LM393 comparator output does not attain 'high'

[earckens]

In the schematic attached, comparator LM393 (IC2A) does not attain more than 3.4V when driven high (Vcc = 18Vdc).

The output (pin 1) does turn to 0V when pin 3 voltage is lower than pin 2.
However, in the other case the output does not even attain 1/4 of Vcc.

How to remedy?

Edit: the output on pin 1 also serves as a sink for LED1. Which must turn on when pin 3 goes low, and turn off when pin 3 goes high.

 

[Diver300]

I guess that something is oscillating.

Does the LED light when it shouldn't?

 

[Nigel Goodwin]

A little positive feedback round the comparator might help?.

 

[earckens]

A little positive feedback round the comparator might help?.

Say 10k between pin 1 and pin 3?
But pin3 never exceeds +/-6V. Pin 1 would be stuck there?

Oscillation: I check on my scope this afternoon and feed back here.

 

[Nigel Goodwin]

Say 10k between pin 1 and pin 3?
But pin3 never exceeds +/-6V. Pin 1 would be stuck there?

Oscillation: I check on my scope this afternoon and feed back here.

I would suggest higher than that - try 100K or 1M and see if it makes any difference, if not try lower.

 

[Ian Rogers]

Is it me? I don't do analogue, bu!!! Vref path is sitting at 70uA R1 and R2 should equal around 14k Will this work with 10 times lower current?

Sitting down now just in case I get blown out the water!!!

 

[earckens]

I guess that something is oscillating.

Does the LED light when it shouldn't?

I just put the scope on pin 1: indeed, plenty of oscillations: 400ns period, 5vpp on pin 1.

When pin 3 is low (0V), then no oscillations, and the LED turn (correctly) bright.

I set up the LM393 in lab conditions, with LED and series 4k7, and then it behaves as it should: clean on or off. And pin 1 all the way to near Vcc.

Pin 3 (input +), connected to pin 4 (LM723) cannot be interfered with (because pin 4 is crucial to current sensing circuit in the LM723);
There is of course a difference in LED brightness when LM393 pin 1 is low, versus oscillating.

Hard to see for me how to solve this right now.

 

[AnalogKid]

There is no power supply decoupling around IC1. Depending on the IC's internal circuit, a high source impedance for Vcc (and "high' in this case can be just a few ohms) can create either a positive or negative feedback path backwards through the circuit, disrupting normal operation. One chip can put enough noise on a power rail to disrupt another chip.

Power supply circuits need heavy decoupling. Each IC should have a 0.1 uF ceramic or stacked metal film cap between its power pins, as close as possible with short leads. In parallel with this add a 10 uF to 47 uF electrolytic.

ak

 

[schmitt trigger]

A little positive feedback round the comparator might help?.

When using comparators, hysteresis is your best friend.

 

[earckens]

There is no power supply decoupling around IC1. Depending on the IC's internal circuit, a high source impedance for Vcc (and "high' in this case can be just a few ohms) can create either a positive or negative feedback path backwards through the circuit, disrupting normal operation. One chip can put enough noise on a power rail to disrupt another chip.

Power supply circuits need heavy decoupling. Each IC should have a 0.1 uF ceramic or stacked metal film cap between its power pins, as close as possible with short leads. In parallel with this add a 10 uF to 47 uF electrolytic.

ak

SOLVED!! I put a 10µF electrolytic cap from LM723 pin 12 (Vcc) to GND, oscillations gone, LED extinguished.

How did i not think of this bare essentials necessity from the start (in 2016!).

 

[AnalogKid]

The classics always come through.

ak

 

[Beau Schwabe]

Ummm ... the LM393 is open collector and only driven LOW, so you may not "see" any voltage unless it is back driven from the connected circuitry.

 

[Tony Stewart]

It has pullup to 24V err 18V (edit)
1/4 of Vcc. means 3/4 Vcc across LED & 4k7 when oscillating ON/off

 

[Diver300]

It is a pullup to 18 V, not 24 V. Also if the voltage is measured at the comparator output with the output off, the voltage drop across the LED may still be there (depending on the LED characteristics at very low currents and the impedance of the voltmeter)

However, the OP's ambition is to turn the LED off, which has worked with suitable decoupling.

 

[MrAl]

The problem when using a DC meter to measure a voltage is that if it happens to be switching between +Vcc and ground it will appear to be a constant voltage somewhere between 0v and +Vcc. The voltage will depend on the pulse width, and that is what PWM is.

 

[Tony Stewart]

maybe when I wrote 24V I was thinking the 18V was full load, which is around 24V with no load.

This uA723 was a 1968 upgrade to the vintage Bob Widlar uA 709 design

This design could use some OTP and OCP protection.

I was thinking, well if it oscillates, I wonder if anyone had used the uA723 as a buck regulator and then my 1st search hit was yes, Fairchild had used it and published all the details about Linear circuit design and why this solved most of the industry problems at the time.

It may be a great read for anyone interested in historical linear regulator design from Fairchild.

https://www.repeater-builder.com/astron/pdf/uA723_Application_Notes_Fairchild_Semiconductor_1968.pdf

 

[MrAl]

Hi,

I've seen it happen before that's why I mentioned it. Long time ago I was measuring something I thought was DC but it was oscillating so the meter could not pick up the changes, it just had shown a value that looked like an incorrect DC level. What a pain in the neck that was. Took a while to figure out, but ever since I remembered to check when problems like that came up and there was no apparent reason for it.

 

[AnalogKid]

This uA723 was a 1968 upgrade to the vintage Bob Widlar uA 709 design

Wait ... what? Are you saying that the 723 was designed specifically as an improved general-purpose opamp but with extra regulator-type-stuff thrown in, or just that its internal opamp section was an improvement over the 709?

ak

 

[Nigel Goodwin]

Wait ... what? Are you saying that the 723 was designed specifically as an improved general-purpose opamp but with extra regulator-type-stuff thrown in, or just that its internal opamp section was an improvement over the 709?

ak

You've not read the post properly - it was uA709 PLUS discretes - basically the 723 was based on an original PSU design using a uA709 and plenty of extra components, the more integrated design giving better performance and using many less parts.

To be fair the 723 is a serious antique now

 

[earckens]

You've not read the post properly - it was uA709 PLUS discretes - basically the 723 was based on an original PSU design using a uA709 and plenty of extra components, the more integrated design giving better performance and using many less parts.

To be fair the 723 is a serious antique now

LOL, I love antiques: watch my logo

I first made a LM723 based PSU during my college years, in the 80's. In 2016 I decided to upgrade it, and at the time I did get fantastic help here (ie a member named 'spec' in particular). A few years and lot's of time constraints later (this year) I decided to design a proper pcb which I am now in the process of finalising.