The 'close to pin decoupling' is usually done with 100nF ceramic caps.
There seems to be some kind of consensus that it should be done like this on dual opamps.
Should the value of the cap be increased in case of a quad opamp like the TL074?
There seems to be some kind of consensus that it should be done like this on dual opamps.
Should the value of the cap be increased in case of a quad opamp like the TL074?
No. The cap function is purely to maintain a low impedance directly across the opamp supply pins, the number of opamps within the package not materially affecting that.
H.Ott tells us that we should select the capacitor package that achieves the lowish inductance that the circuit requires.
Then select the bigest capacitor in that package. How you place the vias to the planes and where you locate the cap+vias relative to the pin makes a big difference to the inductance.
Thus if you decide the circuit needs a 603 SMD capacitor to achieve lowish inductance, then you look for an adequate voltage in that case size and the biggest capacitance be it 22nF, or 680nF, or anywhere in between.
If you have a bit of space to choose a slightly higher voltage rating, then your capacitor will achieve closer to it's specified value after de-rating for applied DC voltage.
Then select the bigest capacitor in that package. How you place the vias to the planes and where you locate the cap+vias relative to the pin makes a big difference to the inductance.
Thus if you decide the circuit needs a 603 SMD capacitor to achieve lowish inductance, then you look for an adequate voltage in that case size and the biggest capacitance be it 22nF, or 680nF, or anywhere in between.
If you have a bit of space to choose a slightly higher voltage rating, then your capacitor will achieve closer to it's specified value after de-rating for applied DC voltage.
Last edited:
This is one situation where hand-soldered surface mount assembly is useful. You can put the capacitors on the opposite side of the PCB, directly beneath the opamp chip. There's room under there to mount four caps if you wish; two per rail.
Too bad i placed an order for parts yesterday.
I could have ordered surface mount caps.
....let's see what aliexpress has to offer...
I could have ordered surface mount caps.
....let's see what aliexpress has to offer...
btw, am i right that 0805 package should fit quite good onto strip (-matrix) board?
like this:
https://de.aliexpress.com/item/Free..._1&btsid=9453fcd4-154c-41c1-a9cf-a4d3e8e3c3cf
like this:
https://de.aliexpress.com/item/Free..._1&btsid=9453fcd4-154c-41c1-a9cf-a4d3e8e3c3cf
Last edited:
805 can be fitted between two adjacent holes/pads of a 0.1" grid.
It can also be placed diagonally between two holes/pads.
A thin 805 (lower capacitance) can be placed under an IC to minimise the trace length to get to the other power pin.
It can also be placed diagonally between two holes/pads.
A thin 805 (lower capacitance) can be placed under an IC to minimise the trace length to get to the other power pin.
So you suggest to put them between the power pins?
That's easy with the pinout of a tl074.
In an other thread i was told to put the decoupling between each power pin and gnd-rail (center tap of two batteries).
So many truths....
That's easy with the pinout of a tl074.
In an other thread i was told to put the decoupling between each power pin and gnd-rail (center tap of two batteries).
So many truths....
The output current flows through the load and back to power ground. Decoupling for that current should be to the power ground. Since the output current alternately flows from +ve pin to PG and -ve pin to pg you need two decouplingroutes to resolve that.
Power pin to power pin is a third route.
Power pin to power pin is a third route.
Is PG the center tap or lowest potential in the power circuit?
Does two 'decoupling routes' mean that if the center tap is used as signal ground and power ground at the same time, there should be two different routes towards the center tap: one for signal, one for power?
...that is how i actually did the layout of the last active x-o i built.
Does two 'decoupling routes' mean that if the center tap is used as signal ground and power ground at the same time, there should be two different routes towards the center tap: one for signal, one for power?
...that is how i actually did the layout of the last active x-o i built.
the signal input will come from a two wire connection The signal flow route and the signal return route.
The output from a push pull amplifier has two routes internal to the amplifier. From+ve supply to output pin and the return back to the +ve pin. Then ther eis the output that comes from the -ve pin comes out of the output and has to return to the -ve pin.
These two output routes operate alternately on opposite halves of the AC waveform.
It's these two routes that use the decoupling caps. The caps have to be located so that the return route actually goes back to the decoupling cap.
The output from a push pull amplifier has two routes internal to the amplifier. From+ve supply to output pin and the return back to the +ve pin. Then ther eis the output that comes from the -ve pin comes out of the output and has to return to the -ve pin.
These two output routes operate alternately on opposite halves of the AC waveform.
It's these two routes that use the decoupling caps. The caps have to be located so that the return route actually goes back to the decoupling cap.
The most important decoupling for op amp stability is from rail to rail, make this path fast (low inductance). Use as much capacitance as possible.
Rail to GND decoupling is much less critical (opamps don't even know about the GND), but you need it too. The more rail-to-rail decoupling you have, the less important is rail-to-GND because any ripple on the supply from load currents will be undistorted, supply ripple is a) identical on both rails and b) a linear copy of (the sum of all) load current(s), opamp PSRR will handle that gracefully at audio frequencies.
If you have zero rail-to-rail capacitance, then the rail ripple is distorted (rectified) copy of audio current once the opamp enters class B region and/or cross currents are appearing (like with LT1365) and then PSRR will have a hard time to cope with the high frequency content of the ripple.
When you have any rail-to-GND decoupling it is most important that the center node between the caps is established first and then you go from there to your GND (plane). Otherwise, with GND attached to each cap at seperate locations, the half wave rectified currents don't combine before they the hit the GND and thereore will induce half-wave rectified residual voltage drops along the connection path in the GND plane/trace, polluting it.
Rail to GND decoupling is much less critical (opamps don't even know about the GND), but you need it too. The more rail-to-rail decoupling you have, the less important is rail-to-GND because any ripple on the supply from load currents will be undistorted, supply ripple is a) identical on both rails and b) a linear copy of (the sum of all) load current(s), opamp PSRR will handle that gracefully at audio frequencies.
If you have zero rail-to-rail capacitance, then the rail ripple is distorted (rectified) copy of audio current once the opamp enters class B region and/or cross currents are appearing (like with LT1365) and then PSRR will have a hard time to cope with the high frequency content of the ripple.
When you have any rail-to-GND decoupling it is most important that the center node between the caps is established first and then you go from there to your GND (plane). Otherwise, with GND attached to each cap at seperate locations, the half wave rectified currents don't combine before they the hit the GND and thereore will induce half-wave rectified residual voltage drops along the connection path in the GND plane/trace, polluting it.
- Status
- Not open for further replies.