Decouple, decouple and do it again..

[Steve Anderson]

Dear all,

One thing that I generally omit from my circuit diagrams is the supply decoupling...this to me is assumed. Now others here have had problems and this is (probably) down to this.

If a circuit misbehaves in some manner beyond simple errors in wiring it's probably supply decoupling. Stock up on a shite-load of 10nF disc ceramic capacitors AND FIT THEM. In the rush "To get things working", you're likely to be disappointed.

There is no fixed 'Rule of thumb', but in a HC logic circuit (unless at really high speed) about one cap per four chips should be enough. Analogue op-amps one per supply rail will not hurt.

A small effort and small cost makes it all much more rewarding...

Also fatten up earth/ground leads, remember HC chips will respond to stimuli of only some 5ns wide. AC even less...the HEF/CD series, no problem if you don't mind waiting a week for a result...

Steve A.

[AncientBrit]

That's a fair point to make, Steve, about decoupling being assumed and not shown on circuits. Useful advice for those just setting out on circuit development.

I must admit I tend to be fairly economical with rail decoupling, then add a few extra caps as (and when!) problems show.

Like you say HC devices can produce the odd spike and cause havoc in feedback counters.

Phase lock loops and gated oscillators tend to be very sensitive, with timing modulation showing up as a wavy edge to scan edges.

I've also found that derived mid-rails (for single rail op amps) need special attention.


Regards,


Graham

[Steve Anderson]

...Useful advice for those just setting out on circuit development. I've also found that derived mid-rails (for single rail op amps) need special attention. Regards, Graham

A very valid point Graham, something that is often glossed over in texts and electronic courses...most of us find out the hard way. Here's the nub of it...

Decoupling is used (also called bypassing sometimes) where a supply from a battery or a power supply can have unwanted noise on it which upsets in some manner another part of the system using the same supply in some way. A bit like interference on AM radio. A NBTV specific example is the use of a brushed DC motor (standard cassette-type) to spin a disc. Decoupling (perhaps 'uncoupling' would be better) is a bit like seperating the coaches or cars on a train, so that one cannot affect another.

Quoting Horowitz & Hill from "The Art of Electronics"..."In general it is a good idea to bypass power-supply leads to ground liberally throughout a circuit, using a combination of ceramic types (10-100nF) and electrolytic or Tantalum types (1-10uF)." In fact there are no less than ten references to bypassing in the index scattered throughout the book.

Good advice, one ceramic capacitor for four logic chips, and an electrolytic per board is the rule of thumb I go by unless the chip datasheet says otherwise, mainly A-D & D-A chips or hi-speed analogue circuitry (not NBTV!).

The other reason...don't groan...is stability, this mainly applies to analogue circuits, but is not unknown in simple logic too, especially Schmitt-trigger inputs. A simple short wire from a battery has an inductance, small, but it's there. Combine that with its opposite number (positive or negative as the case may be) and you've got a capacitor. Throw in some board capacitance, PCB or Vero track resistance & inductance and chip input capacitance and you've got one rather nasty resonant circuit awaiting to bite on the...wherever.

Even expensive regulated power supplies don't escape this. Often people use a simulator to evaluate a circuit on a PC, all fine and well, but the simulator assumes (usually) that all components are 'perfect', no lead inductance, no board capacitance and the power source is perfect and noise-free too. Sorry, not real life! They are very useful, but you must remember they don't work in the real world.

So stock up on disc-ceramic 10 & 100nF caps (50V)..cheap, under a penny in 100-off quantities and a load of 47uf electrolytics (16/25V).

Attached is one of the boards for "The Brute" undergoing testing (hence that's why it's a mess) but you'll note no less than seven ceramics and one electrolytic. Also note the short fat green ground leads to the chassis.

Steve A.

[Steve Anderson]

Phase lock loops and gated oscillators tend to be very sensitive, with timing modulation showing up as a wavy edge to scan edges. Regards, Graham

As for phase locked loops especially the CD4046 used so often for NBTV I can only again quote H & H, "Note: the CD/HEF series of 4046 PPLs have a chronic and severe supply sensitivity disease; check the graphs in the datasheet."

Hence the rather unmentioned and glossed-over 7V Zener included in the package (Pin 15) to provide local regulation. This is omitted in the HC/HCT versions which are far better.

Now I will admit that this primarily applies to the VCO (unused in the usual NBTV motor control circuit), but it doesn't bode well.

Steve A.

[AncientBrit]

re CD4046

Too right Steve, I've been there!

Cheers,

Graham

[Steve Anderson]

re CD4046 Too right Steve, I've been there! Cheers, Graham

It does work and millions have been put into all sorts of goods, but it is severely compromised in a number of areas. It's fine with nice stable inputs, frequency generation and wide-band FM demodulation for example, but used for NBTV motor control it is like trying to get a dog to walk on its hind legs.

The problem stems from using a single Silicon process for both the digital parts (flip-flops and gates) and the analogue parts (VCO and input amplifier). The process used in later chips of the HC/HCT/AC/ACT series is better suited to this but does limit the supply voltage to 6V.

It's on my (rather long) list of 'To Do' things for NBTV. I hope to get around to it before the next millennium. I might shift up the ladder in terms of priority.

Steve A.

[dominicbeesley]

On this decoupling subject I've tended to use foil caps rather than ceramics - I had a box of them which I've now exhausted. Do ceramics have an advantage other than cost? [I'll be buying some next time I can get together some money for an order!]

Now I know what the zener is for in the 4046 - they do tend to gloss over why you would want regulation in the datasheet so I never used it I think I may in future!

Cheers

Dom

[Steve Anderson]

Do ceramics have an advantage other than cost?...Now I know what the zener is for in the 4046 - they do tend to gloss over why you would want regulation in the datasheet so I never used it I think I may in future! Cheers. Dom

Disc ceramic caps generally have lower self-inductance and this helps in 'mopping up' voltage spikes on the supplies caused by logic or some other form of switching (DC motors). The lower self-inductance also helps in analogue circuits in retaining stability.

However their initial tolerance and temperature coefficient is lousy unless specifically made for resonant circuits. The usual 10 and 100nF jobbies should really only be used for decoupling.

Yes, they are rather coy in confessing to the function of the Zener in the CD/HEF4046, it's primarily for the VCO (unused in the NBTV motor control circuit) and if used for the rest of the device you end up with awkward 7V logic pulses...so it's not much of a help at all.

Steve A.

[Viewmaster]

Although this is an old thread I can tell a recent story.

Terrible glitches and spikes on the NBTV video going into the sync seperator and video amplifier gave eratic picture. When motor was off all was OK. New motor and big C made no difference.

After much probing and searching realised that the earth return from the seperate power supply to the source connector of the IRF510 motor control was via other earth strips and leads on the veroboard supplying the 4046 etc.

I then connected the earth return straight onto the source and lo and behold the spikes have gone.

When will I learn?

[Harry Dalek]

Although this is an old thread I can tell a recent story.

Terrible glitches and spikes on the NBTV video going into the sync seperator and video amplifier gave eratic picture. When motor was off all was OK. New motor and big C made no difference.

After much probing and searching realised that the earth return from the seperate power supply to the source connector of the IRF510 motor control was via other earth strips and leads on the veroboard supplying the 4046 etc.

I then connected the earth return straight onto the source and lo and behold the spikes have gone.

When will I learn?

Albert i had a problem like this stopping my stepper motor due to vibration earthing or grounding from the power supplies to each possible earth in the circuit i tracked it down to my 555 clocks ground and that fixed it ...I was using a battery operated soldering iron so must of not soldered correctly.

If something should work and doesn't the reason is more likely your own fault than the circuits .

[Steve Anderson]

If something should work and doesn't the reason is more likely your own fault than the circuits .

Sage advice from someone who has obviously "Been there and done that."

Albert, a timely reminder regarding this, so hopefully newer members will take heed of the advice.

Earth returns can be as tricky to deal with as supply noise and dry-joints.

If you are new to electronics read through the whole thread above, who knows it may stop you taking up golf instead!

Steve A.