Gyrators, do the opamp matters?

Just wondering if the opamp used on a gyrator of a EQ circuit do matter.
Like, if I use a 4558 or a TL072 or a OPA604, which parameter would it change??
How close, by experience, do you guys think it will sound from a real inductor?

:guinness:
Fabio

If the thing is part of a fairly high frequency resonator then decent gain-bandwidth is important. High input bias current will affect noise if your impedances are fairly high. Voltage noise will often be dominant though, as these one op-amp synthetic inductor circuits have fairly high noise gain.

Work at the lowest Z's practical to keep resistor thermal noise contribution small, and at as high of levels as possible---but keep an eye on the actual opamp output voltage swing to make sure there is headroom.

I'll defer the question about sound to those fortunate enough to have been able to afford to design with real inductors! They will always be lower noise at least.

Brad

Just some strange speculation, but since Ampeg has been part of a larger company and the tapped inductor still seems to be in the mid-EQ of the SVT-pre I guess there's a reason for it :wink:
Otherwise one would expect on command of the bean-counters (as I've seen them being called here) it would have been replaced with an emulated inductor. But who knows, it's already gone by now.

How's that for an useless answer ? :?: :!: :wink: :grin:

Sometimes no one tells the bean counters, and it is a good thing.

Right ! And who knows, they feel obliged to keep the real inductor to keep it calling a SVT-preamp. I don't know why yellow LEDs come to mind now... :wink:

a true gyrator design benefits from high bandwidth, fast(precision) opamps, however look at the circuit and see if there are also summing amps sprinkled in there too.. those are the ones you want to focus on first.

[quote author="bcarso"]Voltage noise will often be dominant though, as these one op-amp synthetic inductor circuits have fairly high noise gain.

Work at the lowest Z's practical to keep resistor thermal noise contribution small, [/quote]

Thermal resistor noise is the main problem of single opamp gyrator L's.
The RC that feeds the noninverting opamp input will appear as a parasitic in parallel to the simulated inductor, so in order to make the parasitics small, you'd want this RC to be high impedance. That will inevitably mean a lot of noise. A GIC circuit built from two opamps is much better.

and at as high of levels as possible---but keep an eye on the actual opamp output voltage swing to make sure there is headroom.

Click to expand...

Also important: The opamp's input voltage swing (common mode range!), and the opamp's (hopefully lack of) phase inversal when common mode range is exceeded.
The problem here is that you have to go for a noninverting configuration, which means that your signal level = common mode voltage. Most opamps do very nasty things when their CM range is exceeded, like making a hard switch of their output voltage from negative rail to positive rail. That's much worse than just "hard clipping" when you overdrive it.
So you're in another dilemma: make th esignal large to get out of the noise, and make the signal small on order to avoid phase reversal at occasional clipping.

JH.

Thank you all for the inputs! I was going for this one:

http://www.forsselltech.com/schematics/Gyrator1.htm

JH, can you point me to a 2 opamps Gyrator design??

cheers!
Fabio

[quote author="Bauman"]Thank you all for the inputs! I was going for this one:

http://www.forsselltech.com/schematics/Gyrator1.htm

JH, can you point me to a 2 opamps Gyrator design??

cheers!
Fabio[/quote]


I gotta run - look for "GIC filter" on google

Another good 2-opamp solution is shown in Orban parametric EQs.

Also, look for state variable / leapfrog stuff.
You can even emulate floating inductors with this.
See the following link for an (untested) example:
http://www.oldcrows.net/~jhaible/scanner_vibrato/jh_leapfrog_untested.pdf

The Orban EQs use the same method, resulting in 2 opamp integrators (one representing the voltage across the C, one representing the current thru the L of your LC filter).

JH.

to fabio or anyone who is interested in this circuit, be sure to read the "evolution of an EQ design" whitepaper that goes with it:

http://www.forsselltech.com/Evolution%20of%20an%20EQ%20Design2.pdf

forssell recommends a gyrator design that uses 3 amps. does anyone have the old studio sound article referenced in this whitepaper?

ed

Good point about resistor noise jhaible. OTOH most of the time the circuit is optimized for the total equivalent circuit, where the equivalent R across the synthesized L is figured in---that is, it doesn't have to be huge, usually. It is a nuisance, because a simple dip equivalent RLC becomes a dip with a shelf, although this can be readily compensated.

Also good point on all of the overload issues. Phase reversal when exceeding the CMR is ugly. And rail-to-rail input op amps are low-voltage curiousities at this point.

A true gyrator would best be realized imo with two OTA's. Unfortunately no one makes a decent integrated OTA that I know of. The 13700/5517 stuff is terrible.

Brad

> a true gyrator design benefits from high bandwidth, fast(precision) opamps

> rail-to-rail input op amps are low-voltage curiousities at this point.

But the grounded-choke gyrator doesn't need an op-amp. Just a follower. Millions of consumer EQs used emitter followers. The Jung Buffer is a nice option. And I've sketched-up a tube limiter with "choke" loading emulated with cathode followers: not great performance but maybe better than buying iron.

Brad and PRR, sometimes I forget that I'm coming from building synthesizer electronics. (;->) When I used gyrators, that was for filter banks where high Q factors, high channel separation, and high precision for center frequency are important.
Musical EQs are not so demanding, so you can live with high parasitics (i.e. lower impedance at the buffer input, less of a noise problem), and relative low Qs also allow to use a simple emitter follower in some applications. (For high Q, the output impoedance of an emitter follower starts to become relevant.)

As for OTAs, try the CA3280. It's much better than the LM and NE stuff in many ways.

JH.

j, you are nonetheless correct in your identification of the likely predominant noise source with the "parallel" R (the one from follower input to common). It had been a while since I looked at the fundamental limitations of that circuit.

There is a strategy for simulating a lower-than-thermal R for that location, but I doubt that the result is any more desirable than the two- or three (Riordan etc.) op amp based gyrators.

Yes, the 3280 is among the better integrated parts. One of the major limitations of the integrated parts is the noise from the current mirrors, and iirc the 3280 is a bit better in that regard.

Maybe I will do a discrete low-noise gyrator block for the forum---unless I get more urgent paying work first of course ;-). I did a very-high-Q resonator for another instrument (not musical) and it worked pretty well. I think I was getting Q's of about 3000 with adequate stability. In the end there were other techniques to achieve the same overall result.

Brad

Thanks for all the info guys. Will digest and I'm sure I'll get back with couple questions more....

:guinness:
Fabio

> Musical EQs are not so demanding, so you can live with high parasitics

Yes, I used an ARP with Q up to 100 (so the knob said). Useful for that world, pretty useless for stuff that is already "music". Boost with Q over 2 tends to be musically annoying, and dips narrower than Q of 3 are inaudible except for notching garbage (hum and tones) or room acoustic tuning.

For a "bad" yet mass-mass-produced 1-transistor gyrator, see a LA3607. On page 3, note that it is an op-amp plus 7 transistors with built-in 1K2 and 68K resistors. Put a couple caps on each transistor, connect 7 sliders, wrap with the opamp, 7-band graphic equalizer. Yes, the Q is only about 1, and yes the distortion sneaks up when boosting because the emitter-followers are imperfect. This was the guts of many-many low-end home hi-fi EQs. I still use one in cassette transfers.

The cathode follower "load choke" was harder. The series resistance, tube plus a swamping resistor, gave plenty of DC drop. With practical value shunt resistor, Q of 10 or 20 was about the limit. (This isn't a resonant Q, but ratio of DC to AC impedance.) And of course it ate twice the supply voltage, and extra heater power.

[quote author="PRR"]
Yes, I used an ARP with Q up to 100 (so the knob said).[/quote]

An ARP with scaled knob for Q values - must have been an ARP 2500 then, right?

JH.