EEVblog #1115 - Traps In Chips - And the 7660

[EEVblog]

Dave looks at some traps in chips and their modes of operation. In this case the TI LM2776, a look a the classic 7660 charge pump voltage inverter, and output ripple and ways to reduce it.
Parametric searching, and a look at some interesting rail inverters.
The uSleeve: https://www.eevblog.com/product/usleeve/

[JS]

You did spoke of capacitor voltage dependancy, but for ceramics in EEVBlog #626, a fundamentals fridays which should be back... Maybe electrolytics is the topic for it.

I live a link just in case someone doesnt know how to use the search in YouTube.
https://youtu.be/2MQyQUkwmMk

JS

[invzim]

Another thing to be aware of with the LM2776 which I don't think is in the data sheet, is failure to start if there is a positive voltage on the output rail when it's powered up, enabling/disabling the IC has no effect.  I had to install a low vf diode on the output to fix this.  For TI chips I have now finally learned to google the part number and "problem" to catch the gotchas.

[tszaboo]

OK, so I recently got a UNI-t sound level mater. UT352 to be precise. It had a Chinatek 7660 in it. So I could hear the switching frequency of the 7660 very faint, but definetly audible, which would be OK but this was a sound pressure level meter. I got a linear LTC1144, which is compatible, and for good measure I enabled the "boost"pin on it, which changes the frequency to 100KHz. The end result was about 3-4dB less noise floor on the meter.
I think a tool becomes really yours after you mod it.

[djQUAN]

A 7660 from a Chinese online seller got me on a few when I bought some a while back. It worked but barely. Other times, it didn't so I investigated and did some tests. These were my results: http://www.quan-diy.com/misc/fake/icl7660.htm

As with the reply above, I also had an adventure with a 7660 in a test equipment (craftsman dc clamp multimeter) with its 7660 not working when the battery voltage gets low.
http://quan-diy.com/misc/multimeter/craftsman.htm

[RomDump]

Don't know if Dave noticed he could filter by Negative output type to get a better list on Digikey.

[thm_w]

Wonder if there is some kind of application they expect it to be used for that we are not aware of. They even provide a load step in the datasheet from 10 to 100mA output. So presumably, the circuit would be "idling" at a low power level and not in use, and then when you actually care about the negative voltage rail it is active and drawing 100-200mA.

That or its just a poorly designed chip where someone wanted to boost the efficiency/current consumption numbers for marketing.

[Bud]

Such behaviour is not unique, many converter chips have it, some have it controlled via a pin. If the designer does not want it it can be disabled via setting the pin.

[Peabody]

I've posted about this problem before, but don't want to be a pest, so won't post again about it after this.  In the video, the datasheets look ok, but the websites (Digikey, Mouser) were truly tiny on my 15-inch laptop, fullscreen, and essentially unreadable.  If there's any way you could adjust your browser so everything looks bigger, it would be helpful to us small-screen types.

[langwadt]

related, https://www.eevblog.com/forum/projects/sot23-buck-converter-ic-without-pulse-skipping/

[David Hess]

Another thing to be aware of with the LM2776 which I don't think is in the data sheet, is failure to start if there is a positive voltage on the output rail when it's powered up, enabling/disabling the IC has no effect.  I had to install a low vf diode on the output to fix this.

This problem is more common than not with any IC which relies on a junction isolated process.  Every NPN collector and MOS drain (?) is connected to the substrate via a parasitic diode which is part of the isolation structure.  The substrate must be maintained at a voltage equal to or lower than the lowest operating voltage (1) to prevent these parasitic diodes from conducting and injecting minority carriers into the various nodes.  On a switched capacitor voltage inverter, this substrate connection is connected to the negative output.  The ancient Linear Technology LT1026 datasheet describes this issue.

With dual and quad operational amplifiers on built on junction isolated processes or ICs that contain multiple logic gates, you can get very strange results if a pin is pulled below the negative supply injecting carriers into the *other* function units which is why you sometimes see external schottky clamp diodes used despite the presence of a "free" internal clamp diode.

For TI chips I have now finally learned to google the part number and "problem" to catch the gotchas.

Problems like these are often overlooked in the datasheets unfortunately.  TI is one of the greater offenders.

Back when this sort of thing mattered, they released a "new and improved" dual-slope integrating converter which specifically listed having a unipolar zero instead of +/- zero which is usually inconvenient.  Guess what?  It still returned +/- zero just like earlier implementations and the problem existed for years; I do not know if the documentation or part was ever fixed.

TI also has a history of misleading specifications like noise in their datasheets and they were called out on this many times with no positive results unless you count them buying up the companies who complained.

(1) This is why the 7905 and 337 negative regulators in the TO-220 package have their negative inputs on the center pin; the TO-220 package always connects the center pin (and tab) to the substrate.  And the 7805 uses the center pin for common and the 317 uses the center pin for adjust for the same reason; those are the most negative applied voltages to the positive regulators.

[firehopper]

I've posted about this problem before, but don't want to be a pest, so won't post again about it after this.  In the video, the datasheets look ok, but the websites (Digikey, Mouser) were truly tiny on my 15-inch laptop, fullscreen, and essentially unreadable.  If there's any way you could adjust your browser so everything looks bigger, it would be helpful to us small-screen types.

hit shift + to make the font bigger or shift - to make it smaller..

[Brumby]

It's Ctrl + and it does not apply to a video.

Dave would have to do this during his recording - or do a zoom in post-production.  I can tell you now, he won't be keen on additional post-production faffing around.

[Peabody]

It's Ctrl + and it does not apply to a video.

Dave would have to do this during his recording - or do a zoom in post-production.  I can tell you now, he won't be keen on additional post-production faffing around.

He would have to change the effective DPI for his browser before recording.  This may not be correct, but my impression is that this started with his new monitor.  I'm sure he can see it just fine when he's recording, but unless you're viewing the video on an equally large display, or close to that, you just can't read it.  At least I can't.

Well, all I would ask is that he watch this episode on Youtube on a 15-inch laptop.  If he's ok with how it looks, then so be it.

[David Hess]

I only had a little trouble reading it on a 1080 display.

[Brumby]

Sorry.  I was just making the distinction that it would have to be Dave that takes remedial action.  Viewers of the video are rather limited, especially on small screens.

As for reading the browser detail myself - I didn't really bother.  I've used the parametric searches before and know they aren't perfect, so I was just listening to what Dave was saying.

[David Hess]

Distributor web sites like Mouser and Digi-key are not the only place where parts may be difficult to find; the TI and Analog Devices web sites have the same problem.  I have had to look up parts on both by part number instead of going through the selection guides because they are misfiled or just completely missing.  TI is particularly bad about this for old and inexpensive parts which are still in full production; if you do not know the part number, they might as well not exist.  But new, improved, and expensive alternatives are easy to find, by coincidence I am sure.

Has anybody else noticed that TI does not sell any linear regulators which are *not* classified as low dropout?  Who knew that the 317 and 7805 were low dropout parts this whole time.

[Clear as mud]

How long have you had those microsleeves?  I was looking all over the place for something like that back in 2013/2014 when I was trying to make a living doing electronics stuff, and I thought I would have to make my own because I couldn't find any anywhere, except stupid small ones from a multitude of different suppliers, but I wanted a big one like yours.

[T3sl4co1l]

Note that all of this stuff about ripple is just the nominal intended waveform -- quirks in timing (shoot-thru) and dV/dt and dI/dt generate just as much noise, and it's more pernicious -- hard to filter.  Worst of all, power supplies are usually measured in a 20MHz bandwidth that misses this!

Tim

[thm_w]

How long have you had those microsleeves?  I was looking all over the place for something like that back in 2013/2014 when I was trying to make a living doing electronics stuff, and I thought I would have to make my own because I couldn't find any anywhere, except stupid small ones from a multitude of different suppliers, but I wanted a big one like yours.

Its been out for about a year, so not available back then.

[Towger]

Sounds familiar 

https://www.theregister.co.uk/2018/09/11/raspberry_pi_poe_hat_issue/

Upton continued: "Because the regulator operates at a fairly low frequency, each time it switches it moves quite a large chunk of energy into the three USB reservoir caps via the current limiting switch: this large instantaneous current is fooling the switch into thinking that a genuine over-current event is occurring."

It was, Upton observed, "dumb luck" that heavy load testing was done with one brand of switch while lighter testing occurred with the other. Thus the PoE HAT passed product validation.

[nctnico]

I stumbled across this video by accident. Coincidentally I just used the LM27762 in a design. One thing I don't agree with is to be extremely focussed at component cost. Especially with jelly bean parts (or cheaper) versions it is easy to run into unpredictable behaviour which in turn wastes a lot of time (which also costs money). Let's call that a trap for both young and old players. Even 10k units isn't high volume and an extra week spend on fixing a design easely adds 50 dollar cents of cost and lost revenue to each unit.