silly power supply

Oh, the 1N821 through 1N829 are still somewhat available (fifty year old parts, shocking prices). Even the old uA723 had a good temperature compensated zener. And, it's so thoroughly designed-in that it's still available.

How much temp. coeff. are we talking about? For example, if the voltage is 9V1 at 20C, what is the voltage at 30C?

There's a blocking capacitor.

Cheers

Klaus

10nF at 230V/50Hz is 0.7mA, above regulatory limits for isolated systems. I am guessing you use Y caps?

Cheers

Klaus

Always is.

This was back before component data-bases. We had a stack of distributor catalogues, and deals with Philips for some high volume components, which the engimeers were expected to keep in mind. The accounts got quite stroppy about an extra few pence, and held up payments to critical supplies to milk a few pence in interest out of the money they delayed paying them.

<snip>

That often meant "until the oscillation was too small to see on the oscilloscope I was using to look for it".

On one occasion I had to cleans up a ten year-old design that worked well enough in the field. The clown who had done the work had tried to drive a long cable directly from an op amp and had chosen to swamp the cable capacitance with a 100nF ceramic capacitor to ground.

There's a standard circuit for dealing with that problem

at figure 3. The link is dated 02.12.2008 but the idea has been known for a lot longer - I applied in in 1987, and it had been well known for at least ten years back then.

You had to do tolerably careful design to get the performance out of the 1N829 you were paying for.

You only got the low temperature coefficient - +/- 0.0005% /C- you were paying for a diode current of 7.5mA and getting 1mA away from that would change the temperature coefficient that much. The actual output voltage ranged from 5.81 to 6.51V - +/- 6% - and if if you used the usual scheme to boost that to an exactly 10.000V with a trim pot to take out the voltage variation, and derived the zener current from that output with a 505R E96 single resistor, you'd have a range of +/--0.7mA on that current. What you needed to do was to fit 564R E96 resistor and leave space to fit another in parallel to get the current close to 7.5 mA when the output had been sent to 10.000V - the most likely value is 5k1 and you could probably get away with an E12 part.

One machine I ran into exhibited exactly this over-sight - somebody had bunged in a 1N829 and not gone to the trouble of getting out all the perfomance that they had paid through the nose for. The guy who had done it went on to better things - Mike Englehardt knows him too - and probably did it because his boss had overloaded him.

Well, the 1N829 comes in at 0.0005 % per C tempco, according to this

but it's unknown what test protocol they use; maybe it's fully tested (explains the price), but not likely over the whole temperature range.

The uA723 and its successors are a more reasonably priced alternative, similarly stable but included amplification so you could get buffered output (in the shunt configuration) at adjustable voltages. the 1N82x series isn't adjustable, the physics sets the output at about 7V. Figure 12 here is the shunt 'voltage reference' that gets best accuracy.

So, it's between 0.01%/C to 0.0005%/C. Even mid-range 0.005%/C should be good enough for me, as long as all of them in the same group acts the same. If my target goes over much more than 40C, i have other things to worry about anyway.

On a sunny day (Tue, 27 Dec 2022 08:34:23 -0800) it happened John Larkin snipped-for-privacy@highlandSNIPMEtechnology.com wrote in snipped-for-privacy@4ax.com:

How do you store those? Escpecially the first yype? In foam? Must take a lot of space if hundreds.

Yes multi-source is an interesting factor.

I sometimes like to build things from what I find in my scrap-box.

On a sunny day (Wed, 28 Dec 2022 02:22:33 +0100) it happened Klaus Vestergaard Kragelund snipped-for-privacy@hotmail.com wrote in <tog5oo$k3u$ snipped-for-privacy@gioia.aioe.org>:

If you mean C60 C61 output caps that does nothing to limit the drive current.

The ADR420 series is pretty good.

The ADR425 seems best suited to Ed Lee's application, though he'd have to do a bit of work to get what he seems to be after.

huh? what needs 3V?

There's no zener like that.

Is there a way to remove the bonding agent? I managed to soften the glue on some flyback transformers with IPA, but the black stuff used in some computer PSU is different. what's the tradename of that, and what solvents are effective?

Mostly yes. Deutsche Telekom switched off their ISDN BRI ports 2019/2020. I still had an ADSL/ISDN line until 1 month ago, when Vodafone also switched that to All-IP. I'm not sure about other countries, but I would guess similar developments there.

We have been making routers with ISDN ports for >20 years - our latest devices still have ISDN ports, but NT-mode only (so you can hook up your existing PBX or phone), and demand is declining fast.

So yes, I would not recommend using ISDN transformers for new designs with a long expected product life. We had multiple vendors drop these parts, and some versions (with increased insulation and mechanically bigger housing, IIRC required for Norway?) have been difficult to source for some years now.

cu Michael

On a sunny day (Wed, 28 Dec 2022 14:13:10 +0100) it happened Klaus Vestergaard Kragelund snipped-for-privacy@hotmail.com wrote in <tohfd5$1j98$ snipped-for-privacy@gioia.aioe.org>:

You are right, 20 uF at 100 kHz is about 80 mOhm

1 uF would be 1.6 Ohm..., would be too high 1/(2*PI*100000*20E-6) = 0.079577 1/(2*PI*100000*1E-6) = 1.591549

I dunno how that chip internal circuit drives those transistors. I personally prefer a system with a cycle by cycle current sense and limit. Then you only need one transistor... No coupling caps... PIC 18F14K22 has that, hardware comparator that can switch the PWM output to zero. Or am I missing something here? Most PC power supplies use only one power transistor. For the thing bottom left I'd use a LM317 with transistors to switch the sense lead... No zeners...

With 15k/470pF the switch frequency is 100kHz. 24V applied, divide by 2 due to C60/C61, so 12V for 5us@100uH is 600mA magnetizing current.

DRQ127 100uH has 3.6A saturation, but that's at 25 degrees. Probably

2.5A at high temp.

With 20uF series cap (which is way too high), seems there will be large inrush currents also, but still far away from 2.5A saturation.

Cheers

Klaus

Assuming that i float the LM3914 at 9V and power supply it with 3V, then take the high end measure range between 9V and 12V. Doable but not simple.

The 'R' of the RGB turn on at 11.5V. Below that, nothing is on and thus no power draw.