Hi,
I am just finishing a classic class AB amp with a sensitive
triple-darlington OS with 2 pairs of transistors at the end
and it has some oscillation and I found out by chance that
when I touch the heatsink with all of my palm then the
oscillation dissapears completely at all output level...
Just touching it with 1 finger is not enough.
My 1st question: why does this "human touch" solve the oscillation?
What could be the schematic here to model this phenomenon?
2nd: how should I simulate myself in an electric way to solve this?
(Connecting the heatsink directly to the PS GND or to the 230VAC protecting earth just make the oscillation even worse...)
Thx!
I am just finishing a classic class AB amp with a sensitive
triple-darlington OS with 2 pairs of transistors at the end
and it has some oscillation and I found out by chance that
when I touch the heatsink with all of my palm then the
oscillation dissapears completely at all output level...
Just touching it with 1 finger is not enough.
My 1st question: why does this "human touch" solve the oscillation?
What could be the schematic here to model this phenomenon?
2nd: how should I simulate myself in an electric way to solve this?
(Connecting the heatsink directly to the PS GND or to the 230VAC protecting earth just make the oscillation even worse...)
Thx!
Try a heavy ground wire from the heat sink to the chassis ground. Also make sure the tab on your transistors aren't conducting to the heat sink.Hi,
I am just finishing a classic class AB amp with a sensitive
triple-darlington OS with 2 pairs of transistors at the end
and it has some oscillation and I found out by chance that
when I touch the heatsink with all of my palm then the
oscillation dissapears completely at all output level...
Just touching it with 1 finger is not enough.
My 1st question: why does this "human touch" solve the oscillation?
What could be the schematic here to model this phenomenon?
2nd: how should I simulate myself in an electric way to solve this?
(Connecting the heatsink directly to the PS GND or to the 230VAC protecting earth just make the oscillation even worse...)
Thx!
Last edited:
Your body is like a capacitor.
My class A/B design would oscillate too. I wound up adding a 500pf cap across the network that performs the bias current function. I also know Doug self always puts a small cap 10pf? from collector to base on the negative side transistor of his class A/B complimentary designs. He admitted never figuring out why only that it always seemed to be necessary.
Oscillation is usually indicative of an AC transfer function whereby the gain of the circuit increases at some high frequency. This is because there will always be some phase shift at some high frequency at that point where the transistors begin to drop off in frequency response. Note that in a feedback circuit the phase shift is ideally 180 degrees (totally negative). Your circuit may not have a global feedback by design, however, even things like a 0.1 ohm resistor in the emitter leg of an output transistor is an example of (local) negative feed back. Since there's no such thing as a transistor that has infinite frequency response one has to make sure to roll off the circuit gain well before you hit that frequency point. There's a general rule of thumb that goes like this:
At that frequency in the response of the circuit whereby the phase shift is greater than 40 degrees (away from the ideal 180), the closed loop gain of the circuit at that frequency point needs to be less than one. This is because if we continue to phase shift to say 180 + 40 equals 220 degrees (due to lack of transistor speed) we begin to approach 360 degrees. When this happens your negative feedback becomes positive and guess how one makes an oscilator - with positive feedback.
There will always be some equivalent phase shift where the response starts to roll off. One just has to make sure the entire closed loop gain is less than one at that point.
Easy - add a capacitor to attenuate the gain well before that point. That pont may be 1meg hz for example so who cares if you roll off the gain at 100Khz
My class A/B design would oscillate too. I wound up adding a 500pf cap across the network that performs the bias current function. I also know Doug self always puts a small cap 10pf? from collector to base on the negative side transistor of his class A/B complimentary designs. He admitted never figuring out why only that it always seemed to be necessary.
Oscillation is usually indicative of an AC transfer function whereby the gain of the circuit increases at some high frequency. This is because there will always be some phase shift at some high frequency at that point where the transistors begin to drop off in frequency response. Note that in a feedback circuit the phase shift is ideally 180 degrees (totally negative). Your circuit may not have a global feedback by design, however, even things like a 0.1 ohm resistor in the emitter leg of an output transistor is an example of (local) negative feed back. Since there's no such thing as a transistor that has infinite frequency response one has to make sure to roll off the circuit gain well before you hit that frequency point. There's a general rule of thumb that goes like this:
At that frequency in the response of the circuit whereby the phase shift is greater than 40 degrees (away from the ideal 180), the closed loop gain of the circuit at that frequency point needs to be less than one. This is because if we continue to phase shift to say 180 + 40 equals 220 degrees (due to lack of transistor speed) we begin to approach 360 degrees. When this happens your negative feedback becomes positive and guess how one makes an oscilator - with positive feedback.
There will always be some equivalent phase shift where the response starts to roll off. One just has to make sure the entire closed loop gain is less than one at that point.
Easy - add a capacitor to attenuate the gain well before that point. That pont may be 1meg hz for example so who cares if you roll off the gain at 100Khz
None of the transistors are conducting I checked it!
Connecting the heatsink directly to the GND / Earth just does not stop the oscillation.
New experiments:
- The oscillation dissapears when I touch the heatsink AND the input wires at the same time.
- If I touch the signal generator's + OUT then the ripple on the square increases...
- The oscillation is the biggest when the input potentiometer is in the middle.
At "0" and at max level there is no oscillation at. The pot is 10k.
I guess this should be something capacitive...?!
Connecting the heatsink directly to the GND / Earth just does not stop the oscillation.
New experiments:
- The oscillation dissapears when I touch the heatsink AND the input wires at the same time.
- If I touch the signal generator's + OUT then the ripple on the square increases...
- The oscillation is the biggest when the input potentiometer is in the middle.
At "0" and at max level there is no oscillation at. The pot is 10k.
I guess this should be something capacitive...?!
campsquire: thanks for the detailed reply but if it is possible I would like not to change in the OS if the oscillation can be eliminated with just a simple "touch".
It would be great to: 1) understand how my touch works and 2) create a similar solution without me...
OS is simple triple darlington with 2 pairs of transistors.
Grouding: basically star like and the input GND goes to the pot and then further to the AMP and from there to the common PS GND.
New results: oscillation disappears when I touch the HS with my left hand and just "shield" the pot around with my right hand about 3-4 cm (without touching it...)
It would be great to: 1) understand how my touch works and 2) create a similar solution without me...
OS is simple triple darlington with 2 pairs of transistors.
Grouding: basically star like and the input GND goes to the pot and then further to the AMP and from there to the common PS GND.
New results: oscillation disappears when I touch the HS with my left hand and just "shield" the pot around with my right hand about 3-4 cm (without touching it...)
First, make sure that the heat sink is actually grounded with an Ohmeter.
Connect a 0.1uF/100V low inductance film capacitor to the input socket common, and then going directly to the chassis ground, with short lead lengths.
If this doesn't work, try connecting the capacitor from the input ground to the heat sink (using a sheet metal screw), with the wire only as long as necessary.
You likely have more problems than just the oscillation. Build up the circuit properly before any more testing, and rethink the "star" grounding,
which often causes more problems than it solves. Is there a pcb or hf compensation?
If you just designed an EF3 "out of the book" it will do what you say.
Without predriver/driver decoupling and predriver shunting it WILL oscillate,
You touching it may be canceling out a parasitic feedback loop involving
the PS ripple.
EF3 is not like an EF2 - I was "scared" of them for years
.EF3 is even "picky" about having a fast pre/fast driver - and not too slow OP.
Post your circuit !
OS
The PCB is very compact and the prototype with just 1 pair of transistors already worked with the same heatsink and layout.
So even if this sensitivity could/should be solved in an other way I am just curious how this phenomena works and why my touch solves it...
And if it is not necessary I wouldnt change my PCB just add the same solution (but with electrical components) what my hand does because it solves it completely.
So this "mistery" could be modelled this way:
- we have an amp like a black box unit and it has an input and an output
- the input is 1kHz square wave (2Vpp)
- without me it doesnt oscillate at the 2 ends of the pot
- but it does in the middle
- and now if I touch the heatsink and "shield" the pot with my palm the oscillation goes completely away
So even if this sensitivity could/should be solved in an other way I am just curious how this phenomena works and why my touch solves it...
And if it is not necessary I wouldnt change my PCB just add the same solution (but with electrical components) what my hand does because it solves it completely.
So this "mistery" could be modelled this way:
- we have an amp like a black box unit and it has an input and an output
- the input is 1kHz square wave (2Vpp)
- without me it doesnt oscillate at the 2 ends of the pot
- but it does in the middle
- and now if I touch the heatsink and "shield" the pot with my palm the oscillation goes completely away
Attachments
Last edited:
That's the problem. You are a mix of all these elements and no way to know how much of each.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.