Voltage controlled bipolar variable resistor

[Skimask]

Trying to reduce parts count for my project.

Original plan was to use a couple of ADG5408's to switch in 16 different resistor values in series with the D.U.T.
(see DDS R(s) Select):   https://www.eevblog.com/forum/projects/project-pocasa-power-supply-pcb/msg670556/#msg670556)

The ADG5408/ADG5409 will handle up to about +/- 20V across the inputs/outputs.  I'm planning on using it up to about +/-15V.

Looking around for a decent circuit that'll handle the same function with fewer parts, and be repeatable (apparently that's the biggie).
Not really sure what I'm needing.  A digital pot would likely handle it, if I could find one that could handle a fair amount of current, say 500mA or so at most.
The term "voltage controlled resistor" comes up here and there, but in general, won't handle + and - voltages across the circuits, and if they do, the circuits I've seen indicate they have a fair amount of crossover distortion.

Ideas?

[SArepairman]

switched capacitor circuits are kinda like frequency controlled resistors,  you can do V-F converter that controls SC.

with sensing you can make a constant current, resistance or power (switch mode) load. you use a DAC to control the gate voltage in a transmission gate or a single mosfet/bjt (or power op amp) thats inside of a stabilization loop.

of course your loopless design is gonna be alot faster and cleaner with much less resolution. you need to make a choice as to what you need. a discrete design like you propose is ALOT more capable of higher frequencies at less distortion etc.

you can eliminate crossover distortion by using a DC bias. that has its own problems though..

[Skimask]

You saying I'd be better off with the 16 resistors rather than trying to fudge a handful of FETs into making it work...sort of?

The 16 resistor / 2x ADG5408 design using all 1206 parts (HUGE right!) uses up a little over 2 sq.in. of PCB real estate.  Not horrible.  At least I know it works.

[SArepairman]

You saying I'd be better off with the 16 resistors rather than trying to fudge a handful of FETs into making it work...sort of?

it depends. I made a current source that uses 5 mechanical relays for ranging because electronic switches had leakages and other problems. If you want more advice i require a better description of your problem.

 One of my designs uses MDACs (in op amp loops), relay's, digipots (digital highly stable analog filter control element), direct buffered dac offset/computer correction and digital switches.. room for everything if you wanna optimize your circuit (mine was kinda a source meter deal). Each of the methods was the ideal solution for part of the circuit.
oh, i see a link. i will read that later and get back to you

[Skimask]

It's basically a multi-channel, multi-source, digital octopus... I/V curve tracer...think Huntron 2800S with more options.
DDS chip feeding some opamps for level/offset/coupling/etc, which feeds into the source resistance network.
ADC's pick off before and after the source resistance network to get V's and to calculate I for capture/display.

Looking like it's not worth the hassle of running the numbers and trying to find perfectly matched sets of resistors/FETs/etc.

[SArepairman]

It's basically a multi-channel, multi-source, digital octopus... I/V curve tracer...think Huntron 2800S with more options.
DDS chip feeding some opamps for level/offset/coupling/etc, which feeds into the source resistance network.
ADC's pick off before and after the source resistance network to get V's and to calculate I for capture/display.

Looking like it's not worth the hassle of running the numbers and trying to find perfectly matched sets of resistors/FETs/etc.

if your doing digital pickup with no analog feedback loop after the digitization you can compensate digitally for alot of things.. stability of your switches/resistors is more important then exact value. You might lose % of your full scale on some ranges, but who gives a shit. You don't need to squeeze the percentages of the FS ranges out of it like crazy.

[Skimask]

Yep, been playing that game right along with everything else...look-up tables stacked on top of look-up tables.  Works well enough.
But, the main point I guess was trying to (semi-easily) knock those 19 parts in that particular section (16 resistors, 2x 8 channel analog switches) down to a small handful...something along the lines of 1 channel of a digital pot for a variable voltage, maybe a FET or two as needed, with whatever resistors/capacitors would be required in the middle of it all.

AC In ---> Super-Duper-Digitally-Controlled-Resistance-Value-via-a-voltage-coming-out-of-a-digital-pot ---> AC Out  (simple as that right? )
The solutions I've tried thus far in that Falstad Circuit Simulator have come up short.  (Keep in mind, I know just enough analog electronics to be dangerous, not nearly enough to be considered knowledgeable by anybody in the field)

[TimFox]

Outside the box:  what about a opto-coupler with an LED shining on a photoresistor (CdS or similar).  There is a time lag on the photoresistor, but the overall curve could be calibrated and used in a LUT.  One advantage of the photoresistor for controlling audio is the time lag can even out bumps from a discrete control.
See  http://advancedphotonix.com/optosolutions/products/optocouplers/

[SArepairman]

Outside the box:  what about a opto-coupler with an LED shining on a photoresistor (CdS or similar).  There is a time lag on the photoresistor, but the overall curve could be calibrated and used in a LUT.  One advantage of the photoresistor for controlling audio is the time lag can even out bumps from a discrete control.
See  http://advancedphotonix.com/optosolutions/products/optocouplers/

the disadvantage of this method is that the tempco of a photo resistor is in the order of 3000 ppm/C.

[Marco]

A photoFET or LDR won't have constant resistance over a 20V range.

[SArepairman]

cds is none the less a method i have seen used for a ultra ultra low distortion wein bridge oscillator feedback. like -170dbm

[Skimask]

Optocoupler / LED / photocell / photodiode = no good.
Need/desire repeatable results (well, up to a point anyways).  Looking to get resistances in the range of 10-1M, in a 1-2-5 sequence (eg. 10,20,50,100,200,500, etc.), while being driven from an AC source ranging from +/- 0 -> +/- ~15VAC.  (guess I should've mentioned that from the start huh?).
Thinking the way I've got it now is the way to go..."manually" selecting resistors thru a couple of multi-channel analog switches (eg. ADG5408 / ADG1408, etc).

[Marco]

cds is none the less a method i have seen used for a ultra ultra low distortion wein bridge oscillator feedback. like -170dbm

With tiny voltages across it.

[TimFox]

Actually, photoresistors are reasonably linear for a constant illumination.  In a Wien bridge oscillator, the voltage across one resistor leg is either 2/3 or 1/3 the output voltage.
The -hp- 334 distortion analyzers used photoresistor optoisolators for the auto-tune circuits.

[T3sl4co1l]

Specs?  Range?  Bandwidth? Distortion?  Accuracy, stability?

Tim

[Skimask]

Spec's?  From me?  (If you're referring to the previous post, then ignore the rest...I'll just sit back and shut my hole while the experts discuss...)
We don't need no steeekin' specs!

Resistances as noted before, 1-2-5 sequence (kinda like an o'scope's time base selector), starting at 10 ohms, up to 1M, if either of those are possible.
Bandwidth - 25Hz up to 1Mhz, again, if possible or even reasonable.  10Khz will do.  Shoot for the stars, get a roof top...
Distortion - low as reasonably achievable, no real spec's here.  I guess if I can see it on the 'scope, it's too much.
Accuracy - Shoot for 1% at each resistance, end up with 10%, good enough.
Stability - yes, repeatable results within room temp, figure maybe 60F-80F, after the pieces/parts are warmed up.

[T3sl4co1l]

So....

10%?  At room temperature?  Couple orders of magnitude?

An OTA will do, e.g. LM13700.  Continuously variable, could add selectable ranges or a DAC for steps.  Should work out to a MHz or so, might be less at lower gain.

One OTA can be wired as a VGA, or as a variable resistor to ground.  Or using a dual, a true floating variable resistor.

You definitely don't want it all in one stage ("ohms to megs").  Cascade stages of modest attenuation, while keeping the impedance constant.  That's also what scopes do, the good ones anyway.  Best case is a modest system impedance (50 to 1000 ohms?) using relays (RF type if necessary) to switch a mixed binary/unary sequence of attenuators (e.g., 1, 2, 4, 8, 16dB, for a total of 31dB reachable in 1dB steps, then additional 16dB stages for as much remaining attenuation as desired).  Or using a variable control (pot or VGA, or throwing in a gain term into the DAC) to cover the first few dB while hardly sacrificing any SNR.

Tim

[Kalvin]

Don't know whether this is of any help or suitable for your application:

http://www.edn.com/design/analog/4368893/Use-a-photoelectric-FET-optocoupler-as-a-linear-voltage-controlled-potentiometer

https://www.fairchildsemi.com/datasheets/H1/H11F3M.pdf

[Kalvin]

Tried to collect your requirements:

- Amplitude range +/- 15V
- Maximum current up to 500mA.
- Adjustable resistor range 10ohm ... 1Mohm
- Accuracy 1% but 10% will do
- Good temperature stability
- Frequency range 25Hz ... 1MHz but 10kHz will do

As T3sl4co1l suggested above, try to construct attenuation sections with constant impedance.  I would split this into two ranges:

- Low attenuation / higher current sections using relays.
- Higher attenuation / low current using analog switches.

The relays may be substituted by mosfets. Edit: DPDT-relays may still be simpler solution than driving mosfets with proper biasing. 

[Skimask]

So....
10%?  At room temperature?  Couple orders of magnitude?
An OTA will do, e.g. LM13700.  Continuously variable, could add selectable ranges or a DAC for steps.  Should work out to a MHz or so, might be less at lower gain.
One OTA can be wired as a VGA, or as a variable resistor to ground.  Or using a dual, a true floating variable resistor.
You definitely don't want it all in one stage ("ohms to megs").  Cascade stages of modest attenuation, while keeping the impedance constant.  That's also what scopes do, the good ones anyway.  Best case is a modest system impedance (50 to 1000 ohms?) using relays (RF type if necessary) to switch a mixed binary/unary sequence of attenuators (e.g., 1, 2, 4, 8, 16dB, for a total of 31dB reachable in 1dB steps, then additional 16dB stages for as much remaining attenuation as desired).  Or using a variable control (pot or VGA, or throwing in a gain term into the DAC) to cover the first few dB while hardly sacrificing any SNR.
Tim

I think I'm beginning to smell what you're cooking, but I haven't fully wrapped my head around the concept of an OTA yet.  Unless I'm mistaken, I don't believe I've had the opportunity to implement one yet.
I get what you mean by using multiple stages vs a single stage setup.  That just makes sense.
But, with all that being said, it's still beginning to look like switching in a handful of known resistor values via analog switches might be easier.  At the very least, a lot less math!

[Kalvin]

If I understood your requirements correctly for the attenuation steps, you are looking something like:

1/1 (pass-through)
1/2 (-6dB)
1/5 (-14dB)
1/10 (-20dB = -6dB + -14dB)
1/20 (-26dB = -20dB + -6dB = -6dB + -14dB + -6dB)
1/50 (-34dB = -20dB + -14dB = -6dB + -14dB + -6dB + -14dB)
1/100 (-40dB) ...

Thus, you only need to design and calculate two different attenuations: -6dB and -14dB. Then, use the relays and analog switches to select appropriate attenuation combination.

[Skimask]

No, not attenuation as the word "attenuation" would properly suggest.
I'm looking for pure resistance in the "ohm" sense, not dB drops.
I get where you're going, but that would suggest that dB drops can and would only apply to a known definitive load.

[Kalvin]

Ok, now I got it (looked finally the link in your first message) 

I kind of like TimFox's idea of using photoresistor. You should be able to perform software calibration for the resistor value as a function of the LED current. However, as  SArepairman pointed out, the temp coeff is approx. 3000 ppm/C, so it may or may not be a problem. Maybe you could be able to keep the photoresistor and the LED at the constant temperature with some kind of heater / oven.

[T3sl4co1l]

Added benefit of an OTA is full range, bidirectional, current mode output, so you know precisely how much current you're delivering*.  Measure the terminal voltage and there's your plot thing.

*Assuming compensation for "programming current" input Vbe, and temperature (gain is a function of T_absolute).  Not terrifically bad to deal with.

Making a "really big" OTA would be a bit of a challenge, if needed.  But not intractable.  Still, >40dB+ range selections would be challenging to be repeatable, in which case you'd rather want a switched range (most likely using relays).

Tim

[Skimask]

It's all good.  I'm learning something nonetheless...

I see what you're getting at with the OTA method, although I don't fully understand how it works yet.
Still boiling down to simple vs mildly complicated and PCB space.  The setup I've got now is only going to 'cost' me ~2 sq.in.  If I knew the OTA method might 'cost' an order of magnitude less PCB space, I'm thinking I'd be all over it.

As far as tempco goes, I had a semi-sorta-half-baked plan in the back of my head of putting the whole thing in a semi-sealed box, a few DallasSemi temp sensors, a couple of fat resistors and a fan...to keep everything at one temperature-ish.  Maybe it will work decently, maybe not.  The digital pot's themselves (AD5204's, AD5206's) have a datasheet tempco of 15 ppm/C in my configuration.  Whether that'll hold true is another story.