I made a circuit based on the TDA7294, which is the most basic circuit given in the data sheets (Typical application and test circuit, fig.1). Most of the circuits I've seen, including a more elaborate circuit given in the same data sheets, include a parallel RL circuit in series with the loudspeaker. My question is: can I harm the amplifier if I obviate this network?
Fit input filters.
A low pass to attenuate RF interference.
A high pass to block DC from faulty Sources.
Build an AC coupled amplifier to prevent damage to your headphones.
Read the datasheet and determine whether the "typical" of fig1 is actually the recommended assembly to avoid stability problems.
A low pass to attenuate RF interference.
A high pass to block DC from faulty Sources.
Build an AC coupled amplifier to prevent damage to your headphones.
Read the datasheet and determine whether the "typical" of fig1 is actually the recommended assembly to avoid stability problems.
Thanks for your reply. The circuit in fig.1 says "Typical application and test circuit". Do you think that the "test" part in the title implies I can safely use it without adding an RL parallel network in series with the speaker?
The test and application circuits are not intended to be complete designs suitable for foolproof operation. It almost certainly works as shown (with a single speaker directly coupled to the board, probably via a couple of inches of cable) but as shown isn't how it is used in practice. Unless you can prove (and you should do this to any design) that it is stable under all possible conditions, such as with a 1nf, then a 10nf and then a 100nf load, then you should assume the worst and add the output inductor.
Saying the design works OK with your speakers and leads is not the same as saying it will work with any length and type of lead and speaker.
Saying the design works OK with your speakers and leads is not the same as saying it will work with any length and type of lead and speaker.
OK, but again my concern is with the amplifier safety. I mean, can it be damaged because of the lack of the inductor? Or perhaps this is implied in the word "stability"? I only want to make a little test, just to see if everything is properly connected. The inductor, if mandatory, I can always added later.
If a problem does exist then it could be anywhere in the range of a slight rise in distortion, through to the chip running warm because of HF instability through to it exploding violently. Any of those scenarios is equally possible.
I would test it with a limited supply first to avoid damage. It should drive all three of those cap values I mentioned without complaint. If it doesn't, add the inductor.
For the sake of one 1W 10 ohm resistor with some enamelled wire wound around it, is it work risking omitting it ?
The main purpose of the RL network is to prevent a speaker wire becoming an antenna, and injecting RF into the amplifier. This can cause oscillation and damage. Pretty much every commercial amp includes these components.
The main purpose of the RL network is to prevent a speaker wire becoming an antenna, and injecting RF into the amplifier. This can cause oscillation and damage. Pretty much every commercial amp includes these components.
No the RL network is to make shure the amp is shielded from a a capacitive load. Some cables or for that mater pizio tweeters have a lot of capacitence so at an extreemly high frequency this can cause an amp to have problems.
You have convinced me. The datasheets show a more elaborate circuit using 270 ohms and 5uH. But this circuit can deliver 180W peak power to a 4-ohm load, where as my circuit delivers up to 43 W average power. Do the R and L values depend on the output power?
Not really. Just make sure the wire is thick enough to carry the current. The resistor across the coil is just to damp any ringing, mainly seen under adverse test conditions with fast square waves. Normally it sees little voltage/current and 10 ohm 2 watt would be typical. Don't forget any zobel network goes before the coil, not after.
Stability is not an easy thing to simulate or measure. The oscillations are at around 1MHz, at the most upper limit of the bandwidth.
Not only the load connected that causes problems, but also the feedback loop and bypass caps. I am convinced that if you have a very good PCB layout there is far less need for the output networks.
Not only the load connected that causes problems, but also the feedback loop and bypass caps. I am convinced that if you have a very good PCB layout there is far less need for the output networks.
Well, I have separated the signal ground from the power ground, making the input be the point of union. But mine is not a PCB but a "universal printed circuit board". A rectangular array of copper islands where I soldered the components. All leads are wire using the shortest, save for the ground, possible path. So, the odds are this is not a very good physical implementation.
It sound like the original gaincard prototype.
I remember seeing a P2P LM3886 where the leads of the bypass caps had been extended by 10 cm and he was complaining that it was all getting very hot. I don't think a output filter is going to help here.
Low inductive connections, especially the GND connections and removing the HF at the input should all help to improve stability. You can also solder the feedback resistor directly to the chip pins.
I remember seeing a P2P LM3886 where the leads of the bypass caps had been extended by 10 cm and he was complaining that it was all getting very hot. I don't think a output filter is going to help here.
Low inductive connections, especially the GND connections and removing the HF at the input should all help to improve stability. You can also solder the feedback resistor directly to the chip pins.
- Status
- Not open for further replies.