[time-nuts] HP5328A LEDS driver transistor

Charles P. Steinmetz charles_steinmetz at lavabit.com
Mon Jan 23 05:01:31 UTC 2012


Bob wrote:

>Specifying the digit scan transistor is a bit tricky, and often power
>dissipation is just not a concern. Why? Consider the duty cycle. A transistor
>switching one amp (a LOT for 7+1 segments) with a 0.6V saturation dissipates
>600mW, but only when on. Given a 1/8 duty cycle that's 75mW average. 
>Any stinkin
>TO-92 will handle that power all day long! Only an ultra-conservative design
>would use an exposed tab transistor, as expected for HP!

OK, looking at the schematic 7 of the digit drivers have 7+2 loads 
(DP and annunciator LEDs).  Each segment is driven from 5 volts 
through two saturated transistors and an LED.  Assuming 0.6 + 0.6 + 
1.7 V, each segment current is 0.21 A and the total digit current 
with all 9 segments lit is 1.89 A, for a digit driver dissipation of 
(1.13 W x duty cycle).  With 9 digits, the duty cycle must be less 
than 11.1%, so the power dissipation will be around 125 mW worst 
case.  As Bob said, any TO-92 will dissipate this with no difficulty.

The problem is the current.  You need transistors that can handle 
1.89 A without sweating.  The MPSU51 is rated at 2A continuous, and 
is not specified for pulse duty so there is no way to know how much 
if any more current it will safely handle at an 11% duty cycle.  The 
fact that Chris has two bad ones may be taken to indicate that they 
were not hugely overspecified in this application.  I would be 
looking at transistors rated for a maximum continuous collector 
current of at least 3 A, preferably more.  Only a very few TO-92 
devices fit this bill.  I think my first choice would be the ZTX949 
-- it has low saturation voltage and is rated for pulse current 
substantially in excess of the continuous current rating, and its 
pinout matches the MPSU51.  I have used lots of them and they are 
very rugged.  In stock and priced from $1.12 to $1.24 in singles from 
Digi-Key, Mouser, and Avnet.

Best regards,

Charles












More information about the Time-nuts_lists.febo.com mailing list