Barry Gilbert now works for Analog Devices, and they do produce some rather nice Gilbert-cell-based integrated circuits - I've already listed the AD633, AD834 and AD835 in this thread. There are others - Barry has exploited the idea with some enthusiasm.
If you insist on tubes, look for 7360 beam-deflection tubes for mixers. They were the best of their kind at their time. Also, you'll a hefty sensitivity for magnetic disturbances in addition of the other tube quirks.
If you just need the difference between the fixed and variable frequency oscillator, any envelope detection (AM detection) circuit should work. Feed the detector with both frequencies to the rectifier. There is a beat (amplitude variations) between the two frequencies. Using a half wave rectifier will create a DC biased waveform with the beat (the difference frequency) riding on top of it and some high frequency noise above that. Remove the DC bias with a capacitor and the original frequency and their sum frequency with some low pass filtering.
This should be perfectly sufficient when there are only exactly two frequencies sufficiently higher (say 50 kHz) above the desired audio difference frequency.
In that application it's not such a problem because the anode resistor is very low and it's value is the impedance of that node now. The tube doesn't amplify the signal voltage, it can't.
Links are not broken, click of "Fig. 1" and not "opamps".
If you can't get MC1494 but can arrange one bipolar drive signal, you can make the rest of a Gilbert cell very easily using the matched transistors in a LM3046, which are readily available, cheap and sufficiently retro. You even get a spare transistor you could use for the phase splitter.
Forgot to mention, the Ft of the LM3046 is circa 350MHz, so no problem operating at HF. And there's always the rather faster and more expensive HFA3101 if you need to operate up to 1GHz :). But the LM3046 is one of my jelly-beans for discrete stuff up to 50MHz. It can go higher, but it's less trouble to design using faster transistors.
es.
The transconductance OA has been mentioned already---and if I was insisting on OAs, I'd probably go with that. The Gilbert (or Jones) cell method was mentioned, and is very effective, but it is not an OA.
One comment on your link "complained" about the need for a LPF on the outpu t. I don't fully get the complaint. After all, LPF is always needed unless "image reject" double balanced and linear mixing is done. But even that is too hard to do for obtaining very high rejection.
If considered non-OA methods, a simple analog switch in chopping mode will do it. That would be "singly balanced." IIRC, one can differentially drive four analog switches, that are also differentially chopped, and obtain dou ble balance.
I/O + - O O | | +---+ +---+ | | | | o o o o \...|...|...\.....O LO o o o o SW/Relay control | | | | | \ / | | X | | / \ | | | | | +---+ +---+ | | O O + - I/O
The "X" is simply a wire cross for the wires drawn at 45 degrees.
There's a very simple circuit for a four-quadrant multiplier in the LM13700 application notes. I'm not sure if it will work at a few hundred kHz, however.
The LM13700 is a very useful chip and works at low supply voltages as well. It's unfortunate they don't make better OTAs that would be suitable for much higher frequency applications.
You might find the LM13700 especially useful in your Theremin project for the volume control circuitry. That VCA stage can be a problem area in Theremin design. I find the old RCA design which modulates tube filament power to vary the volume delightfully zany!
piglet
A lot of creative inventor types are bipolar and compulsive and a little schitzoid. Not survival traits. Business types and sociopaths and evil VCs often take advantage of them.
Like I said, you can use a pair, if you don't mind the count of two. They're cheap enough to throw in anywhere, so who cares?
I don't think they made a heptode, at least in the CKxxxx series for example. I see some battery op types (1AE5, 2G21/22), but those might be incongrouous with your system (or an alternate direction to go in, for the whole thing, if you buy a case of filament-type pentodes instead).
Or you can "cheat" the submini theme by tossing in a regular "miniature" socket and use a 6BE6 or whatever. I don't see that as a sacrifice; it'd be like using a SOT-223 LDO (instead of a DFN or CSP variety) in a circuit that's mostly TSSOP and QFN. It's not like it's terrifically different, and if it works, it works. I've got a 6AL5 in my submini SW radio, since I don't have any submini diodes handy.
If you're going to give in and throw SS at it, you might as well do the whole thing in SS, in which case I'd recommend making/using SBMs like the CA3028, or DBMs like the MC1496, etc. Compare with my volume mixer:
BTW, all of your timbre (tone quality) comes from the oscillators and mixer: their distortion/purity, locking behavior (defines how rough the low notes are), and distortion/IMD in the product.
So if you're looking for a "particular sound", that's where you need to do it. SS vs. toob will have some play there.
You can always distort a sine wave more, so you can instead make as clean an audio output as possible, then distort it with add-ons. You can literally use guitar pedals (and I would recommend it), with one caveat: try to get a fixed amplitude audio signal, then distort it, then volume-control it. So you'll be putting the pedal on a "patch" loop on the unit, not the output. Putting the final (variable amplitude) output into a guitar pedal (or some circuit like that) may have undesirable results (attack/decay, change in timbre in the process), depending on the circuit. But that might also be a good thing, so see what works for you (when you get there).
Tim
Which design was that? I can't imagine modulating filament power at low audio frequencies could be very great for it...O_o
No kidding. Running a tube in the cathode-emission-limited condition is for a good time, not a long time. (Rectifiers are an exception.)
Cheers
Phil Hobbs
This is his too.. I've not seen it.
Two-quadrant multiplication with an OTA is more than enough, and it'll work fine at megahertz frequencies, because you don't care about power gain. Feed your low-level signal into the diff inputs, and modulate the bias current with the LO.
They do: TI's OPA861, for example. They just have such opaque datasheets and descriptions, that it's hard to recognize the face under the greasepaint and red nose.
LM3046 is available in heaps of places, and used in lots of designs so is likely to be manufactured for some time. Use the SMT versions. I was thinking of a circuit like this one. The input transformer does the phase splitting and provides the low impedance you need. You can ground either of the other inputs for single-ended, and ignore either of the outputs:
No need to sign in or sign up to Dropbox, just cancel on through to the PDF.
Clifford Heath.
OTAs are just an input diff pair and four current mirrors: one to turn the bias current input (I_ABC in the LM13700 datasheet) into the tail current for the diff pair, and three more to take the diff pair's collector currents and mirror them from the supply rails. (There's a hard-to-avoid asymmetry, because the diff pair collectors always sink current, so the + current source only needs one mirror, whereas the - current source needs two in cascade.)
I still have a couple of dozen VTC VA713 and VA2713s from the early '90s. They're a lot like the LM13700 except a hundred or so times faster. Long gone, of course. :(
The OPA861 isn't an OTA, despite the name--it's intended as an improved BJT. It lacks differential inputs, for a start. TI rented the name for a completely different device, drat them.
Cheers
Phil Hobbs
If you feel like showing me an example of the kind of circuit you have in mind, that would be great... ;-)
Unfortunately, however, it seems like the MC1494, 1496, LM3046, and so on are all non-stock just about everywhere.
Have something to add? Share your thoughts — no account required.
Ask the community — no account required