Measuring forward voltage of diodes?

[Chuck D. Bones]

The designer should have an idea of what part parameters are important. As DIY builders we need to know that too if we want our stuff to perform like a production pedal.

Robert has started putting parametric data (HFE, lkg, Vp, Idss) on some of the Pedal PCB schematics. My understanding is these values were measured on devices from a production pedal. We still don't know how important each parameter is or how much margin we have on the stated numbers, but it's a decent target to shoot for. When I post schematics, I'll indicate critical voltages in the circuit and the notes usually indicate how to dial them in.

 

[Big Monk]

No, I'm not saying that because the datasheets rarely, if ever, provide enough info to tell us how a transistor, diode or opamp will work in a pedal circuit. It gets worse: using a certain part number Ge transistor, Ge diode or JFET is no guarantee of success because there is too much part-to-part variability. Professional pedal builders do one of three things about it:
1. They screen the parts to a criteria known only to them. We know that Finnegan did that. I suspect that smart guys like Brian Wampler do it too.
2. They build a bunch of pedals and scrap or rework the ones that don't sound right. I don't know for sure that anyone does that, but I have my suspicions.
3. They build a bunch and let the customers figure out if they got a good one or not. This was common practice with germanium FuzzFaces back in the '60s.

I test parts, audition them on a breadboard and/or socket them on the PCB. I will also tweak resistor values as req'd to dial-in the bias on JFETs & Ge trannys. Trimmer resistors are not a panacea. If you have to go to the extreme ends of the trimmer, then you have the wrong transistor installed. You've probably seen me write: "not all pedal circuits are plug-n-play." If you want easy, then stick to Si transistor and opamp circuits. IMO, the rewards for getting a Ge or JFET circuit to sound right are worth the effort.

I treat the values measured on my TC-1 for transistors and diodes similar to how I used to treat calculated IBU values for hops when homebrewing beer:

1.) Unless you lab tested the finished beer, you had no actual IBU value.
2.) You could use the calculated IBU value and your taste buds to have a frame of reference for bitterness.
3.) In future batches, you increase or decrease the bitterness to taste based on the new calculated value and the old baseline.

With respect to diodes and transistors, if a measured value of 335 hFE for a BC549B works in a silicon Fuzz Face, then new units tested on the TC-1 will be judged against that.

Furthermore, for diodes in a Big Muff, I know that the ones I like will test around 580 mV on the TC-1 so when I try new ones out, those get compared to that value.

 

[Chuck D. Bones]

The problem with doing that is a single data point does not make a curve. The diode currents in a BMP peak out at around 120μA in the first clipping stage and around 55μA in the 2nd clipping stage. Unless the TC-1 measures at those currents, it's not clear to me how its measurement is relevant. FD100, FDH400, 1N4937 & 1N914 all measure close to 580mV at 5mA, but they're all very different down around 100μA.

When you measure HFE in a BC549B, do you have a pass-fail criteria? IOW, how close to 335 does a transistor need to be to pass muster? This is akin to a question I used to ask a lot at work: "once you have collected this data, what are you going to do with it?"

 

[Big Monk]

The problem with doing that is a single data point does not make a curve. The diode currents in a BMP peak out at around 120μA in the first clipping stage and around 55μA in the 2nd clipping stage. Unless the TC-1 measures at those currents, it's not clear to me how its measurement is relevant. FD100, FDH400, 1N4937 & 1N914 all measure close to 580mV at 5mA, but they're all very different down around 100μA.

When you measure HFE in a BC549B, do you have a pass-fail criteria? IOW, how close to 335 does a transistor need to be to pass muster? This is akin to a question I used to ask a lot at work: "once you have collected this data, what are you going to do with it?"

I’m heading to coach my sons basketball practice but I’ll add more detail and expand on this when I settled in later.

 

[Big Monk]

The problem with doing that is a single data point does not make a curve. The diode currents in a BMP peak out at around 120μA in the first clipping stage and around 55μA in the 2nd clipping stage. Unless the TC-1 measures at those currents, it's not clear to me how its measurement is relevant. FD100, FDH400, 1N4937 & 1N914 all measure close to 580mV at 5mA, but they're all very different down around 100μA.

I openly admit that my methods are as crude as the trial and error methods for estimating bitterness units often used in homebrewing.

I want to stay on this parallel for a minute.

When I started homebrewing, there were a few crude methods of IBU (bitterness unit) estimating using various package data for the hops we were using.

Over time, I fell in with a crew of people who were basically brewing macro quality beer at home, along with all the analytical and equipment based bells and whistles.

We began modifying the popular Tinseth equation for IBU estimation using real time data from my buddies HMI for his brewhouse. We began to make the equation dynamic across the boil to track volume and its effect on bitterness.

When we sent the control beer away to the test lab, we ended up estimating the actual IBU to within the margin of error of the test equipment.

When we published this data, most people on the homebrew forums thought it was cool but preferred the trial and error method because it was easier and it’s what they were used to.

To them, the results they got were in step with their methods, such that the finished beer had the proper bitterness to taste, and it didn’t really matter to them if it measured to that value.

I’ll admit to being in this camp with respect to the diodes I’ve settled on for the Big Monk Pi. Having tested something like 2 dozen different part numbers of diodes, I found it hard to land on the 540-580mV vF I had targeted. Most of the common types like 1N914 and 1N4148 all tested > 750 mV on my TC-1. Other units I procured from Mouser tested < 450 mV vF on the TC-1.

Not very scientific at all and if I had applied the same level of analytics as I did in the homebrew realm, and much to your ultimate point here, I may have saved myself time and effort.

What I did find is that settling on the Diotec units I’m using now was like a “BAM!” moment for the particular circuit I settled on.

And those measure as 565 mV +/- 15 mV vF on the TC-1. In the end that “tasted” right to me.

When you measure HFE in a BC549B, do you have a pass-fail criteria? IOW, how close to 335 does a transistor need to be to pass muster? This is akin to a question I used to ask a lot at work: "once you have collected this data, what are you going to do with it?"

The Diotec BC54x stuff is super consistent for the most part.

I originally went with Diotec because they were the only modern production offering all three gain buckets.

The higher hFE units tend to have tighter tolerances, to the point where I’m not really bothering with the BC549A.

The C gain bucket tends toward 450 hFE +/- 15-20 hFE. The B gain bucket is in the 335 +/- 20-25 hFE. All measurements are on the TC-1.

I’m good with those tolerances as I don’t really start hearing a difference in response in the Fuzz Faces I’ve used them in until about 50 hFE off a specific target value.

The A gain bucket exhibited an almost 50/50 split between 215 +/- 25-30 hFE and 285 +/- 20-25 hFE.

I eventually abandoned buying those because the 50/50 split was not consistent enough for building a ton of the same pedal with.

They were slated for the same Big Muff circuit and I ended up using (don’t quote me on this offhand) the KSYC1008 from Fairchild (?) that was an insanely consistent 215 +/- 10 hFE.

The three transistors I’ve settled on buying in bulk, Diotec BC549B/C and the Fairchild units, are consistent enough so far that I have not gone outside the +/- 50 hFE failure criteria, although I openly admit that I have not as yet bought enough to test the extremes.

I have something like 500 of each in stock and random sampling has show the tolerances and hFE values to be bang on as tested on the TC-1.