I've got a number of old Haflers on hand that I can scavenge MOSFETs from if I take a notion to do so. At this point I'm unaware of any particular advantage to the older MOSFETs versus newer parts, so I'll probably just buy new laterals if I decide to go that route.
I find myself in a curious gray area regarding the software I use. Back several years ago I was using PC cad or Win cad or whatever they call it. The stuff was a nightmare, totally, completely, and absolutely. There were a number of problems, but the one that finally broke the camel's back was the inability to export files in any reasonable format. Geoff Moss, a member here, undertook the tedious task of evaluating the available (free) software available in an attempt to find something that wasn't crippleware to such an extent that it was unusable. He finally came up with two suggestions, one of them being Circad. I don't remember what the other was. I adopted Circad and began the laborious process of transferring things over to their format. Some things never made it and are lost forever due to a disk crash. Now I'm stuck in Circad, which has more bugs than my beehive and I am as disgusted as ever.
The problems fall into several categories. One is pin limitations. Another is PCB file formats. There are others, but the tradeoff I made by going to Circad was that I got unlimited pins, but lost the ability to export Gerber files. I can export PCB files in the BMP format, then convert them into something else, but I'm not sure whether it's possible to turn a BMP into a Gerber.
If I had enough money to drop on software, I'd probably buy a speaker measuring program first, but even that's unattainable at this time. Schematic/PCB software would be second or third on the list. Translate this as "It just ain't likely to happen."
Bummer.
Grey
I find myself in a curious gray area regarding the software I use. Back several years ago I was using PC cad or Win cad or whatever they call it. The stuff was a nightmare, totally, completely, and absolutely. There were a number of problems, but the one that finally broke the camel's back was the inability to export files in any reasonable format. Geoff Moss, a member here, undertook the tedious task of evaluating the available (free) software available in an attempt to find something that wasn't crippleware to such an extent that it was unusable. He finally came up with two suggestions, one of them being Circad. I don't remember what the other was. I adopted Circad and began the laborious process of transferring things over to their format. Some things never made it and are lost forever due to a disk crash. Now I'm stuck in Circad, which has more bugs than my beehive and I am as disgusted as ever.
The problems fall into several categories. One is pin limitations. Another is PCB file formats. There are others, but the tradeoff I made by going to Circad was that I got unlimited pins, but lost the ability to export Gerber files. I can export PCB files in the BMP format, then convert them into something else, but I'm not sure whether it's possible to turn a BMP into a Gerber.
If I had enough money to drop on software, I'd probably buy a speaker measuring program first, but even that's unattainable at this time. Schematic/PCB software would be second or third on the list. Translate this as "It just ain't likely to happen."
Bummer.
Grey
GRollins said:
Nope, there's no going back from the raster BMP image to the stroked Gerber... and that's leaving aside the drill map and any other odds & ends that I'm forgetting right this moment. I assume this is an intentional limitation of a demo version?
I'm not sure what the current state of the Windows port is, but the GEDA tools are at least usable on Linux. Compared to various free/demo Windows proggies I've met, the GEDA suite is currently lacking a [working] overall manager, and requires some manual running of scripts from time to time. But it is free and under active development, and as I said I've found it usable... after some getting my head around their approach to things.
http://geda.seul.org/
-- Martin
Okay...finally made it back to this thread...
I used to run Linux at home, but I had a disk crash. At the time, I had a story deadline to meet and couldn't afford all the craziness to get Word (you'll not be surprised to hear that the publishing industry is pretty well standardized on Microsoft Word) up and running in a dual-boot system, so I threw a copy of Windows into the fray, dropped in Word, and met the deadline. Never went back, unfortunately.
I have never tried the technical software available for Linux, so I have no opinion. Given that this was all ten years ago or more, any software I used then would be an antique by now anyway. I'm on the verge of setting up an older spare computer down in the Dungeon where I do my fiddlings. It'd be nice to have a spreadsheet handy to speed some calculations along. As it is now, I keep a calculator on the bench, but that gets tedious sometimes. Then I have to trot upstairs and calculate, jot down the numbers, go back downstairs...you get the idea. It's a pain.
I've been threatening to use Linux for that PC, just to exercise my Unix braincells. They've accumulated quite a bit of dust. Aside from schematics and PCB art, what's a good Linux spreadsheet program to look for?
Grey
I used to run Linux at home, but I had a disk crash. At the time, I had a story deadline to meet and couldn't afford all the craziness to get Word (you'll not be surprised to hear that the publishing industry is pretty well standardized on Microsoft Word) up and running in a dual-boot system, so I threw a copy of Windows into the fray, dropped in Word, and met the deadline. Never went back, unfortunately.
I have never tried the technical software available for Linux, so I have no opinion. Given that this was all ten years ago or more, any software I used then would be an antique by now anyway. I'm on the verge of setting up an older spare computer down in the Dungeon where I do my fiddlings. It'd be nice to have a spreadsheet handy to speed some calculations along. As it is now, I keep a calculator on the bench, but that gets tedious sometimes. Then I have to trot upstairs and calculate, jot down the numbers, go back downstairs...you get the idea. It's a pain.
I've been threatening to use Linux for that PC, just to exercise my Unix braincells. They've accumulated quite a bit of dust. Aside from schematics and PCB art, what's a good Linux spreadsheet program to look for?
Grey
Open Office is a pretty complete clone of the Microsoft products,
very functional, and works fine on Windows or Linux.
For the latter, I recommend Ubuntu.
For free schematics and pcb on Linux, check out Eagle.
very functional, and works fine on Windows or Linux.
For the latter, I recommend Ubuntu.
For free schematics and pcb on Linux, check out Eagle.
Does Eagle hit you with pin limits? That seems to be the sticking point. It's either export capabilities or pin limits in most cases.
I may face a hardware limitation once I get down to actually trying this. I can't remember with the processor is. Given the age of the PC, I think it's one of the early Pentium variants. (I still wish Intel had shown the guts to name their next chip the Sexium...) That may or may not prove to be a show-stopper.
My wife just brought home a new Dell of some description, and is thereby releasing the one she's been using. The problem there is that it's dying. It will spontaneously drop and reboot for no apparent reason. If it's a software bug (that machine's running Windows 98, which was never worth a toot), then flushing the hard drive and putting Linux on it will cure at least that one ill. Unfortunately, it has other problems as well, which is why she got the new PC.
I have no love for Circad, but I'm leery of shooting myself in the foot if I jump before looking.
Grey
I may face a hardware limitation once I get down to actually trying this. I can't remember with the processor is. Given the age of the PC, I think it's one of the early Pentium variants. (I still wish Intel had shown the guts to name their next chip the Sexium...) That may or may not prove to be a show-stopper.
My wife just brought home a new Dell of some description, and is thereby releasing the one she's been using. The problem there is that it's dying. It will spontaneously drop and reboot for no apparent reason. If it's a software bug (that machine's running Windows 98, which was never worth a toot), then flushing the hard drive and putting Linux on it will cure at least that one ill. Unfortunately, it has other problems as well, which is why she got the new PC.
I have no love for Circad, but I'm leery of shooting myself in the foot if I jump before looking.
Grey
oops retracted see post above ...
GEDA: GPL'd Electronic Design Applications. Not as slick as Eagle no doubt, but no restrictions either.
http://geda.seul.org/index.html
Ryan
GEDA: GPL'd Electronic Design Applications. Not as slick as Eagle no doubt, but no restrictions either.
http://geda.seul.org/index.html
Ryan
I used to get such a kick out of playing with PCs...nowadays I just wish the blasted things would set themselves up. While I'm arm wrestling with my wife's new box, I may go ahead and try to shoehorn Red Hat into my old box.
(Yes, current computers are far easier to set up than they used to be, but I seem to be the sort of person who triggers a Murphy's Law toggle somewhere down in the code. Something nearly always goes wonky. Curiously, it's only digital electronics that behave this way--my analog stuff is generally pretty well behaved, even when I do things that should cause problems.)
I've heard of Open Office, but have yet to investigate it. My one Absolute Rule is that I've got to be able to squirt stories up the line and have Microsoft Word read them in New York. But that's assuming I ever get another story written...so who am I kidding? Okay, let me phrase it this way--IF I ever get another story written and IF said hypothetical story should be accepted, it would be really groovy, nifty, keeno-cool if it was in a format that Word could deal with. I tried a couple of other programs way back when that claimed compatibility with Word. No dice. Claims were as far as they got.
The electronics stuff can be encoded in any old format as long as I can get it back out into the real world in some reasonable manner. I'd like to be able to post stuff here, print, and possibly do something really odd like have Gerber in case I ever decide to farm out the whole PCB thing instead of doing them myself.
Did I mention that I etched six circuit boards today? Having a commercially made board is nice, but I've gotten fairly good at doing my own. I like being able to conceive of a board at breakfast, have the artwork done by midday, and etch the thing before sitting down for supper. That's pretty neat.
At least there appear to be decent options for Linux. That's encouraging.
Grey
(Yes, current computers are far easier to set up than they used to be, but I seem to be the sort of person who triggers a Murphy's Law toggle somewhere down in the code. Something nearly always goes wonky. Curiously, it's only digital electronics that behave this way--my analog stuff is generally pretty well behaved, even when I do things that should cause problems.)
I've heard of Open Office, but have yet to investigate it. My one Absolute Rule is that I've got to be able to squirt stories up the line and have Microsoft Word read them in New York. But that's assuming I ever get another story written...so who am I kidding? Okay, let me phrase it this way--IF I ever get another story written and IF said hypothetical story should be accepted, it would be really groovy, nifty, keeno-cool if it was in a format that Word could deal with. I tried a couple of other programs way back when that claimed compatibility with Word. No dice. Claims were as far as they got.
The electronics stuff can be encoded in any old format as long as I can get it back out into the real world in some reasonable manner. I'd like to be able to post stuff here, print, and possibly do something really odd like have Gerber in case I ever decide to farm out the whole PCB thing instead of doing them myself.
Did I mention that I etched six circuit boards today? Having a commercially made board is nice, but I've gotten fairly good at doing my own. I like being able to conceive of a board at breakfast, have the artwork done by midday, and etch the thing before sitting down for supper. That's pretty neat.
At least there appear to be decent options for Linux. That's encouraging.
Grey
GRollins said:
The following limitations apply to the EAGLE Light Edition in general:
* The useable board area is limited to 100 x 80 mm (4 x 3.2 inches).
* Only two signal layers can be used (Top and Bottom).
* The schematic editor can only create one sheet.
The Freeware version of EAGLE Light adds these limitations:
* Support is only available via email or through our forum (no fax or phone support).
* Use is limited to non-profit applications or evaluation purposes.
As best I recall, the one sheet limit didn't seem to be too serious, though you might need larger than letter paper to get a legible printout if you push it. The board size limit would have been an issue, and the reasonably priced version shares that with the demoware. I tried using Eagle to capture the schematic and FreePCB (which runs well under Wine), but this was pretty cumbersome because a good deal of mapping was needed to carry the netlist over. So I finally got around to looking at gEDA again. I wasn't expecting too much, as it hadn't impressed me a couple years ago, but it had improved quite a bit, and it was obvious it was at least better than trying to make the Eagle/FreePCB lashup work. Now that I'm more or less accustomed to the gEDA way of doing things I'm reasonably happy.
Oh, and I hear that at least some vintages of Word and the other Office tools run well in the Wine environment these days. Haven't had any reason to try it myself. Wouldn't have been the case ten years ago, probably not even five.
The bad news is that I don't want to think about running *any* modern GUI system on an early Pentium - it would be an interesting question whether the slow CPU or the hardware limit on how much RAM you could stuff into the box would be the more annoying limitation. A PII might be tolerable; a PIII certainly would be.
Oh yeah, Ubuntu is a pretty safe choice for a desktop. They'll even send you install CDs. 🙂
http://www.freepcb.com/
That 4" x 3.2" limit is painful. The front end board alone for this amp is 6" x 6" although I could pull that down a little. Worse yet, I'm working on a 6" x 9" board at the moment.
Grey
Grey
PART I
So…
I got an e-mail saying something to the effect of, “How the &*%$ does the thing work? The @#!^ing thing’s got too many inputs, too many outputs, and more twists than a Moebius strip!” Then I got another, and another, and…and, well, I figured I’d better post something before it got to the point where I was receiving death threats.
I’m going to break this down into pieces, starting at the input. It may take a while to get this done so for those who are lost, don’t despair. I’ll post more as soon as I can.
The place to begin is with a differential. Just one differential, not the complementary one in the GR-25. Most amplifiers have at least one differential in them and you won’t have to search too hard to find examples with two, three, four…let’s just say that differentials are pretty popular. In fact, you can make an entire amplifier out of nothing but differentials if you take a notion to.
A differential is a pair of gain devices, back to back. You can build a differential with tubes, bipolars, JFETs, or MOSFETs. In the GR-25 I used JFETs, so I’ll use JFET nomenclature so as not to have to shift gears later in the game. Suffice it to say that all differentials work the same way, so it doesn’t really matter what kind of device you have on hand, you can pretty much get the same sort of action, regardless.
When you put a signal into the Gate of a JFET and take the output at the Drain, it’s called a Common Source connection. When you put the signal in at the Gate and take the output at the Source, it’s called a Common Drain connection, although it’s generally known by its alias: Source follower. And if you put the signal in at the Source, then take the output at the Drain, that’s called a Common Gate connection. Before all is said and done, we’ll be using all three of those options.
I’m attaching a file showing a series of differentials. I’ll be using that to try to keep things on an even keel.
The first differential on the left (A) is the bare-bones sort. We’ll assume the signal enters the Gate of Q1, the JFET on the left-hand side. That signal then appears, out of phase, at the Drain of Q1, where it meets R1, and is available for use somewhere else in the circuit. But—and this is something that a lot of people don’t realize when they first look at a differential—there is also a signal at Q1’s Source. If you view Q1 as a Common Source amplifier, then your focus is on the signal at Q1’s Drain. If you view Q1 as a Common Drain amplifier, then your focus is on the signal at Q1’s Source. It’s not an either/or proposition; Q1 is quite happy to perform both functions at the same time. (This is intuitively obvious to people who work with tubes, because the split load phase splitter is used in precisely the same way. Those who want an example of the analogous solid state circuit can look at the JLH amplifier.)
The signal coming from Q1’s Source is faced with two choices—it can travel down Rbias, or it can enter Q2’s Source. Which route it chooses will depend on the relative impedances of those two legs. Q2’s Source is a low impedance. Whether that’s good or bad depends on your application. A low impedance means you’re going to have to have enough current to drive it. In this case, everything is going to work out just fine because, from Q2’s point of view, Q1 is acting like a follower, and if there’s one thing followers are good at, it’s delivering current. The other possible route for the signal from Q1 is Rbias. You want the majority of that signal to go to Q2, so to encourage that you make Rbias much larger than the impedance of Q2’s Source. Let’s say that Q2’s Source presents a load of 100Ω. This is higher than the JFETs in the GR-25, but it’s a convenient, round number. If you sit down and tickle a calculator for a minute, running a few iterations of Ohm’s Law, you’ll find that it doesn’t take nearly as much resistance as you might think to convince the signal coming out of Q1 that Q2’s Source is the path of least resistance. (Ahem.) A few thousand Ohms will do quite admirably, although you see a lot of people using current sources instead of resistors in an attempt to provide a reasonable approximation of an infinite impedance. It’s not necessary to go to those lengths, so we’ll just say for the moment that Rbias is something like 10k, which is obviously much, much greater than our hypothetical 100Ω impedance at Q2’s Source. As a result, something like 99% of the signal slides right on over to Q2.
So now we’ve got a signal at Q2’s Source, and in turn that signal becomes available at Q2’s Drain. So that makes Q2 a Common Gate configuration. Ta da! As promised, we’ve used all three possible hookups in one fell swoop.
Now, the thing about a Common Gate connection is that the signal at the Drain is in phase with the input signal. It has to be, because they’re in series. And since the output of a follower is always in phase with the input signal, that means that the output at Q2’s Drain is also in phase with the input. If you’ll recall, the signal at Q1’s Drain was out of phase. So now we have our choice of an in-phase signal or an out-of-phase signal, depending on our mood. As it happens, the GR-25 uses both at the same time, but that’s not required. A lot of amplifiers get by quite nicely with using just one output from the differential.
Circuit (B) is exactly the same as circuit (A) except for the fact that it’s rendered in P-ch JFETs; it’s a mirror image of (A). As a differential, it performs the same way.
Circuit (C) combines (A) and (B). The resulting circuit only needs one Rbias to make things come together. Note that both the N-ch and P-ch JFETs on the left side (Q1 & Q3) share the same signal input. Likewise for Q2 & Q4 on the right. This is called a complimentary differential and it forms the heart of the GR-25 (as it did for the JC-3). When a positive signal is applied to Q1 & Q3, they both shift downwards in response; Q1 conducts more heavily, driving its Drain down, and Q3 conducts less heavily, also causing its Drain to move downwards. The entire circuit acts like a see-saw, which leads the JFETs on the other side, Q2 & Q4, to move upwards.
(This is the kind of circuit that gives tube folks feverish dreams. Much good could be done if only someone would devise a P-tube that didn’t require anti-matter. For obvious reasons, the anti-matter sort isn’t something you’d want in your listening room.)
If you look at the schematic for the front end of the GR-25, you’ll see that the equivalent resistor to Rbias is 47.5Ω. What happened to the 10k we were just talking about? It’s not needed. The cool thing about a complimentary differential is that the two differentials work in tandem, both from side to side and from top to bottom. As one side draws more current, the other draws less and at all times the current draw sums to a constant DC value. One of the consequences of this is that the differentials almost act like current sources for each other.
If you’ve managed to keep up with the changes so far, the next step is a fairly simple one—add cascodes to all four JFETs. A cascode is a gain device that sits on the shoulders of the one below it. It operates in Common Gate mode, with the signal entering through its Source and exiting through its Drain. The Gate is set to an arbitrary voltage using a voltage divider, although there are other ways to accomplish the same ends. Given that there are two cascodes sitting side-by-side, we can save parts and gain some subtle benefits by using the same resistive divider to set both cascode devices’ Gates at the same time.
Cascodes are useful in several ways. One is that they cap the voltage that the gain device (the one underneath) sees at its Drain. In the original iteration of the GR-25 I used 2SK389s and 2SJ109s as the input devices. Neither of those parts can take much voltage, but if you add cascode devices set to take the brunt of the rail voltage, those parts will never know it’s there.
Another benefit is that cascodes reduce the effects of the inherent capacitance at the Gates of the input devices. This improves bandwidth and distortion, both desirable things if you’re trying to pull off an amp that doesn’t use any feedback. And if you do intend to use feedback, there’ll be less for it to do.
As before, output is taken at R1, R2, R3, and R4, with the sole difference being that the Drains driving those resistors are those of the cascode devices (Q5, Q6, Q7, Q8) instead of the gain devices (Q1, Q2, Q3, Q4).
I’m approaching the character limit for a single post, and have covered all of the first stage to some degree. I’ll try to whip up some stuff covering the folded cascodes and their cascodes. Depending on how things go, I may be able to get the bias circuit in, too.
Grey
So…
I got an e-mail saying something to the effect of, “How the &*%$ does the thing work? The @#!^ing thing’s got too many inputs, too many outputs, and more twists than a Moebius strip!” Then I got another, and another, and…and, well, I figured I’d better post something before it got to the point where I was receiving death threats.
I’m going to break this down into pieces, starting at the input. It may take a while to get this done so for those who are lost, don’t despair. I’ll post more as soon as I can.
The place to begin is with a differential. Just one differential, not the complementary one in the GR-25. Most amplifiers have at least one differential in them and you won’t have to search too hard to find examples with two, three, four…let’s just say that differentials are pretty popular. In fact, you can make an entire amplifier out of nothing but differentials if you take a notion to.
A differential is a pair of gain devices, back to back. You can build a differential with tubes, bipolars, JFETs, or MOSFETs. In the GR-25 I used JFETs, so I’ll use JFET nomenclature so as not to have to shift gears later in the game. Suffice it to say that all differentials work the same way, so it doesn’t really matter what kind of device you have on hand, you can pretty much get the same sort of action, regardless.
When you put a signal into the Gate of a JFET and take the output at the Drain, it’s called a Common Source connection. When you put the signal in at the Gate and take the output at the Source, it’s called a Common Drain connection, although it’s generally known by its alias: Source follower. And if you put the signal in at the Source, then take the output at the Drain, that’s called a Common Gate connection. Before all is said and done, we’ll be using all three of those options.
I’m attaching a file showing a series of differentials. I’ll be using that to try to keep things on an even keel.
The first differential on the left (A) is the bare-bones sort. We’ll assume the signal enters the Gate of Q1, the JFET on the left-hand side. That signal then appears, out of phase, at the Drain of Q1, where it meets R1, and is available for use somewhere else in the circuit. But—and this is something that a lot of people don’t realize when they first look at a differential—there is also a signal at Q1’s Source. If you view Q1 as a Common Source amplifier, then your focus is on the signal at Q1’s Drain. If you view Q1 as a Common Drain amplifier, then your focus is on the signal at Q1’s Source. It’s not an either/or proposition; Q1 is quite happy to perform both functions at the same time. (This is intuitively obvious to people who work with tubes, because the split load phase splitter is used in precisely the same way. Those who want an example of the analogous solid state circuit can look at the JLH amplifier.)
The signal coming from Q1’s Source is faced with two choices—it can travel down Rbias, or it can enter Q2’s Source. Which route it chooses will depend on the relative impedances of those two legs. Q2’s Source is a low impedance. Whether that’s good or bad depends on your application. A low impedance means you’re going to have to have enough current to drive it. In this case, everything is going to work out just fine because, from Q2’s point of view, Q1 is acting like a follower, and if there’s one thing followers are good at, it’s delivering current. The other possible route for the signal from Q1 is Rbias. You want the majority of that signal to go to Q2, so to encourage that you make Rbias much larger than the impedance of Q2’s Source. Let’s say that Q2’s Source presents a load of 100Ω. This is higher than the JFETs in the GR-25, but it’s a convenient, round number. If you sit down and tickle a calculator for a minute, running a few iterations of Ohm’s Law, you’ll find that it doesn’t take nearly as much resistance as you might think to convince the signal coming out of Q1 that Q2’s Source is the path of least resistance. (Ahem.) A few thousand Ohms will do quite admirably, although you see a lot of people using current sources instead of resistors in an attempt to provide a reasonable approximation of an infinite impedance. It’s not necessary to go to those lengths, so we’ll just say for the moment that Rbias is something like 10k, which is obviously much, much greater than our hypothetical 100Ω impedance at Q2’s Source. As a result, something like 99% of the signal slides right on over to Q2.
So now we’ve got a signal at Q2’s Source, and in turn that signal becomes available at Q2’s Drain. So that makes Q2 a Common Gate configuration. Ta da! As promised, we’ve used all three possible hookups in one fell swoop.
Now, the thing about a Common Gate connection is that the signal at the Drain is in phase with the input signal. It has to be, because they’re in series. And since the output of a follower is always in phase with the input signal, that means that the output at Q2’s Drain is also in phase with the input. If you’ll recall, the signal at Q1’s Drain was out of phase. So now we have our choice of an in-phase signal or an out-of-phase signal, depending on our mood. As it happens, the GR-25 uses both at the same time, but that’s not required. A lot of amplifiers get by quite nicely with using just one output from the differential.
Circuit (B) is exactly the same as circuit (A) except for the fact that it’s rendered in P-ch JFETs; it’s a mirror image of (A). As a differential, it performs the same way.
Circuit (C) combines (A) and (B). The resulting circuit only needs one Rbias to make things come together. Note that both the N-ch and P-ch JFETs on the left side (Q1 & Q3) share the same signal input. Likewise for Q2 & Q4 on the right. This is called a complimentary differential and it forms the heart of the GR-25 (as it did for the JC-3). When a positive signal is applied to Q1 & Q3, they both shift downwards in response; Q1 conducts more heavily, driving its Drain down, and Q3 conducts less heavily, also causing its Drain to move downwards. The entire circuit acts like a see-saw, which leads the JFETs on the other side, Q2 & Q4, to move upwards.
(This is the kind of circuit that gives tube folks feverish dreams. Much good could be done if only someone would devise a P-tube that didn’t require anti-matter. For obvious reasons, the anti-matter sort isn’t something you’d want in your listening room.)
If you look at the schematic for the front end of the GR-25, you’ll see that the equivalent resistor to Rbias is 47.5Ω. What happened to the 10k we were just talking about? It’s not needed. The cool thing about a complimentary differential is that the two differentials work in tandem, both from side to side and from top to bottom. As one side draws more current, the other draws less and at all times the current draw sums to a constant DC value. One of the consequences of this is that the differentials almost act like current sources for each other.
If you’ve managed to keep up with the changes so far, the next step is a fairly simple one—add cascodes to all four JFETs. A cascode is a gain device that sits on the shoulders of the one below it. It operates in Common Gate mode, with the signal entering through its Source and exiting through its Drain. The Gate is set to an arbitrary voltage using a voltage divider, although there are other ways to accomplish the same ends. Given that there are two cascodes sitting side-by-side, we can save parts and gain some subtle benefits by using the same resistive divider to set both cascode devices’ Gates at the same time.
Cascodes are useful in several ways. One is that they cap the voltage that the gain device (the one underneath) sees at its Drain. In the original iteration of the GR-25 I used 2SK389s and 2SJ109s as the input devices. Neither of those parts can take much voltage, but if you add cascode devices set to take the brunt of the rail voltage, those parts will never know it’s there.
Another benefit is that cascodes reduce the effects of the inherent capacitance at the Gates of the input devices. This improves bandwidth and distortion, both desirable things if you’re trying to pull off an amp that doesn’t use any feedback. And if you do intend to use feedback, there’ll be less for it to do.
As before, output is taken at R1, R2, R3, and R4, with the sole difference being that the Drains driving those resistors are those of the cascode devices (Q5, Q6, Q7, Q8) instead of the gain devices (Q1, Q2, Q3, Q4).
I’m approaching the character limit for a single post, and have covered all of the first stage to some degree. I’ll try to whip up some stuff covering the folded cascodes and their cascodes. Depending on how things go, I may be able to get the bias circuit in, too.
Grey
Thanks, for us less cerebral that are in the early stages of understanding( that was meant for me, if anyone else resembles that remark feel free to join in.)
Regards, Elwood
Regards, Elwood
Thanks Grey,
Q2 driven from its source lit the bulb in understanding the differential pair.
Jess
Q2 driven from its source lit the bulb in understanding the differential pair.
Jess
hola you all,
what's a good source for k246/j103 jfets in the front end???
Need the info...
Regards, Elwood
what's a good source for k246/j103 jfets in the front end???
Need the info...
Regards, Elwood
I had cause to touch base with Nelson and he pointed out that I had allowed myself to slip into a bad habit. My wife and I are fond of puns and such and have a habit of working the compliment (meaning to say something flattering) vs. complement (in this case meaning N-ch and P-ch parts working in tandem) furrow of the garden. That has unfortunately worked its way into my writing and if you read the exposition above you'll notice a certain ambiguity in spelling as I play both sides against the middle.
Mea culpa.
I promise it won't happen again...until next time.
(In another thread someone had requested that I put spaces between my paragraphs. I responded that I didn't want to do so because it would be a bad habit for the other side of my life, meaning my stories. In the real world, people use tabs to indicate a new paragraph. Now, within the very same week, I find that a different habit has crept in the back door. Very embarrassing.)
I'm not done with PART II yet. I've been detained by real-world things this weekend. I'll get it done as soon as possible.
If Part I answered questions, good. If I didn't cover something in enough detail...ask. I'll try to attend to it ASAP. If you have question that I need to deal with in a future installment, ask. I'll make sure to cover them.
Incidentally, if you have any Chateau Gloria '95 on hand, I'd suggest drinking it now, as it's starting to fade. I'm on my second bottle...okay, we're on our second bottle and they're both coming off the far side of the peak. Yes, they were stored properly. Perhaps I should pop the last bottle and get it over with.
Grey
Mea culpa.
I promise it won't happen again...until next time.
(In another thread someone had requested that I put spaces between my paragraphs. I responded that I didn't want to do so because it would be a bad habit for the other side of my life, meaning my stories. In the real world, people use tabs to indicate a new paragraph. Now, within the very same week, I find that a different habit has crept in the back door. Very embarrassing.)
I'm not done with PART II yet. I've been detained by real-world things this weekend. I'll get it done as soon as possible.
If Part I answered questions, good. If I didn't cover something in enough detail...ask. I'll try to attend to it ASAP. If you have question that I need to deal with in a future installment, ask. I'll make sure to cover them.
Incidentally, if you have any Chateau Gloria '95 on hand, I'd suggest drinking it now, as it's starting to fade. I'm on my second bottle...okay, we're on our second bottle and they're both coming off the far side of the peak. Yes, they were stored properly. Perhaps I should pop the last bottle and get it over with.
Grey
All I got is a bottle of Mondavi Cabernet and five gallons of mead in the secondary fermenter sitting beside my bench where I can keep an eye on it.
Regards, Elwood
Regards, Elwood
I'm trying to get a clean culture of yeast out of an Affligem bottle to brew...you guessed it...mead. The yeast is viable, but I keep getting contamination.
Dammit.
Grey
P.S.: What? You think I keep bees because I went to Georgia Tech? Au contraire! Those of us who went to UNC have our priorities straight. As Winnie the Pooh says, "The only reason for bees is to make honey."
(Them what went to Ga need not beat me up. I realize I'm playing fast and loose with the bugs...)
Dammit.
Grey
P.S.: What? You think I keep bees because I went to Georgia Tech? Au contraire! Those of us who went to UNC have our priorities straight. As Winnie the Pooh says, "The only reason for bees is to make honey."
(Them what went to Ga need not beat me up. I realize I'm playing fast and loose with the bugs...)
For reasons that are not clear to me, most of the communication regarding this amplifier is coming in via e-mail directly to me instead of as posts in this thread.
Today's entry:
"Who does your PB boards? How much do they cost?" (That's paraphrased.)
The answer is that I do my own. (This is DIY, you know...)
I sit down with a piece of paper and scratch out a few alternate layouts. I wait a few days, then (because I've forgotten the details of how I did it before) sit down and do it over, hopefully coming up with a couple of new possibilities. Then I choose the most promising candidates and start transcribing them into the computer.
One of the things that gets overlooked is the importance of letting things sit for a while. You will rarely find yourself in a circumstance where someone is holding a gun to your head, threatening to blow your brains out if you don't finish by dawn. Don't be in such a hurry. You'll be (unpleasantly) surprised when you look at your artwork a day or two later and find that some malevolent demon has rearranged all your "perfect" artwork since you last looked at it. You'll find errors, traces that can be rerouted to accomplish the same thing more elegantly, and better parts positions. Fix those things, then let it sit for a few more days. Fix. Repeat. Fix. Repeat.
There's no way I can tell you when to stop. There comes a day when you look at the board and it just looks right. Don't try to rush that day. The single best piece of advice I can give you is not to fall in love with any particular layout. Distance yourself from the board so that if a complete rework becomes necessary due to some really horrendous error (not something you did, of course...it was those gremlins messing with it during the night...), you can start over with a minimum of grumbling. When you can no longer find things to improve, it's time to etch.
I use three methods depending on the complexity of the board. The simplest boards, the ones with big, blocky patterns like rows of capacitors straddling from rail to ground, I do using ordinary clear packing tape. Cover the board with tape, then put a sheet of paper over the board with your pattern on it. Cut with an X-acto knife. Peel off the cut portions of tape. Etch. Peel off the rest of your tape and you're got a finished board. Simple. Fast. Cheap. Easy. Just the way I like things. There are a few refinements, but that's the basic idea and it works like a charm.
For more complex boards I use an optical method. This allows me to pull of stunts like ground planes and tight clearances. This takes a little more fiddle factor, but the results are very nice.
The third method? The third method is the toner transfer method. I DO NOT RECOMMEND THIS METHOD TO ANYONE UNDER ANY CIRCUMSTANCES. PERIOD. EVEN IF THERE'S A GUN PRESSED TO YOUR HEAD. I should have known when the people who sell the stuff refused to endorse a brand of toner that there'd be trouble. You might also look closely at the pictures on the toner peoples' own website. At least as of the last time I looked, the demonstration boards had pinholes in the remaining copper. And that's from the guys who sell the paper. You'd think that if anyone could get the stuff to work, it'd be them. They can't. You can't. It's a flawed system. I could go on at length about the shortcomings of this method, but I won't. If you're determined to hit yourself over the head with a 2 x 4" then go ahead, but don't say I didn't warn you. You've reached the edge of the map and even the people who sold you the toner paper can't guide you. Having been to the edge of the world and back, I will give you the strategy that worked best for me.
Take two ceramic tiles. Get flat ones and when I say flat, I mean flat. Take a known-good straightedge and go through as many tiles as necessary to find two that are flat to the point that you can't see light under the edge of your straightedge. Place a brick in the oven. Place one of your ceramic tiles on the brick. Place the other tile and bricks, pans of sand, whatever in the oven also. Preheat the whole mess to 350 degrees. When it's all good and hot, place your blank PCB with the toner sheet taped to it on the tile that's sitting on the brick. Then place the second tile, face down, on the board. On top of that go the other bricks, pans of sand, etc. that you're using to weigh the whole thing down. How much weight? Oddly enough, it works out to about 1 pound per square inch of board; a nice, even number. Yes, that requires that you calculate the proper weight for every PCB you do. If you use more weight, the traces will squish. If you use less, the toner won't stick at all. Bake the board for 3 1/2 minutes. Take it out. Let it cool under a small amount of pressure, otherwise the paper tends to draw up as it cools and it will ruin some of the traces. Or you could just skip the whole mess and go to another strategy, like rub-on traces and pads. At least they work. Yes, I've done rub-ons, too, but it's time consuming. I don't do boards that way anymore.
I've still got a little of that toner crud left. I'm slowly using it up (Aw, Mom...do I have to?). Once it's gone, I'm never going back. Yes, you see posts from people extolling the virtues of this method. All I can say is that they obviously have low standards. Set your sights higher.
The young 'uns are up to mischief, so I need to go restore order to the universe. The whole PCB topic could go for pages. Treat this as an overview.
Grey
Today's entry:
"Who does your PB boards? How much do they cost?" (That's paraphrased.)
The answer is that I do my own. (This is DIY, you know...)
I sit down with a piece of paper and scratch out a few alternate layouts. I wait a few days, then (because I've forgotten the details of how I did it before) sit down and do it over, hopefully coming up with a couple of new possibilities. Then I choose the most promising candidates and start transcribing them into the computer.
One of the things that gets overlooked is the importance of letting things sit for a while. You will rarely find yourself in a circumstance where someone is holding a gun to your head, threatening to blow your brains out if you don't finish by dawn. Don't be in such a hurry. You'll be (unpleasantly) surprised when you look at your artwork a day or two later and find that some malevolent demon has rearranged all your "perfect" artwork since you last looked at it. You'll find errors, traces that can be rerouted to accomplish the same thing more elegantly, and better parts positions. Fix those things, then let it sit for a few more days. Fix. Repeat. Fix. Repeat.
There's no way I can tell you when to stop. There comes a day when you look at the board and it just looks right. Don't try to rush that day. The single best piece of advice I can give you is not to fall in love with any particular layout. Distance yourself from the board so that if a complete rework becomes necessary due to some really horrendous error (not something you did, of course...it was those gremlins messing with it during the night...), you can start over with a minimum of grumbling. When you can no longer find things to improve, it's time to etch.
I use three methods depending on the complexity of the board. The simplest boards, the ones with big, blocky patterns like rows of capacitors straddling from rail to ground, I do using ordinary clear packing tape. Cover the board with tape, then put a sheet of paper over the board with your pattern on it. Cut with an X-acto knife. Peel off the cut portions of tape. Etch. Peel off the rest of your tape and you're got a finished board. Simple. Fast. Cheap. Easy. Just the way I like things. There are a few refinements, but that's the basic idea and it works like a charm.
For more complex boards I use an optical method. This allows me to pull of stunts like ground planes and tight clearances. This takes a little more fiddle factor, but the results are very nice.
The third method? The third method is the toner transfer method. I DO NOT RECOMMEND THIS METHOD TO ANYONE UNDER ANY CIRCUMSTANCES. PERIOD. EVEN IF THERE'S A GUN PRESSED TO YOUR HEAD. I should have known when the people who sell the stuff refused to endorse a brand of toner that there'd be trouble. You might also look closely at the pictures on the toner peoples' own website. At least as of the last time I looked, the demonstration boards had pinholes in the remaining copper. And that's from the guys who sell the paper. You'd think that if anyone could get the stuff to work, it'd be them. They can't. You can't. It's a flawed system. I could go on at length about the shortcomings of this method, but I won't. If you're determined to hit yourself over the head with a 2 x 4" then go ahead, but don't say I didn't warn you. You've reached the edge of the map and even the people who sold you the toner paper can't guide you. Having been to the edge of the world and back, I will give you the strategy that worked best for me.
Take two ceramic tiles. Get flat ones and when I say flat, I mean flat. Take a known-good straightedge and go through as many tiles as necessary to find two that are flat to the point that you can't see light under the edge of your straightedge. Place a brick in the oven. Place one of your ceramic tiles on the brick. Place the other tile and bricks, pans of sand, whatever in the oven also. Preheat the whole mess to 350 degrees. When it's all good and hot, place your blank PCB with the toner sheet taped to it on the tile that's sitting on the brick. Then place the second tile, face down, on the board. On top of that go the other bricks, pans of sand, etc. that you're using to weigh the whole thing down. How much weight? Oddly enough, it works out to about 1 pound per square inch of board; a nice, even number. Yes, that requires that you calculate the proper weight for every PCB you do. If you use more weight, the traces will squish. If you use less, the toner won't stick at all. Bake the board for 3 1/2 minutes. Take it out. Let it cool under a small amount of pressure, otherwise the paper tends to draw up as it cools and it will ruin some of the traces. Or you could just skip the whole mess and go to another strategy, like rub-on traces and pads. At least they work. Yes, I've done rub-ons, too, but it's time consuming. I don't do boards that way anymore.
I've still got a little of that toner crud left. I'm slowly using it up (Aw, Mom...do I have to?). Once it's gone, I'm never going back. Yes, you see posts from people extolling the virtues of this method. All I can say is that they obviously have low standards. Set your sights higher.
The young 'uns are up to mischief, so I need to go restore order to the universe. The whole PCB topic could go for pages. Treat this as an overview.
Grey
Nothing realated to this amp but this should not be too hard. Make some basic agar plates. Melts in a microwave so nice and sterile.
http://en.wikipedia.org/wiki/Agar_plate
Use sterile technique and douse your hands in Isopropanol.
Use a pressure cooker to sterilize all your stuff.
Take your sample and do a serial dilution using sterile water. Then plate the last few and look for isolated colonies. Yeast (BIG cells) should be easy to identify vs fungi or bacteria. May be nothing new to you.
Kris
Nothing new. I've been doing this sort of thing for a long time. The thing I'm up against is--I think--airborne contamination. Something that springs from baby poop, for all I know. I've lost something like eight or ten yeast cultures in a row now and I'm starting to get peeved. Short of building an isolation box (and don't think I haven't threatened to do so), I'm pretty much at the mercy of whatever is floating around. I don't have a lab, I have a house...and it's very thoroughly lived in. It's a nuisance, but there it is. The problem is that you can't nuke the cultures once they're inoculated or you'll kill the yeast, too. You can arguably acid wash a yeast culture, but it weakens the yeast and it's not all that effective against the real nuisance critters, anyway, as they're just about as acid-tolerant as yeast is. Couldn't grow in wort if they weren't.
Maybe I should be haunting hospital dumpsters looking for medical hardware, because I sure can't find electronic stuff around here.
I was trying to get PART II ready last night but got clobbered. I'll try again later tonight. I've got a simplified schematic showing the circuit turning the corner into folded cascodes, just need to finish the words to go with it.
Grey
Maybe I should be haunting hospital dumpsters looking for medical hardware, because I sure can't find electronic stuff around here.
I was trying to get PART II ready last night but got clobbered. I'll try again later tonight. I've got a simplified schematic showing the circuit turning the corner into folded cascodes, just need to finish the words to go with it.
Grey
- Status
- Not open for further replies.