Re: FET Switching

[quote author="mediatechnology"]
I remember testing this in a modified configuration where the series FET was within the feedback loop of the op amp. It looked like this:


[/quote]

Looks like S/H device.

I used something similar in an instrument that checked quality of coating of coils for electric motors: I measured it's resistance before applying huge DC that heated them, and after that, then compared results. The results were stored by devices wery similar to what you drew.
The same S/H device I used in analog synthesizer to remember the key pressed so the frequency stays relatively stable during decay phase...

Speaking of FET switches, I used them in electronic instruments, but would NEVER EVER use in microphone and Hi-Fi paths.

Your schematic is incomplete, as most applications would be switching between two or more parallel input/feedback networks. As drawn when the inner FET is open, there is no DC path for the opamp so it will drift to one supply or the other.

This could be a sample and hold if the capacitor was reasonably large, opamp bifet or equivalent, and hold interval modest. For my application I used a 1M or larger R across the opamp so there was always DC feedback with no inputs selected.

To answer Wavebourn's implied criticism of FET switching for audio quality, I used a variation of this in a recording console back in '70s/80's with no measurable signal degradation using test gear of the day (Sound Technology). The FET in this example is only passing the minute current it takes to drive the very high impedance opamp input pin and small currents to drive any inner R and C feedback components. Scaling the actual outer feedback networks to be orders of magnitude lower impedance than the inner network values holds all FET non linear conduction related errors miniscule.

YMMV

JR

[quote author="mediatechnology"]
Thanks for your flippant response Wavebourne - we've come to expect it.[/quote]

What should you expect else if you draw a schematics where when a switch is on it inverts a polarity charging a capacitor, but it leaves an integrator with huge time constant when a switch is off? Sample and Hold, of course!



[quote author="mediatechnology"]

Yes Waveboune all SSLs are shit and everything ever recorded with them is crap. I'll take FET over an oxidized Schadow switch every day of the week.[/quote]

I did not say that. I said that I am superstitious sometimes.

PS: Sorry if you find my humor to be rough...

[quote author="mediatechnology"]

Does the "enhanced version still look like a sample-hold?

[/quote]

The enhanced version looks like a switchable summing amp. However, when all inputs are off output noise is higher...

Don't berate me, because I know I must be spoon fed.

But I pulled some chips earlier today PMI MUX08's, and then I found some spoon feedy ap notes...

:razz:

JFET TUTORIAL PROJECT

MUX08

=FB=

Going back to S/H and 1970'th, I was punished once for my creaticity... Playing with different parts / topologies / capacitors for S/H of that motor coil QA system I connected output of it to VCO and listened how it sounds. Dielectric absorption of capacitors and other errors were audible well. It was late and dark, when The Boss come to the laboratory from some banquet... Hearing what's going on he started his angry lection about valuable parts, tools, and stupid young engineers who don't respect all values of The Laboratory... I had to listen and keep silence untill he finished and gone...
However, I used later what I've found in my musical synthesizer, but that time I was innocent working on The Theme! :roll:

...hopefully, he knew nothing of baking potatoes in muffin ovens (used to bake resistors in hybrid ICs) and of beer in expensive precision Peltie thermostatic cameras)... We vere really guilty...

Put 2 matches between 2 razor blades, connect wires to blades, fix it together using a sewing thread, connect it to 220V and put the whole construction to a glass bowl with water. It is called "Laboratory Coffeemaker". However, 120V will will not work since such a coffeemaker is not UL listed...

Interesting topic :thumb:

If I scan & post those two Douglas Self articles from EW+WW about electronic switching,
will you all become friendly to each other again then ? :wink:

This "virtual-ground"-switching is also used - in form of a 4053 - in most parts of the Calrec UA-8000, e.g.:

http://www.gyraf.dk/schematics/Calrec_UA8000_mic_input.gif

Jakob E.

[quote author="mediatechnology"]Sure Peter I thought we had though.[/quote]
OK, sounds like the sun is shing again here. Must admit I already responded before completely reading the thread in detail, wasn't that ... :wink:

I'd like to see them.

Click to expand...

As a note to myself I post the issue-info here, gimme a few days for
finding these EW+WW-issues & scanning:

Douglas Self:
Analogue Switching: Part 1 Jan 2004,
Analogue Switching: Part 2 Feb 2004

Bye,

Peter

[quote author="mediatechnology"]Peter - Are they still publishing? Last time I looked it appeared that the site had changed ownership and was under construction.[/quote]
About a few months ago we had some talk about this and then I posted that I had received info from the new owner, you may remember that update. They're 'back', but 'they' are partly somebody else. I've seen a few new issues already.

BTW thanks for the Bateman. I may push the exposure on your jpgs and pdf the series.

Click to expand...

Nice, feel free of course. Note that there may be an error somewhere, I got a PM from (sorry I forgot who) that some scans were double. Haven't found time to check that.

and it works really well for doing scans in jpg needing repair.

Click to expand...

Pardon me ?!? :evil: :wink: :green:

I must admit my motivation for posting this stuff is a bit selfish: It's the first time in weeks that I've actually been able to see the scanning table. Things are starting to look a little more organized and filed in nice square piles. But - there are about 5 more boxes of Ideas for Design. Don't mean to be a thread hog...

Click to expand...

Hey that's not too selfish I'd say, we all can benefit.

Cheers,

Peter

[quote author="mediatechnology"]

I was punished once for my creaticity

Click to expand...

And why am I not surprised by that? :grin:

I just plugged hot dogs into the wall. Stick some nails in the end, apply 120V and in a few minutes a cooked Weenie. Those lab coats always looked like Chef uniforms to me...

Dielectric absorption can be deadly. In RF transmitter facilities we always shunted the HV caps in storage. It doesn't take much DA (even after years) or field to charge them to deadly potentials.

With sample and holds as you know DA can be a bear to deal with.[/quote]

Many of us have electrified hot dogs, and there were even commercial hot dog cookers that used that principle in simpler consumer safety times.

I am not picking on you, but fear you may be thinking of something other than DA. You're musings are highly regarded so I hold you to a high standard.

Dielectric absorption in capacitors (AFAIK) is best modeled by a number of larger and smaller RCs in parallel. A somewhat simplified description of the mechanism is that there are real impedances between the terminal leads and the bulk pure capacitance distributed inside.

This is clearly seen in sample and hold applications as driving the terminal briefly from a low impedance voltage, and then removing that driving voltage will show a settling of the terminal voltage to the average internal charge. The larger the step size and briefer the application the more evident the DA error. Different capacitor dielectric and construction/lead attachment approaches result in different amounts of DA.

I believe this somewhat over rated as a problem for audio blocking applications, but in any audio frequency pole this is a valid error source.

Unterminated caps picking up charge in a high RF environment suggest rectification to me, and could be a shocking development indeed.

JR

[quote author="JohnRoberts"]
Unterminated caps picking up charge in a high RF environment suggest rectification to me, and could be a shocking development indeed.
JR[/quote]

I'd suspect that a high power RF environment is likely to have some serious field emission going on from areas with high d.c. voltage, which might be more likely responsible for the cap charging.

Most of my RF experience is chasing it out of audio circuitry. A DC field sounds a little like static charge or being inside a big vacuum tube..

I'll defer to your judgement about what it is, I am more confident about what it wasn't.

JR

[quote author="mediatechnology"]Back OT.

JohnR: You wrote in another thread...

To reduce crosstalk, additional switches, shunting the unselected nodes to ground prevent any capacitive coupling through open switches.

Click to expand...

OK that's what was done in the original cite and the inside-the-feedback version you did and also in the one I tested a quarter century ago. Did you use complimentary JFETS or inversion in the gate lines? Just curious. Seems like complimentary JFETS e.g. J112/J175 might save parts. I suppose in a "one of two" selector the compliment would already be available but in a "multi-position" selector it make make sense.[/quote]

You will need diodes and resistors to prevent currents through open PN junctions. Diodes are Ok, but resistors will leak control signals...

Studer also used the virtual-ground/4053 switching in their 800-series.

In difference to the electronics article the 4053 was powered with +/- 7,5V. I must have a look for the service manual and do a scan.

analogguru

I used some digital multiplexers in guitar pedals, I don't remember which chips... They were Soviet copy of the 40'th series. I've found their internal schematics and was happy to see that they had no logics between inputs and outputs, just plain FET keys...

My recollection is that I used CMOS transfer gates in that application. I liked JFETs better for shunt mutes due to lower on resistance. I don't recall if it was simple 4016/66 or one of more complex matrix switches. Probably the simple uncommitted quad TGs. Likewise latching logic and control tricks were done in CMOS level logic. I could switch channels individually or as a bank. The channel just did the last thing it was told.

Sending large bi-phase JFET drive signals around a console becomes another noise issue (clicks) to deal with. Better a smaller edge limited logic signal that could be cleaned up by per channel latching.

An additional benefit of shunting unused inputs is that it holds even deselected inputs close to ground so CMOS can be driven from low voltage supplies. In high performance consoles, crosstalk is huge consideration after low noise/distortion.

I recall switching between tracking and mix down so off really needed to be off if a track was being ignored.

JR

Running the CMOS TG off lower supplies may somewhat compromise their on resistance but it also reduced the charge dumped when they switch. I think they may be quieter when switched faster rather than slower, and the on resistance is not much of an issue inside the loop.

If really bad you could design a 2-pole MFB filter into that circuit to reduce clicks further but I don't recall it being a problem that couldn't be managed with a small feedback cap.

-----

Yup DA does explain the recovery in discharged caps and some of that may be more complicated than my simple model.

Any who have played with high voltage supplies know the danger of not discharging twice or longer than a few seconds. While you won't end up with 5% of 5%, all of the internal paths are exponential (like RCs), so 2x will get you closer to a useful discharge.

JR

[quote author="mediatechnology"]
Back OT: I just charged a 1000u/50V Panasonic SU to 10V, shorted the leads for maybe 10 seconds, and then connected a voltmeter. That was maybe ten minutes ago. It now has ~470 mV and still rising. As a control I briefly connected an "uncharged" identical one to the voltmeter and it read, continuously, less than 1 mV. So the voltmeter bias current is indeed low and not charging the cap. It's DA or dielectric memory.

Now if that cap's HV cousin (with the same DA) had been charged to 10 KV, and held there continuously for say a year on the air, and briefly discharged by the interlocks and ground "magic wand" it would be around 500V right now and still heading Northward. I wouldn't want to grab the un-shunted terminals. Would you?

We didn't shunt the pf-range air caps, neutralization or tuning ones just the big boys plate and screen filters.[/quote]

Lytics are terrible for DA/memory. Plastic film, oil, etc. are much less bad, although as you point out 1% of 10kV is still a fair number of volts.

I had a friend who worked in a physics lab come to my office door one night. When I opened the door he looked a bit like Wiley E. Coyote after one of the Acme device malfunctions. Seems he was working alone on a laser power supply he had designed, which had failed again. It had two different safety discharge mechanisms to ensure that it didn't kill people working inside the cabinet. Outside, you were on your own, since it did at least need to work---it was a 100mA current generator with 20kV compliance! It used a big luminous sign transformer, phase-control SCRs, a floating stack of transistors for the fine regulation, etc.

He cursed when it wasn't working, and opened up the cabinet and was about to reach in. A little whisper told him it might not be safe, despite the interlocks, crowbars, etc. So he did a quick order-of-magnitude calculation in his head about likely energy stored, and decided that a 2W resistor taped to the end of a stick would do; I think maybe it was 100 ohm Allen-Bradley hot-molded one.

Anyway, he gingerly reached in with it and applied to one of the big HV cap terminals.

It exploded with such force and noise that his hearing loss was immediate---he'd had a radio playing in the background and he suddenly couldn't hear it anymore. He had a cut on his face where the shrapnel had hit, missing his eye but near. A bit later he could see a janitor across the room who had come in to see what the big bang had been, and was talking to him but he couldn't hear him.

So he came by to tell me all of this, his hearing slowly recovering. He wanted me to know how close he'd come to killing himself, and suggested not working on such equipment alone.

He got the degree eventually and after a post-doc in Utrecht went on to TRW, where he wound up working with a laser power supply that required a whole substation, where 3 phase 23kV was directly rectified for the d.c. The currents were in the 100A range---this for a DF laser with an output power of 1MW.