555 Class D Power Amplifier [130144-I]

With this project, I like to build a small class D amplifier based on a 555 timer IC. I also like to use only “common” components, so this project has great educational value and is easy to build.My goal is to produce about 5W RMS power into 4ohm. Distortion level should be less than 1%.Circuit description VERSION1 (see circuit diagram below)

My goal is to produce about 5W RMS power into 4ohm. Distortion level should be less than 1%.

Circuit description VERSION1 (see circuit diagram below)

In order to build a class D amplifier, we must convert the analog signal to a digital signal. In fact we like to use a pulse with modulated signal that follows the analog audio signal. This way, we can switch 2 power MOSFETs on and off in a very efficient way.

I use the NE555 as the heart of the PWM encoder. At the input, we have a standard NPN transistor input (Q4) biased by R10 and R9. C1 blocks all DC at the input. The idea is to modulate the way the charging capacitor C3 is loading with the analog audio signal.

In most standard 555 timing configurations, the timing capacitor is charged with a constant voltage, resulting in severe non linearities, especially at high output levels. To improve linearity, I charge the timing capacitor with the constant current. Therefore I use a current source at the input (Q3, R2 and R7) and a voltage to current convertor in the feedback loop of the 555. (Q1, Q2, R1 and R3) This way we get a real triangle wave on C3. (see image below “signal on C3 no signal.jpg”)

Oscillation frequency of the 555 is around 300 kHz.

Output signal is 0V – 18V PWM signal that switches on and off M1 and M2.

D1, D2, R4 and R5 are dealing with “on” and “off” timings and avoid that M1 and M2 are “on” at the same time (and avoiding to short circuit the power supply). R12, R14 and Q5 form a feedback circuit that improves linearity and distortion levels even more. L1 and C4 is a LPF around 25 kHz, removing all high frequency switching components.

I use a SMPS laptop power supply of 18V to test with (cheap). Initially I had switching noise in the speaker. Therefore I add a coil of 100µH to filter this out. Working fine!

I built this circuit on a standard testing board, and I’m quite happy with the result. Audio sound “direct” and “clear” to me, with tight basses! I get about 6 W RMS power, what is fine for my first attempt J I do not have any distortion measurement equipment, so I’m not sure if my goal of <1% is reached.

About efficiency, the heat sink is cool at all levels, so this should be fine.

I attached some images and of course the schematic and also some scope prints of the output square wave and charging capacitor wave. I have also added a short video of the amp @ work. Don’t bother the bad sound quality of the movie; it’s the poor mic quality of my camera J

Possible improvements (VERSION2 tbc)

Power MOSFETs must be controlled with 18V to be fully saturated (and “on”), if I raise the supply voltage on the MOSFETs and control with 18V from 555, MOSFETs are running hot as they don’t fully switch on. A voltage level shifter circuit is recommended here.
I will to do some tests with IR2110 IC to achieve this, maybe there's a discrete solution? Maybe with a "charge pump"?
Now we use “half bridge” 2 MOSFET configuration, therefore needing a big output capacitor, I did some tests with full bridge configuration (4 MOSFETs) in SPICE but I get a lot of cross over distortion. Need more investigation on this….

Any help, suggestions, or hints are welcome to improve this project!

Project Elements

Signal on C3, no signal (JPG, 206.28 KB)

Signal before LC filter, no signal (JPG, 223.63 KB)

Board (JPG, 290.09 KB)

Board detail 1 (JPG, 249.13 KB)

Board detail 2 (JPG, 239.56 KB)

Board detail 3 (JPG, 268.27 KB)

Power supply (JPG, 395.08 KB)

555 20ok fb print.pdf (PDF)

Circuit diagram

video1.zip (ZIP)

Zipped MP4 video

Discussion (23 comments)

Bobbylebob 11 years ago

Hi frederik

I've just had a look to your circuit description. You are talking about efficiency matter and you're right when you say that a strongest drive give shortest rise and fall times and reduces loss.

You tried to increase to 18V but you've reached the maximum voltage, that's a good idea. However, you have to consider MOSFET's gates as small capacitors to properly drive them. When a transition from low to high occurs, the capacitor is at 0V and is charged through the 220R resistor. The peak current is then 18/220 = 82mA.This current is easily supplied by NE555's output but is well too low to give a fast transition

During High to low transitions, 220R resistors are bypassed by the diodes and the current is then only limited by NE555, and is probably greater than it can handle safely...

I guess that it also explains your interrogation about the assymetric waveform you found

Usually, it's recommanded to add MOSFETS drivers which can handle bigger currents (many amps !)

I suggest you to add a 18/0.2 = 100 ohms resistor in series with the diode to limit the falling current to the maximum allowable of 200mA.

You could also try to reduce the 220R to something like 150 or 120 which could speed up commutation, but pay attention to cross conduction...

Regards,

.David.

0 Comment(s)

NuttyProf 11 years ago

Hi Frederick

I had a bit of thought. I have added a simple extra pair on the end to produce a bridge output in order to give extra power. This is purely a theoretical concept and has not been tested but I see no reason for it not to work. If the duty cycle is exactly 50/50 then in theory there will be no voltage across the Speaker A and Speaker B terminals.

The original had lousy FETS in the positive side (IRF9530 I think) with a high on resistance. Using the IRF4905 means a lower on resistance and better efficiency for the high side of the totem pole. The 555 only has to drive the one pair and the output of this pair drives the second pair. In my sketch I have left out the protection diodes across the FETs to the supply and also left out the full filtering circuit that goes on the output of the FETs before the speakers.

I would be keen to see if anyone takes up this idea. It should lead to a very efficient compact 25 watt RMS sine wave into 4 ohm output if my calculations are correct. Or 2 such amps (stereo pair) running from the PC PSU to power a pair of decent sized speaker boxes from the PC (maybe have to get a larger PC PSU to get the extra amperage for the amps). Maybe even design it to go into a 5 & ¼ or 3&1/2 rack drive straight into the PC. Or even 2 such amps (stereo) and a 12v to 19v supply all in one small box to use with a notebook PC on the 12 volts of a cars battery on summer camping trips. Here in South Africa this would be great for a party pack when the power goes out at a party.

Anyway the idea is there for anyone to take up. Please keep me informed should there be any takers.

Regards Harry Bartelink

clasdamp-1.jpg (36kb)

frederik 11 years ago

Hello Harry, Great to hear you like to use the amp on South African parties :-) The full bridge version might be possible to build, i did some spice tests during the developement of this project but did not go on with it (see http://www.elektor-projects.com/contribution/new-idea.13031.html ) Reason for this is that the full bridge version generates a lot of cross over distortion. Also it is very difficult to avoid short circuit when both fets are accidently on. Timing management is very difficult to do without complex electronics! What I suggest to do when you have a powercut on a party: use 2 car batteries in series (so you get 24V) and use more amps and small speakers! This amp distorts like a tube amp, so you can drive it further to the limits without anoying distortion! All the best and greetings from Belgium! frederik

1 Attachment(s) 1 Comment(s)

TonGiesberts 11 years ago

We build our prototype and did some measurements. But let’s get to the most important and interesting part of the circuit first before getting to the cold hard figures. The hart of the pulse width modulator is a 555. There are many applications to make a PWM signal with a 555. One is using the control voltage in a monostable mode with an external clock; another is feedback the output with a potentiometer and two diodes to a capacitor connected from threshold en trigger input to ground. The circuit here is a variation of the second example mentioned. Instead of a potentiometer the voltage across de capacitor is controlled by de LF signal. In an ideal case the triangular waveform should have perfect straight slopes. Any deviation will produce distortion of the modulated LF signal. Simply feeding back the output through a resistor would produce an exponential curve. To linearize this two transistors (T3/T4) are used to keep the current for charging and discharging the capacitor (C2) constant. This improves linearity a lot but it is not a perfect solution. The current through the capacitor is a sum of the current determined by the base emitter-voltage across R6 and the base current of T3 or T4. Since the collector current of T3/T4 varies with the current through R7 so does the base emitter voltage (and the base current). This relation is not linear nor are the two transistors perfectly complementary causing the two slopes to be different. Even a slight deviation from the ideal triangular waveform will have consequences.

Since there is no feedback to counteract distortion at the output the final distortion will increase with every stage. However with large distortion the different stages can counteract each other partially and as a result give less distortion as certain levels. This effect can be seen in Plot A (THDvsInputLevel.emf, 1kHz THD+N at B = 22 kHz). Here total harmonic distortion can be seen as function of input level. The blue trace is the THD at the output of the 555 (pin 3) after filtering. The red trace is the THD at the corresponding output of the amplifier. Between input levels of 7 to 40 mV THD is less at the amplifiers output than at the output of the 555. Part of the extra distortion can be a result of the deadtime in the half-bridge power stage (similar to the distortion in a linear class B stage) but also the passive output filter.

Plot B shows amplitude of the output as a function of frequency at 1 W with an 8 Ω load (AMPvsFRQ.emf). The low cut-off frequency is 11 Hz and the high cut-off frequency is 38 kHz. At the low-end the -1 dB corner is at about 21 Hz and at 19 kHz there is slight rise of 0.66 dB. Loss in amplitude due to the output filter is only 0.1 dB with an 8 Ω load. To reduce the rise in amplitude around 19 kHz the capacitor in the output filter (C12) can be reduced to 390 nF or so. This however will reduce the suppression of the modulation frequency by 4 dB or so.

Plot C shows total harmonic distortion plus noise as a function of frequency with an 8 Ω load (THDvsFRQ_9th_order_filter_2.emf). Two traces are shown. Trace A (red) is measured at the output. To have an idea what the influence of the output filter is; we also measured the signal at the input of the output filter (trace B, blue). For both measurements we used our Measurement Filter for Class D (100540-1). This way all harmonics up to 100 kHz are almost unaffected, attenuation at 100 kHz is 1 dB. Measuring bandwidth of our analyzer was set to 80 kHz. The distortion at the output is slightly less at the lower frequencies due to the smaller bandwidth. At the highest frequencies the distortion is largely due to intermodulation products that are not filtered out by the extra measurement filter.

Plot D shows a FFT of a 1 kHz signal of 1 W at an 8 Ω load (FFT_1W_8ohm_1.emf). The 5 harmonics that determine the distortion (THD+N = 0.23 %) are at a level of -60 dB. But also visible are intermodulation products (at a distance of 326 Hz) of the two channels caused by the difference in clock frequencies of the two modulators. These products are all below -85 dB. Plot E shows that if we disable the other channel these products are no longer visible (FFT_1W_8ohm_2.emf). To prevent these products from getting inside the audio band the two clock frequencies should at least be 40 kHz apart. Intermodulation products of a 20 kHz signal in the other channel will cause an extra 20 kHz shift.

Measurements

Input sensitivity 580 mV (THD+N = 1 %)

830 mV (THD+N = 10 %)

Input impedance 11 kΩ

Sine-wave power 6.3 W (THD+N = 1 %)

10.8 W (THD+N = 10 %)

Bandwidth 11 Hz…37.5 kHz (-3 dB)

21.5 Hz…31 kHz

Signal to noise ratio 69 dB (1 W/8 Ω, B = 22 Hz…22 kHz linear)

Harmonic distortion plus noise 0.23 % (1 kHz, 1 W/8 Ω)

Channel separation 42 dB (100 Hz, Pmax)

54 dB (1 kHz, Pmax)

60 dB (20 kHz, Pmax)

Current consumption 0.8 A at 2 x 6.3 W/8 Ω

Bill of materials 130144-1 v2.0

Resistor

R1,R15 = 100 kΩ, 5 %, 0W25

R2,R16 = 15 kΩ, 5 %, 0W25

R3,R17 = 82 kΩ, 5 %, 0W25

R4,R8,R18,R22 = 1 kΩ, 5 %, 0W25

R5,R19 = 10 kΩ, 5 %, 0W25

R6,R20 = 680 Ω, 5 %, 0W25

R7,R21,R29 = 2k2, 5 %, 0W25

R9,R23 = 18 kΩ, 5 %, 0W25

R10,R24 = 470 Ω, 5 %, 0W25

R11,R25 = 100 Ω, 5 %, 0W25

R12,R26 = 330 Ω, 5 %, 0W25

R13,R27 = 8Ω2, 5 %, 1W

R14,R28 = 2k7, 5 %, 0W25

R30 = 56 kΩ, 5 %, 0W25

R31,R32 = 56 Ω, 5 %, 1W,

P1,P2 = 2 kΩ, 10 %, 0W5, trimmer multiturn (Vishay Sfernice T93YB202KT20)

Capacitor

C1,C15 = 2µ2, 5 %, 50 V, lead spacing 5/7.5 mm (Panasonic ECQV1H225JL)

C2,C16 = 330 pF, 1%, 63 V, lead spacing 7.1842 mm (LCR Components EXFS/HR 330PF +/- 1%)

C3-C6,C8,C10,C17-C20,C22,C24,C30,C32 = 100 nF, 10 %, 50 V, X7R, lead spacing 5.08 mm

C7,C9,C21,C23 = 1 nF, 10 %, 63 V, MKT, lead spacing 5 mm

C11,C25,C29 = 1000 µF, 20 %, 35 V, diam.12.5 mm, lead spacing 5 mm

(Rubycon 35YXF1000MEFC12.5X25)

C12,C26 = 680 nF, 5 %, 250 V, PP, lead spacing 15 mm (Panasonic ECWF2684JAQ)

C13,C27 = 220 nF, 5 %, 63 V, MKT, lead spacing 5 mm

C14,C28, = 2200 µF, 20 %, 35 V, diam.18 mm, lead spacing 5/7.5 mm

(Panasonic EEUTP1V222)

C31 = 10 uF, 20 %, 100 V, diam.6.3 mm, lead spacing 2.5 mm

C33 = 47 uF, 20 %, 35 V, diam.8.5 mm max., leadspacing 2.5 mm

Inductor

L1,L2 = 47 µH, 10 %, 21 mΩ, 8.5 A, radial, bobbin type (Murata Power Solutions 1447385C)

L3 = 100 µH, 20 %, 35 mΩ, 5 A, radial, toroid (Würth Elektronik 7447070)

Semiconductor

D1-D8 = 1N5819

D9,D10 = Led red, 2x5 mm rectangular,

D11 = Led red 3 mm

D12 = 15 V, 0W5

D13 = 1N4148

T1,T3,T10,T12,T19 = BC547C

T2,T4,T11,T13 = BC557C

T5,T14 = BD139

T6,T15 = BD140

T7,T16 = 2N2222

T8,T9,T17,T18 = IRF530

IC1,IC2 = TLC555CP

IC3 = 7815

Other

K1,K2 = 2-way pinheader SIL, 2.54 mm spacing

K3,LS1,LS2 = Screw termination block, PCB, 5.08 mm pitch

RE1 = Relais 24 V, 8 A, 1.2 kΩ, DPDT-CO, Finder 40.52.7.024.0000

F1 = Fuse holder, 20x5 mm

F1 = Fuse holder cover

F1 = 2 A fuse, antisurge, 5 x 20 mm

TP1,TP2 = test pin (1-way pinheader)

HS1 = 1…2 mm aluminum plate, 130 x 50 mm

4 x SIL-PAD, K-10, .006', TO-220 (Bergquist) or something similar

4 x Bush 3 mm, TO-220

Heatsink for IC3 30 K/W (Fischer Elektronik SK 12 SA 32)

Misc.

PCB 130144-1 v2.0

A few remarks about building the PCB

Everything is pretty straight forward. Only a few details need some clarification.

To keep the temperature coefficient of current sources T2 and T11 as small as possible leds D9 and D10 (rectangular) should be mounted against the corresponding transistor. That’s why on the PCB D9 is directly alongside T2 and D10 alongside T11.

We used ceramic (aluminum oxide) wafers for insulating the power transistors in the two half-bridges (T8, T9, T17, T18), but it can be any other electrically insulating material (mica, kapton etc.). The power dissipation is minimal. Just beware not to put a sideway force on the leads of the power transistors. Bend the leads in such way that they align with the corresponding holes in the PCB and the body of the transistors with the heatsink. For this the exact placement of the plate aluminum for the cooling of the power transistors should already be figured out. Two brackets (custom made) are used to fix the aluminum plate (3^rd photo). This ensures also that this heatsink is grounded.

For the 15 V regulator (IC3) we used a small heatsink from Fischer (SK 12 SA 32, 30 K/W) but it can also be a little aluminum plate. We had to cut out a little piece to make room for C33.

In the 2^nd photo R31 and R32 don’t look like 1 watt resistors, but they are. We used a series from Welwyn type MFP1-56R JI. The footprint also accommodates larger types.

Topoverlay of PCB 130144-1 v2.0 (106kb)
Top copper and bottom copper of PCB 130144-1 v2.0 (72kb)
Plots of measurements of PCB 130144-1 v2.0 (28kb)
130144-1-555-class-d-power-amplifier-schematic-v200-1.jpg (1459kb)
130144-1v20-001.jpg (2018kb)
130144-1v20-004.jpg (1669kb)
130144-1v20-005.jpg (1373kb)

frederik 11 years ago

Tank you very much for the tests om my amp! I think I reached my initial goal: less than 1% distortion and about 5W of power. Some more remarks and idea's: - I was wondering if you have had any oscillations during the test? - My amp is now fully stable but I need to run both channels on a different pulse width (with no signal on input) : left = 55% and rigtht = 45% Maybe it has something to do with the intermodulation products you mentioned? Maybe i need to change the 555 capacitor in one channel? - About distortion: when you overdirve the amp, I distorts like a tube amplifier! Its not clipping like a class AB amp but more "compressing" the signal. It sounds pretty nice to me :-) - Output bandwidth is now a bit too high (37kHz) Maybe we need to increase ouput coils value to get aroud 20K? - About overall feedback; maybe we can get even better distortion figures, but I believe an amp without overal feedback is sounding much more "relaxed" and direct. I built once a Vanderveen valve amp without any feedback and ever since then, im a fan of such designs ;-) - Do you have any suggestions to improve the design? Maybe higher supply voltage to gain more power? Tnx, Frederik

7 Attachment(s) 1 Comment(s)

TonGiesberts 11 years ago

Here’s our first schematic and PCB design. Now that the design from Frederik is finalized we’ve decided to design a PCB and made some changes in component choices we think will improve the design. The PCB design is a bit more compact (131.22 x 99.47 mm). Components we think could be improved on are the electrolytic capacitors for power supply decoupling (C11/C25/C29) and the output signals (C14/C28). Also the capacitors for the pulse width modulation (C2/C16) are high quality polystyrene versions (radial types). The footprint also provides a double pitch for the use of silver mica types. The dielectric is critical for the linearity of the modulation. As an output inductor a bobbin type from Murata Power Solutions is used that (hopefully) has a more constant inductance versus current than the types from Pulse in the original design, thus less distortion in the output filter. This is also the reason it is a bit overrated (8.5 A). The footprint also has a double pitch for the larger 100 µH version. Should anyone one want to change the electrolytic capacitors in the output, the footprint has a double pitch of 5 and 7.5 mm. The capacitors in the LC-filter in the output (C12,C26) are polypropylene capacitors which are better suitable for this high current and high frequency application. To make measuring of the PWM signal from the 555 more easily we added two test points (TP1 and TP2). After filtering this signal first the distortion of the modulator can be measured and compare it to the total distortion at the output. That’s it for now and we’ll get back when we’ve had some first test results with our version.

130144-1-555-class-d-power-amplifier-schematic-v200.jpg (1458kb)
PCB design from the lab for a first prototype (106kb)

frederik 11 years ago

Hello Ton, Tank you so much for the prototyping and advice on my design. I'm happy my amp is getting "elektorized" now :-) Meanwhile, my version of the amp is running daily for about a month now, without any problems. Sound is good, amp is cold (the powersupply is producing most of the heat) and no smoke or other accidents so far :-) Grtz, Frederik

2 Attachment(s) 1 Comment(s)

frederik 11 years ago

some images of the final result!

dsc01295.JPG (483kb)
dsc01321.JPG (635kb)
dsc01322.JPG (305kb)
dsc01323.JPG (432kb)
dsc01325.JPG (293kb)

5 Attachment(s) 0 Comment(s)

frederik 11 years ago

I did some more tests on the frequency response.
With the 100µH + 680hF output filter I get a -3db point ant 15kHz at 8ohm load.

I changed the output filter inductor to 47uH, what corrects the problem.

Also I solved the problem of the 'speaker pop' when the relay is switching on. (due to the output capacitors storing DC) I add resistor of 15ohm to the 'off' side to the relay, solved.

I will finalise this project by building this amp in a metal case. I will use it in my living room now. Pictures will follow soon.

Frederik

555d V5 schematic.zip (97kb)

1 Attachment(s) 0 Comment(s)

frederik 11 years ago

I did some more tests and measurements to define output power and efficiency.
Please shoot if i make any mistakes :-)

1. POWER

I connected 4ohm 50W resistor to output and adjusted input signal to maximum output without visual clipping on scope. I got 15V peak to peak sinus on output, input signal is 1.6V peak to peak.

To calculate power:

P = U*I = (U*U)/R = (7.5V * 7.5V) / 4 = 14.06W peak
P rms = 0.707 * 14.06 = 9.94W rms

2. EFFICIENCY

I connected a current multimeter in the power supply line.
I disconnect input signal and short circuit inputs to ground.
Both speakers are connected; I measure 140mA current.

power use without any signal = 24V * 0.140A = 3.36W (good I think!)

I connect L and R 1kHz signal and adjust to max power output without viisible distortion. I measure 980mA

power use with L and R signal to full power = 24V * 0.980A = 23.52W

Efficiency = ((Prms left + Prms right) / P powersupply ) * 100
= ((9.94W + 9.94W) / 23.52W) * 100 = 84.52%

If my calculations are correct, Im a happy man!

0 Comment(s)

frederik 11 years ago

Hi Ton,

Attached you'll find the new pcb, version 10.

Hopefully this will solve the oscillation problem. (i can't test this myself, as I can't re order it :-) )

grtz,

Frederik

555d V10.zip (53kb)

frederik 11 years ago

Here's the list; Weerstanden: R1, R15: 100K 1/4w R2, R17: 15K 1/4W R3, R16: 82K 1/4W R4, R8, R19, R22: 1K 1/4W R5, R18: 10K 1/4W R6, R20: 680R 1/4W R7, R21, R29: 2K2 1/4W R9, R23: 18K 1/4W R10, R24: 470R 1/4W R11, R25: 100R 1/4W R12, R26: 330R 1/4W R13, R27: 8R2 1W (1565416) R14, R28: 2K7 1/4W R30: 56K 1/4W P1, P2: 2K multiturn trimmer (9353232) Condensatoren: C1, C15: 2,2µF MKT (107422) C2, C17: 330pF ceramic (9411763) C3, C4, C5, C6, C8, C10, C16, C18, C19, C20, C22, C24, C30, C32: 100nF ceramic (9411887) C7, C9, C21, C23: 1nF MKT (1685470) C11, C25, C29: 1000µF 50V elco (9693009) C12, C26: 680nF MKT (1166042) C13, C27: 220nF MKT (1166039) C14, C28: 2200µF 50V elco (9693017) C31: 10µF 63V elco (9451480) C33: 47µF 63V elco (9451501) Spoelen: L1,L2,L3: 100µH 5A (2215971) Halfgeleiders T1, T3, T10, T12, T19: BC547C (1467871) T2, T4, T11, T13: BC557C (2317584) T5, T14: BD139 (1015770) T6, T15: BD140 (1574351) T7, T16: 2N2222 (9206884) T8, T9, T17,T18: IRF530 (8648263) LED1, LED2, LED3: standaard rode led (1208853) D1, D2, D3, D4, D5, D6, D7, D8: 1N5819 (9801219) D9: 1N4148 (1612346) ZD1: zenerdiode 15V (1612374) IC1, IC2: NE555 (1467742) of ICL555 (1103036) IC3: 7815 (1087092) Diverse RL1: Dubbelpolig Relais 8A 24V Stroom: 27mA 900ohm DPDT http://www.gotron.be/schakelmateriaa/relais/printrelais/dubbelpolig-relais-8a-24v-stroom-27ma-900w-dpdt.html Powersupply: Meanwell 100W 24V http://www.gotron.be/industriele-voeding-meanwell-100w-24v.html Heatsinks are recycled from an old PSU…. Must be replaced with commercial ones!

TonGiesberts 11 years ago

Splitting the input signal ground planes will create a ground loop when an external stereo source is connected. Better would have been if the input circuits around T1/T2/IC1 on one side and T10/T11/IC2 had been placed closer together. Now the ground planes of the two inputs are split by the power supply connection and it will cause interference. Also the driver stages T5/T6 and T14/T15 must be placed as close to the output stages as possible to reduce signal loops. All traces with larger (switching) currents should be as short as possible. This will be difficult to realize in the present design. So a compromise would be to move the screw terminal block more inward and move the input circuits more toward each other and the drivers more toward the output stages. We can change the PCB design and build a prototype. That will give us also the possibility to do or own measurements. In any case we would like a detailed BOM of all parts including exact manufacturer numbers and distributor order codes (don’t forget the heat sinks for IC5 and the output stages).

1 Attachment(s) 2 Comment(s)

TonGiesberts 11 years ago

A few things can cause the oscillation. Better is to put driver stages T5/T5 and T14/T15 in the same ground plane as their respective output stages (see graphically changed PCB). This way the charge and discharge currents from the driver to the output stage have the shortest return path. The current from the 555’s to the drivers is considerably less. Also the screw terminal block for the power supply should not be in the input signal grounds. Better is to place is it as close to C29 as possible and only from there a thick connection to the input ground plane is to be made. At C29 the two output ground planes should also come together, not by way of the signal ground plane.

555d-v9-pcb.jpg (460kb)

1 Attachment(s) 0 Comment(s)

frederik 11 years ago

I built a stereo version of the and i must say i'm impressed of the sound quality! I like amp without any feedback circuits, it's sounds very direct and tight....

I don't use the 555 buffer any longer, beause I had the same timing problem as mentioned before. I changed it to BD139/BD140 totempole buffer. This is working fine on the PCB.

About the PCB, I made some small improvements, some hole were not right sized, I forgot one track. Fixed now.

I think my goal is reached: Class D amplifier based on a NE555 and only basic common components used. I'm happy all other semiconductors are discrete! :-) I'm planning to build this amp for living room use.

Right now I use the CMOS version of the 555 (ICL555) what gives a more clean square wave. I don't hear any difference with the NE555,I think we can use this one anyway.

Possible improvements: relais output; when the output switches on, I get a 'plop' in the speaker, maybe the output resistors (2K7) should be smaller?

When I adjust potmeter so I get 50% duty cycle, I get a small oscillation around 15kHz. As soon I change dury cycle to 55% problem is solved.... Not shure how I can fix this, maybe the Elektor engineers know hw to sort this? :-)

Please feel free to post your reactions!

I have add pictures, schematic, pcb and a small movie of the amp at work!

dsc00557.JPG (450kb)
dsc00559.JPG (520kb)
dsc00561.JPG (432kb)
555d V4 - Schematic and pcb.zip (150kb)
movie version4.zip (3870kb)

5 Attachment(s) 0 Comment(s)

frederik 11 years ago

update with the missing link added !

555d V8 pcb.zip (138kb)
555d V3 sch.zip (96kb)

2 Attachment(s) 0 Comment(s)

frederik 11 years ago

update with the missing link added !

555d V8 pcb.zip (138kb)
555d V3 sch.zip (96kb)

2 Attachment(s) 0 Comment(s)

frederik 11 years ago

Attached the final schematic ( I think) and a PCB for a stereo version.

130144.zip (84kb)
130144 schematics and pcb designspark.zip (125kb)

frederik 11 years ago

Hello, Tnx for the remark, indeed there's a link missing ! Ill make the changes tonight and will update schematic and PCB ! I use LT spice for simulation and Designspark for schematics and PCB. Best regards, Frederik

JD 11 years ago

Hi Frederik, Nice project, I inspected your schematics and noticed still a missing connection. The connection between P1 (Tr3 and Tr4 base , emitter) and the connection between TRIG and THRS which is in your spice-schematics is missing in the new schematic. Further more I should like to know what programs do you use for simulating and PCB design, so I might have a look at those files too. I surely intend to build your design and do some experiments of my own based on this design, because its relative simplicity. Thanks for your work.

2 Attachment(s) 2 Comment(s)

frederik 11 years ago

Hi,

I made some small modifications; the current source around T2 is changed now. This should be more temparature stable.

I also add a speaker protection, speaker turn on 2 secs after power on and switches off directly after power off.

I'm planning to draw a pcb of a stereo version with protection on it.

Schematic attached!

Any suggestions, feel free to reply !

Frederik

555 38 ok.pdf (34kb)
555d V2 - Schematic.pdf (89kb)

2 Attachment(s) 0 Comment(s)

frederik 11 years ago

Attached you'll find the new version, this time again with the extra 555 as buffer.

I increased the timing c to 330pF what brings the oscillation frequency back to 250kHz. (It was 360 kHz with 220pF)

This way i don't push the 555 to its freq limits.

All is working fine now, no oscillations and a good sound again.

I'm off for a week holiday now, I'll start designing PCB when im back.
Any remarks or suggestions are of course very welcome!

grtz,

Frederik

555 34 ok.pdf (31kb)

1 Attachment(s) 0 Comment(s)

frederik 11 years ago

I did some more tests; and made some changes.

The 555 buffer uses the reset pin as input. This gives some timing problems, input and output are not the same....
I changed this by using BD139/140 pair as totem pole buffer. Solved.

By doing this, the amp gives a little oscillation around 15kHz, music sounds good, but i get sometimes oscillation effect.

Maybe somebody can give some advice where to add a little C or R? :-)
I use plenty of decoupling caps so this can not be the problem. Maybe I need to take VCC at 15V instead of 18V for the totempole?

About the feedback circuit I did some tests too. Purely by listening to music, I think the feedback circuit is killing a lot of 'dynamics' and 'fastness' of the amplifier. To me with the feedback, it sounds like youre driving a car with the handbrake on.

So I removed the feedback, and the buffer 555 and now i think we have a good sounding amp. (Removing feedback circuit did not affect oscillations at all!)
As i can't measure any distortion, I know this is very subjective, something I have build myself allways has something magic and emotional. I'm very curious about the facts and figures of my amp :-)

About power, I measured 100mA current with no signal on input. This raised to about 250mA with 1kHz at clipping level.

I'm planning to build this amp in stereo (for my living room). I'll draw a pcb for this.

I'm curious if some engineers @ Elektor are interested in this project, or maybe can give some further advice?

Regards,

Frederik

dsc09498.jpg (299kb)
555 32 nfb ok.pdf (30kb)

2 Attachment(s) 0 Comment(s)

frederik 11 years ago

Hello,

Version4 is ready and.... working!
Great sound, no heat at all !

I add an extra 555 used as high speed buffer to drive the charge pump and mosfets. The pwm 555 was acting strange if I use it for driver too. Probably the load was affecting proper working of it....

The buffer is made by connecting input to 'reset' pin, this way output follows input. (amazing what a 555 all can do, isn't it .... i start believing you can make a microwave oven with it too if you like :-) )

I have 24V pwm on the output, mosfets don't get to much drive voltage, charge pump is not running hot.

I changed powersupply to a proper 24V Mean Well SMPS. I add 7815 to get 555 Vcc.

And most importantly; i get good sound at a proper level for (my) living room use. (about 15W @ 4 ohm)

I add some more images incl scope image of output before filter, and i'll post a movie of the amp @ work soon.

UPDATE: movie attached! (Arno - let's go to heaven )

Frederik

dsc09482.JPG (250kb)
dsc09483.JPG (254kb)
dsc09484.JPG (285kb)
dsc09485.JPG (349kb)
dsc09486.JPG (276kb)
555 31 ok.pdf (31kb)
version4.zip (3250kb)

frederik 12 years ago

Luc, Thank you so much for all the support in this project! You're my mentor here :-) I'm still working on improving the circuit, see version 5. Grtz, Frederik

lucdm 12 years ago

Hi Frederik. Now THAT is terrific good news! I am so glad that your efforts payed off in the end... And the power is exactly as predicted too. I calculated it to be 8 watts but that is because you have a load of 8 ohms in the schematic. That is exactly right with the 15 watts at 4 ohms you are getting. I would be interested to know how much current it draws at full power on the 24 V (Icc?) How is the distortion ? OK now ? I'm glad to have been able to help you along the way! Happy testing. Regards Luc

7 Attachment(s) 2 Comment(s)

Baule 11 years ago

Hi Frederik !

You are making at least 2 fundamental mistakes:

You are Steering the MOSFet Output-Stage with +-24Volts although the Gate-Voltage of theese Devices are only specified at +-20Volt maximun Value. You can be lucky, that the MOSFets are still working, because the +-4Volt Overvoltage hasn`t destroyed the MOSFets yet!!!!

The other minor Mistake is, that you are recoupling at the input of the Mosfets. This is not where the Distortions expire. You should recouple at the output of the Mosfets(Input to L1), because this is the direct spot, where the Distortion expires. So in order to minimize harmonic Distortions, You should to connect your discrete Integrator-Device to this point!!!

Beside this your steering-Stage has no Dead-Time Protect on the MOSFets. So the MOSfets by Switching cascade with full short-cuircuit Current shortly on each other, which may destroy these Devices sooner or later. Furthermore you can minimize the quiescent Current and maximize Efficency by implement a Dead-Time Control!!!

I myself has worked on a switching Mosfet-Amplifier for Battery-Operation within several Years and now my prototype is working well within 3 up to 18 Volts Operation Voltage, 7,83mA quiescent Current at 9Volts and is able to get at least 9Watts into 4Ohms with an efficency of 94%!!!
The heart of my Version is also the 'allmighty' 555(as ICM7555) combined with an Operational-Amplifier IC as Interagtor(LT1351).
So if You want to become a little more 'professional', don`t hesitate to contact me!
I will then post my Concept onto this Side with a fully explanation!!!!
You can contact me using paul.wernicke@freenet.de....

Your`s sincerely

Baule

PS: If you can understand german or even are a german, we should communicate in german, which is making the whole Issue much easier to me.

Switched Amplifier_MOSFet.pdf (57kb)

KENT 12 years ago

Frederic, To me the simplicity of your design is the great attraction of your amplifier - even it this means that the efficiency is not optimum. I hoop that you will keep working towards developing your current design, with NE555 at the hart of the circuit and as few as possible addtional components. regards,

frederik 12 years ago

Hello Paul, Thank you for the suggestions! My initial idea is to construct a simple small class D amp with common components. around the NE555. (goal is now to get about 15W into 8ohm) About your remarks: - I measured the Ugs on both mosfets, i get max 10V so this should be fine i think - The feedback loop is now correct in the last version (555 30ok.pdf) - About short circuit pulses, I can limit this by proper selection of gate resistors and shottky diode. I get now spikes of 8A max in Spice. I'm building version 30 now so i'll post my findings soon! regards, Frederik

1 Attachment(s) 2 Comment(s)

frederik 11 years ago

I use 2x N chanel mosfet now.

High side fet is switched with a 'charge pump' as this one needs higher gate voltage than Vcc to conduct completely.

This way, we get 24V square wave on output with 15V pwm on the gates. NE555 is no longer pushed to its Vcc limits.

I spice, it looks promising, but again I need to build it and test.

Hopefully this time without burned fingers :-)

frederik

555 29 ok.pdf (28kb)
555 30 ok.pdf (29kb)

lucdm 12 years ago

Hi Frederik, about R4, true... with the low power output from this amp (calculated to be approx 8 Watts RMS), this resistor will dissipate approx, 0,25 Watts RMS using a test tone of 20 kHz at full power). You should be OK with a 0,5 or 1 Watt resistor. The 0,5W could get fairly hot under test conditions but should indeed be OK., the 1W is more than enough. I had much higher amplifier output levels in mind. Usually class-D amps are 100-200 watts (but then they use + and - 50 volts Vcc or so) and then it becomes a different story. The maths correct me here. Strange that the LPF adds distortion.. Something to be measured I guess... Success building this circuit. Luc

frederik 12 years ago

Hi Luc, i forgot output cap indeed :-) The output resistor has a 220n cap in series, so no need for high power one. In my new schematic, I have add feedback again. I have a feeling when i add C to it (LPF filter 1nF + 8K2) I get more distortion. Maybe I do something wrong there. I left the cap out for now. Anyway, current spice setup works, let's hope the real world will follow soon. soon more news! frederik

lucdm 12 years ago

Hi Frederik, Nice to see you try the original design again. I think this is an easier way to get to your goal. Now you can increase the voltage in the output section for more power. This circuit might be OK (almost). Because you may not burn your fingers, but your speakers won't survive very long, I think. You use a single-ended power supply and a normal (not bridged) output section again in this version-3, so it appears to me you have forgotten to put the output capacitor 2200 uF (used to be C5 in the previous version) in this new circuit. Also R4 at the output RC circuit will have to be a very high watt (say 5 or 10W) as it will get a lot of power to dissipate being only 8 ohm! OR you may still get your fingers burned ;-) I am curious to hear about the results... Regards, Luc

2 Attachment(s) 3 Comment(s)

frederik 11 years ago

Hello,

I played around in SPICE with a full bridge version.

I used 2 extra NE555's to invert the PWM signal (bottom one) and buffer (top one) This way I get enough power to drive 4 mosfets.

With 2 extra 555's timings between buffered and inverted signals are ok.

I skipped the feedback loop, to make things easier at this design stage.

What I still miss: charge pump at the high side output to create the high gate voltage....

Any suggestions, comments or idea's to improve are welcome!

frederik

555 25 ok.pdf (31kb)

lucdm 12 years ago

hi frederik, Just wondering how you guarantee dead time using this circuit. I see you drive one leg from the top 555 and the other from the bottom one. I haven't checked the timing but i hope you have checked that it is absolutely impossible that both legs are 'on' because that will be a very nice short circuit. If the second one is triggered by the first then that will work on the leading edge but not on the trailing edge. Then the bottom one is still conducting, while you already switch on the top one again. You must build in guaranteed dead time eitherway. And this must also be as short as possible as a long dead time causes lots of distortion. In good designs, this is 10 to 15 nanoseconds. Please also note that Fairchild specifies max power voltage of the 555 to be only 16 volts! regards Luc

1 Attachment(s) 1 Comment(s)

frederik 11 years ago

Hello,

I have a problem with version2.

I simulate it in LTSPICE and it seems to work fine.

555 oscillates and outputs square wave of 15V.

With the voltage level shifter, i can drive the 2 mosfets in 24V.

Current trough 2N2222 and BS170 seem to be within limits.....

I tried to build it in the real world and.... 2N2222 is heating up and smoking after a minute, burning base emitter junction.....

Can anyone advise me on what's wrong ? Is there a mistake in the voltage level shifter schematic? It would be great to create a class D amp without any IC exept the 555, so I keep on searching... but if anyone can advise it would be great!

I attached the spice files and special .lib files

tnx a lot for your help!

LTspiceIV.zip (3280kb)

frederik 12 years ago

Hi Luc, Thank you so much for the advice, it pushes me to go for version 3 :-) I'm about to do some tests with the IR2110 driver IC. I have to give up N-chanel / P-chanel mosfets too I think. I'll keep on posting how it gets. thank's again. frederik

lucdm 12 years ago

Yes, I see what you mean... I just edited my reply to explain the added phase shift when taking it from the output behind L1... that is not the best of ideas.. Better to take it where you did but then must add a C so you get an RC filter I think. Your shifter was underrated and Q6 still a question mark. But read this link that I added. https://www.coldamp.com/store/medi/pdf/Class_D_audio_amplifiers_White_Paper_en.pdf It contains maybe good info for you to start again with a better design. I'll keep watching this space. Your version 1 looks promising, but you would need to improve the version-2. From what I can see, this should be do-able... Regards Luc

frederik 12 years ago

Hello Luc, Reason for that is that the output is inverted due to the level shifting. In my tests, I disconected the feedback (to make things more easy to troubleshoot) Anyway, version 2 is in the bin now because it's not a good design :-) All I got was heat and smoke, and a burned finger ..... and no sound at all. In my first version, I fed back the non filtered output signal too. When i feed back the filtered signal, I got more (visible) distortion in SPICE due to phase errors (I think). The square wave works fine, I think because the input signal is directly converted to a triangle/square wave at the collector of Q4. Anyway, this version is working very well (ice cold and good sounding), only the low power output should be better..... still looking for that. grtz, Frederik

lucdm 12 years ago

Hi Frederik, I do have another question, going over your version-2 schematic. On the input side you seem to have a fairly normal long-tail pair analog differential input. What I don't get is that your feedback from the output is the 300khz switched drive signal (coming from the exit of the level shifter emitter-Q6 through R14). If you pick up the feedback here, you would want an RC low pass filter to filter out the switching component from the feedback. I would not want to feed the 300khz chopped squarewave into the feedback (I think)... I would expect you to pick up the analog feedback at the output of the amplifier at the real output after L1 being at point L1-C4-C5 because that is the filtered analog output signal which then makes sense to feed back into the analog differential amp at the input. You need to watch out for the phase shift introduced by L1 then though. Have a look at this document and especially the page regarding "Feedback". https://www.coldamp.com/store/media/pdf/Class_D_audio_amplifiers_White_Paper_en.pdf Luc

lucdm 12 years ago

Hi Frederik I am interested in this project but dont have distortion measuring equipment either so I cannot help you there. Regarding the version 2 circuit, I THINK your problem is definitely in the level shifter. First of all, I do not understand the design around Q6... why have you shorted the base to the emitter? It looks to me that this circuit cannot work like this... I would expect at least a resistor between point R17-Q6-base and R18-D5.... Maybe you wanted to put the 1n5819 there but the schottky-diode does not have enough Vf to allow the 2n2222 to conduct as it requires at least 0.6 V Ube. So unless I do not understand the design, you may want to look at this first... Furthermore, the dimensioning of the components should be better, I think Look at Q6, D5, R18, M3 : voltage drops are Usat Q6 (no other resistor in the Q6 section) + Ufwd D5 + Usat M3 This means there is (I guess) about 20 Volts over R18 which is only 22 ohm according the circuit: I = U /R = 20V / 22 ohm = approx 0,9 - 1 AMP... much more than the 2n2222 can support (= 800 mA PEAK but depending on the type and manufacturer sometimes only 600 mA) So it looks like you continuously run the 2n2222 far over or too close to maximum specs. You may also consider that the 2n2222 is a small-signal transistor. It is not very well suited to switch high loads at high rates... Perhaps you should use a switching-type transistor. This also benefits the distortion of D-class amplifiers that benefit from very fast switching times. The faster the feedback loop is, the less distortion you should get. The same goes basically for M3, The BS170 has a max continuous Id of 500 ma. It says also 1200 mA pulsed, but that is not at 300kHz continuous 50% on/off pulse rate. And furthermore look at the specs at http://www.fairchildsemi.com/ds/BS/BS170.pdf and look at the Safe Operating Area specs on figure 13... If you draw a load line from 1 Amp (Id) to 20V (Uds) then you will come very close to SOA max (where Id = 0,5 and Vds = 2 V). This means every time the BS170 is conducting, you go very near max SOA. That is a problem waiting to happen. I think that "as designed" the reliability of the level shifter will be very poor. In my opinion, you must allow for a much wider safe operating design... Do not run your components over (say) 50 percent of their maximum capacity. Be generous. These standard components are cheap anyway. Regards, Luc

1 Attachment(s) 5 Comment(s)

Thilo Opaterny 11 years ago

Hi Frederik,

once I used a 555 as heart of a power supply. Like you I controlled the charge/discharge currents to get the feedback for the regulation. First I though it might be necessary to have an exact linearity, but after some testing I learned that this does not matter at all. The second thing I figured out that it worked like an amp! This was more accidentially. When I had solded everything together I connected an oscilloscope at the capacitor at the end to see the ripple at different loads. (I developed a power supply :-) Now my finger touched the wire to TRIG and instead of a ripple I saw a perfect sinus. When saw your schematic I remembered it.

My 'left' side was much easier. I don't think that is necessary to have a real triangle wave. Ok, I know the standard application of the 555 connects a resistor to VCC and a second one to DIS. But if you want to have a 50% duty cycle, then connect just one resistor to OUT. You can see this trick in Elektor July/August Page 89 in the german version. The scematic has the number 120141-11.

Starting with the idea that we have a 50% duty cycle. And we have only one resistor between OUT and TRIG and C3 down to ground. Now add a second resistor of round about the same value (and like your C1 a capacitor to get rid of the DC) and couple in an analog signal over this resistor. As long as the input is high more current will load C3. This means that the output will be longer low. When the input is low we additionally unload C3 and the output will be longer high.

Do we need a perfect triangle? With the proposed OUT -- TRIG resistor we get an expotential curve on both sides. Since loading and unloading C3 is symmetric, the time we are too early during load we are too early during unload. The on time should be the same as if you try to do this with a perfect triangle. I do not think that you can measure a kind of phase shift.

May be someone could model this with spice -- I'm a bit old school for this.

frederik 12 years ago

Hello, Thank's for the comment and advice. About the linearity, I'm not a big theory man, so I did some tests on my first attemps to build a class d amp around a 555. I connected a DC voltage to control the pulse width in a constant voltage charging setup. Here I can see clearly a "compression effect": higher voltages give less change in the pulse width than small voltages. This acts exactly like the exponential curve of a capacitor charged with a constant voltage! Therefore, I chose for the constant current charging settup. regards, Frederik

frederik 12 years ago

Hi, About the triangle wave over the charge capacitor, i think it is important to charge with constant current to avoid distortion. I did some tests ( my early experiments on this project) and these are my findings: Instead of using an AC signal, i did the test with a DC voltage, changing the pulse width. There it was obvious that changes in the DC level did nog change pulse width in a linear way. It was behaving "exponentional" exactly as the capacitor is charging with a constant voltage ! I would be great is someone @ Elektor could measure distortion levels, I'm very curious how good we can get this :-) grtz, f

2 Comment(s)

frederik 11 years ago

I found a way to create a voltage level shifter between the 555 and the mosfets!

I'm happy it's a discrete version, so the only 'IC' is the almighty 555 ! :-)

I tested the circuit in SPICE and all is working, I use 24V powersupply and I can get about 15W power in 4ohm. This week I'll build it on the prototype, more news and images soon.

Attached the schematic.

555 20ok fb 24V.pdf (31kb)

1 Attachment(s) 0 Comment(s)

555 Class D Power Amplifier [130144-I]

Project Elements

Discussion (23 comments)

Bobbylebob 11 years ago

NuttyProf 11 years ago

frederik 11 years ago

TonGiesberts 11 years ago

frederik 11 years ago

TonGiesberts 11 years ago

frederik 11 years ago

frederik 11 years ago

frederik 11 years ago

frederik 11 years ago

frederik 11 years ago

frederik 11 years ago

TonGiesberts 11 years ago

TonGiesberts 11 years ago

frederik 11 years ago

frederik 11 years ago

frederik 11 years ago

frederik 11 years ago

frederik 11 years ago

JD 11 years ago

frederik 11 years ago

frederik 11 years ago

frederik 11 years ago

frederik 11 years ago

frederik 12 years ago

lucdm 12 years ago

Baule 11 years ago

KENT 12 years ago

frederik 12 years ago

frederik 11 years ago

lucdm 12 years ago

frederik 12 years ago

lucdm 12 years ago

frederik 11 years ago

lucdm 12 years ago

frederik 11 years ago

frederik 12 years ago

lucdm 12 years ago

frederik 12 years ago

lucdm 12 years ago

lucdm 12 years ago

Thilo Opaterny 11 years ago

frederik 12 years ago

frederik 12 years ago

frederik 11 years ago

Embed Code