Some time ago I built a low distortion SVF oscillator using LM4562 dual op amps (with sample & hold ALC using a FET).
It works pretty well, although I recently noticed a spurious low level HF oscillation at around 60 MHz superimposed on the main output signal.
There is a short (about 4->5 inches) length of ribbon cable that connects the dual gang frequency control pot to the main PCB.
If I wrap my hand around the cable, the spurious signal stops.
I ended up wrapping an earthed screening braid around the ribbon cable.
That seems to have worked but I am interested in discovering the source of the spurious oscillation.
I didn't think an LM4562 would even work at 60 MHz (FT is 55MHz).
Do the op amps need some external frequency compensation?
The +ve and -ve 12V rails appear to be adequately decoupled with 0.1uF and short tracks to the op amp supply pins.
The braid solution looks a bit of a bodge.
I am considering trying other options such as using a ribbon cable with more more wires arranged such that every alternate wire is earthed.
Could the problem be stray capacitance between the wires connecting the two halves of the dual pot?
Maybe it would be better to split the ribbon into two separate cables or replace it with a pair of twin core screened cables?
I just wondered if anyone else had experienced a similar issue?
It works pretty well, although I recently noticed a spurious low level HF oscillation at around 60 MHz superimposed on the main output signal.
There is a short (about 4->5 inches) length of ribbon cable that connects the dual gang frequency control pot to the main PCB.
If I wrap my hand around the cable, the spurious signal stops.
I ended up wrapping an earthed screening braid around the ribbon cable.
That seems to have worked but I am interested in discovering the source of the spurious oscillation.
I didn't think an LM4562 would even work at 60 MHz (FT is 55MHz).
Do the op amps need some external frequency compensation?
The +ve and -ve 12V rails appear to be adequately decoupled with 0.1uF and short tracks to the op amp supply pins.
The braid solution looks a bit of a bodge.
I am considering trying other options such as using a ribbon cable with more more wires arranged such that every alternate wire is earthed.
Could the problem be stray capacitance between the wires connecting the two halves of the dual pot?
Maybe it would be better to split the ribbon into two separate cables or replace it with a pair of twin core screened cables?
I just wondered if anyone else had experienced a similar issue?
The LM4562 is damn fast in order to get low distortion (lots of open loop bandwidth still at 10kHz). It needs proper attention to decoupling and layout as you have discovered. You cannot get away with just a single 100nF between V- and V+ with this chip for instance, both need decoupling to ground.
Not sure how you would add external frequency compensation to a dual opamp in an 8 pin package!
Not sure how you would add external frequency compensation to a dual opamp in an 8 pin package!
Many fast OpAmps will benefit from a small value C0G/NP0 cap across the feedback resistor from output to inverting input, or sometimes one from non-inverting input to ground, or maybe both - normally a value somewhere around 50 pF give or take, it depends on the circuit...the particular implementation and layout can have a lot to do with instabilities like this.
Mike
Mike
Last edited:
If the layout has too much capacitance from output to +ve input you'll need extra capacitance from output to -ve input to stabilize it. 50pF being required suggests poor layout, 5 to 15pF is usually plenty. This is the reason the standard dual opamp pinout has output next to -ve input rather than next to +ve input.
Through-hole (and breadboard) designs can be more problematic as there is more stray capacitance and inductance compared to SMT when high frequencies are involved, SMT tends to lead to natural layout which probably helps. One pitfall of poor high-frequency layout is the temptation to place neat rows of through-hole components all in the same direction like on a parade ground - layout needs to follow the circuit topology and keep parasitics low - human aesthetics aren't important to electrons!
Through-hole (and breadboard) designs can be more problematic as there is more stray capacitance and inductance compared to SMT when high frequencies are involved, SMT tends to lead to natural layout which probably helps. One pitfall of poor high-frequency layout is the temptation to place neat rows of through-hole components all in the same direction like on a parade ground - layout needs to follow the circuit topology and keep parasitics low - human aesthetics aren't important to electrons!
