I was saddened to hear of the passing of Don Lancaster. His innumerable contributions to the field of electronics prompted many people all over the world to take up the hobby or even to start their careers.

@Enigma758

Жыл бұрын

Sadly, we lost another legend :(.

@dalenassar9152

Жыл бұрын

Oh NO!! I hadn't heard. In the 80's, I spoke with him on the phone several times. We became friends, exchanging information many times...

@eevibessite

7 ай бұрын

kzread.info/dash/bejne/fqaVxsSBhK-8gLw.html

@TomLeg

11 күн бұрын

I learned a ton from The TTL & CMOS Cookbooks ... and his descriptions of how digital circuits might be implemented with transistors taught me some about transistors.!

4:43 A resistor is required parallel to the cap in the feedback path. The reason is the circuit is acting as an "ideal" integrator, but a practical integrator requires the resistor to provide DC stabilization.

@gblargg

Жыл бұрын

A "leaky" integrator.

@therealjammit

Жыл бұрын

I missed that. The op-amp bias current will eventually put a voltage across the capacitor and latch it up.

@russellhltn1396

Жыл бұрын

I think the real issue is what is in the input? I think it will work ok with a +/- input, but not a 0/+ input. The input resistor should deal with the offset currents.

@BowlingSuperior

Жыл бұрын

There is a variable rate input control to the signal generator - the fixed integrator time constant around the opamp may not work well for all of the possible oscillator frequency rates.

@russellhltn1396

Жыл бұрын

@@BowlingSuperior It would certainly change the output amplitude.

4:23 input bias current needs to flow in the 741, which otherwise will slam into one of the rails otherwise. With some, like LM380, it will work, because the designers put internal biasing, so all inputs can be directly AC coupled, and in fact you need to AC couple, as the input bias is set by the output devices to put them at mid rail, even when used in a single ended supply. 8:00 your output will have a big DC offset, and it will vary with temperature, as your amplifier will be amplifying any drift in the transistor Vbe, plus it will clip to negative rail very easily, as it has too high a collector resistor compared to the emitter resistor, so will basically ignore the input biasing, just using the transistor as if it was biased hard into class A, with the opamp input probably sitting at around 2V off the ground. 10:48 the only opamp you can connect directly to 110VAC is one of the INA series of differential amplifiers, which have input resistors that are happy with up to 200V on the inputs. 13:00 using depletion mode Jfets, with positive gate bias, so they actually will always be on, as you need negative gate bias to turn the device off. I have abused them that way, using a dual Jfet to act as an AND gate (negative logic, to use the pulses on a VFD display to allow me to tell if a segment on a multiplexed display was on) to operate a small low current relay, to allow selection of an input based on the display showing a particular segment. Display ran on -30V, so a pair of 560k 1/16W resistors in series with the gate, and another to ground, made it work, keeping the gate biased correctly negative, and not causing much ghosting of the rest of the display. Then grabbed a convenient 12V power supply for the sensitive relay, and used it. 14:00 The bad part is the resistor values for TTL, not low enough to have correct input current, and for the wired OR no pull down resistor to handle the input bias current that flows out of the TTL input. LED drive direct off TTL was fine enough, though your logic level was not going to be able to drive another gate input, unless you used a resistor to limit current, and also ran the LED from Vcc, as the gate sink current capability is much higher than source, typically 24mA versus 1.6mA when sinking versus sourcing, because TTL is ground current driven. Also bad one seen but not talked about was driving input with negative voltage, or voltage much higher than supply, without using both a series resistor and a set of clamp diodes to the supply rails. for TTL give enough current into the input, around 50mA, and there is a good chance of turning the parasitic SCR in the substrate on, and if your processing at the fab was having an iffy day the overall gain of each transistor in the parasitic structure was over 1, making it a pretty good crowbar across the supply. If the power supply was current limited, it would just clamp the 5V rail to the current limit, and if it was over 1A current, the IC would get hot. If you had a 5A or better supply, then the crowbar IC would get nice and warm, till it either went short circuit from the die cooking, or it blew up the power pass transistors, if it did not have foldback current limiting. If it was capable of over 20A the ceramic packages would pop the lid, and the plastic DIP would blow a hole in the top, letting the TI/NS/Motorola/RCA special smoke out. I framed the one board that did that....

@user-yl3nu3zq9p

Жыл бұрын

13:00 Not only the depletion mode JFET needs a negative voltage across the Gate-Source junction for turning off, but also a positive voltage on this jonction would destroy the device if the current is not limited, as this jonction actually acts as a diode, with a voltage limited to around 0.65V 🙂

@eevibessite

7 ай бұрын

kzread.info/dash/bejne/fqaVxsSBhK-8gLw.html

I love this book and have used it a lot but it has the same problem as all academic textbooks in not providing answers or explanations. It even calls these "self evident" which is hilarious because I've never figured out most of them 😅

@arnotek

Жыл бұрын

Not having the answers to the exercises in the book is a HUGE short coming. It does no good to work an exercise and not know if your answer is correct. If you get feedback that your answer is correct, you build confidence in that your knowledge is increasing. If you get the wrong answer, then you can rework the exercise to figure out where you went wrong. This is especially true if you are self-teaching without the benefit of an instructor. If they published answer guides for the three books, I would gladly purchase them.

@p_mouse8676

Жыл бұрын

This is an issue with most theoretical books or other books they use at school and university. You'll get a bunch of examples and problems that only exists in theory. Or worse won't ever work in practice or are even dangerous.

@arvetemecha

Жыл бұрын

@@p_mouse8676 In that case AOE is full of very practical circuits as well (and specs, and some data sheets...). Some exercises are similar to detailed analysis of some "must be known" circuits. But some of them are also quite tricky and require further study. There is in fact another book "Learning the Art of Electronics" which is of great help when one wants to study the "AOE" reference book.

@fixfaxerify

Жыл бұрын

Yes that is pretty annoying. If they really were "self evident" why put them in the book in the first place!

@andrewallen9993

Жыл бұрын

That makes two of us :)

Gx100 gain audio amp. The problem is that the opamp input is not DC decoupled so it amplifies the DC x 100 going into saturation.

@davidkclayton

Жыл бұрын

You beat me to it 😊

@p_mouse8676

Жыл бұрын

Also non inverting stage should have 99k feedback resistor, since the gain is (R2/R1) +1. In general it's just an odd circuit that doesn't make any sense, so I get that he was confused about it.

@mihalyponyiczki1855

Жыл бұрын

ahhh thanks it's so obvious but I never would have noticed it lol

8:00 (G)- the op-amp is DC coupled to the collector. It's quiescent voltage is well above ground. The op-amp output will be railed.

The integrator circuit to create a triangle wave will end up getting stuck at one or the other of the rails due to nonidealities of the op amp. You need to have something in integrator circuits to control the constant of integration. The Schmitt trigger needs to use positive feedback not negative.

@VintageTechFan

Жыл бұрын

It happens with a perfect integrator as soon you have a duty-cycle or symmetrie slightly off 1:1, or more exactly, slightly unequal areas below and above 0V. The closer they get, it takes just longer to drift to one of the rails, but after infinite time, it always will. The offset of the opamp and so on just makes it worse. You can either put a resistor parallel to the capacitor to limit the DC gain, but this does hurt the linearity of the triangle .. or, as done in triangle generators, use the triangle to produce the square wave.

@argcargv

Жыл бұрын

@@VintageTechFan Any periodic signal that passes through a capacitor will technically integrate to zero. The real problem is that the op-amp has a small offset voltage that will also be integrated. Putting a resistor in parallel with the capacitor is adding a high pass filter to the integrator which effectively removes the "DC" error. This can be ok solution for some applications, but deciding the roll off for the filter will be application dependent

@VintageTechFan

Жыл бұрын

@@argcargv Not really. If the input signal has a DC component, it will integrate that one, too. Periodic doesn't mean DC-free.

@argcargv

Жыл бұрын

@@VintageTechFan but all you have to do to remove the dc component is capacitive coupling. My point is that even if you do that, any input offset error of the op amp will still cause problems

@VintageTechFan

Жыл бұрын

@@argcargv Putting a capacitor in front of it will also distort the triangle shape. But yes, you are correct. It all condenses down to the fundamental problem of all integrators. ANY DC component, wherever it may come from, will have them drift to one of the limits sooner or later. An infinitely small DC component will just take infinite time to do it, but it will happen. You need to implement measures to counter that drift. Making it a low pass instead of an integrator (which is what adding a resistor parallel to the capacitor does) is the easy way and often sufficient, since a low pass behaves as integrator if you a far above its edge frequency. If you have to deal with very low frequency signals, it gets more complicated. Also, with this solution, you still set a DC gain by the two resistors and will get the input offset voltage amplified by that amount at the output .. if you set it low, the capacitor will get large. You could for example track the DC offset developing on the output over a long time (many, many periods) and apply it with an appropriate polarity to the + input. After you chose a low offset opamp (maybe even one that you can trim to near zero).

Excellent video! Circuit H + and - reversed. When input goes positive it will make the threshold higher, not lower.

This is incredibly interesting listening to you examining the schematics and talking about how they will not function properly. I have to do this myself when I design my own electronics as a hobbyist and I will literally sit there and talk to myself pointing these things out on paper and scratching my head but you managed to do it for an audience which is valuable in that showing how the process works in our minds as we analyze these things. I've done a few things like this with my own videos describing how my circuits work and how I modify them over time but I don't think people really appreciate it until they are faced with doing it themselves. It's like getting inside the teachers head and seeing how they think.

The circuit with the transistor upside down works in reverse beta mode. That’s handy sometimes - usually for hard switching though. They saturate at a much lower voltage than in the forward beta mode. The reverse beta mode also gives you a beautiful knee so the transistor can be used as a pretty good diode just by going between forward and reverse beta. In reverse beta there’s so little collector current compared to same base current in forward beta that a sharp knee around 0V results - a much better switching diode than a diode-connected transistor. The latter has a very good exponential response - better than many discrete diodes, especially when the base is wide (high voltage types).

G is actually fairly obvious. The positive input of the op-amp is likely not around ground so it has a positive offset, yet is DC-coupled to the first stage. This DC-offset sill be amplified by a factor of 100. At a supply voltage of +-15V this means that unless the offset is far below 0,15V the OP-Amp will simply clip. One might think that if you adjust the resistors correctly, and connect them to the negative rail instead of ground, it might work, but keep in mind that temperature drift will easily change the offset voltages enough to clip again.

13:25 (F) it's a J-FET. Recall that they are depletion mode only so gate between 0 and +5V isn't useful. This is a hint for (E0, as well. 4:30: offset errors will lead to the output running to one rail or the other. 6:35 (F-- the Diefenderfer circuit) the current through the 45 ohm resistor will pass through tthe load (neg side of the load is at the op-amp output). Standard transconductance configuration. BUT, the current for the 45ohm resistor is sourced via the 240ohm. The voltage divider action will set the cathode of the zener at about 2.4V, giving an ACTUAL current of about 50mA.

I've got it.... I think H&H got this from a compendium someone made over the years of circuits that kind of worked - but needed improvement. Nowadays practice of course the product is sold with errors and then things fixed in "Software Updates". I've always considered Horowitz&Hall more really as an all-in-one source of inspiration rather than for instructions to follow

The trick for the 'bad circuits' in H&H is that they have good circuits for the same purpose right next to them - compare them and figure out what's wrong! (I think they mention that in the introduction, but who reads that?)

@Umski

Жыл бұрын

Now if someone had told me that 20 odd years ago 😆

Just watched this one and noted your mention of an answer guide. I’m not aware of one but Thomas C. Hayes did author a Student Manual for the Art of Electronics along with Paul Horowitz. I have a plastic ringed binder version of that first published by Cambridge University. Press in 1989, with updates at least until 2002. Full of lab exercises (23 in fact), loads of examples, worked problems, circuit diagrams and references to related readings in the Art of E. Highly recommended. Thanks for your content. I find your videos often touch on current projects. Keep it up!

@IMSAIGuy

7 ай бұрын

milesdai.github.io/TAoE3Solutions/data/taoe3-solutions.pdf

12:24 The analog switches: A will burn out the gate by applying +5V without current limiting. When 0V is on the gate, it will clamp negative voltages when it should be passing ±10V. B won't work because the MOSFET's parasitic diode will conduct when the input goes negative. Two MOSFETs in a row with the diodes in opposite directions would fix it. In C, when the JFET gate goes to 15V, the gate diode will be forward-biased and it will inject ~14.4V into the input and output. Depending on the surrounding circuit, it might also exceed the gate current rating and destroy the JFET.

Hmm, for the triangle wave generator it has two fixed value components for R and C, but a variable rate square wave at the input. So the shape at the output will be heavily dependent on the frequency of the input - could be anything from a square wave with a slewed edge at low frequency (with reference to the RC-determined slew rate), through a reasonable triangle wave, to a triangle with diminishing amplitude as the frequency rises. I think!

Nice to see someone going through some of these. I still scratch my head at many of them!

Actually, while long and varied I enjoyed watching this video over several days. Thank you, it was actually great for me. :)

MY FAVORITE BOOK!!! I literally wore-out the book!! I spoke to Paul several times (he warned me about fake 2nd editions selling cheap! We had some very interesting conversations over time. I remember some very good discussions concerning the AD549. He sent me a very interesting book: "STUDENT MANUAL FOR THE ART OF ELECTRONICS" written by Paul and Thomas C. Hayes...lot of other information there. I spoke to Thomas, who sent me the "answer key" to the manual (this wasn't published), I seemed to have misplaced it, but he was kind enough to redo the special problem I was interested in concerning a comparater circuit....this manual contained the best information on comparators I have ever seen. ...really nice fellows... THANKS FOR THE VIDEO!!!

The wired-or circuit needs something to pull the input to ground. The diodes only pull high. Left on it's own, most TTL gate inputs float high.

Triangle-wave generator: When the rate is turned down low (or at DC) the gain on the op-amp is huge since there is no feedback resistor.

@altebander2767

Жыл бұрын

Yes, or to put it in another way: The op-amp is an integrator. It will integrate any DC-voltage on the input... as well as it's own offset voltage. So in any case this will most likely integrate towards one of the voltage rails.

The problem with the integrator (4'38") is that the opamp has an offset voltage. This creates a current through the R that is integrated by the cap. The voltage at the opamp's output goes to VCC or VEE, depending on the offset voltage sign.

@spcct

Жыл бұрын

Another issue could be with the variable rate on the square wave generator - the output amplitude will change with the input frequency instead of staying constantish, which may not have been intended? And it would end up clipping if the duty cycle isn't exactly 50%, probably.

I know I am late to this party. But, this one made me smile. 8:15 - DC bias on opamp input makes output nearly 900 volts. Or saturates at +15V. Whichever comes first. :)

Non inverted gain 10 circuit (4') the wrong thing is that the non inverted pin of the opamp is not biased. There's always a bias current, even in the FET opamps. The capacitor integrates the current and the output follows this voltage (with the gain).

The zero crossing one is fine. When the line is at zero volts the indicator light built into the opamp will extinguish. When it is not at zero volts it will be illuminated (briefly)

Schmidt trigger op amp inputs need to be swapped. In order to operate as a Schmidt trigger positive feedback is needed to create "snap-on, snap-off" characteristic.

The reason 741 was bad idea to build a Schmitt trigger was that 741 has a slow slew rate which makes it sub mega hertz devices. Besides the feedback loop suggested was a linear feedback not an exponential one which could make it to swing faster.

so good! I have learned a lot from your videos！

The 200mA current source - the problem is the 200mA has to flow to the +15V supply and the 240 ohm resistor is way too large to support that.

i learn pretty well from mysteries, plus the way these are formulated, it's like an endless list of things that are actually wrong with them, not just one. everything is so squishy, things can work but not for the reasons you think and whatnot, things can end up vestigial or in a product just to detect clones and stuff. the opamp ones are the most interesting, one of those will oscillate because the inputs aren't tied down, and they show 2 unconnected opamps, if the part or routing is changed when it coincidentally didn't oscillate before, it can start oscillating and stuff

It was fun for me. I like 'always look at good circuits' as a precept, though

The issue with the 555 @11:34 seems to be that the output pulse is dependent on both the RC time as well as the duration of the input. As soon as pin 2 goes low the output goes high. The output will remain high as long as pin 2 is held low and then start the timer when released. If the goal is a multiples of seconds or more output pulse, it is probably fine. If the input is a push button, the output duration will be imprecise especially if it is set for fractions of a second.

For the integrator, when the surface area of the square wave signal above and below ground isn't the same (due to a mismatch in the magnitude of the positive and negative peak value AND/OR duty cycle) the output will pull to either rail. Adding a resistor in parallel with the capacitor fixes this issue :-). Sometimes the leakage of the capacitor is big enough to compensate for this already, but I guess it's good practice to add a high value resistor anyways?

7:14 There is the error that should be obvious, but was also overlooked by me, namely the 240R resistor being a factor 10 too large for the operating current. If scaled to 24R (or maybe 27R) the numbers add up. But the other thing is, max. zener currents are rather large at over 200 mA. It is good practice for the zener to be able to handle the case of no load / load not drawing current, but for standard 0.5W or 1W zeners this won't be possible. For instance absolute max. for a 1W zener at 9 volts is 1W / 9V = 111 mA, around half of what is needed. I would say at least the diagram should specify a power zener able to handle the max. current.

The authors state in the intro that aside from the Bad ones, they are not showing circuits intended for practical use, but to illustrate or isolate a concept. Or some such thing IIRC. The book is littered with errors, much like the Numerical Recipes books - which have code that works and derivations that are wrong. If you want practical analog stuff "How to Build and Use Electronic Devices Without Frustration, Panic, Mountains of Money, or an Engineer Degree". The triangle wave generator will drift due to the input offset voltage until it sits at the + or - rail or be chopped off.

I thought that you would build the bad circuits and then show why they are bad. Now that would be a educating video.

I feel for you. I could never understand these newfangled solid-state thingies. Give me a vacuum!

@GodmanchesterGoblin

Жыл бұрын

Not to mention gas filled voltage regulators, thyratrons, decatrons and the like... 😉

@gdj6298

Жыл бұрын

​@@GodmanchesterGoblinNot forgetting the spectacular mercury arc rectifier.....

@GodmanchesterGoblin

Жыл бұрын

@@gdj6298 Indeed! I thought of that just after I commented.

For an LM358, x100 gain is more like telephone audio bandwidth so that’s at the edge of what the part can do. Even if the biasing was right. LM358 should be biased to (VCC-1.5V)/2 for low voltage applications where output is expected to swing to ground. Then there needs to be a 1mA current source load on the output so it can work in class A for light loads, cutting down on distortion. For anything high-bandwidth, 358 really needs a BJT follower on the output, in the loop of course. Its output stages are wimpy and kill the part’s gain with any sort of load.

I bought this book because you recommended it and it's been great help for my work, I'm a physicist but my degree didn't teach me nearly enough about electronics! This book is great though

I think that at 9:04 the plus and minus are swapped again. Since the transistor acts as a (current) buffer, so its output is non-inverted.

7:18 The idea here is that the current through the load equals the current through the 45 ohm resistor, which is set by the negative input being fixed at 0V and the 9V zener. So the 0 current into the op-amp input is actually part of the design. The mistake they made in this circuit was assuming the voltage at the zener will always be 9V, but the circuit as designed will pull that voltage below 9V, thus it will not actually drive 200mA current through the load. If you put an ideal 9V voltage source there, it will work.

@Noxxet

10 ай бұрын

Yup! They should have added a 30 Ohm resistor in series with the 15 V supply, not 240 Ohm.. The circuit is now limited to 25 mA

By Circuit G: there is a coupling Cap missing in the Input.

Zero crossing: (as there's no information of the supply) there are a lot of opamps that can't apply voltage greater than ~0.6 V or so directly to the inputs. You should use a comparator or add a limiting resistor.

Some more observations :D 4:27 - the triangle wave will vary significantly in amplitude, because the RC time constant is fixed and the incoming square wave frequency is variable. If the square wave isn't exactly 50% duty cycle, the output will drift to a rail too.12:30 and 13:20 both have the same problem, they forward bias the JFET's junction to ground which is basically guaranteed to let out smoke.

7:18 (F). Assuming that the diode is reverse biased (only then there will be 9V at that node) then 240Ohm has a current of 25mA. And we want 200mA through the 45Ohm. Thus, the Zener must Source a current of 200mA-25mA=175mA. But for reverse bias the Zener must "Sink" current from hat node, instead of sourcing. Thus Zener will not turn ON (neither in Forward nor Backward direction). Thus, it can be removed from the circuit.

I believe these came from actual real errors, even the experienced make them

Of the first three circuits the biasing should be a potential divider as per circuit 2 (notwithstanding the transistor emitter should be to ground) with an emitter and collector resistor. The 4 resistor network provides stable biasing (depending on how stable you want it) and protect against temperature effects. One of the first books I bought for my electronics degree in the early 1980's. From what I can remember the book never had answers/solutions which was a real shame.

Circuit F 7:10 - Current will flow from left to right, only problem is the drop across bias Res 240 Ohms is 48V for 200mA (assuming 0A bias current for Zener). Hence, supply should be atleast (9+48)V or Bias Res can be at max 30 Ohms

From the title, I assumed that you were going to build some of them, then watch them fail in real time. I guess I've seen too much "Electroboom". 🤣 I'm glad I'm not the only one who likes to have the "bad examples" explained in a textbook! I also tend to agree that, while explaining why doing something a certain way is bad, it is probably more useful to only show the student the right way to do things, so that's what their brain memorizes. But, then, I'm no education expert.

2:08 the right circuit doesn't need an extra resistor since the -15V is lower than 0V GND. But the transistor should be NPN and not PNP.

Horowitz and Hill (The Art of Electronics) was the bible in our electronics design lab (in the UK) to such an extent copies disappeared regularly (under grad trainees?). I actually bought myself a copy when I retired over 20 years ago and it's on my shelf as I type.

In the third example the biasing is just fine -- but the transistor is of the wrong type. With the correct transistor (NPN) the emitter would sit at -0.7V and be able to follow the input signal on the base to around 27V peak-to-peak. The transistor would definitely be in class-A (linear) mode of operation. Of course this would only be an emitter-follower amp so the amplitude of the output signal would be the same as that of the input signal so (if the value of the emitter resistor was suitably chosen) it would be just a current amplifier suitable for altering impedance betwee input and output.

14:12 Even nowadays that's still a bad circuit. You need a resistor or something on that LED to limit the current. Even if the output of that gate can drive the LED, and even if it somehow has some built-in current limiting so it won't fry the LED or the gate, it'll still end up pulling the output level down to the LED's forward voltage, which (depending on the LED) probably means it won't be high enough to drive any of the "other gates" down the line properly (assuming they're also TTL)...

I think the first AC coupled follower with a resistor to ground might work if the emitter resistor would be connected to a negative supply (much lower -0.6 volts, ofc)

Unlike what you said which it wouldn't bean interesting clip, a lot of people watching your will continue from what you left off and will complete your answers. I think it was a very good start. I highly suggest you study the circuits before hand and think about them before making your clips. No matter how experience you think you are you cannot come up with a pefect answer immediately

@IMSAIGuy

Жыл бұрын

I hate polished videos. I think it demotivates people when the KZread guys know everything and they try and fail.

@bayareapianist

Жыл бұрын

@@IMSAIGuy that unpredictablity makes your show like Kardashians 😉

I don't want to go through all that unexplained oddities in all that bad circuits, because it's much better if you find that yourself, but there is one thing catching my attention. Swapping E and C of BJT is not necessarily bad thing. The only what swapping E with C does is that this changes the parameters of that transistor to something, that is typically ugly, but despite the fact it is ugly, it might be useful in some particular cases.

L will actually fail spectacularly on many op-amps with protection diodes on the inputs. It'll short those 110V AC directly to ground. Depending on the impedance of the voltage source this can make your op-amp catch fire or explode. Also those 60Hz depend on your input frequency. That circuit does make a bit of sense for some applications, for example when using a motor as an input device. Depending on the motor you might get high voltages and high frequencies. However in that case you should at least AC-couple it and add external protection diodes.

I'm curious... in your previous video you mentioned you liked Horowitz&Hall first edition better than second. I have their second and wondering if it is beneficial to purchase a first edition. I really enjoy and learn, even at my age, from your videos and look forward to each new posting. Thank you!

@IMSAIGuy

Жыл бұрын

My love of the first edition is probably nostalgia. The cover and binding is falling apart from so much use. The first edition has more digital circuits which no one does anymore. If you feel a need to buy another, get the third edition. That is actually the one I pull off the shelf if I want some info.

None of those bad emitter followers will work - but the middle one is close. Put the resistor on what is nominally the collector lead, and as long as the Vcc is less than the base emitter breakdown voltage (typically about 7 volts) the emitter junction will be reverse biased and function as a collector, and the collector junction will function as an emitter. Try it with a 5 volt supply. It will work. Current gain of a transistor put in backward like this is about 1/10th of what you get when you use it as intended, but as an emitter follower that doesn't matter very much. Even if the gain is only 10, that's enough to be useful. Using a transistor this way is not recommended - except for one specific purpose. Running a transistor backwards like this gives you a very low saturation voltage. The Vce sat is far lower with the transistor inverted than when used the normal way. If you want make a chopper, this inverted connection is a trick that lets you make a superior switch that gets close to ground. An obsolete bit of trivia, as FETs work even better. But once upon a time, this was a clever trick.

...cool vid...you look at them the way I would...;-)

1:58 - The bias is actually ok here, since the base resistor goes to GND, while you have a +-15 V supply. The transistor is still placed upside-down though :) 8:10 - The transistor bias is ok, the base is around 1.5V. The Q point for the collector is at 9 V (also good) and around 1.2 mA ((1.5 V - Vbe) / Re). This first stage has a gain of 10. However, we now amplify the total signal (AC + DC) by a 100, which will obviously saturate the Opamp. We need a blocking cap between the output of the transistor stage and the non-inverting input. We also then need a pull-down resistor from the non-inverting input to not have a floating input with the cap in place. Then the gain of the opamp should be lowered to 10x, for the total gain to be 100

G: The only thing I can think of is the transistor output is connected directly to the op-amp and might DC bias the op-amp and send it into saturation. H: The op-amp is biased to half of 15v, or 7.5 volts. If the op-amp is an open collector output the output will swing from 0v to 5v. The 5v swing won't "flop" the schmitt trigger at 7.5v

C. the integrator will not produce triangle wave out for the variable square wave in for all frequencies. R would need to be varied as the rate varies. G. A coupling capacitor is needed between the transistor collector and non inverting op amp input. Also the op amp gain is probably too high. H. The Schmitt trigger needs a positive feedback not negative. L. The 110 Vac needs to go into a coupling capacitor then 10 Meg resistor into inverting in with a 1 M to ground.

I have a first edition and a third edition of AoE. I really miss those bad circuits in the third edition ! Aside from being educational, they are often very hilarious as well.

Bad circuit #2 is exactly what made it to ZX Spectrum +2 issue 3. It has two 2N3904s marked and soldered incorrectly; the error comes from switching over from a different type transistor which had reversed pinout.

Love it!

In G. There will be a tremendous DC offset on output. Assuming transistors are biased to 7.5 volts of output, then ther will be 750 volts on opamp outptut. Looks like there is a capacitor missing in fron of opamp.

4:30 that triangle generator output will float randomly anywhere between GND and the suppy, tried that and it didn't work. a system feedback loop is needed to detect the desired high point and reset to GND. 8:53 only comparator like lm339/lm393 need a pullup at output, the 10K feedback is really small and since all are 10K it would offset the voltage divider extremely, the voltage divider is producing 5V and via the feedback (5V - 10K + 4K) is offsetting the voltage divider and probably also cause behavior like a bad configured non inverting opamp acting more like a "voltage follower" rather than a binary output, they use 10M in feedback on lm339/lm393 to slightly offset the voltage divider once triggered and prevent getting erratic behavior around the tripping point. I got the 3rd edition, pretty heavy book with 1200 pages, you probably want someone to hold it for you. : )

At 1:45 "the emitter needs to be down here at the bottom". No, because that would no longer be an AC coupled follower, per the caption. You must use a pnp, and it will not work "either way". At 1:56 "we just aren't biased correctly; we need to have an extra resistor in". No, you have split power supplies, so you can bias the base from ground without an extra resistor. The only problem is that it needs an npn. At 3:59 "what's wrong with this one?" There's no path for the non-inverting bias current. At 4:47 "you tell me". The opamp is operating open-loop at dc, so the smallest of offsets will eventually build up and the output will clip to one rail or the other. You need some dc feedback (a resistor) to stabilise the operating point. At 7:06 "that's not going to work because, errr ... [blather] ... it might be hard to drive this thing correctly". No, the problem is purely that the 200mA which is intended to be flowing through the 45R resistor then through the load has to be sourced via the 240R resistor, which would then have just 6V across it - impossible. Either replace the 15V supply with one greater than 57V (=240R x 200mA + 9V) or replace the 240R resistor with one less than 30R. At 7:56 "I don't know". Check the gain: the npn discrete stage will have an approximate gain of 7.5K/75R = 10, while the opamp has a gain of 101. That's a combined gain of around 1,000, not the x100 in the caption. At 8:35 "plus is right ...". No it isn't. The positive feedback on a Schmitt trigger goes to the non-inverting input. At 10:05 "bunch of stuff wrong with that one". Not really, it just needs the opamp supply to be taken from Vin, assuming that the 741's drive of 25mA is enough to drive the base of the pass transistor at the maximum desired output current. At 10:22 "this can't be referenced to ground". Actually, it can. The LM358 includes its negative supply in its common-mode range, but you're right that the non-inverting input sets the dc output in this configuration, so it ought to be biased to around half the supply voltage, especially with an intended gain of 100. Incidentally, the 358 can't manage 20KHz at a gain of 100, so would do a terrible job as an audio amp with that sort of planned gain. At 12:40 the jfet in the first circuit must not have its gate driven positive wrt its source (it's a depletion-mode device). The mosfet in the second circuit will have a body diode that will conduct in one direction even when the mosfet is switched off, so it's not suitable for switching ±10V. You can't drive the gate of the jfet in the third circuit positive wrt its source. At 13:08 "ten ohms is a little low" it certainly is a lot low. The circuit is probably expected to run at around Id = 1mA judging by the 10K drain resistor and 20V supply, so it will have a gm of about 10mS. The gain is given by Rd / (Rs + 1/gm) and 1/gm is going to be at least 100R, meaning that it swamps the source resistor used and making the gain a rather variable x100, not the x1000 hoped for. At 13:23 you can't drive a jfet's gate 5 volts above its source. At 14:30 the problem, even "nowadays" is that the LED might be red, so it could clip the ouptut to the other gates below a TTL logic 1. It was great fun. Thank you.

Hmm, it looks like neither the "student manual" nor "learning the art of electronics" cover these directly, so this is a novel exercise, thanks!

I think the square to triangle via integrator would produce a variable amplitude depending on frequency. The DC bias might also drift around with variations in frequency, and that would probably result in clipping. I am guessing that the rate control is what makes circuit "bad" since this would be ok with a constant frequency and suitable R and C values.

Schmitt trigger: the 1 k R does nothing. It is connected to +5 V and the opamp's output.

Increase voltage and "bad circuits" starts run in avalanche operation mode :D

First one is common in a class C amplifier, with brief current pulses meaning low operating current, when driven hard enough. Second one is ofr course an error, you need to have a negative supply, which was common in the days of germanium transistors, all PNP normally, though you did get expensive NPN ones, with typically a negative supply and positive ground. Third might actually work with some transistors, which would operate symmetrically, as the collector and emitter were identically doped, which was often the case with early diffused Ge transistors, which actually had in some cases equal gain both forward and backward, though of course they did not operate at 30V, only around 3V total.

@andygozzo72

Жыл бұрын

i've commented somewhere else here about a transistor used 'in reverse' like in the second one..😉

The Schmitt Trigger needs positive feedback.

Triangular wave generator is missing a high value resistor in parallel with the integrating capacitor, otherwise it will saturate by integrating even the smallest offset; Schmitt trigger should have positive feed-back. i.e. inputs are swapped;

Circuit G has no offset compensation. The output will be stuck high leading to highly clipped output.

Its like having a carbook and illustrate/explain not to put the wheels on the wrong way, or swap gas and brake position 😅

The 100x audio output amplifier, besides there being more than 100x gain, is DC coupled into the op amp-it’s going to amplify that DC and saturate

7:26 Should it have a decoupling capacitor before the op-amp?

The integrator on C. will integrate constant DC offset to infinity. Some sort of 1Mohm (depending on Opamp input offset) must be added in parallel to C.

I think I have the newest version of the Art of Electronics on my Kindle. Packed with good info. They often leave the schematics vague, and you need to read the text carefully.

7:57 Looks like a transcription of an empirically developed circuit attempt. Maybe they added transistor because the opamp was a 741 wasn't fast enough gain*frequency. If the opamp was intended to directly drive a low resistance earphone the superimposed DC voltage on the speaker would make the opamp run warm. That does look like it was someone's actual attempt at one time.

Think for E the x1000 amplifier the circuit is designed to a current-driven transistor, but FETs are voltage driven. So both resistors on the output should be around 10k ohms. I think some kind of feedback from the output may be needed to set up a voltage divider for the FET. So maybe replace the 1M resistor with 1k and place the 1M resistor bridging the output and input. F: logic switch: again FETs are voltage driven and have very high impedance on the input. You need to tie the input to ground (or supply but that would fail on) through a resistor to avoid a floating input. G: is definitely missing a feedback resistor somewhere: can't have a voltage divider with only one resistor.

This brings back some (unfond) memories 😆 Not knowing why they were bad always bothered me 😬

My guess for G would be to put the op amp stage before the BJT stage, ditch Rc, then take the output off of the emitter instead. You would need a coupling cap on the output though. Also, those gain setting resistors aren't the most intelligent choices in terms of noise. Maybe cut those values in half.

For the G, maybe a capacitor is missing between the transistor and the AOP to cut off the DC signal?

Not an expert here but I think some of the FET circuits might be wrong because JFETs are triggered by a negative voltage, so the gate would have to be less than both the source and drain voltages which cannot be always guaranteed

@IMSAIGuy

Жыл бұрын

kzread.info/dash/bejne/pH6Wk8ePgpaWlLw.html

4:00 Isn't the problem the high-impedance input? Even DC will effectively couple through a capacitor if there's no current draw (e.g. floating CMOS logic inputs). You're basically using a high-pass RC filter but R is near infinity. As you mentioned, a resistor to ground fixes that.

12:59 I'm guessing that the resulting 20,000V is a little high as the arc distance would be roughly 20mm.

schmitt trigger + - switched around

Regarding that op-amp with 110V (or whatever) directly into one of the inputs. I first ran into something like this when troubleshooting a digital multimeter from the late 1970s. It had the input + terminal connected almost directly into the op-amp, via a series resistor, so except for that one component there was potentially up to whatever input voltage (1000V or more) without the benefit of the 'normal' input attenuator voltage dividers. Yes, there were reverse-parallel resistors from the op-amp input to ground after that one resistor. The thing that made it work was a bank of negative feedback resistors on the op-amp, all connected in parallel, but each with an FET in series with it, and those FETs were turned on from the automatic range selection logic. So for very low input voltages, a feedback resistor would be selected into the circuit which in contracts with that one input resistor would provide gain, and for high inputs the feedback resistor would provide attenuation. While solving the design problem this was worked fine, it still seemed alarming initially. That one input resistor value was negligible compared to the op-amp input impedance, so it was not by itself really providing any protection for the op-amp (although it DID also serve to limit current through the clipping diodes). So, for the 'bad circuits' example, the op-amp is still connected about the same way, relying on the op-amp's very high input impedance to prevent damage from high input voltages. Even with 100+ volts applied, current into the op-amp is negligible. The circuit is configured as a simple comparator, so it is expected to have its output slam against the power supply rail. Seems like this might actually be OK......I wonder what I am missing? Maybe it is just the op-amp's maximum voltage to ground rating is being exceeded, in that some current path through substrate, or whatever, would blow it up, rather than the 'ideal' op-amp input being OK with the situation.

7:06 -- Give me a break guy... the zener will do nothing as the voltage across the zener is a bit over 2 volts

There are circuits here that cannot be simulated with LTSPICE It's surprising at how some error situations cannot be caught by the software. Some are simple logic errors so you should easily catch them. "Not fun" when you find out how weak your basic theory is.

7:35 (G) .. the gain for DC is also 100x, but the collector voltage is around 8.5V (correctly biased, as you said) so the opamp wants to give 850V DC at its output, which does not work. But an electrolytic an series with the 1k to ground and you are fine. 12:27 .. first and last one. It's an N-JFET .. that GS-Diode conducts for voltages >0.65V or so. Last one also plays that D/S are usually interchangeable, so same problem. +15V even puts current back into the signal source. Second one misses a lot, to make it work you need to build a "transmission gate" .. used all the time in ICs, and whats in the (in?)famous CD4066 and various analog multiplexers. 12:59 (E) .. you don't get 1000x Gain out of that single JFET. With a AC grounded source, the gain is gm*RL (exactly gm*(RL||Ri) with Ri being the differential resistance of the drain, but that is very high) so with a typical gm of 5mS or so around 50x with that 10kOhms collector load. The bias of 10 ohms doesn't matter, so input voltage would be limited to 0.5-1Vpp or so (before the G-S diode begins conducting). (F) same as the clamps before, a N-JFET is always depletion mode and takes negative G-S voltages. Would be fine with a (enhancement mode) MOSFET.

Did Don Lancaster ever do any work on audio circuits using germanium transistors? Maybe he wrote articles on it. I have a 1961 Tranistronics stereo amp that uses germanium transistors. It sounded very good until those germanium transistors did a nose dive. I have the schematic of it, but even the most qualified techs say don't even bother trying to fix it.

now that you have all the answers you have in comments now you can you can redo it : )

Good Grief! I still have my copy, does that make me ancient?

Circuit H schmit trigger. Shouldnt the gain be massive so has minimal rise time and no bounce. It looks like it only has a gain of 1. 1+10k/10k. You would think it wont clamp well enough. I never played around with them so this is a guess but I do understand how an opamp schmit works. Maybe Im wrong too 😂