I've been aways for a few days and the spam activity seems to have gotten worse in the meantime.
Well, IIRC you once mentioned some some bad experience (high latchup sensitivity) with some TI ADCs.
Can you tell us a bit more?
I right now have some pretty weird behavior with an ADS8327
formatting link
(all external causes thoroughly ruled out).
That is, about a 30 LSBs dead band or jump around each of the 2 or 3 MSBs transitions.
Strangely this seemed to depend on the number of power up, but now that all the small power up glitches have been removed the behavior still is there (new ADCs of course).
Not surprisingly, TI support is as responsive and helpful as a petrified mollusk so we too are going to ban the part.
It's a capacitor based SAR so there is a chance that the smallest (and thus most spike-sensitive) capacitors in the DAC feedback have blown. This would indicate ESD damage.
If it really depends on the number of power cycles maybe the input receives hard signals before the power is there. OTOH there is usually a buffer amp and if that blows the whole converter would die. Does the reference get applied (hard) before power to the converter comes up? Even if it's just tens of nanoseconds that is an area to look at with a DSO.
TI support has become the pits IMHO. Interesting to see that it seems to be the same in Europe. They just don't answer. What are they thinking? For a semiconductor company that's almost like suicide in the long run.
If it's any consolation I am waiting for an answer to a couple of not too complicated technical questions for over a month now. Can't believe it.
I don't recall a TI adc problem. We did have horrible problems with the THS3062 opamp early this year. After two months of pounding on TI support, without response, we finally got the VP of Ethics involved, and they finally confessed that the part is indeed broken.
We had latchup problems with an Analog Devices flash ADC, which they fixed by discontinuing the part without notice. And we did have latchup problems with some Burr-Brown octal DACS, before the TI takeover.
And Bob Pease promised me he'd fix the LM45 but, ten years or so later, it still latches.
Do you have his email? I have to write, too. It used to be a great company, has been good to me, but they are about to commit suicide by lack of tech support.
Hey, he's retired ... let him have some fun. Got a call yesterday. A dear biz friend passed away. Great guy, worked pretty much until his last days, never really retired because he had so many more ideas to pursue. Very sad.
on a completely unrelated note, have you seen this? Performance of a Diode Clamped Linear Amplifier, Fujita & Yamashita, IEEE trans. power Electronics vol. 23 no.2 march 2008
--
"it\'s the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
its what you come up with if you study multilevel converters then look at linear amps with multiple output transistors and different voltage supplies. quite cute.
dunno, its all geek to me. have winzip, will travel. And the .rar version was about 15% smaller than the .zip version.
but I gave up and brute-forced it one page at a time :)
Thanks. Reminds me of a circuit we're considering. We have an ecl gate that's driving a gaasfet differential pair. For obscure reasons, we want the Vcc of the gate to be dac programmed from about +4 to -4 volts. When it's above ground, we can source the current from +5. But below ground, that's pretty inefficient, so we'd like to source the supply current from ground. So we need a composite regulator that softly switches over where it gets its input from.
I've seen audio amps that did this too, automatically transitioning between supply rails, tracking the signal swing, to keep power dissipation down.
I've seen the same in audio amps (eg in Ben Jacksons book), but it was interesting to look at it like a multicell converter. Having seen it, its pretty easy to synthesise any arbitrarily stupid arrangement like that.
It just seemed to me you might be able to use such an approach to pump up the efficiency of your gradient amps, allowing more watts per cube.
You should sell those things to audio enthusiasts.
I've toyed with building an in-speaker amp that has a chunk of memory, fits a complex hull to some audio then controls the supply rails according to that. make the hull have continuous 1st & 2nd derivatives and you ought to be able to push it right up to the SMPS CL BW, thereby maximising efficiency.
but I dont care very much about audio amplifiers. I would rather play with my 300KVA RF filter (inspired by that piconics broad-band inductor), or my 3kVA mag-amp.
The gradient amps are genuine current sources, and need ppm settling in microseconds, so that may be tricky. One thing we do now is to have the HV power supplies operate in a "boost" mode: the supply might be stiff at, say +-75 volts, but we overcharge the filter caps from a softer supply, to +-150 maybe. When it's time to make a pulse into the inductive gradient coil, we have extra voltahe available for a millisecond or so, to slam the load inductance, but then it droops down to the base level, so's not to fry the fets.
The big 20 KW-per-axis whole-body imagers just use big switchers to drive the gradient coils, but they're slow and noisy. Switchers make a lot of these analog tricks obsolete.
Yuk! I'd have to deal with the lunatics. And I bet they're a PITA, and don't pay on time.
Just switch and don't tell anybody.
Did you make a huge version of the spiral inductor? Did you use the equivalent of the ferrite epoxy fill? Does it get hot?
yeah, I guess it all depends on how the current commutates. It'd have to be inherently extremely linear, but its probably riddled with transients as diodes switch in and out.
cute. an old trick but a good one.
now that would be interesting to play with.
:)
:)
mostly I forget to use my stereo anyway, which is why I havent bothered.
In this particular application I happen to have large amounts of volume, whether or not I choose to use it. So thanks to Ken, Harry & Jamie, I am trialling an air-core version of my iron-powder filter (200 x E305-2 cores). Normally I cant do this, but like I said, I have literally got cubic metres of space, so I can trade 20dB of volume for 6dB of money (note differing dB scales).
My metalwork guy is working on winding me some now. But I didnt make it a conical spiral, I vary the pitch as that turns out to be far more convenient to fabricate.
I did a bunch of reading on frequency independant antennae, while cracking a problem with sheath helix resonance in a 10W SMPS filter inductor (I know, crazy but true; thanks again Ken). Seeing as its the periodicity that causes these sorts of resonances (I read Brillouins book a while ago. hard work, but very interesting), I figure the dual of gradually varying R is to gradually vary P.
And seeing as I am making LPFs, I am using stupidly low inductance caps as my mounting brackets. I have a tame capacitor manufacturer who is going to make me some monstrous feedthru caps too.
As far as losses are concerned, this ought to run WAY colder. For a start, no core losses at all. Current is ~ 500A 50Hz, and one thing we are trying to do is wind the entire filter assembly out of a single piece of bar. At high currents joints are a real PITA, and can represent a sizeable chunk of the total series R.
Plus its awkward winding very high current chokes on small cores, so I can do a lot better when I allow myself an essentially unlimited winding window. By eliminating the cores I get rid of a very large heatsink - actually what I do is take about that volume of Aluminium, and re-arrange it to form a large, thin enclosure.
Of course measuring this is going to be interesting. But I like the idea of a filter made mostly from solidified electricity :)
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.