Thanks Kean, and thanks to Bob as well.
Will do some comparisons to see the differences.
Well, i dont have doubts about Bob s abilities..
As for the amp being totally different , i wouldnt go as far since
this is a very little variation , but for sure that linearity wise, yes,
it s totally different.
Indeed, IF a darlington is used as VAS , one will conclude that the
apparent component count is still the same !!
Unfortunately, there are not darlingtons that have as good
caracteristics as a BC546+2SA1360.
Overall, i think that the use of a JFET at the input and mosfets
at the output, with the Bjts providing very high gain between these
two stages , is the most logical way to go to achieve performances
that are good enough to make the circuit preferable to an IC amp.
Hi swordfishy,
few months ago I completed a topologically similar amp (main difference is that I used the single rail PSU) so it may be of some use to you and the others. I also tried different types of popular FETs, MOSFETs and BJTs. In the end, this is what sounded best to me :
http://www.diyaudio.com/forums/pass-labs/183482-25w-cl-amp-lateral-mosfets.html#post2474936
few months ago I completed a topologically similar amp (main difference is that I used the single rail PSU) so it may be of some use to you and the others. I also tried different types of popular FETs, MOSFETs and BJTs. In the end, this is what sounded best to me :
http://www.diyaudio.com/forums/pass-labs/183482-25w-cl-amp-lateral-mosfets.html#post2474936
Juma,
Very very nice. Great square wave there too.
Your timing with this post is eerie - Hugh and I were just discussing the benefits of a bjt vas and input. I am actually getting tempted to go this way too. I did actually try it a week or two back, and thought the bjt input and vas was a little more sterile sounding. But I did not experiment too much with it and maybe could tweak some warmth into it.
Can I ask, why would one choose to, or choose not to use a inverted or non inverted configuration? It seems to me it would be preferable to use the more linear NPN device in the vas, but the input isn't exactly non important either.
AndrewT,
Can't see a problem with the inverted version. Please do and let us know how it goes.
Keantoken,
Can't see any signs of a 4khz dip, but in the mean time I've temporarily removed the decoupling capacitors. By the way, my oscillations are occurring at many hundreds of kHz, at least.
Very very nice. Great square wave there too.
Your timing with this post is eerie - Hugh and I were just discussing the benefits of a bjt vas and input. I am actually getting tempted to go this way too. I did actually try it a week or two back, and thought the bjt input and vas was a little more sterile sounding. But I did not experiment too much with it and maybe could tweak some warmth into it.
Can I ask, why would one choose to, or choose not to use a inverted or non inverted configuration? It seems to me it would be preferable to use the more linear NPN device in the vas, but the input isn't exactly non important either.
AndrewT,
Can't see a problem with the inverted version. Please do and let us know how it goes.
Keantoken,
Can't see any signs of a 4khz dip, but in the mean time I've temporarily removed the decoupling capacitors. By the way, my oscillations are occurring at many hundreds of kHz, at least.
Sorry but it's not completely clear to me what you refer to when using terms inverted/non-inverted...
Through the amp, the signal is inverted twice (at Q1 and Q2 collectors) so the whole amp is of non-inverted type (output signal is in phase with the input signal).VAS is NPN BJT (Q2=BC550c) and input (Q1=BC560c) is strongly degenerated and without contribution to OLG (its task being to provide high and linear Zin and Low Z NGFB point).
Last edited:
Juma,
Sorry for not being clear.
What I meant was what a the pros/cons of using a npn input and pnp vas versus a pnp input and npn vas?
As I mentioned in my post, and mikelm too, i assume it is preferable to use the more linear npn device in the vas where most of the voltage amplification takes place (as you have done).
This is the opposite to what the current design of fetzilla uses.
Sorry for not being clear.
What I meant was what a the pros/cons of using a npn input and pnp vas versus a pnp input and npn vas?
As I mentioned in my post, and mikelm too, i assume it is preferable to use the more linear npn device in the vas where most of the voltage amplification takes place (as you have done).
This is the opposite to what the current design of fetzilla uses.
Last edited:
The downside is that it forces the adoption of 2sj74 for the input device if we retain the jFET as the input device.
I am lucky, I have a stock of all of the required devices.
I've sen many circuits that primarily use an N-Jfet input transistor. Ultimately I think it is a decision of which transistor needs the N-silicon more, input or VAS.
SWF, more filtering should be better. A separate supply or heavy filtering for the frontend is better. I thought at first the filter caps were resonating with something but it turns out that is not the case. In fact, if there is a problem it's that there is not enough filtering. Giving the frontend a separate supply or filtering from a capacitance multiplier will go a great way towards lowering the distortion of the amplifier without increasing OLG or adding more active devices in the signal path. It will also decrease the contribution of the filter caps to the output signal, as I mentioned before. I have attached a simulation showing THD and FFT with separate rails (with stability compensation removed).
Although the harmonic profile looks less appealing, the 98% criteria is still fulfilled. Regulating the frontend removes a lot of the input stage distortion, which reveals the distortion of the output stage, which is primarily odd-harmonics but still has little high-order content.
EDIT: Attachment attached!
- keantoken
SWF, more filtering should be better. A separate supply or heavy filtering for the frontend is better. I thought at first the filter caps were resonating with something but it turns out that is not the case. In fact, if there is a problem it's that there is not enough filtering. Giving the frontend a separate supply or filtering from a capacitance multiplier will go a great way towards lowering the distortion of the amplifier without increasing OLG or adding more active devices in the signal path. It will also decrease the contribution of the filter caps to the output signal, as I mentioned before. I have attached a simulation showing THD and FFT with separate rails (with stability compensation removed).
Although the harmonic profile looks less appealing, the 98% criteria is still fulfilled. Regulating the frontend removes a lot of the input stage distortion, which reveals the distortion of the output stage, which is primarily odd-harmonics but still has little high-order content.
EDIT: Attachment attached!
- keantoken
Attachments
Wait a sec, I made a mistake. The distortion benefit doesn't occur at 7KHz, because the vast majority of input Jfet loading is from the capacitance of the VAS and output MOSFETs. 7KHz distortion benefits little from filtering. It is distortion below 4KHz that benefits from filtering, because this is when low PSRR begins to outweigh capacitive loading as the main source of distortion.
500Hz distortion is drastically reduced. Harmonic profile is the same as in the above photo, but distortion and output impedance are both reduced by about a factor of 10.
- keantoken
500Hz distortion is drastically reduced. Harmonic profile is the same as in the above photo, but distortion and output impedance are both reduced by about a factor of 10.
- keantoken
I don't see that NPN BC550C is more linear than PNP BC560C. Maybe you mean N-ch MOSFETs versus P-ch ?
Yes, transfer characteristic shows that N-ch (IRF610) seems more linear at small currents. ZVN3310 look even better, but to my ears the BC550c takes the cake...
Keantoken,
those are valid facts but you have to look at them in perspective:
Early effect (AKA base-width modulation) is when the voltages applied to the base-emitter and base-collector junctions are changed, the depletion layer widths and the quasi-neutral regions vary as well. This causes the collector current to vary with the collector-emitter voltage.
So how big is the Vce variation of Q1 (BC560) ? We are talking tenths of mVs here - negligible...
Also, the hfe linearity (i.e. its dependance of Ic) is non-issue because Q1's Ic is pretty much constant at 1mA.
swordfishy,
Just a thought.
Thinking about your comments and spicing what you have stated as your favourite versions you seems to like very high feedback designs but it is worth bearing in mind that this extra feedback has two different effects. One is to reduce distortion but the other is to increase power supply noise rejection ratio.
If you have a noisy power supply you are very likely always going to choose the higher feedback designs because lower noise almost always sounds better, but if you have a whisper quiet PSU arrangement already, it may be that you choose differently regarding the feedback levels.
mike
Just a thought.
Thinking about your comments and spicing what you have stated as your favourite versions you seems to like very high feedback designs but it is worth bearing in mind that this extra feedback has two different effects. One is to reduce distortion but the other is to increase power supply noise rejection ratio.
If you have a noisy power supply you are very likely always going to choose the higher feedback designs because lower noise almost always sounds better, but if you have a whisper quiet PSU arrangement already, it may be that you choose differently regarding the feedback levels.
mike
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.