Cermet potentiometers of course are not new, but are still relatively unknown. The reason they are used in the Model A is simple: they are sonically clearly superior to any other potentiometer type.
Still, they (apart from being somewhat fragile) have one shortcoming too: low-level inter-channel miss-tracking. I am not aware if this was discussed previously anywhere, so more information on this follows.
As opposed to the conductive plastic potentiometers, in which two channels are nicely balanced at all levels, cermets “officially” have two kinds of inter-channel tolerances: relative and absolute. Typical relative tolerance of 2% is acceptable (conductive plastic pots are similar here), however typical absolute tolerance of 3 Ohm is a potential problem, as it becomes a major contributor at the lowest levels, taking into account that we are operating here with relatively low impedances, namely with series resistance below 1 kOhm.
Also, dealing with cermet pots for some time, and measuring many samples, I have found another, “unofficial” and potentially more critical cause of inter-channel miss-tracking. Due to manufacturing issues, the cermet “audio taper” is not really logarithmic but is the approximation made of two or three segments of differently changing resistivity, which simulates the required logarithmic curve. The problem is, that two channels often do not cross the “borders” between segments fully simultaneously, so the resistance of one channel changes there more rapidly than that of the other. This again happens at relatively low levels, somewhere between 5 and 20 Ohm (Vishay), making the pot possibly useless at and below this point.
So, with such a potentiometer used as a shunt element, and 750 – 820 Ohm series resistance (please see the previous article on Model A volume control), we can usually have 30 – 40 dB usable volume control range. In most cases such a range is in fact sufficient, however different systems have different gains and efficiencies, and customers’ requirements are generally different as well. As a result, and adopting the classic 60 dB attenuation requirement for volume control, this low-level inter-channel miss-tracking may leave us with 20 – 30 dB of the unknown part of the loudness equation.
Now, there is no universal solution to prevent this inter-channel miss-tracking to come into the actual listening level. For the customers it is usually too complicated to intervene in the gain of the system, however there is a simple fix to move this point down, and it is to add some series resistance in front of such a volume control. (Of course, such a fix can be applied to the shunt potentiometer only.) So, by using say 1.5 kOhm instead of 750 Ohm (or simply by adding 750 Ohm – a miniature resistor can be put even into the RCA connector), this point will drop down by roughly 6 dB.
It can look like an inconvenience, and added resistance may compromise a bit the performance of the potentiometer itself (remember, we want low impedance), but things can ultimately be set to operate smoothly.
And the reward of cermet potentiometer sonic properties is high enough. No other volume control sounds that natural and full-bodied. And in my view, it is only digitally controlled resistive networks (sometimes mistakenly referred to as “digital potentiometers”) that can come close to the performance of such a volume control setup.
Related topic:
Model A volume control
Interesting & thank you for sharing this with us. I would have two questions: any type of potentiometer you would recommend? (Bourns model 95 or Vishay P11 / P11L) Problem is unfortunately to source these in 1K with audio (logaritmic) taper…
My second question (and solution to the sourcing problem): did you try a linear pot using a law faking resistor with same results?
Best regards,
Berny
Hello Berny,
Cermet pot availability might be the hardest part here. However you do not really need 1 k, you can use 5 k, or even higher potentiometer values with say 820 Ohm series resistance. Of course, the higher the pot value, the less convenient volume adjustment will be, but so long as you are setting this volume control for your own system only, I believe you can tweak it nicely.
In most cases linear potentiometer will be too rough though, and law faking resistor won’t be of much use in this case.
Regards
Cermet pots from SFERNICE has a good reputation in France in the DIY communauty in France! It’s an industrial part with a good reliability and tolerance !
If I remmember , it can be sourced at selectronic.com
Maybe Worth to try this brand too, but I don’t know the Z range of choice but for sure they have both linear and logaritmic pots.
Ah just seen SFERNICE is now a brand of Vishay…
Hello Pedja,
I have made some experiments with active an passive carbon pots and would like to try now a cermet shunt pot after having tried an ALPS Black Beauty pot in serie (in passive: very good tone but soundstage a little 2D in my setup).
if i understood your paper, there can be a mismatch between right & left channel due to the way cermet pot are made.
So I vould like to use two mono cermet pot, one for each output of my AYA 2014 board. Also in the goal to have a better manual control of the balance in my room (I don’t care to have a double mono volume pots, ok for me)
Having a 100K ohms imput amp & planning to go passive with the cermets pots just after the output of the AYA II 2014 (I use no DC caps, hence the last 100K R leaved out) : what should be please the value of the shunted pots ?
5 K ohms log pot is ok ? output in serie with the ground & input of the pot in shunt position with the serie signal if I understand your shematic ?
Many thanks
Eldam
Sorry, I meant:
signal after the buffer & last serie resitor of the AYA II to the input of the pot (then in serie to the amp)
Output of the pot connected to the Gnd of the pot : both shunted to the ground of the circuit.
Is it ok ?
Apologies for not being able to address all the questions in a timely manner, and sometimes I even overlook to do it later.
But, not to leave this topic floating:
If you want to use the AYA II series resistor as the only series element for this shunt volume control, I’d expect such a load to be too tight, so the AYA will compromise its performance.
And then the shunt element must be very low value, probably lower than anything available. In my experience, the pot value can be up to 5-6 times higher than the series resistance, otherwise, it may get really uncomfortable to set the volume at the lowest levels, even with low or medium sensitivity speakers, assuming 2 V RMS source and 25-30 x amplifier’s gain.
So, it is not necessary, but if any possible, it is really the best to use the pot value as low as the series resistance, or even lower than that. And, to add to, or partially correct my above answer to Berny: in the meantime, I tried a 1k linear taper pot with 820 Ohm series resistance – yes, it may work fine. A higher value linear taper pots will be usually too rough.
Regarding the pot connections: yes, that would be fine. Also, please note that the pot’s low end goes to the signal, whereas its wiper and high end are connected to the ground.
Thank you so much for that Pedja.
Hello Pedja & Eldam,
Yes, thank you for the reaction… I’m still using my compact AYA (without powerswitch 😉 ) on a daily base and very happy! And using the 1K linear pot version with 820 series… integrated in my dac (see pics in the forum here) which is ‘OK’ towards control-range. I can’t find 1pc of a stereo log cermet version unfortunately. I’m finishing currently different attenuators to compare: a cheap autoformer, linear carbon track studio glider, Muses 72320 attenuator with fet buffer and this cermet pot which is currently my reference (luckily as it is the one integrated in the dac, I have 2 outputs with a selection switch). Bass definition of the cermet is very nice, but the autoformer is adding some “airyness” I like… but loosing in the low end. When everything is finished and deeply compared, I will further explore the best solution.. maybe the cermet with some 1/1 transformer in the end. As I want a remote, I might develop a dual RC servo drive for mono log cermet pots.
Hi Berny,
I have a diy Muse volume control kit I plan to try without the oap like some do in their preamp.
I tried a passive 25k ALPS Black Beauty (courtesey from Sumotan fellow) before a 100k input amp…not bad but it colored the sound and finally less good than a cx2 yamaha pre.
I hope the Muse to be transparent but dunno yer if the 150R is the way to load it.
The buffer second oap 861 is perhaps not necessary…I don t know. There are now cool buffer at Texas Instrument or else working 5V but the layout could be a challenge and oscillation not far. The better is maybe again a single op861 as I/V followed by a diferent R value…or a pair of smd jfet from Toshiba…Pedja if I am correct dit the Model S fulll discrete…
Cheers.
Currently, my standard for buffer is still the 2sk170/2sj74 combo at about max12V and that is how I plan to use the Muses (Pass-style: no opamp and potentiometer output driving directly the fet-gate). What buffer from Texas are you refering to? First hand experience?
The thing with my current setup: Aya with the internal 820ohm + 1K cermet pot drives nicely my Vfet amp and I like it to avoid completely the need for a preamp. But sometimes with lower level recordings and me wanting realistic ‘live’ volumes, I’m missing some gain (×2 or a bit more). I’m easily at maximum with the cermet volume control… Dont have the problem with my Takman metal film 32 step relay attenuator, but again missing the low end and the cermet is better overal.
Talking about Cermet… anyone tried the Cermet element of such a trimmer as I/V?
Yes, it worth noting that such a shunt volume control, with series and shunt element having equally the same resistance, will attenuate the signal for roughly 6 dB at the highest position.
I would be cautious with non-fixed resistance for I/V. Also, even if OPA861 I/V may work fine with lower I-V resistance, moreover, the OPA861 is officially characterized by Texas Instruments with 500 or 50 Ohm collector load, I would not expect it to perform consistently over the range required for volume control.
Also, as I was told many years ago by the OPA861 designer, most of the failures they experienced during the testing happened while using the OPA861 collector output.
@ Eldam:
Apart from this, a buffered output is way more compatible with different gear that may be connected to the AYA II/III/4, so I just put another 861 as a buffer. And it is a very good buffer.
Toshiba makes nice SMD JFETs for audio, but there are no P-channel.
All the S DACs have “hybrid” I/V stages (transistors + IC), but their output buffers are IC only. 🙂
Hi Pedja, just saw I still have the 150Ohm series output resistor that I’m going to remove keeping only the 820 Ohm + 1K Cermet. Probably moving also towards 2 mono Log Cermets as these are more widely available. Currently in all my comparisons, the Cermet is still the overall winner.
What would be your take to add an output transformer? Was thinking of a 600:600, directly connected at the OPA861 output and secondary feeding the attenuator. Am also tempted to further reduce the 820Ohm resistor a bit in this setup because of transformer impedances protecting the OPA861 anyway…?
I did not use the OPA861 with coupling transformers (the S DACs output buffer is not the OPA861), but some others did, and reported the setups to work well.
I don’t think it is a good idea to omit the series resistor, though. Coupling transformers will add their own windings resistances to the overall impedance at the path, but 600:600 transformers have low resistance windings, usually less than 50 Ohm at each side.
Also, my experience with line-level coupling transformers, including the nominal 600:600 ones, was that they don’t give their best in the sub 1 kOhm environment. It is not that they can not work well here (finally they are specified to work at 600 Ohm), but at higher impedances they do sound better.