Hugh & Keantoken,
thx for your comments & questions.
I am very pleased with the sound of this amp, I am gradually step by step applying measures to make the sound as good as I can get it. Without exception so far all these measures involve reducing noise in one way or another.
My outlook about this process is extremely simple - I think about three elements that combine to produce the "sound" of an amp.
1) Harmonic distortion ( 2nd, 3rd and perhaps 4th & 5th )
2) Dynamics
3) Noise ( HD from 6th upwards and everything else that isn't signal )
In this design HD distortion can easily be low enough and have a signature for a good sound.
Also dynamics can easily be excellent because so little compensation is needed and perhaps also because the o/p mosfets can switch current so effortlessly.
This means all I have to work on is noise and there are plenty of opportunities to reduce this.
Mostly reducing power supply noise but also heavy o/p biasing so the first few watts operate in class A and using quality components ( also avoiding caps in the signal path & o/p chokes 😉 )
I'm currently building a full choke regulated PSU - The circuit just like in Hughs PSU except the resistors are replaced my 3mH chokes and the second cap more like 32,000uF - not tested yet but should sound excellent
Warning: choke regulation can cause ringing & go considerably over voltage during turn on if not designed properly so beware ! Choke regulation needs very careful design.
One last tip - I have found that EI transformers sound much better that torroids.
Haven't addressed DC drift yet - unlike almost everyone else it does not really bother me that much but I will have a look at it soon.
thx for your comments & questions.
I am very pleased with the sound of this amp, I am gradually step by step applying measures to make the sound as good as I can get it. Without exception so far all these measures involve reducing noise in one way or another.
My outlook about this process is extremely simple - I think about three elements that combine to produce the "sound" of an amp.
1) Harmonic distortion ( 2nd, 3rd and perhaps 4th & 5th )
2) Dynamics
3) Noise ( HD from 6th upwards and everything else that isn't signal )
In this design HD distortion can easily be low enough and have a signature for a good sound.
Also dynamics can easily be excellent because so little compensation is needed and perhaps also because the o/p mosfets can switch current so effortlessly.
This means all I have to work on is noise and there are plenty of opportunities to reduce this.
Mostly reducing power supply noise but also heavy o/p biasing so the first few watts operate in class A and using quality components ( also avoiding caps in the signal path & o/p chokes 😉 )
I'm currently building a full choke regulated PSU - The circuit just like in Hughs PSU except the resistors are replaced my 3mH chokes and the second cap more like 32,000uF - not tested yet but should sound excellent
Warning: choke regulation can cause ringing & go considerably over voltage during turn on if not designed properly so beware ! Choke regulation needs very careful design.
One last tip - I have found that EI transformers sound much better that torroids.
Haven't addressed DC drift yet - unlike almost everyone else it does not really bother me that much but I will have a look at it soon.
Last edited:
Hello KT
I've found a page with some free Systron-Donner service manual, I have low speed internet, so I did not dowload them but two of them are maby similar to your Systron-Donner Datapulse signal generator model, maby you can dowload them and check if you can use one to find some leads about the problems of yours. Or you can ask the guy of this web site about your model manual or schematic.
Systron Donner
Bye
Gaetan
Last edited:
Hello
I think it's because that a EI transformers let pass throught less rf noise from the line than a torroids transformers.
Bye
Gaetan
Last edited:
Hello
To cut down noises, I alway use an rf filter at the line side of my power supply and a Pi filter after the rectifier bridge and also another Pi filter for the rails supply of the vas and input section of my amps.
Bye
Gaetan
To cut down noises, I alway use an rf filter at the line side of my power supply and a Pi filter after the rectifier bridge and also another Pi filter for the rails supply of the vas and input section of my amps.
Bye
Gaetan
Last edited:
Is this because EI cores are lossy at RF or something else? And what rail filter are you using? A simple RC?
Thanks for the link Gaetan, interesting but it doesn't look much like what I have.
I heard that soft-recovery diodes were best for low-noise supplies. Is this accurate?
I'm using this rectifier, it was laying around with convenient screw-terminals:
http://www.alliedelec.com/Images/Products/Datasheets/BM/SEMIKRON_INC/550-0760.PDF
😀
- keantoken
Thanks for the link Gaetan, interesting but it doesn't look much like what I have.
I heard that soft-recovery diodes were best for low-noise supplies. Is this accurate?
I'm using this rectifier, it was laying around with convenient screw-terminals:
http://www.alliedelec.com/Images/Products/Datasheets/BM/SEMIKRON_INC/550-0760.PDF
😀
- keantoken
Last edited:
Hello KT
About you signal generator, check all plugs, contacts and switchs, those old gears get some oxidations of plugs, contacts and switchs after so much years, or years of switch uses do some carbon deposits on the switchs contacts.
Yes, sort of, and toroid transfo do have a much wider frequency response, it's good for tube amp output transfo but bad for rf noises in a power supply.
There was a thread about soft-recovery diodes, here's the link:
Soft recovery diodes
Bye
Gaetan
Last edited:
Actually Juma, and I may be wrong, I think I can generalise about this, and I think my findings are not too dependent on parts selection, though I admit to only trying a limited range of parts.
I think a big thing needed to get this amp to perform is careful juggling of the input and vas gain, and you need a complementary action between the two.
The MOSFET vas has a very different OLG character to the BJT at low frequencies especially (look for mike's simulation results) and I think the low gain jfet has too much trouble driving a bjt for this purpose.
I will post more about this over the coming days but in the mean time feel free to ignore my previous post and draw your own conclusions.
Oh, btw, bjt input + bjt vas is pretty good too!
They have more capacitance between their windings because of how they are constructed, although this is offset in better toroids by having an earthed screen between primary and secondary.
Mike,
Today has been a big day of testing your rail filter between the vas and output. As you know I left space for it on my PCBs and I implemented it as you suggested, 20R and 470uF. I was able to solder a jumper in place to allow fairly rapid switches in and out. I really had high hopes for this mod and actually have been listening with it in place for the last two days. It seemed like better performance was a given and made good sense, even if all it did was remove ps ripple.
Today is the first time I have tried it without the filter and now I am ready draw some conclusions.
First, the reduction in hiss with no input is noticeable. I'm going to be bold and say this is a noise reduction.
However, I significantly prefer the sound without the filter, which is surprising and disappointing! I know I say this with numerous mods I have tried, but it seems to remove some of the foot tapping factor, as does phase lead comp (for me).
Currently I prefer the amp with no compensation of any kind, and no rail filter! This is despite some pretty funky square wave performance into capacitive loads (though no oscillation).
It is possible that I have implemented this differently to you or perhaps my layout is poor, but overall I'm tending to shy away from this mod completely. Maybe now we're just getting into the nitty gritty of our personal tastes?
Today has been a big day of testing your rail filter between the vas and output. As you know I left space for it on my PCBs and I implemented it as you suggested, 20R and 470uF. I was able to solder a jumper in place to allow fairly rapid switches in and out. I really had high hopes for this mod and actually have been listening with it in place for the last two days. It seemed like better performance was a given and made good sense, even if all it did was remove ps ripple.
Today is the first time I have tried it without the filter and now I am ready draw some conclusions.
First, the reduction in hiss with no input is noticeable. I'm going to be bold and say this is a noise reduction.
However, I significantly prefer the sound without the filter, which is surprising and disappointing! I know I say this with numerous mods I have tried, but it seems to remove some of the foot tapping factor, as does phase lead comp (for me).
Currently I prefer the amp with no compensation of any kind, and no rail filter! This is despite some pretty funky square wave performance into capacitive loads (though no oscillation).
It is possible that I have implemented this differently to you or perhaps my layout is poor, but overall I'm tending to shy away from this mod completely. Maybe now we're just getting into the nitty gritty of our personal tastes?
Hey guys, good news. I realized after this thread that ultrasonic noise was probably much more of a problem than I think. As an experiment, I mounted a 1uF MEF (I'm assuming mylar) between the terminals of my rectifier. Switching noise from the diodes was reduced only slightly, but (d'oh!) it had a nice secondary effect of eliminating spurious oscillation in the amplifier. I saw this odd behavior on the scope but did not know what to make of it, though I suspected it was oscillation. Dunno why I keep forgetting film rail bypass as a possible cure for oscillation.
I've seen this spurious type of oscillation with more than one circuit, and sometimes it coincides with ripple (though not this time), so I figure it is sometimes triggered by rectifier glitches.
The midrange grunge seems to be gone, with a darker background, and slightly more subtlety.
I also realized something: noise can also be caused by the amplifier oscillating, especially in the case of spurious oscillation.
- keantoken
I've seen this spurious type of oscillation with more than one circuit, and sometimes it coincides with ripple (though not this time), so I figure it is sometimes triggered by rectifier glitches.
The midrange grunge seems to be gone, with a darker background, and slightly more subtlety.
I also realized something: noise can also be caused by the amplifier oscillating, especially in the case of spurious oscillation.
- keantoken
Hello SWF. I found despite the seeming increase in clarity, the rail filter did decrease "foot tapping". However now I have stopped the oscillation, I will have to go back and check. Just another thing on the list I have to do when the new tip arrives.
Technically speaking, the signal from the rails causes positive feedback. In simulation the MOSFET seemed to suffer less from this effect, but this observation is at the mercy of my models.
Your statements about Jfets and MOSFETs may be too general, if you consider that the Gm of these can vary significantly between parts, I think this has the potential to affect the harmonic profile to a significant degree. For instance, in my experience with the simulation models for the VP0106 and VP2206, the VP0106 undoubtedly has lower capacitances, but the VP2206 seems to have higher Gm which actually makes it faster, and also increases DC gain.
- keantoken
Technically speaking, the signal from the rails causes positive feedback. In simulation the MOSFET seemed to suffer less from this effect, but this observation is at the mercy of my models.
Your statements about Jfets and MOSFETs may be too general, if you consider that the Gm of these can vary significantly between parts, I think this has the potential to affect the harmonic profile to a significant degree. For instance, in my experience with the simulation models for the VP0106 and VP2206, the VP0106 undoubtedly has lower capacitances, but the VP2206 seems to have higher Gm which actually makes it faster, and also increases DC gain.
- keantoken
I already know, and I knew before this thread started - it's not about me.
It's about guys who are not professionally involved in electronics and that's most of the DIYers. They might be mislead by incomplete deductions.
As Hugh said, all of the mentioned input+VAS combinations can be made to sound good (or bad) if a guy knows exactly what he is doing, and I think that is the real point in DIY - to learn how to use different components and topologies according to one's design goals.
Really the only thing that separates Jfet from bipolars is the transfer function (EDIT: okay not the only thing, but it's important anyways). Bipolars are exponential, Fets are square law I think? Capacitances can be managed for both, if you have a good selection.
The interesting thing about bipolars is that by virtue of their transfer function they are scalable in operating point. For a bipolar, If you give it a current bias and modulate that current by say 10%, the distortion will be roughly the same no matter what the bias current, as long as the signal is 10% of the bias and Ib and power distortions are avoided.
For FETs, this relationship does not hold true. With the same test setup, they will always have less distortion with higher current bias.
I hope I'm right about this, I'm pretty sure... Gives something to think about considering the "sound" of both.
- keantoken
The interesting thing about bipolars is that by virtue of their transfer function they are scalable in operating point. For a bipolar, If you give it a current bias and modulate that current by say 10%, the distortion will be roughly the same no matter what the bias current, as long as the signal is 10% of the bias and Ib and power distortions are avoided.
For FETs, this relationship does not hold true. With the same test setup, they will always have less distortion with higher current bias.
I hope I'm right about this, I'm pretty sure... Gives something to think about considering the "sound" of both.
- keantoken
Thanks Kean, glad we agree on the filter affect. Hugh has also noticed this and attributes it to the vas and output devices not having the same voltage reference.
As to the effect of fets, bjts, etc, hopefully Mike can dig up his simulation that shows the vastly improved OLG of the fet in the vas at low frequencies.
Btw, you are right about the different gains of different fets, but I would say that on the whole MOSFETs have at least twice the gain of the best audio jfets.
Juma,
I apologise for being vague but surely my statements can't be any more misleading than the discussions on cables, enclosure materials and what not?
As to the effect of fets, bjts, etc, hopefully Mike can dig up his simulation that shows the vastly improved OLG of the fet in the vas at low frequencies.
Btw, you are right about the different gains of different fets, but I would say that on the whole MOSFETs have at least twice the gain of the best audio jfets.
Juma,
I apologise for being vague but surely my statements can't be any more misleading than the discussions on cables, enclosure materials and what not?
No need for that buddy, you did a great job here
It's only human to get carried away by success 😀
Greg,
Yes, I realised many years ago when going around listening to different friends systems that not only did they all sound very different but also from talking with them I realised that we were not even aiming in the same direction.
So I'm guessing now if we ever got around to comparing our favourite version of this amp they could indeed be very different ! It would be so interesting.
Also I have been thinking that as the DC & AC version progress if would be quite natural that they took different directions.
cheers
mike
Yes, I realised many years ago when going around listening to different friends systems that not only did they all sound very different but also from talking with them I realised that we were not even aiming in the same direction.
So I'm guessing now if we ever got around to comparing our favourite version of this amp they could indeed be very different ! It would be so interesting.
Also I have been thinking that as the DC & AC version progress if would be quite natural that they took different directions.
cheers
mike
Juma, why is it so important? Those who would take discussion out of context have their own lessons to learn, will they really get it "right" even if we bend over backwards for them?
Perhaps more important is to keep to ourselves and worry less about what others think, and then a good student will examine the context and find a useful reference.
- keantoken
Perhaps more important is to keep to ourselves and worry less about what others think, and then a good student will examine the context and find a useful reference.
- keantoken
- Status
- Not open for further replies.