[sdiy] MOSFET cap discharge

[ASSI]

On Donnerstag, 15. Dezember 2005 23:53, Ian Fritz wrote:
> We often hear that MOSFETs shouldn't be used for the cap discharge
> switch in a saw VCO, because they have too much leakage current. 
> Indeed, the spec sheet for the BS170 (for example) lists a max drain
> cutoff current of 500 nA.  Pretty bad.
>
> But if you look in Horowitz and Hill, they say that the channel
> resistance in the off state is usually over 10000 M, i.e., 10^10 Ohm.
>  So with  1V across the channel, leakage should be under 100 pA.  How
> bad is it in practice?  I have heard that some people have had
> success with MOSFET discharge switches.

In a MOSFET device there is a tradeoff between on-current and 
off-current if you keep all other things constant - this is known as 
the "universal curve" (there are other "universal curves" for different 
things of course). Most devices will be designed for maximum 
on-current, accepting a fair amount of leakage. The leakage is split 
between channel leakage and diode leakage drain to substrate. In both 
cases keeping 10-20% margin to the maximum rated drain voltage helps to 
improve the leakage performance.

Diode leakage is not a big problem at room temperature, but it doubles 
each 10-15K, so you need to be careful if the device can get hot. A 
junction breakthrough anytime in the devices life will often ruin the 
leakage performance even if the device is otherwise still functioning 
OK. A breakthrough may even introduce additional channel leakage if the 
gate oxide gets damaged during the breakthrough.

Channel leakage is not normally a significant part of the leakage if you 
stay far enough from the threshold voltage in the off-state - in 
subthreshold the channel leakage falls off exponentially: a decade with 
a certain number of mV of Vgs. This is a characteristic value for each 
MOSFET (the theoretical optimum is around 60mV/decade for a planar 
device), low-voltage discrete transistors will typically show between 
100-200mV/dec subthreshold slope. So if you have a Vth of 1V and 
100mV/dec subthreshold slope, you are ten decades below Idsat at a Vgs 
of 0V. Temperature dependence of the channel leakage is a complicated 
subject, but in most cases the leakage will be dominated by the diode 
leakage anyway.

If the threshold voltage is too low to achieve good leakage performance 
at Vgs=0V, you can and should use negative Vgs to better cut off the 
channel. However, more is not always better in this case: at some point 
the leakage will go up again due to an effect that is known as gate 
induced drain leakage (GIDL).

For example, an Idsat=500mA MOSFET could have less than 10pA specified 
leakage at room temperature. You'd be able to select below 1pA specimen 
from a small batch as the typical leakage would be smaller. If you were 
really trying for best leakage performance I think the low-tens fA 
region is reachable for such a device. Of course, if you have a circuit 
that needs such levels of performance you will probably also have to 
worry about things like gate-drain capacity and kickback charge when 
you switch off the channel.



Achim.
-- 
+<[Q+ Matrix-12 WAVE#46 Neuron microQkb Andromeda XTk sonic heaven]>+

SD adaptation for Waldorf microQ V2.22R2:
http://Stromeko.Synth.net/Downloads.html#WaldorfSDada