It's at:
http://arx.ca/projects/BassAmp
Did I miss anything?
It's pretty much just a copy of the non-inverting part of the one in the app-note.
Don't worry about the lack of decoupling caps and invisible power connections. They'll be there, they're just not on the schematic.
How does the input buffer/inverter look. I have pretty limited experience with opamps, so I could have it completely wrong, or with incorrect resistors or something.
Is the TL072 a decent choice for that purpose?
-Nick
http://arx.ca/projects/BassAmp
Did I miss anything?
It's pretty much just a copy of the non-inverting part of the one in the app-note.
Don't worry about the lack of decoupling caps and invisible power connections. They'll be there, they're just not on the schematic.
How does the input buffer/inverter look. I have pretty limited experience with opamps, so I could have it completely wrong, or with incorrect resistors or something.
Is the TL072 a decent choice for that purpose?
-Nick
Hi,
the tl072 will do, but try to keep the traces to the chipamps short and spaced from their neighbours to keep capacitance low. TL does not like caps to ground on it's output.
This is a bass amp, but you have filtered the input with a 5Hz low pass and again filtered the NFB at the same frequency.
That will cut-off the bass to -2db @ 10Hz and the effects will be heard all the way up to 50Hz.
Aim for about 2Hz @ the input and 1 to 1.5Hz @ the NFB leg.
No RF filtering anywhere!
add an input resistor before Rin1 (1k0 would do)
add a cap to ground (330pf//47k5) giving about 0.3uS.
Add more filtering at each chipamp input.
At pin 10 add 15pF to ground (also 0.3uS).
U7b pin 5 needs 11k to ground, to balance the input resistances.
the tl072 will do, but try to keep the traces to the chipamps short and spaced from their neighbours to keep capacitance low. TL does not like caps to ground on it's output.
This is a bass amp, but you have filtered the input with a 5Hz low pass and again filtered the NFB at the same frequency.
That will cut-off the bass to -2db @ 10Hz and the effects will be heard all the way up to 50Hz.
Aim for about 2Hz @ the input and 1 to 1.5Hz @ the NFB leg.
No RF filtering anywhere!
add an input resistor before Rin1 (1k0 would do)
add a cap to ground (330pf//47k5) giving about 0.3uS.
Add more filtering at each chipamp input.
At pin 10 add 15pF to ground (also 0.3uS).
U7b pin 5 needs 11k to ground, to balance the input resistances.
Ok, the input I can easily change. I just chose that size of capacitor because I have a bunch on hand. I'll swap it for something bigger or just put 2 in parallel.
The feedback loop might be trickier. Those caps are already pretty hefty, so unless I swap them out and use servos, I think I'm stuck with what's there.
Also I should clarify that this is a "Bass Amp" as in bass guitar, and may be used with a ported speaker cabinet, so I may need to put in a fillter to roll off the very bottom end so the port won't get out of phase with the speaker. The 5Hz highpass in the feedback loop probably pales in comparison.
The alternative is to build the speaker cabinet sealed, and boost the low end for a little more bass extension.
Either way I think that the 5Hz feedback loop is probably going to matter a lot less by the time I'm done.
Am I wrong?
If you could give me a quick explanation on a couple of these it would be a big help.
add an input resistor before Rin1 (1k0 would do)
ok, done. What does this do?
add a cap to ground (330pf//47k5) giving about 0.3uS.
ok, this would be in parallel with the 47k5 to ground any high frequencies right?
done.
Add more filtering at each chipamp input.
At pin 10 add 15pF to ground (also 0.3uS).
Are both of these items the same thing, or do you mean even more filtering than that?
15pF caps added.
U7b pin 5 needs 11k to ground, to balance the input resistances.
Ok, done. What does this do? Is precision necessary or would a 10k be close enough?
New schematic will be up on the page in a few minutes. My schematic program has no export except for really bad pdf, so I need to take a couple screenshots and stich them 😛~~~
-Nick
The feedback loop might be trickier. Those caps are already pretty hefty, so unless I swap them out and use servos, I think I'm stuck with what's there.
Also I should clarify that this is a "Bass Amp" as in bass guitar, and may be used with a ported speaker cabinet, so I may need to put in a fillter to roll off the very bottom end so the port won't get out of phase with the speaker. The 5Hz highpass in the feedback loop probably pales in comparison.
The alternative is to build the speaker cabinet sealed, and boost the low end for a little more bass extension.
Either way I think that the 5Hz feedback loop is probably going to matter a lot less by the time I'm done.
Am I wrong?
If you could give me a quick explanation on a couple of these it would be a big help.
add an input resistor before Rin1 (1k0 would do)
ok, done. What does this do?
add a cap to ground (330pf//47k5) giving about 0.3uS.
ok, this would be in parallel with the 47k5 to ground any high frequencies right?
done.
Add more filtering at each chipamp input.
At pin 10 add 15pF to ground (also 0.3uS).
Are both of these items the same thing, or do you mean even more filtering than that?
15pF caps added.
U7b pin 5 needs 11k to ground, to balance the input resistances.
Ok, done. What does this do? Is precision necessary or would a 10k be close enough?
New schematic will be up on the page in a few minutes. My schematic program has no export except for really bad pdf, so I need to take a couple screenshots and stich them 😛~~~
-Nick
Nick, you still need to connect the output of the TL072s to the input nodes of the chips. Change the value of the resistors on the inputs of the 3886s from 20k5 to 1k. You'll then need to increase the value of the 15pF caps to 330pF to keep the cutoff frequency about the same. You may also want to add a 10k (or so) pull down resistor at the output of each TL072.
BWRX said:
The TL072s were connected at some point. I respaced everything to make it look a little less screwed up and accidentally lost those connections. They're back now.
Added some 10k pulldown resistors. (I'm not going to update the schematic on the website yet though)
Will switching to 1k have any other effects? I assumed that it had something to do with the DC blocking cap, since on the appnote it's 1k with the servos, and 20.5k with the caps.
I've updated the schematic, BTW. It's much uglier now. I've added in the servos. I'm not planning to use them right away, but it's probably easier if I put them on the layout anyways, then I can switch easily if I want to.
Anything else needed, or should I start on the PCB layout?
-Nick
Hi Arx,
Bass amp = bass guitar amp. Oh. I thought wide band music amp but dedicated to bass duty.
5Hz on the input is probably OK. You could probably get away with raising this a half octave and still be OK. Keep the NFB at 5Hz.
Single cone drivers will probably also be OK for a bass guitar amp.
Do you intend wiring your multiple drivers to 4ohm or to 8ohm or some other value?
Play safe with heatsink dissipation capacity.
Do keep in mind that if you play hard and loud ALL six amps will be drawing the same power at the same time.
The PSU must be designed to meet this combined and potentially heavy load.
A non regulated PSU is designed for average conditions and peak current ability comes from the smoothing caps.
An SMPS with relatively low value post reg smoothing has to meet the peak current from within it's circuitry. This is a very demanding load and a significant difference in supply philosophy that is often overlooked.
Bass amp = bass guitar amp. Oh. I thought wide band music amp but dedicated to bass duty.
5Hz on the input is probably OK. You could probably get away with raising this a half octave and still be OK. Keep the NFB at 5Hz.
Single cone drivers will probably also be OK for a bass guitar amp.
Do you intend wiring your multiple drivers to 4ohm or to 8ohm or some other value?
Play safe with heatsink dissipation capacity.
Do keep in mind that if you play hard and loud ALL six amps will be drawing the same power at the same time.
The PSU must be designed to meet this combined and potentially heavy load.
A non regulated PSU is designed for average conditions and peak current ability comes from the smoothing caps.
An SMPS with relatively low value post reg smoothing has to meet the peak current from within it's circuitry. This is a very demanding load and a significant difference in supply philosophy that is often overlooked.
Hi Bwrx,
why change the chipamp input Rs from 20k5 to 1k0?
Are you taking advantage of the low input offset/bias currents in the chipamp, or for some other reason?
why change the chipamp input Rs from 20k5 to 1k0?
Are you taking advantage of the low input offset/bias currents in the chipamp, or for some other reason?
Hi,
the DC servo shows a 20k5 resistor feed back into the inverting input.
Is this value a bit low?
The ratio Rs5 to Rf5 will allow half the opamp rail voltage correction the be applied to cancell output offset. Will this chipamp ever need such gross correction?
Would 100k do here?
Is there sufficient advantage in splitting this resistor in two and filtering the junction with a cap to ground to make the extra complexity worthwhile?
the DC servo shows a 20k5 resistor feed back into the inverting input.
Is this value a bit low?
The ratio Rs5 to Rf5 will allow half the opamp rail voltage correction the be applied to cancell output offset. Will this chipamp ever need such gross correction?
Would 100k do here?
Is there sufficient advantage in splitting this resistor in two and filtering the junction with a cap to ground to make the extra complexity worthwhile?
AndrewT said:
It's not necessary to use a high resistance in series with that input. Since he's already using 1k resistors elsewhere, why not? All you need to do is increase the value of the capacitor a bit to keep the same cutoff frequency. For that matter, he could also use 20k5 resistors instead of the 22k resistors on the mute pin.
Hi,BWRX said:
I don't understand this reply.
It seems odd to change to another value simply to make it the same as other resistors in the circuit.
That resistance is only there for filtering the input. The filter inputs are driven by the TL072 and the non-inverting input is already a high impedance so there's no point in using a high value resistor there unless it is because you want to use a very small value cap. As I mentioned, lowering that resistor to 1k (I said 1k because it's a low value and he's already using 1k elsewhere) means the capacitance only needs to be increased to 330pF (still quite small) to keep the same cutoff frequency.
Hi Bwrx,
the chipamps' pin9 are 20k5.
Assuming worst case input offset current of 0.2uA gives an output offset ~=4mV (@Tc=25degC). If the 6chipamps do not have DC servos fitted then these offsets will crossfeed each other or bias the speaker. Any idea how much real input offset currents vary with temperature?
It would be better to retain the 20k5 on pin10 and maintain the output offset at considerably lower levels.
Even with DC servos fitted it is better to balance out the offsets without the servo and then set up the servo to just cope with the varying offset that results from changing operating conditions.
That goes back to my question in post9
the chipamps' pin9 are 20k5.
Assuming worst case input offset current of 0.2uA gives an output offset ~=4mV (@Tc=25degC). If the 6chipamps do not have DC servos fitted then these offsets will crossfeed each other or bias the speaker. Any idea how much real input offset currents vary with temperature?
It would be better to retain the 20k5 on pin10 and maintain the output offset at considerably lower levels.
Even with DC servos fitted it is better to balance out the offsets without the servo and then set up the servo to just cope with the varying offset that results from changing operating conditions.
That goes back to my question in post9
AndrewT said:
Ok, I like that line of thinking. I'll change it back to the 20k5 unless anyone can suggest a good reason not to.
I am going for servos afterall and removed the caps as the schematic I've just uploaded shows.
The space I have for the PCB wouldn't allow both, and the caps are huge (32mm x 50mm)
Now that I understand how it works a little bit better I think servos make more sense anyways, especially since I'm using 0.07ohm resistors on the output.
So, no more glaring issues right? I'm going to start laying out the board now.
Oh yeah, and I know I'm going to need a really big power supply.
-Nick
Hi Arx,
Have you got room for the extra resistor and cap mentioned in post9 (times six = RC pair /chipamp)?
You can decide to omit or include the components at a later date.
Have you got room for the extra resistor and cap mentioned in post9 (times six = RC pair /chipamp)?
You can decide to omit or include the components at a later date.
AndrewT said:
Arx said:
Sorry for confusing both of you. I forgot Nick had the DC blocking caps in the NFB loop. If that is the case then Andrew is correct that 20k5 would be the better choice for balancing the input resistors. If the DC blocking caps are going to be removed the 1k0 would be the better choice for balancing the input resistors. Or am I still confused?
BWRX said:
The DC blocking caps are now off the schematic.
AndrewT: I don't really understand that part of post 9
-Nick
Hi Arx,
the DC coupled version you have decided to layout will have a worst case output offset of 4mV*21.5=~88mV @ Tc=25degC.
This version will have a DC servo injecting a signal (DC and VERY low frequency) into the inverting input.
To correct 88mV at the output you only need 4mV of correction (please correct me if that is wrong).
Your schematic shows the servo output connected with a 20k resistor into the inverting input and this goes via another 20k to output (~=0 impedance). This is a resistive dividing ladder.
If you need 4mV of correction, the opamp has to send 8mV of signal. BUT your opamp will be able to send supply rail voltage less a couple of volts (say about 10V of correction).
You can divide down the opamp output considerably and still have sufficient correction.
100k to 20k will give max correction of about 2V and applying the amp gain that is correcting 40V of output offset. That is why I asked if your 20k was too low.
Next, the servo sends noise and an attenuated output to the inverting input. A higher ladder reduction will reduce both the noise and the attenuated output signal.
But, adding a low pass filter to the opamp output will reduce the higher frequencies much more.
I suggest splitting the 100k output resistor in two and connecting the junction to a cap to take the high frequency to clean ground.
Try to find a few sites that discuss this more knowledgibly than my interpretation.
With the DC blocking removed you now need to reduce the input resistor to the value that minimises the output offset (~<=1k0).
And increase the RF filter cap to restore the filter effectiveness.
the DC coupled version you have decided to layout will have a worst case output offset of 4mV*21.5=~88mV @ Tc=25degC.
This version will have a DC servo injecting a signal (DC and VERY low frequency) into the inverting input.
To correct 88mV at the output you only need 4mV of correction (please correct me if that is wrong).
Your schematic shows the servo output connected with a 20k resistor into the inverting input and this goes via another 20k to output (~=0 impedance). This is a resistive dividing ladder.
If you need 4mV of correction, the opamp has to send 8mV of signal. BUT your opamp will be able to send supply rail voltage less a couple of volts (say about 10V of correction).
You can divide down the opamp output considerably and still have sufficient correction.
100k to 20k will give max correction of about 2V and applying the amp gain that is correcting 40V of output offset. That is why I asked if your 20k was too low.
Next, the servo sends noise and an attenuated output to the inverting input. A higher ladder reduction will reduce both the noise and the attenuated output signal.
But, adding a low pass filter to the opamp output will reduce the higher frequencies much more.
I suggest splitting the 100k output resistor in two and connecting the junction to a cap to take the high frequency to clean ground.
Try to find a few sites that discuss this more knowledgibly than my interpretation.
With the DC blocking removed you now need to reduce the input resistor to the value that minimises the output offset (~<=1k0).
And increase the RF filter cap to restore the filter effectiveness.
Ok, I think I see what you're getting at, but if you're talking about Rs4(and its many clones) That's actually a 205k not a 20k5 😉
Perhaps a pair of hundred k's with a cap to ground in the middle would work well?
-Nick
Perhaps a pair of hundred k's with a cap to ground in the middle would work well?
-Nick
Hi,
I need to go back to school and learn to read.
That 205k gives about 12V*1k/205k=58mV of maximum correction at the inverting input.
Does this result in a maximum correctable output of 58*20.5/1=1.2V
That gives plenty leeway for correcting gross errors.
Cs7 &8, are they 68uF or 68nF?
DC servo near or slightly less than 1Hz is usually about right.
The output filter with 100k & 150nF @ 10Hz. Could you simulate that?
Have you seen Gootee's take on filtering the servo?
I need to go back to school and learn to read.
That 205k gives about 12V*1k/205k=58mV of maximum correction at the inverting input.
Does this result in a maximum correctable output of 58*20.5/1=1.2V
That gives plenty leeway for correcting gross errors.
Cs7 &8, are they 68uF or 68nF?
DC servo near or slightly less than 1Hz is usually about right.
The output filter with 100k & 150nF @ 10Hz. Could you simulate that?
Have you seen Gootee's take on filtering the servo?
- Status
- Not open for further replies.