AoE x-Chapters, High-Speed op-amps section, DRAFT

Thanks. People that want to do that don't seem to understand that it only makes their own lives the worse. Self-centredness shoots their own foot.

NT

Reply to
tabbypurr
Loading thread data ...

Win,

Many thanks for this chapter. Bit surprised to see no mention of the really fast fully-differential amplifiers, see:

See e.g. ADL5565, ADA4927-1. The former has -3dB point at 3GHz and it's not the fastest!

Clifford Heath.

Reply to
Clifford Heath

Aggh sorry, that should say 6GHz.

Reply to
Clifford Heath

AD8130 and THS4303 are interesting fast oddballs.

We're using darlington MMICS, cheap microwave amps, in time domain, which could be a whole x-chapter. There are generally no Spice models so we have to experiment and hack our own. We're talking 4, 8, 20 GHz wideband gain here.

And then there are distributed amplifiers... another chapter.

--
John Larkin         Highland Technology, Inc 

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

I'm sure all this will have been found in editing already, but just in case: ? p. 42: "current conveyer" should be "conveyor" ? p. 47: In the ADA4817 note, you are probably referring to output -> input feedback on the ADA4817-1ARDZ (i.e. the SOIC package), right? "pin-6 feedback" could be a bit more evocative. - p. 66: Note (x) seems to be missing ("non-inverting input"?)

Did you see the preview specs for the OPA818? FET-input, e_n = 2.2

rate. 28 mA supply current, though, and decompensated. I might buy some pre-release samples to test.

The VFB scatterplots are great, but what would be a nice touch would be to have them in digital form to look up individual points. (One can always OCR the table, of course.)

? David

Reply to
David Nadlinger

Am 13.07.19 um 07:18 schrieb David Nadlinger:

I wonder why there is no FET input CFB opamp. Is there a fundamental problem? The inputs are dissimilar already, so one could easily reallocate the inv input FET area to the n.i. side, and its bias current, too. That should be a 2:1 noise advantage. The inv. input could be bipolar and minimum size.

cheers, Gerhard

Reply to
Gerhard Hoffmann

Well.. The usual... most accounts really are confused on what the actually fundamental difference is between CFA and VFA

My take on this is here:

formatting link

To wit, its clear many simply don't understand that the typically slew speed increase is because of the class AB input topology, and has, essentially, nothing to do with the feedback topology at all.

Its worth point out that the name "current feedback" has been hijacked from its historical definition.

-- Kevin Aylward

formatting link
- SuperSpice
formatting link

Reply to
Kevin Aylward

Nope, thanks for the flag.

In my copy, it says "pin-6 feedback" now, your suggestion is?

That's nice. Unfortunately the FET low-V section of the table is full. We'll find something to eliminate.

Yes. A convenient thing would be for me to hand out my spreadsheet, but while useful, it would need a lot of work to be ready for primetime. A considerable bit of handwork went into the as-seen publication-ready tables, so that the spreadsheet no longer matches. An OCR grab of the table might be best. Meanwhile, you can eyeball scatterplot values, and scan for them on the table.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

Here I take issue, and also agree with you. To me it's the nature of the VAS, voltage-amplifying-stage. In the CFB, high error voltages dramatically increase the VAS high-Z node currents, making for fast slewing. By contrast, classic VFB stages are struck with their fixed class-A current. OK, perhaps that's what you're saying. The counter example is what we like to call VFB+CFB circuits. Here the CFB "-" input and "feedback" resistor are buffered with a follower, creating a VFB amplifier, see 4x.6.3 and Figure 4x.61, panels C and D. You might note that we called those out in the VFB table, with note Z. There are quite a few of them, and they have dramatically-faster slew rates.

Yes, the audio guys aren't happy. But it is the new standard, and a useful one to identify two rather dramatically-different types of op-amp architectures.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

One (im)practical consequence of the CFB slew rate advantage is dramatic loading of the input signal. Instead of being civilized and adding power gain to go fast, a CFB just steals power from the signal source.

--
John Larkin         Highland Technology, Inc 

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

?? Usually we drive the "+" input, it's the "-" input that draws current, from the output. Is the output signal what your complaining about?

--
 Thanks, 
    - Win
Reply to
Winfield Hill

The + input steals signal power too. Base currents increase as differential voltage increases, unlike a VFB amp where emitter tail current limits base current... and slew rate.

A VFB amp could be as good as CFB, if it had a huge tail current, but the input bias currents would be embarassing. Diff input impedance would of course drop too.

--
John Larkin         Highland Technology, Inc 

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

Old thread but it could help some younger engineers:

It is all a matter of how much. Works like this: The first BE junction that RF usually encounters inside an opamp is at the first BJT, first stage. IN+ as well as IN-. Unfortunately any rectified RF attacks at full open loop gain.

When replacing it with an opamp with CMOS input any demodulation mechanism is so inefficient that it the problem typically vanishes totally. Done it many times and usually some jaws dropped because that was the simplest EMI fix they ever saw.

GSM phones are particularly bad and I think that is because of the way they negotiate with a cell tower, starting at full power for whatever reason. It results in very low frequency pulsing ... WHOPPP .. POP .. POP .. POP.

The problem is that the rectified signal's spectrum can be in the tens of Hertz, hence full open loop gain. For the LM324 that's a whopping 100dB.

Now look at that at 100dB open loop gain. Less than a millivolt and the thing can rail.

It doesn't. Take a closer look. What's depicted is the substrate diode path. Else the abs max listing would have the usual 8V or 12V limit and a max current rating into the protection.

Try an SD5400. A swipe through the air with an umbrella can already cause damage. Those things are excellent muxes but they are like the princess on the pea.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

It's many orders of magnitude lower. Switching out a BJT opamp against a CMOS opamp is among my quickest EMC fixes. Of course, the increased input noise has to be palatable.

Oh yeah. Without a few ohms or a (very small) ferrite bead in series with the base they can do a tarantella dance and it can be at frequencies where the Federales really don't like it. Especially the Air Force.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

We had an EMI problem with one of our VME modules:

formatting link

I could argue that it was customer abuse: they located it in a slot next to a horrible CPU module, and the rectified offsets were technically still in spec. But they are a good customer so we had to fix it.

We found only one microvolt-offset opamp that seems to be really EMI hard, ADA4522. Some so-claimed EMI hard amps were terrible.

--
John Larkin         Highland Technology, Inc 

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

So colour me stupid. What's so different about offsets due to rectification vs. intrinsic imbalances, such that they get multiplied by A_VOL and not A_VCL like all the others?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

[...]

Imbalances are static, they do not show up as spectrally significant noise. Rectified RF is not constant and that's the problem. GSM phones are an extreme case, several watts on-off-on-off-on-off. If you or a friend have one, turn it off (really off so it has to reboot and seek a tower, not just the screen), lay it close to the selected audio input cable of a big stereo, then turn it on. Brace for a really loud RAT-TAT-TAT sound from the speakers. In the unlikely event that your stereo doesn't do that send a thank-you note and maybe a gift basket to their engineers for an excellent job.

CDMA phones like mine (on the Sprint network) do that to a much lesser extent because their tower negotiation sequence is more efficient.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Understood. But unless I misunderstand, you're claiming that an op amp with A_VCL = 20 dB (say) multiplies the rectified RF by A_VOL (100 dB) instead. That's what I'm wondering about.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

OK, dunno about the 100dB number, but preamps that respond to radio signals are legend. My most recent experience, a modest-gain preamp for electret microphones with relatively high signal levels. Worked OK in my lab, but out on the Institute's deck, where the bee hives were, picked up AM radio stations, especially a Latin music station. Excuse me, we were trying to listen to the bees! My first step, change op-amp to a CMOS type. Arrggh, it was worse! Next step, solder a cap right onto the 0.1-inch mic terminals.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

I don't know, but perhaps the huge disparity of transconductance between BJTs and FETs has something to do with it.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

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.