I have the same kind of aha situations when reading documentation in Spanish :-)
But every time I suggest something like this the firmware guys throw stuff at me. Because they are the ones who have to implement it.
There is another variant of this trick: Use only the one CC register output directly and go into the comparator via a fixed RC. If you don't want to sacrifice an OC port pin you can use a diode to discharge the cap quickly in the other direction.
Then use a 2nd timer for PWM, RC filter that as well with a much larger time constant and use this to feed the other comparator input. Then the SW loop only has to adjust a duty cycle, which moves the comparator threshold.
This is looking even better now! The PIC I'm using has a 16-step adjustable reference for the internal comparators so possibly only a single RC (with a diode across the R) may be needed.
That's the trick, run one of them hog-tied. But put in a governor, one that only comes on if it starts over-revving.
Best to put in a governor. Whenever you use SW loops to control a switcher there should always be a redline limiter anyhow. I have seen it too often, someone whips out a cell phone, turns it on, the uC gets perturbed and loses it, a spindle drive runs the sled past the limit switch, way past ... *KABLAM*
It's been done before. But in the design review it results in initial utterances of disgust :-)
Cool. You might need another resistor across the cap to keep it in this
0-66% band.
However, to make this successful for a production run you have to solve the newly introduced problem that you mentioned, the 15/16th to 0/16th step, without resorting to precision caps and stuff. Note that I bit my tongue and didn't say "... but that's just software" :-)
I wasn't arguing that they didn't exist, merely that if you claim that such a device is available off the shelf, you are obliged to point to at least one example.
Sure. And I can recommend ACE-Wikkeltechniek ( snipped-for-privacy@acewikkeltechniek.nl) to anybody in the Netherlands. Sadly, EPCOS won't wind prototypes for you - as I said, coil winding is still pretty much a cottage industry. Some of the bigger cottages run to machines that can more-or-less mass produce toroidal transformers, but the capital involved still falls a number of orders of magnitudes short of the EPCOSs of this world.
That would be like requiring someone to point out bread on the shelves at Safeway, after claiming that bread is available in America. Or sand on the beaches.
Sure they take custom orders. Epcos is now part of TDK.
formatting link
Quote "Even single items can be ordered! ... Customers can order low quantities for product development, special usages, and development studies including orders for single products. Please contact us about low-quantity orders".
Now how much more clearly could they say it?
Then just buy'em ready to eat:
formatting link
But as outlined above, you can also get custom orders processed there. I personally prefer smaller companies because they can do more cutting edge stuff. End of last year I had one flyback transformer where most everyone said "It can't be done" and we did it anyhow.
Rubbish. Tapped inductors are relatively rare. Pulse Engineering isn't a broad-line distributor - if you'd found something at Farnell/Newark who are the electronic equivalents of Safeway, you might have a case.
s.
o
me
There's a big gap between supplying small quantities of cores and formers for development work, and actually winding a coil onto one of those formers. If you wave a big enough order in front of them they may even make you special purpose ferrites, but they'd have to see the prospect of very large sales before they'd devote much time or resesource to your particular problem.
... - as I said, coil winding
If that's exactly what you want, and you can order a big enough batch to get their attention. It's less than obvious that this would do what the OP wanted, and highly unlikely that you could get anything that wasn't on the data sheet if you weren't buying for a 10,000 unit a year market.
That happens. They mostly mean "it can't be done the way we usually do it". Finding another way is the trick, but as soon as you have done it, the same people will tell you that it's obvious.
With the coupled boost converter solution, you need to be carefull with the FET ratings.
You would assume that you can rate the VDS lower due to the autotransformer action, but leakage inductance ringing forces you to add a snubber. Other solutions is to add a diode from the tapped point to the output to quinch t he ringing and regenerate the leakage energy, but then you cannot use lower voltage rated FETs.
It's not too slow. 25ns delay on the rising edge, 35ns on the falling edge.
I wasn't able to use all 16 steps, the lower 3 were useless because of the slope of the discharge at the falling edge. I've thrown away the top 5 to give 8 reasonably linear steps (and it makes the software easier).
With a 250ns period, I need a value of 0-7 to give a delay of
0-218.75ns to the rising edge of the pulse.
With a 1K65 and a 220pF, and a BAT85 across the 1K65, I get:
Value Delay (ns)
0 13
1 44
2 72
3 100
4 130
5 161
6 193
7 226
I need to change the duty cycle and the comparator reference while the output's low.
I can just map an interrupt input to the pin that's assigned the comparator output, set to trigger on a falling edge (once, when I want to update the period). With a maximum 50% duty cycle, I'll have at around 12.5us to change the values, plenty of time.
An extra 3 bits of resolution for 'free'!
I was hoping to find the reference settling time in the datasheet but no, it'd listed as 'TBD'.
Rubbish? You really need to get out more. Further down I have given another example from Epcos. Now if that's not a company you would consider mainstream then I can't help you.
As for tapped inductors, you can easily buy those off the shelf from Delevan, Bourns, Cooper, Sumida, TT, Wuerth, and on and on. A custom order would have been needed in 1970 but this is the 21st century.
So you want something from Newark? Voila:
formatting link
formatting link
Why were you not able to see those yourself?
Yesterday your story was, quote "Sadly, EPCOS won't wind prototypes for you". And that is obviously wrong.
As for NRE, that is the same with all companies or they'd go under. Either you pay hefty NRE or you'll have to dangle a big order in front of them. What is so unusual about that?
This was merely one example. You can get lots of these off the shelf, and David could just pick one that is for European voltage supplies, then use it in reverse. Has to be big enough to ferry the 4 watts across because at his low frequency of 40kHz the energy has to be "tossed over" in chunks. Most such flyback transformers are optimized for 100-150kHz, for obvious reasons.
If you read up on how flybacks work you'll see that such supplies are very flexible when it comes to turns ratio. There are a few trade-offs but I have designed converters that can be steered more than 3:1 in their output voltage.
Many of them weren't even willing to try a different more modern ferrite even after I looked up all the data for them. In that business, one quickly finds out who are the true leaders in the industry. The ones with the loudest advertising are often not the leaders.
action, but leakage inductance ringing forces you to add a snubber. Other solutions is to add a diode from the tapped point to the output to quinch the ringing and regenerate the leakage energy, but then you cannot use lower voltage rated FETs.
Yeah, most flyback architectures need a snubber, and that wrecks efficiency.
Without a snubber, the energy left over in tne magnetizing inductance gets lost just the same, but it makes a lot of ringing noise and overvoltage, which might be OK.
I guess it might be possible to time the period of the fet drive to synchronize with the magnetizing ringing, to use instead of lose some of the energy.
The OP is running low power, so he can just buy a higher-voltage fet without a big penalty, and let it ring. Or run in continuous mode.
--
John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com
Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators
Sloman is generally both obnoxious and wrong, a stupid combination. Digikey offers tons of dual-winding inductors from several vendors. The duals make nice transformers, too. Most are surface mount. Coiltronics DRQ series, as one example.
He also seems to think that Farnell is the only distributor in the world.
--
John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com
Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators
Probably a good idea. I was going to put a shunt in the FET source to cut the gate drive on overcurrent (if the SW leaves the drive pin high for too long) but I could also do a similar thing if the output voltage rises way too high.
For real fun, you want to try using a uC to control a buck regulator that supplies the power *for* that uC from a much higher voltage rail. Startup is the best bit :-) I breadboarded in once and sort of got it working, but never dared use it.
Yay! That's always a good feeling. A free lunch after all :-)
Might be a good thing to ask their application engineers. With uC I found that manufacturers sometimes have data that hasn't made it into the datasheet because they thought it wasn't urgently needed. Even with lowly device like zener diodes I was often able to get min-max data from their own QC at currents much lower than spec'd.
I've seen worse: 115V AC (!) was supplied to a port pin, via a resistor, cap or something. The parasitic substrate diode acted as a zero-Dollar rectifier. The voltage was not regulated but was kept within an allowed band via executing dummy code during times where the uC would normally be almost idle, so that the uC presented an almost constant load.
The problem: Nobody told me. So here I was, touching the OSC-IN pin with a scope probe, the crystal hicced up ... *BAM* ... had a crater in the uC :-)
An example of a rather specialised transformer for a very specific application. This is the catch with mass-produced transformers - there are a lot of parameters which can vary, so the stuff which is produced in volume is rarely what you want for your specific application.
Well, 750311308 - the WURTH ELEKTRONIK- and b82802a0030a220 - the EPCOS fly-back transformers aren't - technically speaking - tapped inductors, but three or four independent windings on the same core.
The reason I wasn't able to see them for myself is that the Farnell/ Newark search engine is absolutely no help when it comes to finding them. They were there, but they were thoroughly buried needles in a very large haystack.
"Flyback" is the only search term that works, which is odd, because the transformers themselves would work fine in a perfectly conventional Royer inverter, or the slightly less conventional Baxandall inverter, where flyback is either a nuisance - in the Royer inverter - or evaded, as in the Baxanadall inverter
g
ecs.
to
r me
Not for most of us. I'd be very surprised if you could get their attention with less-than-mass-market production quantities
NRE is not the whole story. People like EPCOS won't set up a big prototyping section, that everybody can exploit - they'll limit it to customers who can be expected to buy a lot of cores and formers, and I'd be surprised if a low-volume producer would actually be able to buy its services at any price.
Nothing, but I doubt that even a big NRE will automatically buy into that service - they'll want to keep spare capacity for people who might buy a significant volume of cores and formers.
... - as I said, coil winding
Sure. but if you want high efficiency, flyback usually isn't the way to go. Getting custom transformers wound isn't such a big deal that we should be recommending exploiting off-the-shelf transformers in sub-optimal circuits.
I
That doesn't surprise me either. The people who control the advertising budget tend to be a long way from the people who produce the stuff advertised. Management by walking about is a technique for narrowing that gap, but not every manager is willing to talk to people well down the organisation chart.
John Larkin finds it obnoxous when I prove him wrong, which happens too often to suit him. He reduces the damage to his sensitive psyche by practicing denial, which is indeed stupid, but he can't do better.
uals make nice
They do offer a few. The chances are they won't offer one that will do the job you want.
John Larkin infers this from the fact that I usually use Farnell as my example of a broad-line distributor - I'm used to the web-site, and know my way around it. This doesn't mean that I think that it's the only broad-line distributor - my most recent reference to Digikey as a distributor was on the 17th April this year, in the thread "Grinding Ferrite" but John Larkin's memory is as weak as his logic.
Of course you do. You've got more control over them, but some inter- winding capacitance is inevitable.
Back in the last century I built a couple of electronic ignition circuits to drive the induction coils in my cars. I used SCR's to deliver the multi-ampere currents involved, but transistors were developed for that market.
tandard_3279_A2.pdf
Sure you can, if you don't mind the communications problems. Getting something made by somebody you can talk to face-to-face has its advantages. Development time saved can be worth quite a lot of money.
In custom design, you get what you pay for. Ferrite core powder is quite a way from completed ferrite cores - EPCOS/TDK does add value when it turns them into cores.
Copper density isn't great.
-
It looks as if Baxandall invented his inverter for generating roughly
1kV DC voltages for photomuliplier tubes from 24V DC rails. I can't guarantee this - it based on gossip with a very good engineer who did his electronic apprenticeship at the Royal Radar Establishment at Malvern, supervised by Peter Baxandall amongst others. Jim Williams latched onto it for CCLs much later.
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.