I have tried the circuit with verticals and found that they sounded harsher. Will be interested to see what you come up with.
Onto,
I assume CFA is a TLA related to current feedback. If so, I learn something today.
I endorse Mike's comment; the idea was to make a singleton input stage. That means, inevitably, input on the gate and feedback on the source. If this is what you mean, then the choice was quite deliberate. It also means that the input stage has very low gain since source degeneration is necessary to accommodate the feedback network. In turn, pretty much regardless of what VAS is used, this also leads to limited open loop gain when compared to a stock bipolar input stage, and this in turn means less loop gain.
An amp with less loop gain sounds different to one with high loop gain. For starters, the distortion is much higher, and this discourages most people even before they start. But the sound quality can be very good if the distortion profile is musical.
Greg,
The transconductance of the verticals is around 6 times higher, leading to a loop gain which is a full 15dB higher. Switch on/off behaviour is slightly sharper too, leading to quite a bit more crossover. Your finding doesn't surprise me!
Cheers,
Hugh
I assume CFA is a TLA related to current feedback. If so, I learn something today.
I endorse Mike's comment; the idea was to make a singleton input stage. That means, inevitably, input on the gate and feedback on the source. If this is what you mean, then the choice was quite deliberate. It also means that the input stage has very low gain since source degeneration is necessary to accommodate the feedback network. In turn, pretty much regardless of what VAS is used, this also leads to limited open loop gain when compared to a stock bipolar input stage, and this in turn means less loop gain.
An amp with less loop gain sounds different to one with high loop gain. For starters, the distortion is much higher, and this discourages most people even before they start. But the sound quality can be very good if the distortion profile is musical.
Greg,
The transconductance of the verticals is around 6 times higher, leading to a loop gain which is a full 15dB higher. Switch on/off behaviour is slightly sharper too, leading to quite a bit more crossover. Your finding doesn't surprise me!
Cheers,
Hugh
Last edited:
You're welcome, I hope they serve you well. PS, you really should consider upgrading to a lawn mower or something.
(ask me if you want to know how to speed LTSpice up without too much penalty)I've been wanting to build something, and I lately re-discovered Kenpeter's gratuitous pile of stuff he sent me. I didn't really know what to do with most of it, but I'm learning gradually what all these things are and what I can use them for. I discovered some VP2206, and some 2N5464 monolithic dual Jfets! Hey, there's some good stuff here... I also happen to have the right MOSFETs...
Problem is the input Jfet and CCS. I don't think the 2N5464 is optimal for this, but it could double as the input and CCS since it's dual. Or I could use the VN0106.
Is it reasonable to build it using these parts?
- keantoken
You have to change the gatestop resistors, 500 ohm for the Pfet and 175 ohm for the N fet, I have used 220 ohm and not 175 ohm and 680 ohm and not 500 ohm, make it somewhat lazy, a little saver against oscillations, measure a blockwave with your amp, I am shure it has overshoots, that will dissapear with bigger gate resistors, also look at the rest of your design of overshoot appears, make the blockwave clean will give better sound.
You can also put potentiometers in the gate and adjust as you like, damp the speed of the verticals becasue this is the reason of harsher sound, but with my hybride the sound is full and warm with this verticals, I use driver tube and one CCS in cathode follower on 20 mA, I have to make a vbe multiplier, one with two transistors and one in diode mode to match the fet current stability.
SWF uses none in his latest schematic, so that is what I will do. In any case, this lowers the loop area between the VAS and N-MOSFET, which is a good thing. Photo of my progress attached. I don't know what I am going to do about the CCS but I will probably use a conventional bootstrap and then try upgrading.
- keantoken
- keantoken
Attachments
It's alive.
Well, now I know I shouldn't be powering my speakers with my weak little headamp! Well, I knew that anyways... Um...
I've attached the schematic I used, with the parts I had on hand.
Bias seems to have dropped since starting it. I am running at about 180mA, about 200mA before. I will make the offset control a bit smarter, using two green LEDs, and then assess offset stability.
Hey, how much bias do you think I can get away with using that heatsink?
- keantoken
Well, now I know I shouldn't be powering my speakers with my weak little headamp! Well, I knew that anyways... Um...
I've attached the schematic I used, with the parts I had on hand.
Bias seems to have dropped since starting it. I am running at about 180mA, about 200mA before. I will make the offset control a bit smarter, using two green LEDs, and then assess offset stability.
Hey, how much bias do you think I can get away with using that heatsink?
- keantoken
Attachments
Looking good Kean,
You might like to drop the input drain resistor to get a little more current through your input fet. Also, mike hated the 22uf cap across the bias spreader so it may be worth trying a smaller film cap initially. I have found that it sounds better in some configurations than it does in others, so perhaps you should leave it out and try adding it later.
I recommend changing to a ccs when you can, perhaps a led/bjt type. The bootstrap version will have very nice vocals but if you want some seriously slammy bass than a ccs is the way to go. Guess it depends on what you listen to.
I will be interesting so see how you like it with the MOSFET front end- should be quite good.
And, yes, a nice clean bias arrangement using LEDs for the front is a good idea.
I have not yet tried a rail filter between the vas and output. Mike assures me it sounds better with the noise reduction it provides, however I am waiting on the pcbs to try it as I have temporarily gotten sick of making rats nests! He has mentioned here that he is using 20R and 470uF.
Look forward to hearing your listening impressions!
You might like to drop the input drain resistor to get a little more current through your input fet. Also, mike hated the 22uf cap across the bias spreader so it may be worth trying a smaller film cap initially. I have found that it sounds better in some configurations than it does in others, so perhaps you should leave it out and try adding it later.
I recommend changing to a ccs when you can, perhaps a led/bjt type. The bootstrap version will have very nice vocals but if you want some seriously slammy bass than a ccs is the way to go. Guess it depends on what you listen to.
I will be interesting so see how you like it with the MOSFET front end- should be quite good.
And, yes, a nice clean bias arrangement using LEDs for the front is a good idea.
I have not yet tried a rail filter between the vas and output. Mike assures me it sounds better with the noise reduction it provides, however I am waiting on the pcbs to try it as I have temporarily gotten sick of making rats nests! He has mentioned here that he is using 20R and 470uF.
Look forward to hearing your listening impressions!
Kean,
As an addition to my last post:
Up the bias to half an amp and if you can still comfortably hold the heatsink in an hour you're probably fine still, you're running quite a bit of bias anyway. I actually like it with 1A, but it starts getting impractical. What rail voltage are you using?
Also I would try to run that input fet at whatever current you need to get at least 200mW through it. It's a 1W device I believe.
Can you explain your earlier comment about the vas/output loop area in more detail and the consequences it may have?
Thanks, and good work. Nice to put all those simulations to good use.
As an addition to my last post:
Up the bias to half an amp and if you can still comfortably hold the heatsink in an hour you're probably fine
still, you're running quite a bit of bias anyway. I actually like it with 1A, but it starts getting impractical. What rail voltage are you using?Also I would try to run that input fet at whatever current you need to get at least 200mW through it. It's a 1W device I believe.
Can you explain your earlier comment about the vas/output loop area in more detail and the consequences it may have?
Thanks, and good work. Nice to put all those simulations to good use.
If you are considering filtering the frontend you might look at my "K-multiplier" for ideas. It is essentially a CFP capacitance multiplier, and will be much more effective for a small parts count. Filtering is about 50-60db, and can extend way into the low Hz using a small capacitor.
Increasing the input Jfet current would make things worse, I already had to use 3 green LEDs just to conquer the 4V offset. Even then, I estimate the amp kept drifting about 100mV before it stabilized.
Hey, has anyone ever thought about a servo that uses thermistors? Sort of like an LDR except without the fast response...
Alright, I will up the bias. There seems to be some odd kind of resonance, maybe it's my 2200u filter caps resonating with my 12000u ones right by the rectifiers... And also a rather weird kind of harshness in the lower treble..? But I can't say how much of this I can attribute to my speaker system, and I may have adjusted my other amp to help with the shortcomings of my speakers. To try and make an unbiased description of the sound, I will compare it to a class A single-device IRF250 buffer I made one day (run at 800mA bias), no feedback. Not today though.
BJT output stages run at around 50-150mA as I understand, so it makes sense to me that I would want to run MOSFETs at 200mA or above to exploit their benefits. I have not really thought about this.
In the MOSFET stability articles which were posted several pages back in this thread, it was said that the best layout/circuit for driving them will have a small loop area through which the capacitive currents will flow. The driver of the MOSFET will create a current loop which moves through the gate, all the way around back into the drain. In the VAS of your circuit, this loop is closed pretty small because the source of the VAS is right at the N-channel drain. If you insert a rail filter here you break this loop, and for the current to run through two filter caps to get to the N-channel drain. With large gate stoppers and well-managed parasitics the danger is slim even then, however my experience is that amps tend to sound better when they need less compensation, and more stable output stage will help with this. Even so, it may be moot. Just have to try it I guess...
I should also mention I forgot to add the 47R VAS degeneration (couldn't find the resistor), and I suspect that is an important part! However this will increase the offset voltage, and offset for this amplifier already isn't pretty. I'm considering switching to the 2N5464 Jfet, because the MOSFET part has naturally higher Vgs. But I will do this one thing at a time.
- keantoken
Increasing the input Jfet current would make things worse, I already had to use 3 green LEDs just to conquer the 4V offset. Even then, I estimate the amp kept drifting about 100mV before it stabilized.
Hey, has anyone ever thought about a servo that uses thermistors? Sort of like an LDR except without the fast response...
Alright, I will up the bias. There seems to be some odd kind of resonance, maybe it's my 2200u filter caps resonating with my 12000u ones right by the rectifiers... And also a rather weird kind of harshness in the lower treble..? But I can't say how much of this I can attribute to my speaker system, and I may have adjusted my other amp to help with the shortcomings of my speakers. To try and make an unbiased description of the sound, I will compare it to a class A single-device IRF250 buffer I made one day (run at 800mA bias), no feedback. Not today though.
BJT output stages run at around 50-150mA as I understand, so it makes sense to me that I would want to run MOSFETs at 200mA or above to exploit their benefits. I have not really thought about this.
In the MOSFET stability articles which were posted several pages back in this thread, it was said that the best layout/circuit for driving them will have a small loop area through which the capacitive currents will flow. The driver of the MOSFET will create a current loop which moves through the gate, all the way around back into the drain. In the VAS of your circuit, this loop is closed pretty small because the source of the VAS is right at the N-channel drain. If you insert a rail filter here you break this loop, and for the current to run through two filter caps to get to the N-channel drain. With large gate stoppers and well-managed parasitics the danger is slim even then, however my experience is that amps tend to sound better when they need less compensation, and more stable output stage will help with this. Even so, it may be moot. Just have to try it I guess...
I should also mention I forgot to add the 47R VAS degeneration (couldn't find the resistor), and I suspect that is an important part! However this will increase the offset voltage, and offset for this amplifier already isn't pretty. I'm considering switching to the 2N5464 Jfet, because the MOSFET part has naturally higher Vgs. But I will do this one thing at a time.
- keantoken
I find this file to be good for testing bass and listening for amplifier glitches:
ReMix: The Legend of Zelda 'Zelda Heineken' - OverClocked ReMix
Since I increased the bias to 500mA some of the harshness seems to be gone. Then again it could just be my ears, since I haven't slept for a day (hmm...). Then again the sound is more enjoyable, and I have found that perceived harshness decreases with enjoyment factor; it is easier to accept the harshness as part of the music. In any case, going beyond class A into AB creates a dramatic spray of harmonics, and this may have something to do with it.
- keantoken
ReMix: The Legend of Zelda 'Zelda Heineken' - OverClocked ReMix
Since I increased the bias to 500mA some of the harshness seems to be gone. Then again it could just be my ears, since I haven't slept for a day (hmm...). Then again the sound is more enjoyable, and I have found that perceived harshness decreases with enjoyment factor; it is easier to accept the harshness as part of the music. In any case, going beyond class A into AB creates a dramatic spray of harmonics, and this may have something to do with it.
- keantoken
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.