That is too bad. My box is about +/-2.5% on all values

As explained in the datasheet, you can use any silicon diode if its temperature coefficient and forward voltage drop are specified or can be measured. The 1N457 is convenient because its characteristics allow for a simple ratio of 10 between the two resistors used. To obtain near elimination of the temperature coefficient of the current, you can trim the resistors' values.

An engineer also needs to know the limitations of the SPICE program they are using compared to real life. There's no point in designing a single transistor amplifier using LTSpice, for example, if the dc operating point ends up being highly dependant on the transistor β. LTSpice only simulates one value of β for a given model, whereas that transistor may actually have a β that varies from sample to sample by a factor of 7 or more. You have to know the parts you are intending to use, and design circuits in a way that compensates for variations in their parameters.

@@RexxSchneider Then you run a Monte Carlo sim,...

@@mr1enrollment ... and get the same results in the simulation for the dc operating points every time. The other option is to add new transistor models to the library such as BC547lo (with a β of 110) and BC547hi (with a β of 800). Then swap the parts in LTSpice and observe the effects. You still need to know the parts you're using.

@@RexxSchneider Dude - you include the transistor parameters in the Monte Carlo. The thing is: Spice is just a big calculator. It is faster than I am. But I am smarter, so it is on me to organize the problem and simulate it correctly given the requirements. Your example: re Beta and op point is a bit well,... (add the word). When you have a device, (like all transistors), you best know how to arrange the circuit topology and the values around it so that the beta is less important. Because parts do vary. That's part of the game, and the challenge. Spice is simply a tool and a good one. If you try to drive a nail with a screw driver, you may get it done, but a hammer is better. For a simple common emitter amplifier, I can most times design by inspection, and it will be stable, with predictable input and output impedance - and the correct gain. As long as I am not trying to push the boundaries of the device. The more complicated the circuit requirements the more the tool makes a difference. So whats all this stuff about beta anyway?,...

@@mr1enrollment So, "Dude", when you say _"When you have a device, (like all transistors), you best know how to arrange the circuit topology and the values around it so that the beta is less important"_ did you mean you have to know how to design the circuit so that it compensates for variations in the parameters? Which would be exactly what I explained to you in my first comment, and a direct contradiction of your own first comment "engineers need to know all the parts … Well no not really." Spice is indeed a big calculator, but LTSpice doesn't have any means of setting the transistor parameters in a Monte Carlo simulation. Don't you understand the importance of a parameter like β in designing a common emitter, for example? If I specify a 2N2222 with a collector current of 1mA and a 10V supply to give a gain of 16 with a full output swing, then I know I can't get an input impedance more than about 4K, because the base bias resistors will need to carry about 0.2mA so that the maximum base current of (1mA/50) = 0.02mA remains insignificant as the β varies from a minimum of 50 between samples. Check it with Rc=4.7K; Re=270R; Rb1=47K; Rb2=4.7K. If I'm tasked with producing the same performance but with an input impedance more than 10K, then I would use a BC547C instead. With a β of 400 minimum, the maximum base current is reduced by a factor of 8. That would allow Rb1=120K; Rb2=12K, giving the required input impedance and the variation in collector current across samples would also be reduced considerably as a bonus. I can do all of that because I know the parts that I'm intending to use from 50+ years of experience, not because of any simulation software that does little more than confirm what I can design approximately with a simple calculator. So, yes really, engineers do need to know all the parts.

yes but not very stable.

@@IMSAIGuy For simple uses, a depletion mode JFET with a resistor between source and gate is a pretty decent current source, and is quite stable at a fraction of the price of an LM334. The problem is the repeatability of the JFET current source with different samples because of the inherent range of Vgs(off).

@@RexxSchneider Yes I use J201 with resistor in source (300 ohm), which makes it a current source, for discharging capacitors with a constant current. This method of discharging caps is safer and does creates sparks and does not damage the cap. This works up to high 30s Volts. For higher voltage caps, I use this configuration in the emitter of high voltage NPN with base connected 5V (cascode configuration) and collector connected to positive terminal of cap and negative terminal is connected to source resistor of j201. Remember this method for discharging the high voltage caps, bring their voltage down to 4.2ish volt. After that you should use the original method to discharge the rest of the voltage. I can find NPN transistors that go all the way to 1500V, but I could not find JFET that could go above 200. I guess I can use Vacuum Tube version for higher voltages. If you know of any inexpensive JFET or Vacuum Tubes that can handle high voltages let me know. Thanks.

look for LM134 LM234 and LM334 they are all the same. I think mouser has the LM334

only 3 pins. so not sure what it might be

It's not easy to turn an LM334 into a voltage-controlled (Howland) current source. Best I can suggest is to use a photo-FET optocoupler like the H11F1M in place of the set resistance, or in parallel with 680K. I don't expect it would be particularly linear over a wide range, though.

Perhaps a new video showing a Howland current source? Drive with voltage source to make current source. Using a dac on an Ardiuno to make a v-i curve tracer for components

@@jjoeygold The classic Howland circuit ideally needs a split-rail opamp with the input voltage referenced to ground. That allows the output current to drive into components referenced to the negative rail. The voltage source also needs to be able to supply the full output current, and these two constraints make the circuit less popular. For your application, you only need a linearly rising current and there are probably numerous other simpler solutions such as a pnp current mirror being driven from a linearly falling voltage through a fixed resistor. It would still make an interesting video, though.