When a Precision Pot is Not!!

[mawyatt]

We've all used Precision Pots like the Bourns 3590 Multi-Turn types with the Dial attachments. These can be beneficial for use as precision variable voltage attenuators, but mostly for DC or very low frequencies.

While creating some information for another post about Bode Plots, we decided to give these precision pots a test with such since we've been curious about the response for some time now.

https://www.eevblog.com/forum/testgear/bode-plot-up-to-a-few-100-mhz/

Using the setup above with a Bourns 3590 10K 0.25% Linearity type with Dial, we did responses at 4 (#211), 6 (#210), 8 (#216), 10 (#212), 15 (#213) and 20dBV (#214) attenuation. The Dial settings can be created by:

Dial Setting = 100*(10^(-Attenuation/20), where Attenuation is the desired value in dBV.

Note the frequency response has an inductive nature with rising amplitude response and positive phase shift as frequency rises, then declines as the internal capacitive effects and setup capacitance take over reaching resonance around ~70MHz. The inductive nature is due to the wire-wound type pot construction.

The overall response is reasonably good below ~100KHz, but degrades quickly beyond this.

Anyway, thought folks might find this interesting.

Best,

[2N3055]

We've all used Precision Pots like the Bourns 3590 Multi-Turn types with the Dial attachments. These can be beneficial for use as precision variable voltage attenuators, but mostly for DC or very low frequencies.

While creating some information for another post about Bode Plots, we decided to give these precision pots a test with such since we've been curious about the response for some time now.

https://www.eevblog.com/forum/testgear/bode-plot-up-to-a-few-100-mhz/

Using the setup above with a Bourns 3590 10K 0.25% Linearity type with Dial, we did responses at 4 (#211), 6 (#210), 8 (#216), 10 (#212), 15 (#213) and 20dBV (#214) attenuation. The Dial settings can be created by:

Dial Setting = 100*(10^(-Attenuation/20), where Attenuation is the desired value in dBV.

Note the frequency response has an inductive nature with rising amplitude response and positive phase shift as frequency rises, then declines as the internal capacitive effects and setup capacitance take over reaching resonance around ~70MHz. The inductive nature is due to the wire-wound type pot construction.

The overall response is reasonably good below ~100KHz, but degrades quickly beyond this.

Anyway, thought folks might find this interesting.

Best,

Yes standard wire wound pots are good for audio range and that is it. 
One thing to have in mind is that they are not completely analog, but quantized to the single turn value... They change value in discrete steps, a resolution. OTOH carbon and cermet have nonlinearities and noise and nonuniform surface so their change value for small movements is partially stochastic..

Potentiometers are fascinating devices.. so simple yet full of surprises...

For anybody that did not read this already, a must for a library:

https://www.bourns.com/pdfs/OnlinePotentiometerHandbook.pdf

Pages seem scrambled a bit, funny, it is from Bourns themself..
 

[mawyatt]

Recall some of these multi-turn types utilize a composite conductive structure, wire wound with a conductive plastic coating. This gives a smoother "feel" when turning, rather than the more discrete step type, and also gives a similar effect with the resistance change effects.

The 3590 feels very smooth when directly rotating without the added dial, so suspect it may have the plastic coated wire wound structure, the data sheet doesn't say anything tho. We have a different type 10 turn that's longer and thinner,  WXD3-13-2W 1802Mof unknown source that you can "feel" the discreteness in steps.

BTW nice reference, hadn't seen this before!!

Best,

[Johnny B Good]

 This is a subject that has recently become dear to my heart as a result of finally prototyping a MK III GPSDO with a u-Blox ZED9-T that I've managed to stabilise sufficiently to compare against my other 'pet project' (a thermally stabilised /barometrically compensated Efratom LPR101 frequency standard).

 The need for a betterr ten turn calibration pot than the three Bourns 3590s I'd bought from Banggood a couple of years back had been made all too painfully apparent with this test setup. The first pot went open circuit just over a year ago. Luckily, I still had a couple of spares to hand so was able to repair my RFS there and then.

 A month or so back, I unpacked the third pot to do some tests with only to discover that this too had an open cct track   So with the nagging thought that the one remaining pot might go the same way and the fact that the stepped incremental adjustments had become more of a pain than before, I've been looking to source the conductive plastic embedded wire wound version which Bourns refer to as "Hybriton"(tm) in their 3541 Precision Potentiometer with a resolution they describe in their datasheet as "Essentially infinite".

 I spent a couple of hours scouring the internet for (cost effective) sources of similarly enhanced multiturn pots to no avail, finally landing up on RS's uk site where they were prepared to sell a single 5K 3541 at a delivery and vat inclusive price of just over 47 quid!

 I have to admit that after looking at the astronomical prices of alternatives, that weren't Bourns, on ebay from american sellers taking the piss over shipping charges to the UK ($120 for a 9 dollar item) and likewise mouser's ridiculous £29 one off price on a minimum order quantity of 25 or 1000 - I forget which (one of the main line sellers did indeed have a minimum order quantity of a thousand on this part), I was almost tempted to make a purchase until the utter ludicrousness of shelling out over ten times the price of the ubiquitous 3590 (inclusive of a cheap a ten turns counter) brought me back to my senses.

 In view of the alarming failures of two thirds of my stock of 3590 5K ten turn pots, I need to buy at least another one to keep on hand as a spare. However, I'd really like to replace the existing pot with one that has an "Essentially infinite" resolution to allow the finer adjustment that I now crave.

 The only way I'd seriously consider buying a 3541 pot from RS is if Bourns were offering a ten year unconditional guarantee but I saw no mention of any warranty. Perhaps I should check out details on the warranty just in case.

 Anyhow, this was the first topic thread that even so much as mentioned multiturn pots (the metrology threads proved to be a surprising washout on this topic). I'm tempted to buy another of these 3590s from a UK based ebay seller in the hope that they were actually made in Bourns' Mexican factory rather than some chinese clone manufacturing facility supplying Banggood.

 The pictures used by the UK sellers look identical to the pots I'd bought from Banggood so I suspect no matter from whom you purchase these low priced pots, they're almost certainly going to be of rather questionble Chinese manufacture.

 The thing with these composite wirewound conductive plastic tracks BTW, is that the conductive plastic electrically fills in the gaps between turns with mini potentiometers (as long as you use them as such by connecting the wiper to a high resistance voltage sensing port). They're not only smoother in a mechanical sense they're also smoother in an electrical sense as well as Bourns are at pains to point out in their datasheet.

 An alternative solution is to use a carbon brush wiper, guaranteed to always bridge a minimum of two turns at any one time. It's not as effective as the wire wound embedded in conductive plastic "Hybriton"(tm) setup but an improvement nevertheless over the 3590's reliance on a simple metalic wiper pressed into contact against a wire wound track. Mind you, I've only seen this in a low resistance wirewound single turn pot where I think it was used primarily to gain a low resistance contact to allow the pot to be used as a rheostat without overheating the point of contact.

 Oh, and one final comment, I'd had a sneaking suspicion that I'd already downloaded that Bourns document which 2N3055 had posted a link to so clicked on it to check but gave up when the download stalled. Sure enough though, when I searched my datasheets folder, I found it in my ~/potentiometers/Bourns/ folder. It looks like I'd downloaded it only last week but I'm sure it must have been much earlier than that by several months at least. Regardless of that, it had proved to be a very entertaining and informative read.

P.S. I had another look on RS uk and Mouser and realised I only had to spend just over a pound more to raise the total order value over the £33.00 threshold to qualify for free postage on the Bourns 3549H-1AA-502A pot I'd put into the shopping basket to check what the total delivered cost would be. So, since I'd blown one of my two cd4066 quad bilateral switches, I added three TI SN74HC4066s to the order, qualifying it for free delivery for a total of £40.07 vat inclusive delivered price (exactly six quid less than I'd have had to pay RS for the 3541H version alone - their free delivery threshold is a whopping £50.00!).

 So the deed is done! I'll post a follow up to report on the expected return on my investment after I've completed the MK III GPSDO project and had a chance to upgrade the cal pot and fettle my RFS just a little more.

[PCB.Wiz]

This is a subject that has recently become dear to my heart as a result of finally prototyping a MK III GPSDO with a u-Blox ZED9-T that I've managed to stabilise sufficiently to compare against my other 'pet project' (a thermally stabilised /barometrically compensated Efratom LPR101 frequency standard).

... RS's uk site where they were prepared to sell a single 5K 3541 at a delivery and vat inclusive price of just over 47 quid!

As this is a trim setpoint, on a system that also then locks, you could also look at Digital Pots with EEPROM and DACS with EEPROM.
Digital pots are light on the number of taps, but you can get DAC+EEPROM in the Microchip MCP472x family, that are 12b DAC with low ppm drift specs.
The MCP4728 (quad) is widely available as a pcb module.

[Johnny B Good]

This is a subject that has recently become dear to my heart as a result of finally prototyping a MK III GPSDO with a u-Blox ZED9-T that I've managed to stabilise sufficiently to compare against my other 'pet project' (a thermally stabilised /barometrically compensated Efratom LPR101 frequency standard).

... RS's uk site where they were prepared to sell a single 5K 3541 at a delivery and vat inclusive price of just over 47 quid!

As this is a trim setpoint, on a system that also then locks, you could also look at Digital Pots with EEPROM and DACS with EEPROM.
Digital pots are light on the number of taps, but you can get DAC+EEPROM in the Microchip MCP472x family, that are 12b DAC with low ppm drift specs.
The MCP4728 (quad) is widely available as a pcb module.

 Thanks for the 'heads up' on the MPC4728. The prices on the Adafruit break out boards on Ebay were just over 18 to 20 quid delivered (just slightly more than Mouser's 6 quid one off price with 12 quid shipping). This time it's the Germans taking the piss.

 Still, they look an interesting option for a future project so that's another two PDFs downloaded into my ADCs & DACs datasheet folder.

 The Mouser order was delivered some time today... to my doorstep. I only discovered the parcel when nipping out to my local shops for some biscuits. They could have lain there, unnoticed, for several days so it was lucky I'd had the munchies and no biscuits in the house just then (and still no sign of a belated email advising me of its delivery - no note pushed through the letter box to say they'd left it there after getting no response to the door knocker or the bell)

 After logging back into my mouser.co.uk account just now and following a breadcrumb trail onto Fedex's tracking page, it seems the'd tried to deliver at 09:05 that Wednesday morning and had simply left it behind a plantpot on my doorstep (and like most other couriers, had simply slunk away "job done"). Remarkably, it only took 48 hours to ship it all the way from Texas to my front door step!

 Anyhow, I've had enough time to put Bourn's claim of an "Essentially infinite" resolution to the test and can confirm the truth of this claim. Mind you I'd had to dig out my cheap AD584LH voltage reference to get a stable enough 5 volt to connect the pot up to before I could indeed set it to within the 10μV resolution on the 20V range of my SDM3065X.

 I'd checked the voltage of a 12v SLA I had parked on the bench to power a home made drill that I use for drilling PCBs which seemed stable enough to use directly.  However, I'd neglected the effect of loading it with a 5K pot which caused the voltage to slowly sag at around 10 to 20 μV per second. It took a few minutes for the penny to drop before I dug out that cheap voltage reference.

 The test setup was a bit of a lash up but despite this it bodes well for when it's finally mounted and soldered into my RFS. Now all that remains is to either replace the shitty little Bourns 3266W multiturn cermet trimpot Efratom had chosen as an internal frequency adjustment control in their LPRO 101, with a TL431, assuming it's set close enough to midway when the external cal pot is close to its mid point, or else replace the 100K resistor that connects the wiper to the inverting input of an opamp (a virtual earth point) with a 1 or 2 MΩ resistor to both attenuate the range of adjustment and the effect of variable wiper contact resistance by a factor of ten or twenty.

 To be honest, I'm not sure which of those options would be the best choice. Assuming I still have enough adjustment using a 1 or 2 MΩ resistor, that would still allow me to reset the extl cal pot to its mid position this would be the simplest mod. OTOH, the fixed 2.5v from a TL431 could offer even better stability but I'd have to fit a selected on test bias resistor between either ground or the 5v (from the temperature stabilised μ7805 dedicated to feeding the 5K cal pot alone with a low noise voltage) and the external C input pin.

 Either way, I'll have to strip the RFS down in order to probe around with a voltmeter to ascertain the actual voltage setting (and identify the correct 100KΩ resistor) before I can proceed any further so, for the time being, I'll just be upgrading the cheap 3590 to the expensive 3549H-1AA-502A for now which is a relatively simple upgrade job that will allow me to more accurately test Bourns' claim of "Essentially infinite" resolution.

 However, that is all going to have to wait until I've finished building the MK III GPSDO which is currently a rather fragile test build on solderless breadboard. I'm awaiting delivery of a couple of M74HC4024 7 stage binary ripple counters to replace the dual divide by ten 74HCT390 ripple counter IC I'm currently testing with. I suspect it won't make much difference performancewise but it will shrink the package size by two pins.

[T3sl4co1l]

Not surprising that a large-value pot behaves poorly above audio frequencies.  I'm a little surprised, honestly, that it's still inductive despite the large value; I think (but haven't really tested) even meg range wirewounds can still be inductive, though it must not be over much span, for values quite that large.

What about precision/linearity, is it as good as it's supposed to be?

Tim

[Johnny B Good]

Not surprising that a large-value pot behaves poorly above audio frequencies.  I'm a little surprised, honestly, that it's still inductive despite the large value; I think (but haven't really tested) even meg range wirewounds can still be inductive, though it must not be over much span, for values quite that large.

What about precision/linearity, is it as good as it's supposed to be?

Tim

 I'm guessing that that question was aimed at me.

Interestingly Bourns specify the total resistance of the 3540/3541 to a tolerance of +/-10% with a linearity of +/-0.25%. The higher spec 3549H-1AA-502A that I'd bought from mouser also has the same 10% resistance spec (for the Hybritron element - the wire wound version has a 3% tolerance) but a slightly tighter linearity tolerance of +/-0.2%.

 The resistance tolerance might seem a rather slack one but this is normally of little importance in a potentiometer: linearity trumps resistance tolerance every time. Total resistance tolerance only becomes an (admittedly in this case, minor) issue when used in the LPRO 101 where it is fed from a 12K series resistor connected to the internal (presumably more stable)17v regulated voltage supply rather than using a direct connection to the 5v regulator used to supply the main logic chips. That's one of the reasons why I'm tempted to replace that trimpot with a TL431 and shunt the 12K resistor with a 3.9K (1mA is bit on the low side for a TL431 but I don't want to raise its internal dissipation much beyond the 10mW mark if I can help it).

[Conrad Hoffman]

Long ago I took apart a big multi-turn pot and was surprised to find the resistance element was wound on a piece of heavy enameled copper wire. IMO, that should kill the Q of the winding to a large degree. You're than left with the inductance of the larger spiral for the multi-turn action. No doubt there's stray C all over the place.

[Johnny B Good]

Long ago I took apart a big multi-turn pot and was surprised to find the resistance element was wound on a piece of heavy enameled copper wire. IMO, that should kill the Q of the winding to a large degree. You're than left with the inductance of the larger spiral for the multi-turn action. No doubt there's stray C all over the place.

 Those L & C strays are all part and parcel of any wirewound resistor (whether a fixed value resistor or a potentiometer resistance track, multiturn or not).

 In the case of the latter, the design brief only ever considered extremely low frequencies such as "DC" and were never intended to be placed directly into any signal path as an analogue attenuator such as a volume control in an analogue radio or HiFi setup.

 Despite mawyatt's interesting bode plot results indicating that the 3590 pots could be used as a volume controller for audio, there's really no desire to replace a conventional "log-law" volume control pot with a linear law one, no matter the precision, so the whole issue of their stray inductance and capacitance is, to say the least, rather moot (or, to put it bluntly, a total irrelevance ).

[Johnny B Good]

 When I first joined this topic thread, only 11 days ago now, I'd expressed my interest and concerns for the last remaining of my three Banggood sourced "Bourns 3590" precision 10 turn pots (with turns counter) and my decision to blow some 40 odd quid on a genuine Bourns 3549H-1AA-502A from mouser.uk.

 As it happened, my concern over the last pot going open circuit, just like its two predecessors purchased a couple of years back, was entirely justified since this was exactly what happened sometime between Monday and Wednesday when I'd decided to compare the MK II and breadboarded MK III GPSDOs foregoing monitoring of the RFS. Yet another case of Sod laying down his law - make something go wrong when you've stopped monitoring the thing you fear will go wrong.

 Anyhow, I guess another answer to your question is "When you purchase Chinese fakes from Banggood". They might look indistinguishable from the real thing and even work just like the real ones for the first 12 to 18 months or so (regardless of whether used or simply placed into stock).

 Since they all seemed to have gone open circuit at the 70 to 80 percent mark (close to the clockwise end of the track) and testing the last one showed circa 56MΩ rather than a >100MΩ reading open circuit with a bit of hysteresis on winding back from the point at which the reading had dropped to just over 1KΩ from the CW end, I'm convinced that some corrosive contamination introduced by careless Chinese manufacture of these fakes caused the resistance wire to fail and unwind a turn or two giving rise to the hysteresis effect I'd observed.

 So, there you have it (my tuppence worth opinion on the subject). I guess the whole world market must flooded with these fakes by now so if you want to avoid the eye watering pricing of the genuine manufacturers' product, you'd best pay the premium of using an Ebay seller operating within your own country's borders (in my case, the UK), rather than the likes of Banggood and AliExpress. That way, if you do experience similar failures, you're more likely to be able to make a successful warranty claim or obtain a refund.

 I might have simply been unlucky in receiving a 'bad batch' which had been released for distribution due to minimal or even no quality control (a seemingly typical failing of cheap Chinese goods manufacture).

 Even so, unless you only need them to function over the next 12 months in a prototyping project, you'd best be mindful of my own experience (which may or may not be typical - a sample of three is just too low for a statistically meaningful conclusion to be reached).

 This does raise another issue with these cheap "Bourns 3590" precision 10 turn pots (with turns counter) offerings, namely, how many other buyers have witnessed similar failures within the first 12 to 24 months?

 Yesterday afternoon I cleared my bench of the MK III prototype and its detritus to make way for repairing my RFS which has spent the last 3 or 4 weeks in my unheated garage/workshop undergoing ambient temperature extremes (a 27 deg September daytime peak down to a November nighttime low of 4 deg Celsius) and I can now report on the claimed ("essentially infinite") resolution of the Bourns 3549H-1AA-502A as being effectively true (4ppm setting accuracy despite the use of one of the ten turns counters that had been supplied with the  Banggood sourced 3590 pots).

 Surprisingly, I was able to get the internal trimpot to syntonise the RFS spot on frequency with the MK II after presetting the calibrate pot to the 500 midway mark (2.59998v off a 4.97849v reference) without too much difficulty. I got the impression that the trimpot isn't too far from its midway setting so, depending on the actual trimpot's wiper voltage, I'll try swapping out the opamp's 100kΩ input resistor with a 1MΩ one to improve the fineness of adjustment by an order of magnitude (and also reduce the effect of the trimpot's inherent instability by the same order).

 However, that's a modification I'll leave on the back burner for now since I want to make further refinements to my Arduino baseplate temperature control sketch. It's a job I've been itching to get done and dusted these past 3 weeks which means the MK III project will now be set aside for the time being.

[Terry Bites]

Its pretty unusual to put a pot into an hf circuit for the reasons you have uncovered.