PI controller capacitor

MLCC capacitors are not recommended because of piezo-effect related noise. How bad/serious is it? I haven't used ceramics in this application yet. I need time constant in the order of tens of seconds. Cannot find tantalums with high enough RC specs, film caps are too big. Any suggestions - ? Thank s!

Reply to
Michael
Loading thread data ...

You could try some. I haven't seen serious microphonics in ceramic caps on PC boards.

MLCCs often have terrible capacitance vs voltage behavior; I posted a horrible example recently. NPOs are good but only come in smaller values.

Film caps and high-value resistors (and clean boards!) are a good way to get long taus. 2 uF and 5 megs is 10 seconds. 4.7 uF is an affordable film cap.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

I use a lot of 10 uF Panasonic film caps. (100V?) Big... ~1 in^3. We get 'em by the bagful at ~$2 each. I'm not sure what the market is, some future lifetime buy? I should talk to whomever we buy them from (ITT?).

George H.

Reply to
George Herold

The prototype oscillated until I raised time constant to 470s (10M & 47uF ceramic). I was wrong when I mentioned tens of seconds. It is more like minutes. Higher resistance would mean CLEAN boards.... I do not want to go there. Years ago I dealt with 500M resistor (pA TCA). What a pain! I found interesting app note by MLCC manufacturer after this posting (forgot the name, have app note at work). They mention couple interesting remedies:

  1. cut slot in the board under the cap (do not see how it works)
  2. place caps on opposite sides of the board under each other (neat!)
Reply to
Michael

Ceramic is fine for boring applications.

C0G comes in values large enough to be useful in audio frequency range controllers (like SMPSs), though there, the noise requirement isn't going to be low enough to justify the cost (more than ~1nF, C0G is noticably more expensive). For a low noise controller (~uV?), it would be justified.

X7R is noisy (due to piezo) on the order of 10s of mV. Scope a board that's got caps on it, and tap on the board. See what happens.

Tantalum is fine. Dry slug may be leaky at first (and self-heal-ey). (Still more reason to overrate them generously!) Wet slug of course is notoriously expensive, but about as ideal as you can get, in the fractional Hz spectrum (possibly given a reforming cycle, plus a long absorption time constant that can be annoying).

Electrolytics are good for simple, sloppy controls. I don't know how they compare on noise. Leakage isn't horrible. ESR change with temp basically doesn't matter.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

What kind of system is it? Thermal?

Really shouldn't be. A good PCB should have tons of gohms of IR.

Use a uP!

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

. How bad/serious is it?

ms with high enough RC specs, film caps are too big.

Tantalum specs are determined by what the manufacturer can measure quickly.

Once you get a tantalum polarised, the leakage current decreases steadily f or quite a while, and time constants of the order of a minute are perfectly feasible. Goeoge Kent used them in process controllers for year before mic rocontrollers became available - I was working for George Kent from 1973 to 1976 when this was happening. In fact I had a founding subscription to Byt e, and used to take my copy into work where it always got borrowed ...

--
Bill Sloman, Sydney
Reply to
bill.sloman

Jeeze, I used to do PID controllers like that before 1990. Put a micro in there and be done with it!

--sp

--
Best regards,  
Spehro Pefhany
Reply to
Spehro Pefhany

It is thermal. Analog seems to be the quickest/simplest for me. I might stay analog until I retire (haven't written code for almost couple of decades). As to tantalums leakage going down in time, I read about it, but could not find any data about how long it takes and how high RC one can get after dust settles.

Reply to
Michael

Can you move the sensor to be closer to the heat/ cool source? That will reduce the lag. I use a rosin based solder (Kester 44) for high resistance stuff, 100 Meg or 1 Gig seem to work fine.

George H.

Reply to
George Herold

Right. Thermal systems are diffusive, so the phase shifts are awful. Keeping the sensor close to the heater make them more 1st order, easier to close a loop around.

Yes. Keep the parts away from water soluble flux.

Maximally grungy rosin soldered thing, with fingerprints:

formatting link

Pins the needle on the 1e14 ohm range:

formatting link

You can buy surface-mount resistors up to 1 T. And opamps with fA bias currents.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

The sensor is next to the heater - I figured it out... experimentally. Last time I used high value resistor (500M), the board needed to be baked (used 60W incandescent desk lamp overnight for lack of thermal chamber) after cleaning - doubt our MFG will like it.

Reply to
Michael

We do high-Z stuff with rosin solder and solvent cleaning, no specific bake. If you use a water-wash process, it will be difficult to avoid hygroscopic crud lurking under parts.

Another control loop issue is that if your controller output drives a resistive heater directly, power/heat will go as v-squared, and that makes the loop nonlinear and harder to stabilize. PWM is one way to linearize the heater.

SCR phase control can be somewhat linearized too, with a "cosine modified ramp" circuit.

Simple on-off thermostats sort of always work.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

I use LT/Analog TimerBlox. Bang-bang controller we are using now has ~+-0.15deg swing, apparently not good enough

Reply to
Michael

You can do a gain only loop, if you don't care about a little bit of offset. If things are mostly the same you can even work out the offset as a function of heat delivered.

George H.

Reply to
George Herold

If it's a linear controller, then the controller dissipates power complementary to the resistor. So... Put both on the same heatsink and it's linear again. ;-)

See:

formatting link
Note the 7805 and pass transistor are soldered to the heat spreader, so that heat is linear with control output. The emitter resistor is inconsequential in this case, but could be made more significant if it's necessary to reduce the dissipation of the transistor itself.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

Easiest way to generate a delay is digitally using CD4060 or similar clock cap and binary counter in one chip with rising edge and disable or gates.

Reply to
Anthony Stewart

If the conductors are exposed to dust and %RH, it will eventually become dusty and faster with a fan. 10M is a practical limit for long term.

An air slot reduces chance of creepage currents from above which with HV leads to degradation of BDV from 3kV/mm to 300V/mm in extreme cases .

Reply to
Anthony Stewart

Using control element as a heater sounded very attractive (put whole controller there, no need for temperature stable parts) until I was told that in the future target temperature may be ~75degC+ (vs 35degC now). Oops!

Reply to
Michael

How much power? I just did a FET heater thing that goes to 400 K (~130 C).

George H.

Reply to
George Herold

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.