this without R2, then you can add R2, then follow the astable equation above. By the way, my 555 timers are all CMOS type marked as LMC 555 and they are capable of oscillating up to 3 Mhz at 5 volts. They consumed less in power but with very fast rise and fall times of 15 nanoseconds. Of course any bipolar 555 timers will work as well. It just so happen that LMC 555 is the only available stuff in my cookie jar.

Simulation of the circuit above is available below:

210 Views Since 2011-02

A month ago, I came across a site called DX deal extreme and by navigating through their hobby/DIY section I noticed that they are already selling the most powerful white LED emitter that I’ve been so eager to buy. As a member of candlepower forum this white LED is always on top of the wish list by most of us. MCE means “multi chip emiter” as opposed to single chip that I have in my possession.

Picture of Cree MCE White LEDs

My plan is to change the two head light bulbs of my motorcycle with these beasts and sure this will give a very bright white light shining along my path at night. My only concern is how to figure out in mounting appropriate heat sink at the back of the reflector (yes, this little boy generate heat when driven higher than its specs). Another thing of more concern is how to mount a dome cap to create a shadow on top of the beam for dimming purposes otherwise; this may affect the sight of a fellow motorist.

Close-up of the Cree LED

The single chip LED on the left for size comparison

By zooming-in reveals the quad chip with 4 independent bonding wires for configuring either parallel or in series connection mode. For automotive lighting apps, then series connection is the correct choice because each chip requires above 3 volts supply at 350ma of current. Most LED enthusiast I know from CPF usually drive them up to 1 amp without complain and of course the brightness is astounding indeed. I’m not yet prepared to do these overdriving schemes because if I encounter a bad egg from the group, I will wait for a couple of weeks for a replacement and each cost me 17.50 US$ each. Well, I love this stuff so this is not expensive to me. My only gripe is waiting for door to door delivery is a real “killing me softly” scenario and it took me to wait for 2 months due to out of stock from the supplier.

As an electronic hobbyist, I always wanted to own a sound generator good enough as signal source for my audio test needs. Since most audio test instruments available are expensive and some are bulky to occupy my work bench, my options have been narrowed down when I came across on the internet a program called “sound card scope 1.32” by C. Zeitnitz. This program is free for private or public educational use but please read the license agreement from the author regarding this matter.

This program includes oscilloscope, spectrum analyzer, sound generator and numerous knobs and buttons as well as sliding controls that mimics like a real test gear of your dream. As the name implies, this program can measure frequencies up to the limit of your sound card, from few Hz up to 15 kHz. Also take note that the signal that you will measure is being feed into the ”mic input” of the sound card  having a maximum of 1 mv or up to 500 mv only depending on the type of the card being bundled on your PC. This sensitivity can be adjusted via windows volume control panel. If you use the other sound input such as its RCA jack then its typical sensitivity is 1 V to 2. 5 V (sound card dependent also). Likewise, typical input impedance is around 75 to 600 ohms (sound card dependent also). This very low input impedance can give heavy load to some small signal sources in fact when I measure the sawtooth at pin 6 of the 555 timer for checking, the oscillation simply stop, don’t laugh!. So you will need a series resistor to increase the input resistance and a voltage divider to measure voltages above those mentioned level. Please be careful!

A picture is worth more than a thousand words.

This graphical user interface reminds me of a classical yet expensive test gear I’ve seen during the time when I was working in a semiconductor company. Yumm nostalgia. When I test fire this thingy and measure the output of my 555 timer oscillator in a breadboard, these are the result of my test:

A 427 Hertz generated by 555 timer in the breadboard

This sawtooth waveform is generated by 555 timer circuit via a breadboard. This is done by connecting a 100k resistor in series with 0.1uf capacitor at output pin3.

When I crank-up the frequency to 5 kHz my sound card start to show a pathetic waveform

You may notice that the waveform rise and fall times slowness occur at higher frequency but this is due to my sound card response time and not by the program itself. If you want better sound card experience then you may consider upgrading your sound card from creative labs sound blaster X-Fi series. These potent sound cards extend sampling rate above 80 kHz at 24 bit resolution. Though I have not test this either.

Here are some shots taken from the function generator dropdown list options as sawtooth, squarewave, and triangular waves.

When I put a check on the triangle function, I was surprise how poor my sound card perform even at very low frequency. 1kHz is shown above

My sound card looks so pathetic that it cannot produce decent waveform quality even below 1 kHz for signals as listed above.

Its only good in showing sinewave below and above 5 kHz but the square wave also behave as sinewave as well. Oh well.

Again, Please be careful in measuring audio signals with higher voltages. The last image shown above is taken when investigating the frequency spectrum of the 555 timer circuit being tested. The fundamental frequency contains several noises above and below the main frequency. You can use this to check the frequency response of your opamp when wired as active band pass filter or observing the response curve of your  active cross-over network.  Measuring the frequency of any oscillator within the limits of your sound card is also a good bet because of the added frequency counter too.  Well, at least this is how my sound card performs. How about yours? Have you ever use your sound card with these useful programs yet?

My broken CRT oscilloscope is in the brink of extinction due to burned 1.5KV electrolytic filter capacitor, so my initial attempt is to rely again to LT spice simulation and compare the result in real world breadboard.

The schematic is shown below.

For any given supply voltage to pin 8, the voltage that will appear from pin 5 is always 2/3 of that Vcc. Since pins 6 and 2 are tied together, its oscillating voltage varies at 2/3 max and 1/3 min. At 12 V supply, pins 6 and 2 is at its threshold of 8 volts cut-off and 4 Volts cut-in respectively. If I get the average voltage of this two then I may get the desired 50% duty cycle. This statement is actually correct “if” pin 5 is open or floating. The timer will oscillate in astable mode at desired frequency but the problem usually occurs when that pin is terminated with a sound source. The impedance of the source effectively pulls down pin 5 to ground resulting to a change in frequency and duty cycle. In my breadboard the frequency drop to half, yeah I can live with that but the pain is so hard to ignore regarding the distorted sound with some noise. The circuit is shown below without transistor.

When I transferred the headphone to pin 7 with pull-up resistor, the noise disappear. This means that the source of the noise is the amplified signal from pin 3 being feedback to pin 6. So, this is the suggested output termination if someone wants to amplify the signal.

The circuit shown above has a slight reduction in volume of sound due to the high output resistance of R2 but good enough for amplification due to the absence of noise compared to first circuit.

The necessity for a better sound quality and to improve frequency response is more desirable and can be achieve only by adding negative feedback (NFB). And that’s what 2N3904 comes in.

If pin 5 is not connected the 8 volts on that pin is constant while 2/3 and 1/3 Vcc at pin 6 and 2 behave in a normal astable form with 50% duty cycle.

Waveform of unconnected pin 5

When I connect pin 5 to 2N3904 collector, the waveforms changed to both quasi triangular form with voltages at pin 6 and 2 at lower amplitude while pin 5 is higher.

And the corresponding waveform is shown bellow

It’s quite noticeable that the duty cycle is being affected (red trace). The previous 8 and 4 Volts from pin 6 and 2 was diverted to pin 5 now with both “peaks” of green and blue voltages are functions of T rise and T fall output voltage of pin 3.

Note that by “pulling up” the green voltage i.e. the collector of 2N3904 in this case, that green will return back to its original 8 volts and the corresponding duty cycle at output pin 3 will return back to 50%. The only possible way to do this is to put a pot at the base of 2N3904, oh well here we go again, don’t laugh! I think for the time being, this will suffice until I figure out how to avoid it.

Output waveform when the pot is adjusted to a lower resistance. This means the collector current is reduced to “pull-up” pin 5 to its original 8 volts.

Complete schematic with pot included.

The voltage gain of the circuit above is approximately R4 divided by Rin. With an input of +/-200 mv will modulate pin 5 to +/- 2 volts. Please take note that the voltage on pin 5 should not be driven to more than +/- 2.5 volts. Doing so will produce distortion. Also, you cannot expect 100% modulation here. Based on my simulated circuit, typical modulation is something like 15 to 85% max (centered at 50%). Of course, this is only a class D toy for the beginners and should not expect full swing modulation out of it. Take note that there’s a lot of room for further improvement in this circuit such as the use of opamp that will simplify the biasing schemes.

Improvised circuit

Edited version (October 16, 2010)

The schematic shown above requires minor tweaking on the 555 timer side due to the changes in supply voltage which is at 12 volts now. Here, aside from added  offset adjust pot (50% adjust pot via R3), attention must be focused to pin 5 as well because this is the inverting input of the internal comparator which is tied-up from the voltage divider that requires tweaking also. Adjusting R1 pot will also help to accomplish this. Please refer on the schematic shown below.

555 schematic

The sawtooth voltage variations from pin 6 does not conforms to the required 2/3 VCC of the timer but instead, something less than 2/3 of VCC to obtain low distortion. For best result, adding one opamp as a replacement to 2N3904 will give better result.

Edit:

I include LTspice simulation for you play with. Either R2 or C2 can be adjusted for dead time adjustment to the lowest distortion as possible without cross conduction. You may disregard this RC delaying network and its ORed diode resistor combination from the output and use it only as phase splitter if you like.

LT spice simulation

297 Views Since 2011-02

Schematic of dual 555 circuit

Dual 555 circuit. It is sometimes necessary to tweak the resistance of resistors slightly different from their LT spice model circuit. Oh well

By scrutinizing the behavior of the circuit in LT spice simulator, and listening to their sound thru the breadboard circuit layout, the sound is good and clean. Of course there are some interesting ways in doing this but the most interesting thing for me to satisfy my curiosity is to try to use U1 alone without  the trigger U2 to conserve space and  cost to the lowest value as possible. In this case, by feeding back the output pin 3 to the RC time network on pin 6 & 2 the resulting switching mode is astable with 50% duty cycle.  I tried to apply music to pin 5 as usual but noticed a distorting sound even at lower volume settings.

Oscillating frequency of the circuit at 240khz

Single 555 timer with added 2N3904

Fig. 1  Schematic of 555 circuit with negative FB added. Please take note that 2N3904 varies on their hfe depending on it’s production binning. So you may get different sound quality due to this variations. Better opt for opamp if you want.

By implementing negative feedback thru the help of 2N3904 transistor to self bias it in a conventional way, (see fig. 1) the resulting sound was so good that it seems better than the sound produced by using 2 chips. Why? Well, I don’t have specific answer for this question at the moment and by observing the behavior of the waveforms while the circuit is being modulated reveals a mixture of PWM and FM at the same time. It is quite surprising for me to hear a music in class d amp having a mixture of PWM with FM owing to the fact that most of my class d design works are always  time based as opposed to self oscillating approach.  I think class d amp design based in self oscillating mode experience the same thing like this which makes it (self oscillating) more favored my most audiophiles.                                                                 More circuits to follow.

Thanks for visiting my site

Most hobbyist are already familiar how to bias this chip to function as either astable, monostable, FM or PWM. But recently, I’ve tried to figure out how I can use this chip to function as class d amp source that can drive a headphone good enough to play music (or a class D preamp source).

Enter Ltspice switch cad IV, whom I rely heavily in investigating the behavioral model of 555 timer which is pretty much identical to its real world brother. I started from a basic circuit configuration of a monostable multivibrator being triggered by astable one in order to establish constant frequency. Several discrete parts were added tuned to chosen frequency of operation and quiescent output state as well.

A good schematic is worth more than a thousand words. So here is the simplified schematic that can be easily understood by a newbie as shown in figure 1.

Fig. 1

For a beginning student who hasn’t used Ltspice before, you can download it now or you will miss the trip to enjoyment. Honestly.

My initial attempt was a typical circuit configuration using R1 to charge capacitor C1 for sawtooth generation as shown from the circuit of fig.1. The 277khz clock frequency is generated by U2 at approx 98%  wired as astable oscillator while U1 is wired as monostable or one shot. PWM is produced at pin 3 by modulating an audio signal at pin 5 of U1. Of course Ltspice always behaves precisely as expected. By zooming in and out of the waveforms I noticed that there are some unwanted artifacts on the waves that I’m suspecting that the source of this culprit is the coupling capacitor C2. When I removed the capacitor and retain the offset voltage of 4V on pin 5 the problem disappear.

Fig. 2 Opamp added without coupling capacitor

Another thing that’s not good is the curved voltage across capacitor C1. This is shown in fig.3.

Fig. 3 The humpback sawtooth!

Most professional hobbyist usually stays away from this curved sawtooth because it creates distortion so I replaced R1 with 2N3906 biased as constant current source and so; before you can see it…

Now you don’t!

Fig. 4 linear sawtooth

Figure 5 the improved version of fig.1

Then I immediately grab my breadboard and 555 timer + other parts for verification in real world, (while I’m holding my breath). Lo and behold! I hope it’s not a placebo effect I’m hearing with my two ears.

While my ear cannot distinguish THD precisely, and since I am an ordinary geek guy who usually loves my own discovery, for me the sound is good enough (as usual). Of course there is always room for improvement. Several things to consider in improving the robustness of this chip for audio apps is notably to use voltage regulator for Vcc. Because if small changes occur from the supply voltage, the PWM output is severely affected as well. There’s no negative feedback from input to output; So NFB must be added to improve the overall response.

Did you notice that there’s no coupling capacitor tied to the output of opamp? This is because 4 to 5 volts DC bias to pin 5 is essential to force the upper comparator to sit at 50% of the entire dynamic charge and discharge of capacitor C1. If you attempt to put coupling capacitor there, the carrier frequency will drop to almost half. This is because C1 at pin 6 is being mirrored thru input pin 5 (or vice versa) exactly identical to conventional comparators.

I’m trying to repair my old CRT scope to capture the real word waveforms but I don’t have any luck in procuring 2.5Kv .01uf filter capacitor that leaked on its power supply a week ago. So I ended up like a cameraless paparachi.

So stick around and I’ll be back soon.  Thanks for visiting my site.

Edit: BTW, I include the spice simulated file for you to play with. Just unzip it and open it inside your Ltspice switchcad. You can post a comment for what you’ve noticed or to be added to improve its usability. Good IQ exercise heh?

Simulation below:

301 Views Since 2011-02

Regards

IQspice-4

The basic operating principles of these audio amps are similar to those of switching DC to DC converters utilizing PWM (pulse width modulation). But when used in audio applications, the output is purely in a clock pulses being modulated on its duty cycle in proportion to audio signal as a modulator.

Let’s say at chosen clock frequency of 200 kHz, the quiescent output pulses are at 50% duty cycle without audio but being modulated to +/- 50 when strong audio is present. This will yield an output of from 0 to 100% centered at 50.

Fig. 1 shows the waveform at 50% no audio.

In a real world scenario, audio modulations should not necessarily swing to 100% but practically less than that. In this case, a 10 to 90% centered at 50 are common to avoid distortions.

Fig. 2 shows the waveform at 90% large signal swing.

Fig. 3 shows the waveform at 10% at minimum signal swing.

Fig. 4 shows the composite signal containing 1 KHz audio sine wave as the modulator, and the 200 KHz pulse width modulated carrier.

Fig. 5 shows the “magnified” view of the modulated carrier signal on it’s rising and falling contours of the sine wave.

The waveforms shown above were done in a circuit simulator using LTspice IV with its integral signal generator configured to produce PWM. The next thing I’m going to show you is how to implement it in real circuit but the big question is, where will I get a low cost and readily available chip that can provide a decent sound that can satisfy my taste in music quality?

Well, this question is somewhat subjective and varies from person to person. So, the measurements and data that I may get are strictly for personal evaluation only and not for commercial application purposes.

Another thing I would like to emphasize is to use “non-audio chip” as the primary source of PWM signal as opposed to expensive and hard to get high speed opamps or comparators.  So, as the name of this site implies, I will use exotic circuits but relatively low cost and readily available IC’s that are not intended for audio applications but are specifically made by chip makers for dc to dc converters. Well, this doesn’t mean that the music quality is not good enough but if you will try it, you’ll be surprise. Dont worry guys, this will not necessarily put your learning at risk but it will merely stretch your skill beyond your existing one. So, stand-by please for the following demo circuit soon……….

Thanks for visiting my site…

Class D amp using 555 timer