Selecting a PNP transistor for high-side drive

Go To Last Post
14 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

 

I am using the following circuit except that I am trying different PNP transistors in place of the BSS84 shown (emitter connected to V+).

 

The signal TB0 is being provided by a 100 kHz low-side PWM LED driver (open collector).

 

V+ is 12v. R13 has been adjusted to provide the best rise time on TB0 as observed on a scope.

 

The upper (yellow) trace below is the drive as seen on TB0.

 

The lower (teal) trace is on the collector - R2 junction of F0 (remember, F0 is now a PNP transistor).

 

The part actually in use is a MMUN2113LT3G Pre-Biased PNP transistor because I had some on hand.  This has an internal 47k emitter-base resistor and a 47k base resistor (Datasheet ).

 

Notice that the fall of the collector is delayed by about 2 us from the rise of the base signal (100 kHz PWM).

 

 

 

My question is: How can I reduce this turn-off delay?  It reduces the effective range of the PWM I can use to about 80% of full range of 100 us.

 

Perhaps someone can recommend an SOT-23 PNP transistor (~50v CE, ~100 ma) that has a faster turn-off time?

 

BTW: In this datasheet and others I do not see any references to turn-on or turn-off times as I do for MOSFETs - reason?  What datasheet item would one use to select a transistor with fast response time (higher frequency) when running in digital saturation/cutoff use?  I see "bandwidth product" - is that something I can use for selection?

 

My assumption here is that, because I am driving it into saturation, it is taking longer to come out of saturation due to internal C charge storage.

 

Thanks in advance for your help.

 

Regards,

 

Chuck Hackett

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I just tried adding a resistor from the base/R13 junction to TD0 to reduce the drive current from TD0 (reduce saturation).

 

I can get the turn-off delay down to about 1 us but I loose the sharp turn-on seen without the additional resistor.

 

Sigh ...

 

Regards,

 

Chuck Hackett

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Try googling Baker Clamp and try that. I used one in the 1970's on a switcher power supply I designed to good sucess.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jaydhall wrote:

Try googling Baker Clamp and try that. I used one in the 1970's on a switcher power supply I designed to good sucess.

 

On quick read it looks applicable.  On the current device I have on hand (Pre-Biased transistor) I can't get at the base junction itself so I'll have to order some parts and give it a try.

 

Thanks for the suggestion ...

 

Regards,

 

Chuck Hackett

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Oops. Missed that pre bias stuff. I never used those in design because they were more expensive than the discrete parts and when large quantities were involved, saving 10 cents per board was a huge amount per year. I did think it would be interesting to use. But back then 20% resistors were the thing....

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Sounds like a great indication to run a Spice analysis...

 

Without running any numbers in my head, how much is the R1C1 filter holding up the voltage on the T0 node, which is then holding up the voltage on the Collector R2 node?

 

JC

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ChuckH wrote:
I can't get at the base junction itself so I'll have to order some parts and give it a try.

 

Yeah you will need a plain PNP with base access to use speedup tricks like the "Baker Clamp" (already mentioned) or the "Speedup Capacitor" (see figure 2 here for an example; C1 is the speedup cap https://www.nutsvolts.com/magazine/article/bipolar_transistor_cookbook_part_3 )

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

 

 

Why use a transistor that has a 47k base resistor...not the fastest thing ...Use you own part, where you control the parameters, especially base drive, if you are interested in speed.

It make be specifically made for low speed applications

 

 

The ADC 100K is too high...keep it, preferably, to 10K or less 

 

Notice that the fall of the collector is delayed by about 2 us from the rise of the base signal

Well not the base signal, but a sig going through 47K to the base!! 

 

A schematic wiring connection dot shows wires connecting, not component leads, apply dots to wires, and connect component leads/pins only to wires.  

There are specific types of hidden dots that show component leads connecting to wires.

Never draw a 4 way connection with a dot, use only 3 way connections.

 

 

 

 

When in the dark remember-the future looks brighter than ever.   I look forward to being able to predict the future!

Last Edited: Sun. Apr 26, 2020 - 06:28 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

avrcandies wrote:
Why use a transistor that has a 47k base resistor...

 

Yes, I only used it because it was the only PNP I had on hand for testing.  Will be ordering some BJTs and Schottky diodes ...

 

DocJC wrote:
Without running any numbers in my head, how much is the R1C1 filter holding up the voltage on the T0 node, which is then holding up the voltage on the Collector R2 node?

 

The load connected at T0 varies from a couple of k-ohms down to about 50 ohms (actually railroad track laying on the ground with ties and ballast).  R2 is a "ballast" resistor to limit current when the track is shunted by a train (and the voltage drop detected by the ADC).  The track voltage is processed to remove the effects of leakage (rain), and intermediate contact caused by oxide, leaves, etc. on the track.

 

avrcandies wrote:
The ADC 100K is too high...keep it, preferably, to 10K or less 

 

The 100k and cap (along with other components not shown) is there to isolate the ADC from lightning surges.  I can draw an arc on the "Track In" terminal from a 6,000 volt AC supply and the ATMega ADC survives fine.  It has worked well in Florida - The Lightning Capitol Of The World smiley

 

Regards,

 

Chuck

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

A high side current source might give a sharp ON/OFF depending on the load, but it is more complex. I use parts K1N (MMBT3904), K3N (MMBT3906). I also set the board layout so that the parts thermally track, which causes a sort of foldback current limit should the load be shorted.

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

 Will be ordering some BJTs and Schottky diodes ...

Why don't you order some FETs?  No saturation delay times with fets

 

The 100k and cap (along with other components not shown) is there to isolate the ADC from lightning surges

That's nice, however 100k is kinda too high for the ADC.  Keep it 10K or less, unless high accuracy is not too important.   You can add a zener, TVS, MOV, or other device for  lightning.  Plus you also need to protect the  chip itself!

When in the dark remember-the future looks brighter than ever.   I look forward to being able to predict the future!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

avrcandies wrote:
Why don't you order some FETs?  No saturation delay times with fets

 

I originally had a MOSFET (BSS84 see original ckt above) but this was worse - still have to get rid of the gate charge needing bi-polar drive for fast switching.  Without making major changes to the card I am stuck with open-collector drive (8 channels on the card).  Just to be sure, I'll put the MOSFET back in and see with some other changes but, my understanding is that MOSFETs are better when switching high currents whereas here I am only switching 10's of milliamps.

 

avrcandies wrote:
Keep it 10K or less, unless high accuracy is not too important.

 

I need repeatability but not high accuracy ...

 

avrcandies wrote:
You can add a zener, TVS, MOV, or other device for  lightning.

 

I use TVSs

 

Regards,

 

Chuck

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

still have to get rid of the gate charge needing bi-polar drive for fast switching.

Why are you using 10K for r13?...use something like 750 ohms for R13...will switch much faster...bring the gate to gnd via another 1000 ohms to your open collector driver.  Or use  an active pullup.

 

BSS84 is pretty rotten (10 ohms ??!!)....you can easily get something half that or less (even 0.1 ohm) nowadays, for pennies.

Yeah, you can get BSS84 for 3-5 cents at Digikey...spend 20 cents and get much better performance.

 

Take a look look at:   

SSM3J351RLFCT only 0.130 ohm @ -4V gate, 0.110 ohm at -10V gate, 12 cents @1000 quan

                                      

SI3127DV-T1-GE3  , more like 0.15 ohm  , 17 cents @1000 quan

When in the dark remember-the future looks brighter than ever.   I look forward to being able to predict the future!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

avrcandies wrote:
Why are you using 10K for r13?...

 

Sorry, that value was left over from previous test.  While testing the BJTs I had a variable resistor there in addition to a variable base resistor so I could try different values.  I'd have to check/measure to see what ended up the best.

 

avrcandies wrote:

BSS84 is pretty rotten (10 ohms ??!!)....you can easily get something half that or less (even 0.1 ohm) nowadays, for pennies.

Yeah, you can get BSS84 for 3-5 cents at Digikey...spend 20 cents and get much better performance.

 

Yup, have learned that ... didn't know better at the time (hangs head in shame ...)

 

avrcandies wrote:

Take a look look at:   

SSM3J351RLFCT only 0.130 ohm @ -4V gate, 0.110 ohm at -10V gate, 12 cents @1000 quan

                                      

SI3127DV-T1-GE3  , more like 0.15 ohm  , 17 cents @1000 quan

 

Thanks, will order some to try.

 

Regards,

 

Chuck