SmittyHalibut
commented
1 year ago Source:
http://www.geofex.com/article_folders/potsecrets/potscret.htm is a good page talking about this.
Qucs simulations
I also did some Qucs simulations:
Source impedance is 10R. Pot is 1k (split into R2 and R3, with equations doing the heavy lifting.) The taper resistor is 1/10 of the pot (100R) for a large change in slope. The load impedance is 100k.
The source and load impedances are 100x outside of the pot (source = 1k/100 = 10R, load = 1k*100 = 100k), to minimize the impact of the variable impedance the tapered pot will represent.
Having said that, I also tested with a 10k and 1k load resistance, and the graph barely changed. This makes sense since the load is in parallel with R_taper; as long as R_load is 10x R_taper (or, 1k in this case), it will be dominated by R_taper and not make much effect on the whole graph. Meaning, our load impedance can be a wide range, so long as it's ~10x R_taper or more.
Qucs Result: Output voltage as a function of sweep
Blue line is the resultant voltage, graphed linearly. The red line is the same, graphed logarithmicly. A perfect log pot would result in a straight red line, so curvature in the red line shows how not-logarithmic the graph is.
So, this isn't a great logarithm, but it's WAY better than a linear pot. Notably, it gives a lot of precision at the low end, which is what I really want.
Qucs Result: Input impedance presented to the source, function of sweep
This shows the impedance of this pot as presented to the source. Note that it varies by >90%, from 1k down to below 100R. This is why this architecture isn't normally used.
But, as long as your source can handle driving the minimum presented impedance, then it should be ok.
Analysis
I like the simplicity of a single pot.
Microphone trimmer
The source on the Mic trimmer is our own opamp, so we know the source impedance is very low, plenty low to drive the pot.
The load is the microphone input to the radio, which is likely to be at LEAST 600R, more likely 1k or 10k. I graphed a 600R load and it's still pretty darn good.
So, I'm calling this good.
Headphone/Speaker trimmer
I don't get to specify the source on this one. The source is likely to be:
- Line: 600R
- Headphone: 16R or less
- Speaker: 8R or less
So the worst cast to consider is Line output, 600R. Unfortunately, even with the volume turned all the way up, this results in a voltage divider between source impedance (600R) and R_taper (100R), which results in a 1/6 decrease in voltage, even at full blast.
However, I specify the load here, our headphone amp, which is 10k. So I can go to a 10k pot/1k R_taper safely. That makes the voltage divider 600/1000, or a little less than half voltage.
Going to 100k/10k makes R_taper and R_load equal, effectively dropping R_taper by half. This results in a much flatter graph (not a bad thing), and a full volume much closer to 100%:
Conclusion
- Mic Trim: 1k pot, 100R taper resistor.
- Headphone Trim: 100k pot, 10k taper resistor.
A common problem that all OHIS manufacturers will run into is the lack of availability of audio taper trim pots, for use when matching levels between OHIS and devices.
AI8W suggests putting a fixed resistor in parallel with a linear pot to get a somewhat-not-really logarithmic taper. I'll research that and see what it does. (Easy to simulate, but I suspect someone already has done that.)
The other idea I came up with is to split the 10/90% audio taper into two separate pots stacked on top of each other, one 1/10 the resistance ("fine") of the other ("coarse"). Take the output from the sweep of the Coarse pot, and tie the sweep of the Fine pot to the midpoint between the two pots. For adjustment, turn the Fine pot all the way up, and use the Coarse adjustment to find your volume. If with the coarse all the way down it's still too loud, then start adjusting the fine.