Arrggghh! - injection locking between halves of a 556

[GK]

I've got a prototype thingo on the bench, a small part of which involves heterodyning two basically identical oscillators, one fixed frequency, the other slightly variable +/- some hertz. I'm using an LM556, the dual version of the 555. The idea of having both oscillators on the same chip is that they track nicely for temperature. However whenever the frequency of the variable oscillator gets within cooee of the fixed frequency one, the bloody things injection lock together, and that screws up the function of my system.

It's built on breadboard and I do know that that is not the best, but I've routed all of the grounds properly and have bypassed the thing to death - the DC potentials at VCC and the unused VCO inputs are clean and there is no ground bouncing/spikes of significance at my timing caps with respect to the GND pin.
 
Does anyone have comprehensive experience with this chip? Is this a special feature of the 556? I wonder if the independent oscillators in such close proximity on the same silicon chip pickup each others totem-pole output driver switching spike. Should I attempt a PCB board or would I be wasting my time?


 

[mikerj]

Might be worth trying one of the the CMOS versions (TS556/TLC556 etc.) which don't suffer the output stage shoot-through that the bipolar version does.

[JoeO]

Although you will lose the temperature advantage, it also may be worth trying it with 2 555s from the same production lot.

[Rerouter]

Looking at the die image it seems everything is symmetrical about the supply pins, so you should be able to overcome the injection lock....
http://en.wikipedia.org/wiki/File:STM-NE556-HD.jpg

Looking at how the VCC rail reacts does not mean the ground is innocent, it only means your VCC rail is low impedance compared to the circuit,

As as extreme I would ask you to deadbug the timing components as much as possible, breadboards have adjacent channel capacitance, which could be the injection source (only needs to be a very tiny amount of coupling to make them fall in step) e.g. timing cap across the pins, terminated to the ground pin of the chip rather than through the breadboard,

Equally move your decoupling again hard against the pins, as any change to the supply voltage immediately changes the threshold point, which then slows down your timer a little (Guessing your using ceramic caps)

[GK]

VCC and the VCO pins were probed directly at the IC, WRT the GND pin, and like I said, there is no significant ground bounce/spikes between the GND pin and that of the ground return of my timing capacitors - I have a star right off the ground pin, 220uF electrolytic VCC bypass in parallel with a 100nF COG ceramic and one each of the latter for the VCO inputs. I can't make the leads any shorter. I don't think that breadboard capacitances are to blame for inter-oscillator crosstalk as the separate oscillators are pinned on opposites sides of the package. Sticking with the LM556 the only path from here would be to try a soldered version.......

I think I'll just try a pair of separate 555's on the breadboard for the time being.

[SeanB]

Probably you will be best off with 2 555 timers in a 16 pin DIL socket, and bond them to a common heatsink, or a flat brass or copper strip, to keep them at close to the same temperature. Ground the brass strip to reduce coupling of noise from one to the other.

[David Hess]

I doubt you will get this to work using a dual timer without injection locking.  I have had injection locking occur between two finite gain integrator based triangle wave generators which is probably a worst case situation on opposite sides of a board and ended up adding LC decoupling to isolate them which solved it.

The injection locking was so strong before adding the LC decoupling, that I could have used it to send a clean low bandwidth analog signal from one side of a noisy digital board to the other through the power rails.  Integration is amazing.  In the future, I avoided that specific triangle wave generator design.

[f5r5e5d]

osc can also lock thru their outputs if summed with finite Z - I recall some Allen Variance papers using 2-3 stages of common base stages between the osc to prevent their locking thru the comparator diff input C

[con-f-use]

Btw. the thing that sets the timing on a 555 or 556 is the RC combination. So the temperature argument is in IMHO invalid.

[SeanB]

Btw. the thing that sets the timing on a 555 or 556 is the RC combination. So the temperature argument is in IMHO invalid.

It might in theory but in practise the frequency depends on temperature, voltage supplied and output loading as well. Bonus for using the other device in the package is getting it to work as a PLL or even a divider.

[GK]

Btw. the thing that sets the timing on a 555 or 556 is the RC combination. So the temperature argument is in IMHO invalid.

No it isn't. The 556 has a datasheet-specified temperature-to-frequency dependence of 150ppm per degrees Celsius (typical) in astable mode, independent of the temperature drift contribution of the RC timing components (in comparison: +/- 30ppm from an NPO/COG timing cap and +/-15ppm for a Welwyn - RC55 Series 0.1% timing resistor). The timers threshold voltages and timing pin input bias currents all have a temperature dependency and that effects the operating frequency. 

Anyway, I managed to reduce the injection locking window to approximately +/- 0.1 Hz, but that still wasn’t good enough. I have since moved on to a pair of state variable oscillators, which are working a treat. That didn’t turn out to be much more complicated anyway, as I require a sinusoid quadrature signal pair from the fixed “reference” oscillator, and the SVO produces those by the fundamental mechanism of it operation. Using the 556 I needed additional harmonic and all-pass filtering. Long term temperature drift performance is still adequate.

[moffy]

The most interesting case of injection locking I worked on was a Ring Laser Gyroscope (triangular) as used on commercial aircraft. It depends on two laser beams going counter to one another and mixing the outputs of the beams together. When not locked they detect rotation very accurately. But  at low rotation rates the backscatter from the near perfect mirrors is enough to cause the beams to lock! i.e. no output. They had to add a mechanical dither wheel shacking the gyro at around 300Hz with a randomised amplitude. Took the original designers a while to work out what the problem was.

[coppice]

Btw. the thing that sets the timing on a 555 or 556 is the RC combination. So the temperature argument is in IMHO invalid.

It might in theory but in practise the frequency depends on temperature, voltage supplied and output loading as well.

It doesn't in theory. A 555 works between internal thresholds, and those thresholds have a drift with temperature that is clearly spelled out in the data sheet.