I resized your image to 40x30. There is no way i can etch that.
The closest gap i can make is approx the space between Rb-Rg trace and Cin to Rb.
This means that the following traces are too close:
Intermediate jump island too close to Vp. Just get rid of it completely and make the jump wire curve, or make the jump wire travel under the chip.
The ring around the middle hole of Cs1 and Cs3.
Cs2 and Cs4. Just make the ring around these smaller, but still fat trace going to the jump area.
Make the Vp trace connecting the 2 pins on the chip a bit smaller. It needs more spacing between pin 3 and Vn line going to Rm
Need more spacing between pin 3 trace and Cf-Rf1. What about orienting Rf1 upright?
Move the connecting trace of the two star grounds higher. Its too close to the Out terminal.
And btw, Ci is 33 - 68uf non polarised electrolyic.
Good luck with the changes. Hope to etch this on monday!
The closest gap i can make is approx the space between Rb-Rg trace and Cin to Rb.
This means that the following traces are too close:
Intermediate jump island too close to Vp. Just get rid of it completely and make the jump wire curve, or make the jump wire travel under the chip.
The ring around the middle hole of Cs1 and Cs3.
Cs2 and Cs4. Just make the ring around these smaller, but still fat trace going to the jump area.
Make the Vp trace connecting the 2 pins on the chip a bit smaller. It needs more spacing between pin 3 and Vn line going to Rm
Need more spacing between pin 3 trace and Cf-Rf1. What about orienting Rf1 upright?
Move the connecting trace of the two star grounds higher. Its too close to the Out terminal.
And btw, Ci is 33 - 68uf non polarised electrolyic.
Good luck with the changes. Hope to etch this on monday!
Hi Hurtz
With every revision I'm liking it less and less.
1) there should be only ONE ground on the PCB - signal ground. You can make this happen by returning the chip amp power return from whence it came. The speaker currents are not needing to return at the IC, unless you have substantial capacitance there. All the chip amp does really is route the speaker power from the +/- DC supply rails ie the big caps. Wire the speaker return at the DC supply PCB. The signal ground "heaven" is located at the junction of the 0.1 uF decoupling caps, ideally if you used a ground plane layer or ground pour it would tie there.
2) Use priority routing for the IC supplies and decoupling. If the chip amp is not stable or modulated, every thing is down hill from there. as I suggest in my post above, wire jumpers, save them for signals NOT grounding! The jumpers tend to add small inductance.
3) Again the zobel needs to be tighty coupled to the chipamp ground. Its not related to the speaker at all, IT does NOT need a high power return . The input RF filter s/b designed such that it attenuates pre-amplified signals from reaching the zobel > 2MHz. returning it the PS just increases the loop area and this extra wire adds some series inductance, making it less effective at frequencies where it's needed.
With every revision I'm liking it less and less.
1) there should be only ONE ground on the PCB - signal ground. You can make this happen by returning the chip amp power return from whence it came. The speaker currents are not needing to return at the IC, unless you have substantial capacitance there. All the chip amp does really is route the speaker power from the +/- DC supply rails ie the big caps. Wire the speaker return at the DC supply PCB. The signal ground "heaven" is located at the junction of the 0.1 uF decoupling caps, ideally if you used a ground plane layer or ground pour it would tie there.
2) Use priority routing for the IC supplies and decoupling. If the chip amp is not stable or modulated, every thing is down hill from there. as I suggest in my post above, wire jumpers, save them for signals NOT grounding! The jumpers tend to add small inductance.
3) Again the zobel needs to be tighty coupled to the chipamp ground. Its not related to the speaker at all, IT does NOT need a high power return . The input RF filter s/b designed such that it attenuates pre-amplified signals from reaching the zobel > 2MHz. returning it the PS just increases the loop area and this extra wire adds some series inductance, making it less effective at frequencies where it's needed.
Here's what I'd recommend all beginners to start out with (see attachment). Notice that all grounds have separate conductors back to common (or "star" if you prefer) power supply ground. This will eliminate or at least minimize the possibility of noise and distortion caused by ground loops.
As for the proper value of feedback cap (C i), it depends on value of feedback resistor (R i) as determind by formula; 1/R*C*2*PI), and should be no more than 1/2 the frequency (much less is better) of input filter formed by C in and R in. In other words, if input filter has a corner frequency of 10Hz, the corner frequency of the feedback network should be no higher than 5Hz, but even lower is better. The reason for this is you don't want the amp to recieve frequencies that the feedback loop can't control.
Hope this helps.
Mike
As for the proper value of feedback cap (C i), it depends on value of feedback resistor (R i) as determind by formula; 1/R*C*2*PI), and should be no more than 1/2 the frequency (much less is better) of input filter formed by C in and R in. In other words, if input filter has a corner frequency of 10Hz, the corner frequency of the feedback network should be no higher than 5Hz, but even lower is better. The reason for this is you don't want the amp to recieve frequencies that the feedback loop can't control.
Hope this helps.
Mike
I've never had to separate the input ground from the IC supplies unless there is wonky system (PE earthing) ground loops. And then breaking it there by adding a 10-100 ohm resistor in series instead of routing it back to the main supply. Also the zobel R,C(SN) needs to be tightly linked to the IC (decoupling) return.
Last edited:
Andrew, I totally agree input should be RF filtered, at least five times higher than highest input frequency to avoid early roll-off, just an error of omission. However, since large power supply and output currents are not equal, I prefere to keep them separate, and adding one more wire is cheap and easy.
Mike
Mike
Happy to be corrected. You mention somewhere else about the LM4780 and they are both there on fig 5
Will Rm fit under the chip instead?. Looks like that could free up a lot more space.
Regards
Dave
Infinia, I'm affraid I must respectfully disagree, if you use a common ground conductor for small signal and high current paths, you will get elevated noise and distortion. A very good test for this is to listen to the amp output without any input signal (as in shorted input) through headphones. I do this with all my designs. When it's right, the most you will hear is barely audible thermal noise (high frequency "hiss") and nothing else, no buzz or anything else. Besides, as I said before, adding a wire or two is cheap and easy, and just plain-good design practice.
Mike
Mike
Maybe you haven't read my posts, I'm the one advocating separating high current and small signal grounds too. Decoupling using smallish HF caps to the signal ground works fine, done all the time, even on discrete audio amps. The high currents don't want to return through the Chip amp ground anyways. BTW when you short out the input there are no high currents, so yer testing for something else. Good power testing for noise, distortion, and inter-mods uses a spectrum analyzer with single and multi-tone inputs. I've debugged and solved many more noise and grounding issues than simple power op amp circuits. Been doing it professionally on mixed signal PCBs in communication systems for 30yrs.
Heck I've probably built one of the 1st audio power chip amps from Signetics way back in late 70's. Lessons learned the hard way, it oscillated due to poor decoupling. Ended up using it for a headphone amp, so it worked well and was very quiet.
Heck I've probably built one of the 1st audio power chip amps from Signetics way back in late 70's. Lessons learned the hard way, it oscillated due to poor decoupling. Ended up using it for a headphone amp, so it worked well and was very quiet.
Last edited:
Infinia, sorry if misunderstood, but when I read "I've never had to separate the input ground from the IC supplies unless there is wonky system (PE earthing) ground loops", I took it to mean you were advocating a single return path for grounds. And of course your right about testing at normal current draw, but in my experience, if it's not done right, you will get audible hum and buzz even with no input signal, a usefull test for beginners who typically lack decent test equipment.
Mike
Mike
It's all good, your diagram is useful. I just wouldn't call the chip amp ground "power ground", cause most of the power actually goes through the speaker return. I recommend that the input return should tie to center point of the HF decoupling caps 0.1 uF for the single PCB ground. Sometimes a problem arises when using in a system with 2 different boxes with two different PE's connections, then the noise or hum wants to travel between the two boxes on the input RCA cable shield. (ie ground loop) Then it's useful to use a small resistor in series on the input signal ground, thus forcing the bad hum or noise currents to take another route. Most times all this 50/60Hz hum can be avoided with using a common PE connection. Also the Zobel needs to be shown at the IC ground not at the speaker.
Last edited:
If the Out- were connected to the PS gnd star then all "issues" are pretty much solved.
I will test with the rohde & schwarz audio analyzer how much difference this actually makes. infinia, i dont understand what you are saying about the zobel connection. You say it should go to the chip amp gnd trace instead of its own line to the gnd? I view each ground point and line as a modulating voltage and dont understand why this is better. Afaik, this will make the chip amp gnd modulate more.
And guys, you keep saying bring everything to the PS gnd via their own paths. Lets pretend the PS gnd is the star ground on that board, since it pretty much is as the wire length is 2cm in my case. How should the traces go to this PS gnd?
The way it is now, signal gnd, chip gnd, zobel gnd and speaker ground all go their own path to PS gnd!
I will test with the rohde & schwarz audio analyzer how much difference this actually makes. infinia, i dont understand what you are saying about the zobel connection. You say it should go to the chip amp gnd trace instead of its own line to the gnd? I view each ground point and line as a modulating voltage and dont understand why this is better. Afaik, this will make the chip amp gnd modulate more.
And guys, you keep saying bring everything to the PS gnd via their own paths. Lets pretend the PS gnd is the star ground on that board, since it pretty much is as the wire length is 2cm in my case. How should the traces go to this PS gnd?
The way it is now, signal gnd, chip gnd, zobel gnd and speaker ground all go their own path to PS gnd!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.