HOW TO IMPROVE PSRR (HF) OF THREE TERMINAL REGULATOR (78L05)?

[Hammy]

As the subject states?

I'm running an IR LED driver using a mosfet and driver the driver
transmits 1.25kHz signal on a 38kHz carrier generated with TLC555 CMOS
timers that's detected by the PNA4602M photo detector.

The mosfet and driver as well as the VDD for the FET is taken off the
8V regulator which is a 7808.The timers are getting there Vcc off of a
78l05 that gets its input from the 7808.The 7808 is supplying 1amp
burst at 38kHz 10%duty cycle (400us on 400us off)
The over and undershoots as well as the FETs turn off transient aren't
being rejected by the 78l05.

The cct is done on a bread board I'm using separate returns for the 5V
(top rail) and the 8V (bottom rail). These are both brought back to a
120uf cap at the 12vdc input voltage (bench top supply).

I've been able to improve it, but I'm open to any suggestions. I've
tried several combinations of capacitors at the input and output, and
tried a small inductor scavenged from an SMPS. This schematic is of my
regulator setup so far that gives the best results.

http://i13.tinypic.com/502tjd1.png

These are the waveforms, CH1 (yellow) is the 8V regulator output, CH2
is the 5V output. The carrier is just starting another 400us cycle
after a 400us break.

The measured waveforms are here.

http://i18.tinypic.com/54d8zdj.gif

In case I haven't been clear here is three cycles of the carrier
signal ,with the AC ripple of the 5v rail on ch1.
http://i12.tinypic.com/6h89pgz.gif

This shows the 5v rail and the mosfet Vgs zoom of three pulses. Theres
a small brief undershoot when the switch turns on, then by comparison
a large overshoot then the turn off transient, a deep undershoot then
overshoot. Could the second under and overshoot be caused by the FETS
turn-off transient? I had to use 5 power resistors in series and long
leads for current limiting on the drain, so this probably will be gone
when I transfer it to PCB.

http://i18.tinypic.com/67yd43n.gif

Thanks for any suggestions or advice.

 

[ehsjr]

As the subject states?

I'm running an IR LED driver using a mosfet and driver the driver
transmits 1.25kHz signal on a 38kHz carrier generated with TLC555 CMOS
timers that's detected by the PNA4602M photo detector.

The mosfet and driver as well as the VDD for the FET is taken off the
8V regulator which is a 7808.The timers are getting there Vcc off of a
78l05 that gets its input from the 7808.The 7808 is supplying 1amp
burst at 38kHz 10%duty cycle (400us on 400us off)
The over and undershoots as well as the FETs turn off transient aren't
being rejected by the 78l05.

The cct is done on a bread board I'm using separate returns for the 5V
(top rail) and the 8V (bottom rail). These are both brought back to a
120uf cap at the 12vdc input voltage (bench top supply).

I've been able to improve it, but I'm open to any suggestions. I've
tried several combinations of capacitors at the input and output, and
tried a small inductor scavenged from an SMPS. This schematic is of my
regulator setup so far that gives the best results.

http://i13.tinypic.com/502tjd1.png

These are the waveforms, CH1 (yellow) is the 8V regulator output, CH2
is the 5V output. The carrier is just starting another 400us cycle
after a 400us break.

The measured waveforms are here.

http://i18.tinypic.com/54d8zdj.gif

In case I haven't been clear here is three cycles of the carrier
signal ,with the AC ripple of the 5v rail on ch1.
http://i12.tinypic.com/6h89pgz.gif

This shows the 5v rail and the mosfet Vgs zoom of three pulses. Theres
a small brief undershoot when the switch turns on, then by comparison
a large overshoot then the turn off transient, a deep undershoot then
overshoot. Could the second under and overshoot be caused by the FETS
turn-off transient? I had to use 5 power resistors in series and long
leads for current limiting on the drain, so this probably will be gone
when I transfer it to PCB.

http://i18.tinypic.com/67yd43n.gif

Thanks for any suggestions or advice.

You can help the 5 volt supply this way, by
feeding it from 12V through a diode and with
a big (relatively) filter cap:

+ 12 ----+-----------------------> to 7808
|a
[D1]
|
+---> to 78L05
|+
[47 uF]
|
Gnd -----+---------------------->

That way, any sag on the Vin caused by the 1 amp
transient on the 8V line do not appear on the
timers Vcc.

You may help the 8V supply by increasing that 120 uF cap
to 470 or 1000 uF or even larger. Also, connect a low
current load - say 470 ohms from the +8V to ground.

As you probably know, breadboards can be the source of
grief when battling noise/transients. A nice stiffly
regulated +12 can become ugly on a poor breadboard
connection. The big input cap helps when that is the
case.

Ed

 

[Hammy]

You can help the 5 volt supply this way, by
feeding it from 12V through a diode and with
a big (relatively) filter cap:

+ 12 ----+-----------------------> to 7808
|a
[D1]
|
+---> to 78L05
|+
[47 uF]
|
Gnd -----+---------------------->

That way, any sag on the Vin caused by the 1 amp
transient on the 8V line do not appear on the
timers Vcc.

You may help the 8V supply by increasing that 120 uF cap
to 470 or 1000 uF or even larger. Also, connect a low
current load - say 470 ohms from the +8V to ground.

As you probably know, breadboards can be the source of
grief when battling noise/transients. A nice stiffly
regulated +12 can become ugly on a poor breadboard
connection. The big input cap helps when that is the
case.

Ed

Thanks for the reply,

That was one of the things I tried, running it off the 12V. I never
thought of the dummy load. I took a measurement a cross a sampling
resistor (1.3 ohms) at the 12V input, and it was stable small 1.25 kHz
ripple (small DC component, small ripple) don't remember exactly. I
blew the mosfet driver with all my poking around (shorted something
out), or I'd re-measure it.

I think a lot of it has to do with the breadboard and the long leads I
had to use. So I'm just going to do the layout and put in some extra
pads for caps or inductors in case I need them. I will run the 5V
regulator off the 12V the ripple frequency there is much smaller ,so
the 78l05 should fair better.

Thanks again.

Sorry for posting in caps didn't realise until I sent it.

 

[ehsjr]

You can help the 5 volt supply this way, by
feeding it from 12V through a diode and with
a big (relatively) filter cap:

+ 12 ----+-----------------------> to 7808
|a
[D1]
|
+---> to 78L05
|+
[47 uF]
|
Gnd -----+---------------------->

That way, any sag on the Vin caused by the 1 amp
transient on the 8V line do not appear on the
timers Vcc.

You may help the 8V supply by increasing that 120 uF cap
to 470 or 1000 uF or even larger. Also, connect a low
current load - say 470 ohms from the +8V to ground.

As you probably know, breadboards can be the source of
grief when battling noise/transients. A nice stiffly
regulated +12 can become ugly on a poor breadboard
connection. The big input cap helps when that is the
case.

Ed

Thanks for the reply,

That was one of the things I tried, running it off the 12V. I never
thought of the dummy load. I took a measurement a cross a sampling
resistor (1.3 ohms) at the 12V input, and it was stable small 1.25 kHz
ripple (small DC component, small ripple) don't remember exactly. I
blew the mosfet driver with all my poking around (shorted something
out), or I'd re-measure it.

I think a lot of it has to do with the breadboard and the long leads I
had to use. So I'm just going to do the layout and put in some extra
pads for caps or inductors in case I need them. I will run the 5V
regulator off the 12V the ripple frequency there is much smaller ,so
the 78l05 should fair better.

Run it off the 12V as shown in the diagram I drew,
where the diode & cap isolate the 78L05 from sags
on the 12 volt line. Simply running it directly
from the 12V line is not as good.

Ed

 

[Hammy]

Run it off the 12V as shown in the diagram I drew,
where the diode & cap isolate the 78L05 from sags
on the 12 volt line. Simply running it directly
from the 12V line is not as good.

Ed

I did as you suggested (Thanks), it did provide some improvement. I
also added an RC LPF at the input of the 5V regulator this also
helped; it got rid of most of the switching transient and reduced the
38 kHz and 1.25 kHz ripple. I'll try an RCD snubber across the mosfet
to totally get rid of it (switching transient). I'd like to try some
ferrite beads but I don't have any, next Digikey order I'll get some.

The output of the five volt regulator has a 80mV ripple that decrease
in amplitude over a 400uS cycle to about 40mV, this neglects the spike
from the switch. So it's an improvement from the 220mV ripple and the
large transient spikes from the FET. I'll also try some different
combinations of caps at the output. If I can eliminate the large
portion of the ripple, the 80mV, and the rest of the switch transients
that should be good enough, a careful layout should reduce that even
further.

Thanks again for your help