Hi Cleo,
Thanks for your attention and for taking the trouble to respond - I'd be very interested to hear from the engineering team -don't worry too much about me - I've designed dozens of DC-DC switchers over the years, so I wasn't poking about inside randomly - well, not entirely randomly, anyway!
.
I'll try to explain my current understanding of the problem (though I only spent an hour or two fixing it, and most of that was getting the thing apart and back together!)
.
The main issue is that those ceramic caps (including C847) which are there to bypass (smooth the current demand on) the20v input, are insufficient and as a result, the 20V is bouncing around.
.
In a typical buck converter topology (which I believe the circuit in question is - though I don't have circuit diagrams, obviously),the duty cycle of the PWM is proportional to the load and also to the difference between the input voltage and the target voltage. In a steady-state (i.e. with constant load, output voltage and input voltage) all is well (evidenced by reports that disabling the power management cures the noise issue) - however, since we have a changing output voltage (due to dynamic voltage requirement of the CPU as a result of agile core voltage requirement versus clock frequency), the circuit attempts to move the output voltage up and down depending, ultimately, on the CPU load. When the voltage target changes,this creates a short-term current surge (mainly to charge the output bulk capacitance) - and this places a dynamic current requirement on the input supply - this is all fine, provided the supply can supply it. My observation of the changing voltage across c847 suggests it cannot, or rather, there is excessive impedance between the supply (the battery or the mains adapter) and this point in the circuit.
.
So, as the voltage target is increased, this creates a corresponding transient current demand on the input supply - because a) the input bypass capacitance is too small and/or b) the impedance between the actual supply (mains adapter, battery) is too high/too inductive - this causes the input voltage to sag (by as much as a volt or two), which in turn, changes the duty cycle requirement. But, as the duty cycle changes, so does the demand on the input supply and correspondingly, so the sag changes, resulting in a different duty cycle requirement, and so the whole thing goes 'b'doiiiiinng'. In other words, there is an inadvertent feedback loop causing instability. This would not be a problem per-se except for the fact that inductors can work as loudspeakers (due in part to magnetostriction effects in the core, but often more significantly, due to physical forces created by the magnetic fields on and by the windings within the inductor) where the acoustic output is (approximately) proportional to the rate of change of current through the inductor. So, as the current through the inductor goes 'b'doiiinnng', so do the physical forces in the inductor, and consequently, the acoustic output also goes 'b'doinnnnnggg' at a few kHz resulting in whistles and whines.
.
My 'fix', increasing the bulk capacitance on the input to the switcher, provides a localised current supply and reduces the sag.It's a bit of a brute-force approach, but in the absence of further circuit data, the easiest to apply. Another easy approach is to find a better (less transductic) inductor (L75 is the main culprit).
.
I know about these issues because I've had to fix numerous design problems like this (sometimes my own designs!). This is analog electronics, difficult to get just-right, and a problem in multi-dimensions (due to PCB layout and non-ideal components), but that's why we build prototypes.
.
What is disappointing is that no-one in Lenovo noticed the problem right up to production, and end-users who care about having the best product have paid top-dollar for a machine with such a petty, but extremely annoying issue.
.
When I bought my T510, I wanted the best machine I could buy. Quite frankly, I didn't really care how much it cost - I sit in front of this thing 15+ hours a day and it is my main tool with which I earn my living.
.
With so much fan-fare (no pun intended) about quiet fans inspired by owls wings(!?), silent SSDs etc. it seems a little incongruent that no one cared about the much-more irritating whining noise.
.
My only hope is that a late component source change (for L75) resulted in noise which had not existed during the various prototypes. I know how much components like an extra capacitor cost in quantitiy - and even assuming a20:1 markup by the time the customer pays up, I'd have happily stumped up an extra $1 to avoid this problem.
.
Thanks again for your time and interest. I hope Lenovo can find a workable solution for it's loyal customers. For me, I have resolved the issue sufficiently that I probably can't justify spending any more time on it - I have other problems to solve (ones which I can bill someone for!).
.
Best regards,
.
N.
English Community 
Deutsche Community 
Comunidad en Español 
Русскоязычное Сообщество
