Author Topic: LF RF getting into NAND gate from comparator. Are caps OK on gate inputs? Value?  (Read 2554 times)

0 Members and 1 Guest are viewing this topic.

Offline Chris Wilson

  • Frequent Contributor
  • **
  • Posts: 993
  • Country: gb
  • Country: gb
  • Race car engineer, dog lover, hoarder.
I have a situation where powerful 136kHz RF is getting, I believe into a wire about 10 inches long, between the output of a 393 open collector comparator to the iput of a triple NAND gate. I intend trying screened cable. A high value (50k) pull DOWN resistor on the input pin of the HEF4023 triple NAND gate seems to help, the pull UP resistor is on the comparator and 3.3k. Maybe the resistor is damping the RF? Should I put a cap to ground across this input pin to the NAND gate, and / or the output of the comparator? What value 1nF or higher? Unused input pins are grounded. I can post the schematics if required. Symptom is false partial triggering, perhaps oscillation of one or other when I wind the RF level of this amp up. The comparator reads the output of a Hall effect current sensor, and upon reaching a voltage commensurate with over current to the PA FET's shuts their input off. At high RF power the NAND device has a wobbly (greenhorn technical term.... ;)) Thanks!
Best regards,

                 Chris Wilson.
 

Offline wraper

  • Supporter
  • ****
  • Posts: 10059
  • Country: lv
  • Country: lv
I think you should use some IC with Schmitt trigger inputs instead of HEF4023. Does compactor on LM393 have a hysteresis? Because if there isn't, most likely, problem is there. Did you check actual signal from LM393?
 
The following users thanked this post: Chris Wilson

Online Zero999

  • Super Contributor
  • ***
  • Posts: 12834
  • Country: gb
  • Country: gb
  • Hero999
Try a 10nF capacitor across the gate's input and if that doesn't work, add a ferrite bead with a few turns of wire.
 
The following users thanked this post: Chris Wilson

Offline Chris Wilson

  • Frequent Contributor
  • **
  • Posts: 993
  • Country: gb
  • Country: gb
  • Race car engineer, dog lover, hoarder.
The LM393 has a 3.3k pull up resistor to 12V rail voltage on its output, and it DID have a feedback resistor for hysteresis, but removing it seemed to improve stability.... This evening's job is to `scope the output of the 393 and the input of the 4023. The thing is at a fairly late stage of development, changing the 4023 is something i would like to avoid if possible. I have added extra LED's and flip flops to accurately show what fault triggered the amp's shut down, others have this amp working wothout issue, I am *fairly* sure lead length and RF pick up is the issue. Not being too competent with a `scope makes me not too optimistic about looking for noise accurately, we'll see. Thanks for the reply, all opinion very welcome! Here's a couple of photos of it being bench tested, the wiring is of course very temporary and worst case scenario...
Best regards,

                 Chris Wilson.
 

Offline wraper

  • Supporter
  • ****
  • Posts: 10059
  • Country: lv
  • Country: lv
Try a 10nF capacitor across the gate's input and if that doesn't work, add a ferrite bead with a few turns of wire.
Intentionally making slow slew rate on the CMOS input?  :-// http://www.rcl-radio.ru/wp-content/uploads/2014/11/SCBA004.pdf
Quote
A slow input edge rate, coupled with the noise generated on the power rails when the output switches, can cause excessive output errors or oscillations.
« Last Edit: December 20, 2016, 04:21:07 pm by wraper »
 

Offline Chris Wilson

  • Frequent Contributor
  • **
  • Posts: 993
  • Country: gb
  • Country: gb
  • Race car engineer, dog lover, hoarder.
Try a 10nF capacitor across the gate's input and if that doesn't work, add a ferrite bead with a few turns of wire.

OK, will do, I don't have "stand alone"  ferrite beads but do have leaded ferrites, I can try one of those, thanks again Hero999  8)
Best regards,

                 Chris Wilson.
 

Offline wraper

  • Supporter
  • ****
  • Posts: 10059
  • Country: lv
  • Country: lv
The thing is at a fairly late stage of development, changing the 4023 is something i would like to avoid if possible.
Dunno if prototyping boards can be called "a late stage". You just need to throw different logic IC which will do the same thing, IMO won-t take more than an hour to do this. Do you use all 3 inputs simultaneously on any of the NAND gates.
 

Offline wraper

  • Supporter
  • ****
  • Posts: 10059
  • Country: lv
  • Country: lv
Also, if you don't really need open collector output, you could use LM358 instead, to provide stronger positive drive.
 

Offline wraper

  • Supporter
  • ****
  • Posts: 10059
  • Country: lv
  • Country: lv
BTW, unless they are on the bottom, I don't see any decoupling capacitors (electrolytic does not count). EDIT: also, aren't you, by chance, using the same power for logic and RF part without any filtering in between?
« Last Edit: December 20, 2016, 04:41:02 pm by wraper »
 
The following users thanked this post: Chris Wilson

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 6080
  • Country: de
  • Country: de
Adding caps is likely not a good idea, as this makes the transition slow and thus even more sensitive.

If at all you might consider 50-100 Ohms in series at the comparator side to give it a kind of series termination in the low state. A smaller pull up might also help a little.
 
The following users thanked this post: Chris Wilson

Offline SL4P

  • Super Contributor
  • ***
  • Posts: 2045
  • Country: au
  • Country: au
  • There's more value if you figure it out yourself!
Are there any bypass caps close to your supply pins on the NAND chip?
Don't ask a question if you aren't willing to listen to the answer.
 

Offline Chris Wilson

  • Frequent Contributor
  • **
  • Posts: 993
  • Country: gb
  • Country: gb
  • Race car engineer, dog lover, hoarder.
Also, if you don't really need open collector output, you could use LM358 instead, to provide stronger positive drive.

I don't need open collector, that sounds a fairly simple swap, thanks!
Best regards,

                 Chris Wilson.
 

Offline Chris Wilson

  • Frequent Contributor
  • **
  • Posts: 993
  • Country: gb
  • Country: gb
  • Race car engineer, dog lover, hoarder.
The thing is at a fairly late stage of development, changing the 4023 is something i would like to avoid if possible.
Dunno if prototyping boards can be called "a late stage". You just need to throw different logic IC which will do the same thing, IMO won-t take more than an hour to do this. Do you use all 3 inputs simultaneously on any of the NAND gates.


Prototyping board is late stage for me ;)  It won't go to PCB I am afaraid. I do use all three inputs, the circuit is attached, the red cross shows where the comparator is fed in. I initially had the 10K pull up shown by the NAND (IC6) but then removed it and moved it to the comparator output instead. I have added flip flops at the pink stars, to change and hold state as outputs go high, and via an inverter where they go low, feeding 3 LED's to show what tripped. These are not shown, nor is the comparator for over current being discussed. Thanks again.
Best regards,

                 Chris Wilson.
 

Offline Chris Wilson

  • Frequent Contributor
  • **
  • Posts: 993
  • Country: gb
  • Country: gb
  • Race car engineer, dog lover, hoarder.
BTW, unless they are on the bottom, I don't see any decoupling capacitors (electrolytic does not count). EDIT: also, aren't you, by chance, using the same power for logic and RF part without any filtering in between?

Each IC has a 220nF decoupling cap right by the VCC pin, as do the comparators. The power is from a linear supply at 17.5V DC to two IC regulators, both on the RF heatsink. These feed 5V solely to the Hall effect current measuring mini PC, and 12V to the rest of the stuff. The 12V and 5V regulators have Tantalum caps on their output pins right by them. The 12V is shared with RF and logic, the power for the 4 output FET's is from a separate 100V supply. There is no filtering between the baords, no :( What could I add....? Thank you!
Best regards,

                 Chris Wilson.
 

Offline wraper

  • Supporter
  • ****
  • Posts: 10059
  • Country: lv
  • Country: lv
I do use all three inputs, the circuit is attached, the red cross shows where the comparator is fed in.
Actually you don't, unless you use 3rd NAND gate not shown on the circuit. Both of the 3-input NAND gates have one of the inputs connected to positive rail, which effectively makes them 2-input NAND gates. You could just throw in 4093 (Quad 2-Input NAND Schmitt Trigger) instead.
http://www.ti.com/lit/ds/symlink/cd4093bc.pdf
 

Offline wraper

  • Supporter
  • ****
  • Posts: 10059
  • Country: lv
  • Country: lv
The 12V is shared with RF and logic, the power for the 4 output FET's is from a separate 100V supply. There is no filtering between the baords, no :( What could I add....? Thank you!
You could throw in inductor or ferrite bead + decoupling capacitors  to their power rails, and only then connect both of them to the PSU.
« Last Edit: December 20, 2016, 05:22:50 pm by wraper »
 

Online Zero999

  • Super Contributor
  • ***
  • Posts: 12834
  • Country: gb
  • Country: gb
  • Hero999
Try a 10nF capacitor across the gate's input and if that doesn't work, add a ferrite bead with a few turns of wire.
Intentionally making slow slew rate on the CMOS input?  :-// http://www.rcl-radio.ru/wp-content/uploads/2014/11/SCBA004.pdf
Quote
A slow input edge rate, coupled with the noise generated on the power rails when the output switches, can cause excessive output errors or oscillations.
I know but it's a slow CMOS HEF4023 IC so I doubt 3k3 & 10nF would be a big deal. If it is, reduce the size of the capacitor.

The 12V is shared with RF and logic, the power for the 4 output FET's is from a separate 100V supply. There is no filtering between the baords, no :( What could I add....? Thank you!
You could throw in inductor or ferrite bead + decoupling capacitors  to their power rails, and only then connect both of them to the PSU.

Be careful, that could be prone to ringing, if the Q is high. It may be necessary to add a resistor in parallel with the inductor and in series with C1 & C2. It needs to be critically or over damped.
 

Offline Chris Wilson

  • Frequent Contributor
  • **
  • Posts: 993
  • Country: gb
  • Country: gb
  • Race car engineer, dog lover, hoarder.
I think you are on the nail re the power supply. I had chance to do some testing last night and today and looking at the outputs of the two regulators, idle, or with the control circuitry working and no power to the amplifier's 4 power output FET's they are rock solid 12V and 5V. As I add power to the FET's and the amplifier starts amplifying RF they saty solid until I wind some power on, then the 12V one goes from 12.00 to 12.23 and the 5V regulator output drops from 5.00 to 4.13. The 5V one just feeds a current sensing module, but the 12V one feeds everything including the comparators. When I see the false LED trip it's on a circuit that gets its input from a comparator measuring the output of the current sense module.

This is normally 2.5V with no current draw, and rises or drops according to the polarity of the current draw. So with the 5V regulator feeding the current sense module going low that affects one input to the comparator, and with 12V regulator going high, that affects the resistor divider network for this and another comparator. .... Would scrreened power leads help, or do I need to look into chokes as above? Maybe my physical build methods for the circuitry are at issue? Thanks.
Best regards,

                 Chris Wilson.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf