Need FAQ on tolerances

[Mark Hammer]

The TS808 substitution thread  prompted me to look in the DIY-FAQ document for something about component tolerances.  There were two mentions of "tolerance" found.  Pleasing ones to be sure, but not enough comfort or direction for newbies.  Consequently, I'd like to solicit information from people here with a little more experience with components for the purposes of building up a knowledge base for a decent FAQ insert on tolerances.

What kind of information should this be?

Well, certainly things like your experiences in actual variation in component values.  For instance, how much variation is there really between 1N914's, whether within or between brands?  When you buy a pot that *says* 250k on it, what is it possibly going to be for real?  Resistors are spec'd at 5% tolerance - how far do they actually fall from the stated value?  How much variation is there going to be in value for different types of caps?  Will swapping a plastic "greenie" cap for a ceramic one with the same value printed on the side result in MORE than simply a change in cap type?  How much variation is there in commonly used types of transistors, again whether within or between manufacturers?  Does heat applied in the soldering process run the risk of changing the value?  How important is use of 1% components or matching of components?

A second aspect to this concerns parts substitution.  There are many misconceptions, or rather question marks, about when another component value may be used "safely" (i.e., it won't change the basic functioning or viability of the circuit).  I understand, and endorse the attention to detail and adhering to instructions on the part of newbies, but it is also the case that sometimes their projects are held up for reasons of component values that actually make little difference to the working of the circuit.  For instance, some of the old Anderton projects will call for cap values like 1.5uf or 2.7uf.  Just TRY finding those anywhere!  The novice has no idea what can go in place of these rather exotic values and may falsely perceive a project as hitting a roadblock.

So, please post whatever advice you can offer here, and we'll see if it can be woven into a useful FAQ insert to help recent converts out of a jam.  Thanks in advance.

[aron]

For myself, I have found most 1N914, 1N4148 to be the same. Pots generally are right in the ballpark too.

About the only component that's really all over the place are J201s. I haven't measured my caps but I suspect the smaller values are all over the place.

[gez]

I haven't measured my caps but I suspect the smaller values are all over the place.

It's usually the other way round, large caps tend to have wide variance and things get better as capacitance gets smaller.

[stm]

In my experience I've seen pots with very loose tolerances. For instance, I got some 1 Meg Audio pots that actually measured 750k!

Also, your circuit may specify 250k for a pot but you can also go along with 200k and 220k which are standard values. The same is true with 470k and 500k (and of course the multiples of ten of the cited numbers).

As a general rule, most of the time you can use nearby component values with little impact. For instance, a 1k8 resistor could be replaced with 1k5 or eventually even 2k2, depending on where it is in the circuit. If there are capacitors nearby that will set a cutoff frequency, it will be shifted somewhat, but not necessarily for bad. Also, sometimes you can be clever enough to compensate. For instance, if you have a filter formed by a series 10k resistor and a 15nF cap to ground and you don't have 15nF but 10nF, then you can replace the cap and readjust the resistor to 15k, thus keeping the original cutoff frequency.

As for certain values, you can always put components in series or parallel in order to obtain a value you don't have.

Voltage dividers with equal resistors are much more tolerant. A 10+10k voltage divider could be setup easily with anything between 3k3 and 47k with basically no impact on the circuit (assuming it has large bypass capacitors).

Also, the typical 1 Meg input resistor could be perfectly replaced by 820k, 680k or even by 470k in case of emergency with little impact on the sound (unless the input capacitor is real small).

Bipolar transistors can usually be replaced quite freely, like 2N2222 and 2N3904, and eventually 2N5088 (which is higher gain) depending on the circuit.

On the other hand, JFETs are quite tricky to just replace one by another, since biasing networks rely heavily in the Idss value of the FET. Also, as it has been demonstrated by many builders, the type of FET alters the sound of a circuit due to its gain and clipping characteristics, as is well documented for the BSIAB II.  The general rule is to look for a device with similar Idss and VgsOFF specs, but anyways FETs have very loose tolerances on these parameters.

In particular, whenever you need a JFET as a voltage controlled resistor (phaser, compressor, etc) it is recommended to find a device with a high VgsOFF for best dynamic range and lower THD.  For instance, if you consider the family of the 2N5457-5458-5459 a quick look at the datasheet will show you the best candidate is the 2N5459, followed by the 2N5458, and last the 2N5457.

In terms of opamps, you have to be careful with certain intrinsic characteristics, like input impedance, input noise and gain-bandwidth product, among others. In some applications an RC4558 may be substituted by a TL072 or NE5532. In others the former may be preferrable to some tastes (TS?). Here it is more a matter of trying and listening.

The bottom line is to feel free to experiment and use what you have around with CRITERIA. I love cooking feel comfortable changing and/or adding ingredients from a recipe. Most of the times the result is good/better, but of course eventually I screw up and learn--don't fry mayonnaise!

[Paul Perry (Frostwave)]

The only place I have ever seen matched diodes required, is in a passive (transformer style) ring modulator.

[Mark Hammer]

The variation in diodes is not so much pertaining to matching them, but rather with reference to the impact of varying voltage drops on signal level.  If you want a "hotter" Distortion+, and have a pile of diodes, knowing that there IS variation would lead you to measure the diodes in your parts bin and select those with perhaps the highest measured voltage drop - e.g., 580mv instead of 510mv - just to get a little more output signal.  Likewise, if your were making a sidechain-controlled device with a half-wave rectifier, you might want to select a diode with the lowest voltage drop so as to "conserve" more envelope signal.  Knowing that there IS natural variation provides one with the possibility of doing these things more strategically.