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

Interesting. Trying to get kids interested in Real Electronics.

But this is outrageous:

The word "latchup" is not mentioned.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

No, it's not a Wilson mirror, and really not at all like it, each mirror output would take lots of parts, and would fail to work up close to the rail.

--
 Thanks, 
    - Win

And the output-saturation problem we've been discussing, which is certainly an overvoltage situation, or maybe an excess voltage (beyond-range), is also not mentioned. Not to mention integrator windup that occurs. I dunno if these guys have much real experience.

--
 Thanks, 
    - Win

uff

the one above the Wilson, "11.7.1 Buffered Feedback current mirror" it's just upside down

Yep. 1972? :)

NT

You just have to add a resistor so that it sources some Class A current.

I use LM358s routinely in super-low-cost circuits. One of my favourite sports is doing amazing things(*) with next to zero apparatus.

Cheers

Phil Hobbs

(*) For sufficiently permissive definitions of 'amazing'. ;)

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

A 1-MHz 1/f corner in a GHz-class op amp is probably OK for a lot of applications. I did an ultralow-noise preamp for Samsung's research lab that used an ATF38143 pHEMT, whose 1/f corner was about 10 MHz but whose flatband noise was about 0.35 nV/sqrt(Hz). It was a big win.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

If you dump enough current, you can almost make an LM324 into an acceptable audio amplifier. Almost. Sort of. For AM radio sound maybe.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

I've done just that. Sounds decent, though P_out is not too good. Substituting lower power versions of 324 always degraded the sound (LP324?).

NT

I use them for stuff like temperature control, slow local feedback networks, bias generators, that sort of thing. (Of course I mostly use LM358s because it makes the layout a bit easier.)

My daughter Magdalen has just gone to work for Cirrus Logic, so we need a new layout person.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Is she going to do PCB layout? The three best PCB layout people that I've worked with were all women.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Yes, she's going to be doing demo boards for them. Cirrus seems is a great place to work, by all accounts.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Good products at great prices, too.

[about LM358 and LM324 crossover distortion]

The resistor trick is fine for audio, but there's lots of control applications (like voltage regulation) where no addition is necessary. Realistically, any application where a human could twiddle knobs to keep something steady, the kilohertz-and-up signal handling deficiencies are... well masked.

If zero volts has cross-over, do all class A, +24, -5 supplies. I'm using a TCA3072 that way. Diode on the output, the 'plant' is a resistor.

George H.

Only a few amps like LM324 were brain damaged this way. It's actually class B, because all three (3!) output transistors have zero quiescent current. There is three junction drops of deadband.

Silicon malpractice.

The TCA0372 data sheet shows three output bias diodes, but they drew the upper darlington wrong.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Are you really trying to tell me that a thing with a square law transfer function cannot demodulate? That it takes a PN junction? I've seen it on BF862 and the opa140 on my table (admittedly open) also dislikes my cell phone.

The loop gain is not needed for the RF signal. It's enough if it exists for the demodulated result of the input square law device.

100s of mV offset between the input pins of thing designed to have near infinite gain. OMG.

The data sheet has a different opinion.

The idea of unprotected MOS gates died mostly with the RCA 40600 (verbatim). It had a nice bronze wire wrapped around its leads, to be pulled away after soldering. It was soon away from the market and made place for the 40673, 40841 and 3N140. I still have a few; they will probably survive me unused in their box.

regards, Gerhard

Am 27.04.19 um 00:37 schrieb snipped-for-privacy@gmail.com:

Hm. It definitely needs better output biasing.

In a previous life I have made a LM124-like op amp on an analog array, and also a LM139-like comparator.

Not a bad architecture when all the good PNP transistors you have are substrate transistors. I.E. their collectors are preconnected to the substrate. A extra mA and a VBE multiplier were essential.

Current mirrors with pnp lateral transistors are a joke.

Nevertheless it was surprising what could be done with

1 metal layer and a few predefined resistors and transistors. And there was a 30 pF cap per cell. That one was huge.

regards, Gerhard

Right, it's an overly broad generalization, based on limited experience rather than theory and regulation.

It is supported by these, but not in the strict way that it is presented above.

To wit: if IEC 61000-4-3 says you need to handle 3 V/m at basically any frequency, then expect on the order of 3V to get into your circuit (assuming cables of about 1m length -- typical for a wired product). More with cable resonances, less with short cables, and all other sorts of specifics that vary with frequency and circuit.

Apply that 3V directly to a signal MOSFET, and it's sure as hell going to demodulate it. Even small power transistors may; although not necessarily at the frequencies tested, since gate impedance is so much lower. Again, it's a matter of the whole circuit and environment, not of a single trifling aspect of it.

Noteworthy, too, that when it comes to 61000-4-2 ESD testing, or -4 EFT, anywhere you're getting ~3V on a gate in radiated, you're getting ~kV of these surges. Maybe you put ESD diodes on there, but those just made radiated worse (diode junctions!), if not at 3 V/m (say your supplies are

5V), but at some point they will detect it as well.

In this case, filtering is indicated (ground plane construction goes without saying), or outright shielding.

Say you've got filtering, or short (or nonexistent) cable length, so that your device generally isn't susceptible. Now you're getting good attenuation, like 40dB+. Well, that's still 10s of mV of RF entering the circuit. Apply that to a diode junction and you get meaningful demodulation (poor performance, maybe 10-30%, but definitely detectable). Apply that to a FET gate, and you probably won't get much (the quadratic response has a much softer curve and looks flat for this arc length).

Is filtering or shielding still indicated? Depends. If the FET/MOS amp is too noisy, or expensive, to meet the product's overall goals, then another solution will be needed. If all requirements are still satisfied, then it is certainly acceptable.

This is the theory behind the observation, and also illustrates the limits where one expects to see variance from said observation.

Yes, even vacuum tubes will detect RF; it's just that it's very rare for -- well, for one it's very rare for audiophiles to submit their products to IEC testing _at all_, but there are few circuits quite sensitive enough (phono amps only deal with ~mV, not ~uV), for which typical grid bias (~volts) is surpassed by enough ambient RF (~V/m), to cause perceptible detection. That's even taking into consideration the shitty [RF] grounding that most such amplifiers have. Rectification itself is also fairly slow (ordinary receiving tubes do have bandwidth to the low GHz, but physics itself is against them*).

*For example, I built a VHF oscillator with a 6C4 that generates harmonics above the noise floor (RBW = 300kHz) out beyond 800MHz, and presumably at lower RBW, to beyond 1GHz (I just didn't check at the time). This is far out in the tail (-40dB/dec?) though, and I would expect similar performance with respect to RF detection.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/