whatever you choose - poison is yours
even if I can't agree that one tiny Jfet (or bjt , with same Iq ) is able to properly drive that brick on output
while yuo at it , you can try something in line with this - I didn't calculate source resistors for buffer stage , but I had on mind (as I already wrote ) 10 times greater current than in input stage
Zen,
Concerning your buffer, I hope the following reasoning is correct
From the +Ve rail to the gate of Q3 the voltage drop should be 3.6v = 1.3 A x 0R47 + Vgs(on).
However, on my amp I'm measuring 4.5v drop from the +Ve rail to the gate of Q3.
If a source resistor for your buffer is 51R ohms - that would set I = 88 mA = 4.5v / 51R.
Hope that's enough current.
When I saw your buffer - I rushed out to buy some MOSFET's.
However, all there was were IRF520's and IRF9540's - hope that's OK.
A/B amp comparison - Here is my test set up.
H/K 7600 CDP as a transport ==> modded Xindak DAC-5 ==> Roksan Caspian ==> F5 ==> 10" DC Tannoy.
all plugged into a PS Audio P1000 watt regenerator.
This is the original version of the Caspian Integrated and can be used a preamp or integrated amp.
Its also a MOSFET amp and this is a well respected integrated.
In the A/B comparison, the F5 bettered the Caspian's amp section in every way imaginable.
The F5 has more bass and dynamics.
With the F5, a Piano has an inciting quality that the amp inside the Caspian simply does not have.
The only way to tell if the F5 is on - is when music starts.
All this being said - from memory - the Aleph I modded bettered this version of the F5.
Think I have been spoiled by the Aleph.
.
Uunderhill,
Have you tried cascoding?
jacco says the mosfets you've got are similar to driving 9 pairs of regular ones, Zen says cascode, you pick your poison as far as using BJT or MOSFET for the cascode goes, NP says in the F5 turbo that cascoding is needed to drive additional pairs.
The solution has cascode written all over it.
Have you tried cascoding?
jacco says the mosfets you've got are similar to driving 9 pairs of regular ones, Zen says cascode, you pick your poison as far as using BJT or MOSFET for the cascode goes, NP says in the F5 turbo that cascoding is needed to drive additional pairs.
The solution has cascode written all over it.
cascoding of input Jfets will not increase input stage's drive capability ;
cascodes have two major roles - one is to serve as voltage umbrella for tiny Jfarts , second one is to further decrease capacitance issues (no more Herr Mueller in input stage)
so - for better drive of output bricks , one need either paralleling (even multiple) of input Jfarts (then cascoding blahblah Mueller becomes handy) ....... or introducing buffering interstage
another solution is to use small outputs (say 510/9510) with 75-100mA Iq , then connecting bricks as standard (final ) source follower stage
pick your poison
cascodes have two major roles - one is to serve as voltage umbrella for tiny Jfarts , second one is to further decrease capacitance issues (no more Herr Mueller in input stage)
so - for better drive of output bricks , one need either paralleling (even multiple) of input Jfarts (then cascoding blahblah Mueller becomes handy) ....... or introducing buffering interstage
another solution is to use small outputs (say 510/9510) with 75-100mA Iq , then connecting bricks as standard (final ) source follower stage
pick your poison
Big E,
Thanks for your response ... I'm still learning here and have never looked at cascode voltage amps.
However, on F5 Turbo V3, R25 and R27 create a voltage divider and then there is a Vbe drop.
This creates simple voltage regulation for Q1.
With a Cascode, its true that the BJT will lower the voltage drop across drain to source of Q1 - which will lower the power dissipation.
However using Kirchoff - its clear the current flowing though the BJT and Q1 is the same.
So I think Cascoding will help with voltage regulation and lower the power dissipation of Q1,
but it won't help with the current.
The maximum's for those jFET's (in a TO-92 case) is 20mA and 400mW - but I would Not go beyond 1/2 of either of those values.
Right now I have 3 jFET's in parallel with 18mA flowing through the 3 of them.
A bit nervous about increasing the 18mA.
Also, note with Turbo V3 - a Cap is intentionally connected between the gate and drain of the power MOSFET's.
But the suggested value to add is only 1nF.
Ciss ~ 12.6 nF for the "bricks."
Regardless, I would like to tinker with a Cascode stage if there is time.
.
nuff said
for bricks , you need driving capability - relatively low impedance of preceding stage ..... which only current potent stage can give
so ...... no need of tinkerin' with cascodes , if staying with one complementary pair of input Jfets , without altering PSU voltage
Zen Mod,
I tried your follower buffer.
Got a bit nervous and used 100 R for the source resistors.
When the amp was switched on - BOTH voltage rails dropped from 25 v to 21.6 v - yikes !!!
I have 6A fast blow fuses connected between the power supply and the amp
and they didn't cook ???????
Fast blow fuses should have cooked in 30ms.
I put the circuit back the way it was.
To lower the RC value of the gate - the gate resistors were lowered from 51 R to 22 R.
I couldn't see any singing (oscillation) on the scope and IMO the amp sounded softer.
The F5 power supply shows a 3.3nF cap across the hydro.
But these caps need to be rated for hydro use - a 400 V MKP is Not rated for hydro.
Hydro rated caps are difficult to find, so instead I added 0.1uF 250 V across both secondary windings.
I think with this addition, the amp sounded softer.
As it stands right now, when this F5 warms up, it sounds pretty natural and relaxed.
.
I tried your follower buffer.
Got a bit nervous and used 100 R for the source resistors.
When the amp was switched on - BOTH voltage rails dropped from 25 v to 21.6 v - yikes !!!
I have 6A fast blow fuses connected between the power supply and the amp
and they didn't cook ???????
Fast blow fuses should have cooked in 30ms.
I put the circuit back the way it was.
To lower the RC value of the gate - the gate resistors were lowered from 51 R to 22 R.
I couldn't see any singing (oscillation) on the scope and IMO the amp sounded softer.
The F5 power supply shows a 3.3nF cap across the hydro.
But these caps need to be rated for hydro use - a 400 V MKP is Not rated for hydro.
Hydro rated caps are difficult to find, so instead I added 0.1uF 250 V across both secondary windings.
I think with this addition, the amp sounded softer.
As it stands right now, when this F5 warms up, it sounds pretty natural and relaxed.
.
Yes, the maximum voltage drop across most small signal transistors is 40 V.
To solve the problem of 70 V rails, the BJT and Q1 jFET divide up the voltage.
But I do wonder how much better the amp would sound with a Cascode voltage stage with + / - 25 v rails.
.
Last edited:
poison have been mentioned so many times now, it just makes me wonder if it hasnt already been introduced from the very beginning
so why worry about it
hey, btw, they use snake poison to cure snake bites
poison have been mentioned so many times now, it just makes me wonder if it hasnt already been introduced from the very beginning
so why worry about it
hey, btw, they use snake poison to cure snake bites
The name of this website is DIY Audio - its not IEEE or JEDEC Audio.
I thought this website was for hobbyists to share knowledge and experiences.
Thanks to the help of a few people here,
I now have an amp that easily sounds better than the amp section in a Roksan Caspian Integrated.
Turns out N.P. himself said "... jFETs in sets of 4 parallel, and 30 pairs of output devices, we should be able to get ... "
I didn't know it until recently, but N.P. solved a large Ciss problem by paralleling jFET's
There is nothing that will suppress creatively faster than negativity,
So for a moderator to make a comment like this, I have to ask, are you involved in British or European banking system ?
.
Last edited:
Always a great idea to look at other designs - and take their best ideas.
But I have to ask - why did he mount the electrolytics so close to the heat sinks ?
.
Hi,
I'm very interesting your experimental on IXTN power MOSFET; can you give out some more review of the sound in comparision to the original IRF244?
THANKS
I have not heard an F5 with IRF 244's - so I can't comment.
Gamut amps are known for their sonics and they may use an IXTN MOSFET's.
I'm not sure if IFR makes a p channel version.
Here is a discussion about cloning a Gamut amp, on DIY Audio, and there is quite a difference of opinion.
http://www.diyaudio.com/forums/solid-state/66334-gamut-d200-clone.html
I prototyped an IXTN180n10n in a Aleph style amp with a single 2N5087 or a j177 jFET driving them.
The sonics were jaw dropping.
This version of the Aleph may have sounded so great because a current source was used in the output stage,
which improves the linearity of the MOSFET.
However, if I recall correctly, a 30 watt Aleph is biased at 3 amps, so it dissipates 300 Watts just sitting there.
This is beyond reason - so I did not build an Aleph.
Other people on this website have much more experience
and prefer MOSFET's with more linearity such as paralleling 2sk2013 and 2sj313's.
I would go with their experience and consider using these MOSFET's in your F5.
.
Last edited:
Listening to different types of resistors clearly indicates that metal oxide and metal film resistors sound miles
better than carbon resistors.
I didn't understand why because of the classic text book equation for thermal noise caused by resistors
THERMAL NOISE
V thermal = SRQT (4 * k * T * B *R)
where k = Boltzmann's constant
T is in kelvin
B is the bandwidth
R is resistance in ohms
However, resistors also have Contact Noise and Shot Noise
CONTACT NOISE
V contact = (K * I) / ( f * physical size)
where K is a constant based on the type of material a resistor is made from
This equation says - a 2 W resistor will create less noise than a 1/4 Watt - all else being equal.
Turns out the predominant noise in a carbon, metal oxide or metal film resistor is the contact noise.
This also, says wire wound resistors do not have contact noise.
However, its been my experience that most commercial grade wire would resistors sound like hell.
Mills wire wound resistors sound OK - I haven't heard the double wound Mundorf's
SHOT NOISE
V shot = K * I
where K is another constant.
Looking at these equations - it seems the 100K input resistor on the F5 is quite important
With the F5 I've used $0.20 metal film resistors and some metal oxide resistors - and they sound miles better than carbon.
Here are 2 references.
http://www.aikenamps.com/ResistorNoise.htm
http://www.eetimes.com/design/autom...egies-for-minimizing-resistor-generated-noise
better than carbon resistors.
I didn't understand why because of the classic text book equation for thermal noise caused by resistors
THERMAL NOISE
V thermal = SRQT (4 * k * T * B *R)
where k = Boltzmann's constant
T is in kelvin
B is the bandwidth
R is resistance in ohms
However, resistors also have Contact Noise and Shot Noise
CONTACT NOISE
V contact = (K * I) / ( f * physical size)
where K is a constant based on the type of material a resistor is made from
This equation says - a 2 W resistor will create less noise than a 1/4 Watt - all else being equal.
Turns out the predominant noise in a carbon, metal oxide or metal film resistor is the contact noise.
This also, says wire wound resistors do not have contact noise.
However, its been my experience that most commercial grade wire would resistors sound like hell.
Mills wire wound resistors sound OK - I haven't heard the double wound Mundorf's
SHOT NOISE
V shot = K * I
where K is another constant.
Looking at these equations - it seems the 100K input resistor on the F5 is quite important
With the F5 I've used $0.20 metal film resistors and some metal oxide resistors - and they sound miles better than carbon.
Here are 2 references.
http://www.aikenamps.com/ResistorNoise.htm
http://www.eetimes.com/design/autom...egies-for-minimizing-resistor-generated-noise
Last edited:
Thank you for your information, but I don't fully understand your meaning
I worries about Ciss and linearity of those kind mosfets may cause bad effect to the hf sonic range.
BR
Big E,
Thanks for the info.
Clearly, keeping the noise figure down at the first stage is important, since that noise gets amplified all through the later gain stages of an amp.
Its interesting one author says contact noise is the predominant noise in carbon and metal type resistors.
Yet its thermal noise that always gets discussed.
I've found metal oxide resistors to have a softer sound than metal film.
Metal film seem to have a more focused and dynamic sound.
Metal oxide are a layer of Tin Oxide on a glass or ceramic tube
While metal film are NiCr on a ceramic tube.
But then wire wound resistors typically use NiCr (toaster) wire
Thank you for your information, but I don't fully understand your meaning
I worries about Ciss and linearity of those kind mosfets may cause bad effect to the hf sonic range.
BR
A number of years ago, I read this artical by N.P. where he basically
takes the ideas from an Op Amp and builds it using discrete transistors to make a preamp.
DIY Op Amps
It soon became clear that this pre amp could be changed into a power amp if large MOSFET's were used.
But I didn't like the idea of trying to match numerous IRF240 MOSFET's.
Then I had heard about how great the Gamut D200 mk3 power amp sounds and they used 2 power MOSFET's per channel.
So I prototyped this amp biased at 0.4 amps and it sounded amazing.
The current sources for the first 2 stages could be improved.
Without the current mirror the amp sounds warmer.
Output protection needs to be added.
However, a 30 watt class A amp is biased at around 3 amps.
It soon became clear the power dissipation was unreasonable.
The amp would dissipate 300 watts just in idle ... this would be a space heater.
BEFORE the Turbo F5 was released, I though I would increase the output of the original F5
by adding the IXFN180n10 power MOSFET's.
.
Attachments
Last edited: