Hi Greg
Looks like the oscillation was caused by the CCS after all.
Your air sculptures actually aren't bad. I was more worried about different parts of the circuit being connected together with test leads lying around all over the place.
Looks like the oscillation was caused by the CCS after all.
Would be nice to try tame the other one anyway, for educational purposes. I suspect you can stabilize it with the mods shown below. R3 should probably be a couple of hundred ohms to a kilohm. C1 can probably be much lower than 100pF, maybe even left out. Looks suspiciously like the mosfet CCS you started with. Actually, I think your mosfet CCS still has that resistor.
Looks like an ancient photo. Is that something Nelson made?For entertainment I've attached another such layout - that the keen eyes of someone famous on this forum will recognise
That's half of it, yes. Also, for minimum stray capacitance and inductance, you want short leads, and the shortest route between two points generally doesn't involve a copper track running around a circuit board.
Your air sculptures actually aren't bad. I was more worried about different parts of the circuit being connected together with test leads lying around all over the place.
Actually, no. Have you got a website somewhere? CanAm stuff would be interesting too, but probably more appropriate in the "(not) F5" thread.
Dunno, hopefully not.
That's what I'm thinking too. In the meantime, maybe it would help to put a resistor at the output of the buffer to isolate it a bit from the wiring. Maybe a small capacitor to ground too, before or after the resistor.
Absolutely, and learning along the way too.
Attachments
Thanks Godfrey, I tried the modifications you suggested. And yes, it does look a lot like where we left off The following shots are where its performance now stands.
Boy do I have some cool scope shots for you
Pic 1 is the amplifier's 20kHz square wave output on the bottom plus the base voltage of the MJL21194 on the top. As you can see the square wave isn't too bad. But look at how fast that base voltage is moving! I zoom into the fuzzy part as much as possible (0.2uS) and it _just_ becomes apparent that there is a waveform there, but the period of the waveform is still much less than 0.2uS. So we are talking about tens of MHz oscillation.
I guess this is to be expected and is just the whole process of the CCS loop operating - sensing current, turn off, current drops, turn on, repeat, etc. But I have no idea how much you would expect to see and how much is excess oscillation. Looks neat though!
Pic two is the same again, but with a sine wave output.
Pic three is what the fuzzy bottoms of the square wave output actually are - baby oscillations. Again, very fast at 0.2uS/division. There's no way my inductance-rich lash up is going to help at those frequencies.
Anyway, as far as the design goes, I think pursuing it further with the crude construction may be pointless as it just introduces too many variables and stray inductance. It's time to make a PCB I think.
So, do we take the plunge and use a BJT CCS for the output stage, or do we stick with the FET topology we're familiar with?
To help answer this, I pose the question: Is the modelled decrease in distortion caused by the BJT CCS likely to be real, or just an artifact of poor models for the parts? Maybe the 2N3055 model is a better/worse model than the IRFP240 one?
If it is likely to be better in practice, then why? Maybe due to better linearity of the BJT? But I wouldn't have thought linearity would matter much in this instance as it kind of self corrects anyway?? (just thinking out loud here).
Another thing is maybe the BJT turns on more gently - base current increases slowly rather than having to bring the gate up to voltage? I don't know.
I would model it with Cordell's models, but they don't appear to work with the latest LT spice - unless I'm doing it wrong.
Sorry, this is going somewhere
My goal is to make the lowest distortion zero global feedback circuit I can (including the preamp - next project). I was going to conclude by saying that if the BJT CCS really is capable of lower distortion then it would seem prudent to use it, despite the added complexity. Especially as I just happen to have 25 MJL21194s sitting here.
But then just as I was typing I thought "Hmm, maybe the FET CCS is good enough, and any gain in performance is offset by its simplicity".
Geez, I don't know now.
Anyway, going to sleep on it. It's been an educational weekend.
Oh, and yes, that B&W photo is one of Nelson's attempts from the late 70s.
The following shots are where its performance now stands.Boy do I have some cool scope shots for you
Pic 1 is the amplifier's 20kHz square wave output on the bottom plus the base voltage of the MJL21194 on the top. As you can see the square wave isn't too bad. But look at how fast that base voltage is moving! I zoom into the fuzzy part as much as possible (0.2uS) and it _just_ becomes apparent that there is a waveform there, but the period of the waveform is still much less than 0.2uS. So we are talking about tens of MHz oscillation.
I guess this is to be expected and is just the whole process of the CCS loop operating - sensing current, turn off, current drops, turn on, repeat, etc. But I have no idea how much you would expect to see and how much is excess oscillation. Looks neat though!
Pic two is the same again, but with a sine wave output.
Pic three is what the fuzzy bottoms of the square wave output actually are - baby oscillations. Again, very fast at 0.2uS/division. There's no way my inductance-rich lash up is going to help at those frequencies.
Anyway, as far as the design goes, I think pursuing it further with the crude construction may be pointless as it just introduces too many variables and stray inductance. It's time to make a PCB I think.
So, do we take the plunge and use a BJT CCS for the output stage, or do we stick with the FET topology we're familiar with?
To help answer this, I pose the question: Is the modelled decrease in distortion caused by the BJT CCS likely to be real, or just an artifact of poor models for the parts? Maybe the 2N3055 model is a better/worse model than the IRFP240 one?
If it is likely to be better in practice, then why? Maybe due to better linearity of the BJT? But I wouldn't have thought linearity would matter much in this instance as it kind of self corrects anyway?? (just thinking out loud here).
Another thing is maybe the BJT turns on more gently - base current increases slowly rather than having to bring the gate up to voltage? I don't know.
I would model it with Cordell's models, but they don't appear to work with the latest LT spice - unless I'm doing it wrong.
Sorry, this is going somewhere
My goal is to make the lowest distortion zero global feedback circuit I can (including the preamp - next project). I was going to conclude by saying that if the BJT CCS really is capable of lower distortion then it would seem prudent to use it, despite the added complexity. Especially as I just happen to have 25 MJL21194s sitting here.
But then just as I was typing I thought "Hmm, maybe the FET CCS is good enough, and any gain in performance is offset by its simplicity".
Geez, I don't know now.
Anyway, going to sleep on it. It's been an educational weekend.
Oh, and yes, that B&W photo is one of Nelson's attempts from the late 70s.
Attachments
Godfrey...... Unfortunately, I've been so busy with work (recently retired) and building things, I've yet to build any websites for my cars and amateur radio stuff. But they are on my list. Agreed, too--this will end up on my "Far away from F5" website! I'll let you know when I get my domain set up. Will not attempt to hijack this or other diyAudio threads (at least, not knowingly!)
Great continuing comments on this amp build, and great origional work in the amp prototype, too!
Great continuing comments on this amp build, and great origional work in the amp prototype, too!
That's pretty much how it works. There should be no oscillation at all though. It only happens when the circuit keeps over-reacting to changes in it's own output, as you described.
If you build the BJT version, I'd strongly suggest allowing for the extra resistor and capacitor shown in the schematic in post 61. Also, I just remembered (duh!) it's a very good idea to include a base stopper resistor for the MJL21194 - similar idea to gate stoppers for mosfets, but much lower value.
With them, it should be easy to stop the oscillation. Without them, probably not.
Your call, but I'm leaning towards the BJTs if they sim better and you already have a lot of them. I'm fairly clueless with LTSpice (normally use SIMetrix), but had no problem using Bob's models.
Here's what I did:
a) Download the zip file below (originally posted by Bob here).
b) Unzip it to a folder somewhere
c) Open the folder and double-click on one of the asc files to open it.
Presto! It opens with LTSpice and the sim works, and uses Bob's models (which were included).
As far as I can make out, the trick is to have the "Cordell Models.txt" file in the same folder as the sim file, and have the ".include Cordell Models.txt" directive in the sim. There's probably smarter ways to do it but that's what worked for me.
For the most sensible results, use Bob's models for the IRFP240, MJL21194, and the other BJTs.
p.s. Today's outrageous suggestion: Design the PCB so it can accommodate either version!
I think if you make the PCB for the BJT version and then decide to switch to the mosfet version, it'll be easy; you just leave out a couple of components (e.g. the extra transistor), and put in a wire link.
Attachments
I am delighted that at least one other person is getting something out of this thread!
Hi Godfrey,
Sorry for the late reply...I'm trying not to post unless I have made progress to prevent diluting the thread too much.
Thanks for your suggestions - I have gone with the BJT design with provisions for using the FET one if necessary.
I am currently designing the PCB for the so called "SWORDFREY" single ended power follower. I hope to post the design in the next day or so - it's nearly complete.
I have added base resistors to both the buffer and output CCS transistors. Just to settle it down a bit.
My PCB design skills are pretty terrible. I don't know how everyone makes such nice layouts while still keeping the traces short. It is a real skill.
I look forward to your comments on the layout when I get it finished. One annoying thing is I have had to use a fairly spread out design due to the heatsink required for the BD139. I have also allowed for bypassing of all electrolytics.
Oh, a quick question. I have designed for a 2.2uF input cap. Do you think that is a good size?
I will try Cordells models tonight - though I have been doing some reading which also tends to agree with the notion that a BJT CCS is better than a fet one.
Edit: I Just saw your zip file. Thanks for that, it works perfectly!
Cheers,
Greg.
Last edited:
OK,
Here is how it's going so far. Attached are the final schematic and initial PCB design. PCB keeps getting refined every time I look at it, but if I'm completely on the wrong track it would be nice to have some feedback.
Edit: actually, scrap this design. I've just seen a way to improve it significantly.
Cheers,
Greg.
Here is how it's going so far. Attached are the final schematic and initial PCB design. PCB keeps getting refined every time I look at it, but if I'm completely on the wrong track it would be nice to have some feedback.
Edit: actually, scrap this design. I've just seen a way to improve it significantly.
Cheers,
Greg.
Hi Greg
Final schematic looks nice. I haven't looked at PCB yet, just getting into panic mode now with the move. Hopefully back to normal next week. If you're lucky, AndrewT might comment on the PCB too.
Maybe the "Fishy Follower"
2.2uF puts the low frequency roll-off down at 0.7Hz. IIRC, you're using monster output coupling caps too. Something to think about is what happens when you switch the amp on and off. I'd recommend checking that in the simulator, no input signal, just the power rail going up and down. You might need to play with various cap values to avoid a big "thump" through the speakers. Also check if any transistors get reverse voltage across them, especially when switching off.
btw, There was another similar design awhile ago in this thread. Maybe there's some relevant comments there.
Cheers - Godfrey
Final schematic looks nice. I haven't looked at PCB yet, just getting into panic mode now with the move. Hopefully back to normal next week. If you're lucky, AndrewT might comment on the PCB too.
Umm... Can't you leave "frey" out of the name?
Maybe the "Fishy Follower"2.2uF puts the low frequency roll-off down at 0.7Hz. IIRC, you're using monster output coupling caps too. Something to think about is what happens when you switch the amp on and off. I'd recommend checking that in the simulator, no input signal, just the power rail going up and down. You might need to play with various cap values to avoid a big "thump" through the speakers. Also check if any transistors get reverse voltage across them, especially when switching off.
btw, There was another similar design awhile ago in this thread. Maybe there's some relevant comments there.
Cheers - Godfrey
Thanks Godfrey,
Don't bother with the PCB, or anyone else for that matter. The more I look at it the more I realise how terrible it is. I'm going to try to reconfigure it with a better and more efficient layout. That said, I think most of the important traces are probably short enough, and I'm sure it would work.
I'm also considering changing the bias control to the one shown below (from Nelson's latest creation). Seems to be a worthwhile modification to get a more stable bias voltage, though a higher value zener than the one shown would be required.
OK, will do. Just thought you might like some recognition for all your help with the project.
Good luck with the moving. I just moved myself and it was a painful experience. I will check out the thread you posted and look into the on/off transients.
Don't bother with the PCB, or anyone else for that matter. The more I look at it the more I realise how terrible it is. I'm going to try to reconfigure it with a better and more efficient layout. That said, I think most of the important traces are probably short enough, and I'm sure it would work.
I'm also considering changing the bias control to the one shown below (from Nelson's latest creation). Seems to be a worthwhile modification to get a more stable bias voltage, though a higher value zener than the one shown would be required.
OK, will do. Just thought you might like some recognition for all your help with the project.
Good luck with the moving. I just moved myself and it was a painful experience. I will check out the thread you posted and look into the on/off transients.
Attachments
Hi Godfrey,
I'm afraid not. I've been away for a while and only got back yesterday. I'll start again tonight. Before I do though, what are your thoughts on the biasing scheme I presented in the last post? Do you think it would be a worthwhile improvement?
I still need to sim the circuit again to test those possible transients you mentioned. I hope to get back into it all this week.
I trust the move went well?
I'm afraid not. I've been away for a while and only got back yesterday. I'll start again tonight. Before I do though, what are your thoughts on the biasing scheme I presented in the last post? Do you think it would be a worthwhile improvement?
I still need to sim the circuit again to test those possible transients you mentioned. I hope to get back into it all this week.
I trust the move went well?
Hi Greg
The move was totally chaotic but I'm here now and settling in. Mental health is improving rapidly now I'm out of the city and staying in a small fishing village on the coast (or as one of the locals described Pringle Bay - "a small drinking village with a slight fishing problem")
Anyway...
The bias scheme you showed looks good, so long as you remember to flip polarity since you're using a negative-earth arrangement. One Possible caveat: If your mains supply isn't as stable as it might be, you may be better off with the old circuit, which will tend to track slow changes in PSU voltage, and keep the output biased about half-way between earth and the "voltage-of-the-day".
Cheers - Godfrey
The move was totally chaotic but I'm here now and settling in. Mental health is improving rapidly now I'm out of the city and staying in a small fishing village on the coast (or as one of the locals described Pringle Bay - "a small drinking village with a slight fishing problem"
)Anyway...
The bias scheme you showed looks good, so long as you remember to flip polarity since you're using a negative-earth arrangement. One Possible caveat: If your mains supply isn't as stable as it might be, you may be better off with the old circuit, which will tend to track slow changes in PSU voltage, and keep the output biased about half-way between earth and the "voltage-of-the-day".
Cheers - Godfrey
Thanks Godfrey,
Glad to hear the move is over with. Sounds like you're in a nice area! I recently moved to the country myself. Commuting isn't much fun but the fresh air is good.
Anyway, I will stick with the old bias scheme - mostly out of laziness. I suspect our line voltage is pretty variable given my location anyway.
I have tried to sim the turn on transients - the speaker one is attached. Pretty nasty.
I also looked at the base and gate voltages and there are indeed a few possible issues. The Vbe of the transistors does go up to about 10v I think, which is too much, but it is for fractions of a nanosecond. Not sure if that is significant or not?
I'm trying to get the motivation to redesign the PCB.... I seem to spend all day on the computer at work these days and spending a few additional hours in front of the computer at night is really unappealing at the moment
Glad to hear the move is over with. Sounds like you're in a nice area! I recently moved to the country myself. Commuting isn't much fun but the fresh air is good.
Anyway, I will stick with the old bias scheme - mostly out of laziness. I suspect our line voltage is pretty variable given my location anyway.
I have tried to sim the turn on transients - the speaker one is attached. Pretty nasty.
I also looked at the base and gate voltages and there are indeed a few possible issues. The Vbe of the transistors does go up to about 10v I think, which is too much, but it is for fractions of a nanosecond. Not sure if that is significant or not?
I'm trying to get the motivation to redesign the PCB.... I seem to spend all day on the computer at work these days and spending a few additional hours in front of the computer at night is really unappealing at the moment
Attachments
Godfrey,
Hope you're well.
I designed a new PCB and sent off the order today. Not going to post the layout here because it's well, embarassingly terrible really. But the critical traces are short at least, and all gate/base resistors are hard up against the fets/transistors. It will be better than my "air sculptures" at any rate.
I went back to the FET ccs as I wanted to keep it small enough to fit on the express PCB miniboard size (2.5" x 3.8"), and I was having trouble getting it to fit with the additional heatsink and power resistors needed for the bipolar version.
Will hopefully post some pictures of a completed channel and some scope shots in a few days.
Hope you're well.
I designed a new PCB and sent off the order today. Not going to post the layout here because it's well, embarassingly terrible really. But the critical traces are short at least, and all gate/base resistors are hard up against the fets/transistors. It will be better than my "air sculptures" at any rate.
I went back to the FET ccs as I wanted to keep it small enough to fit on the express PCB miniboard size (2.5" x 3.8"), and I was having trouble getting it to fit with the additional heatsink and power resistors needed for the bipolar version.
Will hopefully post some pictures of a completed channel and some scope shots in a few days.
Update
Got the PCBs back today and assembled one tonight as best I could with the parts I had available (fig 1).
Plugged it in...no smoke.
Adjusted the bias pot...output varied appropriately. Set it to just under 1/2 the rail voltage. Plugged in the sig gen and scope.
First the 20kHz square wave test (fig 2). Rise and fall times not too bad, but still quite a lot of ringing (fig 2 & 3).
I'm not overly concerned about the ringing though, as I'm fairly sure larger gate stoppers would prevent it - the waveforms on at the 240 fet gate is dirty whereas the waveform at the gate stopper is clean. It seems some oscillation is happening in the few mm of lead between the fet and the gate stopper...amazing considering how close I have them.
There probably isn't any point using larger gate stoppers though as audio waveforms would never cause this, right?
Anyway, it sounds quite nice with my little preamp. Pretty happy with it really. The treble is especially sweet sounding, but overall frequency response is very good.
It's running very hot at about 2A per device though!
Got the PCBs back today and assembled one tonight as best I could with the parts I had available (fig 1).
Plugged it in...no smoke.
Adjusted the bias pot...output varied appropriately. Set it to just under 1/2 the rail voltage. Plugged in the sig gen and scope.
First the 20kHz square wave test (fig 2). Rise and fall times not too bad, but still quite a lot of ringing (fig 2 & 3).
I'm not overly concerned about the ringing though, as I'm fairly sure larger gate stoppers would prevent it - the waveforms on at the 240 fet gate is dirty whereas the waveform at the gate stopper is clean. It seems some oscillation is happening in the few mm of lead between the fet and the gate stopper...amazing considering how close I have them.
There probably isn't any point using larger gate stoppers though as audio waveforms would never cause this, right?
Anyway, it sounds quite nice with my little preamp. Pretty happy with it really. The treble is especially sweet sounding, but overall frequency response is very good.
It's running very hot at about 2A per device though!
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.