Simple gyrator circuit

[bcarso]

Sure you can do a Forsell op amp. One caveat: if you use the Tosh output devices you may wish to measure the P devices for breakdown voltage, as they are technically right at their limits at quiescence. Fred is aware of this and has never had a problem, but your mileage may vary. Since Tosh is probably one of those Kaizen companies and not a "goal-post-mentality" one, it's likely most of the 2SJ74's are quite a bit higher breakdown that the stated max.

I'm glad you posted this because I hadn't looked at this synth L variant in any depth, and I find that this one doesn't have a parasitic R that's in the alternative circuit. When the other circuit is used in resonator boost-cut applications, the effect of the parasitic is to come out of the dip (for example) at a lower level than you went in---it's as if the synth inductor has an R in parallel as well as one in series. This can be compensated for, or just taken into account in an overall EQ, but it's a nuisance.

If you want really high Q's a GIC type circuit is recommended, but requires two op amps---and ones with pretty decent CMRR too.

 

[clintrubber]

[quote author="bcarso"]EDIT: A really good, albeit necessarily terse, slide show tutorial on active filters including a bunch of OTA topologies I ran across:

http://msic.ee.ncku.edu.tw/course/AdvIC/slides/ch8.pdf

[/quote]
Nice links !
Right, there we have the floating L, use two back to back (page 26 etc)

 

[bcarso]

[quote author="clintrubber"]Nice links !
Right, there we have the floating L, use two back to back (page 26 etc)[/quote]

Yep---hairy but possible :grin:

 

[peter purpose]

Thanks Brad,
I might make it a hybrid fet in bjt out, because dem Toshibas are as rare as rocking horse shit and expensive.
My original goal was to make a 1H synthetic L as simply as possible and discrete. I'm going to fiddle around with the sim and see what I come up with.

Cheers
peter

 

[clintrubber]

[quote author="bcarso"][quote author="clintrubber"]Nice links !
Right, there we have the floating L, use two back to back (page 26 etc)[/quote]

Yep---hairy but possible :grin:[/quote]
Agreed. :thumb:
Integrated it's easy, but must admit I've never used them otherwise than that.

 

[bcarso]

[quote author="peter purpose"]Thanks Brad,
I might make it a hybrid fet in bjt out, because dem Toshibas are as rare as rocking horse shit and expensive.
My original goal was to make a 1H synthetic L as simply as possible and discrete. I'm going to fiddle around with the sim and see what I come up with.

Cheers
peter[/quote]

Tosh doesn't support the JFETs well at all in most locales. If you need some SK170's I have a bunch of an exact equivalent, plus a bunch of SK389 duals. These latter are nicely matched but have slightly lower transconductance and a little part-to-part substrate capacitance that can be annoying in some apps. Don't know if the Linear Integrated Systems people ever got their dual equivalent out or not, though they are shipping the "LSK170" I think. I don't think LIS sees a market for P channel parts at this point.

Actually, a dual input in this app is of limited advantage since the d.c. offset is of little importance---it doesn't appear in the rest of the circuit. A FET/PNP hybrid voltage follower, maybe with a current source load, would be plenty good, even better in some ways perhaps.

 

[bcarso]

[quote author="clintrubber"][quote author="bcarso"][quote author="clintrubber"]Nice links !
Right, there we have the floating L, use two back to back (page 26 etc)[/quote]

Yep---hairy but possible :grin:[/quote]
Agreed. :thumb:
Integrated it's easy, but must admit I've never used them otherwise than that.[/quote]

Most of the filter books I've seen finish the discussion of gyrators by saying that, despite the appeal of using the umpteen tabulated LCR filter designs of yesteryear with synthetic inductors replacing the real ones, there are other ways to get there.

 

[clintrubber]

[quote author="bcarso"]Most of the filter books I've seen finish the discussion of gyrators by saying that, despite the appeal of using the umpteen tabulated LCR filter designs of yesteryear with synthetic inductors replacing the real ones, there are other ways to get there.[/quote]

It's been a while (that sounds old...) but IIRIC for integrated filters that route from LCR-prototype to nothing but Q's & C's is not unelegant.

 

[Crusty2]

I have a good article on the subject that I can scan if anyone's interested. It's from Electronics circa '77 and is a good primer. It's called "The right gyrator trims fat off actice filters" by Thomas H Lynch. I haven't found it online anywhere.

 

[bcarso]

[quote author="Crusty2"]I have a good article on the subject that I can scan if anyone's interested. It's from Electronics circa '77 and is a good primer. It's called "The right gyrator trims fat off actice filters" by Thomas H Lynch. I haven't found it online anywhere.[/quote]

I'd like to see that. I doubt that McGraw-Hill would mind it being posted since that magazine has been gone for several years now.

 

[peter purpose]

Brad,
This looks like it may work. Am I on the right track?

Thanks for the offer of fets. I'll pm ya.

Cheers
peter

 

[Crusty2]

Here it is Brad:

http://groupdiy.twin-x.com/albums/userpics/10071/The%20right%20gyrator.pdf

 

[bcarso]

Thanks. That is a good article.

 

[bcarso]

[quote author="peter purpose"]Brad,
This looks like it may work. Am I on the right track?

Thanks for the offer of fets. I'll pm ya.

Cheers
peter[/quote]

Sort of. But the circuit shown has problems. I tried to find Self's page on complementary feedback pairs but his site seems to be in the middle of revision and unavailability.

Basically, all you need is to remove the FET used for the inverting input on your first FET-bipolar circuit, and run the PNP collector directly to the source of the one that's left. Thus you have a CFP. It needs a pulldown resistor or current sink to work, sized to carry the FET and the bipolar current. That node (source-collector-I sink) is the follower output and drives the capacitor.

 

[PRR]

Do we have a schematic for the 553, so we know what the inductor really has to do?

 

[peter purpose]

 

[PRR]

> what the inductor really has to do?

Thanks, Peter.

Well, dang: this is just another swingin-inputs EQ.

They draw it with the inductor floating, but it will work just as good if you wire it (from the pot-wiper and in/out switch) as C-R-L-Ground.

Now you have a grounded inductor with a series resistance. The "main faults" of the simple gyrated inductor are: it's grounded and it has series resistance. No problem.

You want a unity-gain stage. It does not have to be EXACTLY unity gain, as long as it is close and consistent.

You don't really have to care about DC offset, since there is a C in series.

The maximum current in the coil is, I think, 2mA (assuming +/-24V output signal and control cranked full).

Given those specs, use an emitter follower with about 5mA idle current, 4K7 emitter resistor. Set the series resistor to {edit}1.8KΩ. Set the shunt resistor much higher than 10K; 100K may be workable.

Oh, with active parts in there, you may want to bridge the in/out switch "off" connection with ~100K to keep the capacitor charged, so it don't THUMP when you switch.

Let's see what SPICE sez.....

 

[peter purpose]

Cheers Paul,
I had already started on the 'braductor', but will have a pop at a 'paulductor' in a mo.

 

[peter purpose]

Brad
http://www.dself.dsl.pipex.com/ampins/discrete/cfp.htm

 

[PRR]

HERE is a $1 Grafic EQ chip which is the same topology as this API, but with active "coils" made out of a BJT and Rs and Cs. Although the impedance values are different, and the opamp is not as spiffy, it is a handy reality check.

> Let's see what SPICE sez.....

Here is
* the API sim, strapped unity-gain, assuming zero resistance in the inductor which is NOT realistic
* the same but with an emitter follower fake inductor

Here are overlaid response curves:

Because of the 150K shunt resistor, there is a part-dB rise in the treble, and maximum gain would be less except I reduced the series resistor. The final fit is quite good.

The shunt resistor size is limited by base current issues. Putting an FET in front of the BJT (or using a small power FET) (or a feedback pair) would allow allow a larger shunt resistor, which would allow the series resistor to go back closer to the API value, and also lead to a smaller gyrator cap. Of course if it is too small the gyrator input picks up buzz and RF.

I have not simulated high signal levels. The level at the emitter seems to be ~1/8th of the signal at the input so it should be cool.

Are we sure of the cap values shown in the latest API blueprint? Maybe I mis-copied them, but they seem to have huge mid-high interaction.