I tried using Hitachi/Renesas lateral MOSFETs in F5 but they lack transconductance and can not provide sufficient gain as drain followers without harming the sound.
Since they perform great as source followers it's obvious that another gain stage is needed between input JFETs and output laterals. The choice fell on 2sk2013/2sj313, well known and respectable pair (estman, thanks for providing j313s
).
Schematic resembles Buzquito but feedback network is F5-like. I think of running 50-100mA through k2013/j313 and 0.5 to 1A through each lateral.
Thermal stability shouldn't be an issue for output MOSFETs (negative tempco) but k2013/j313 should have adequate heatsinking too - although they don't drift too much - I measured about 17mA increase of Id for change of temp. of 30 degrees Cels. and that's not a problem, they are easy to cool down since they dissipate only 1.5 to 2.5 Watts.
Since OLG is higher this way I can increase the amp's gain to 16 V/V - enough to avoid preamplifier. I built one channel point-to-point and it behaves and sounds impressive. Serious build will follow as soon as I get rid of previous two amps
Here are the schematic and PCB layout (both open for discussion ) :
Since they perform great as source followers it's obvious that another gain stage is needed between input JFETs and output laterals. The choice fell on 2sk2013/2sj313, well known and respectable pair (estman, thanks for providing j313s
).Schematic resembles Buzquito but feedback network is F5-like. I think of running 50-100mA through k2013/j313 and 0.5 to 1A through each lateral.
Thermal stability shouldn't be an issue for output MOSFETs (negative tempco) but k2013/j313 should have adequate heatsinking too - although they don't drift too much - I measured about 17mA increase of Id for change of temp. of 30 degrees Cels. and that's not a problem, they are easy to cool down since they dissipate only 1.5 to 2.5 Watts.
Since OLG is higher this way I can increase the amp's gain to 16 V/V - enough to avoid preamplifier. I built one channel point-to-point and it behaves and sounds impressive. Serious build will follow as soon as I get rid of previous two amps
Here are the schematic and PCB layout (both open for discussion
) :Attachments
Last edited:
Of course I did - I played with a bias from 0.1 to 1A per MOSFET and although you cannot consider single channel testing a "critical listening" it's obivous that with small bias current the amp sounds a bit "thin". With 0.5A it's really good and marginally better with 1A per MOSFET.
BTW, PCB pic shows copper side, dimmensions are 160 x 26 mm and I made room for caps parallel to feedback resistors (150R 2W) in case of oscillations (p2p model didn't show any signs of instability but...)
Of course, old 2sk135/2sj50 and similar laterals can be used as output devices.
Last edited:
Hi juma,
I can't help myself - If I see a cool circuit like this I want to build it, even if I don't understand it competely... Of course, I don't know whether I'll end up doing this, but can I throw in a couple of beginner questions?
Can you elaborate a little? I realise it's a different type of mosfet on output, but what is the principal advantage you see in this design?
"OLG" is open loop gain? (sorry if it's clear to everyone except me...)
What sort of size of amp is this? Output? Are F5 size heatsinks more-or-less adequate?
Cheers
Nigel
I can't help myself - If I see a cool circuit like this I want to build it, even if I don't understand it competely... Of course, I don't know whether I'll end up doing this, but can I throw in a couple of beginner questions?
Can you elaborate a little? I realise it's a different type of mosfet on output, but what is the principal advantage you see in this design?
...Since OLG is higher this way I can increase the amp's gain to 16 V/V - enough to avoid preamplifier. I built one channel point-to-point and it behaves and sounds impressive. Serious build will follow as soon as I get rid of previous two amps
"OLG" is open loop gain? (sorry if it's clear to everyone except me...)
What sort of size of amp is this? Output? Are F5 size heatsinks more-or-less adequate?
Cheers
Nigel
Unfortunately, yes...
The only AB cl. amp I can listen to more than a couple of days is SuSy with opa1632 and inverted opa549 - somehow it manages to sound good for months
I use scope just to check on oscillations and noise. I avoid to indulge in 100kHz square wave shapes - it tends to influence one's brain when it comes to amp's true evaluation i.e. how it soundsI agree - this is neat.
It would be interesting to see.
Thank you.
Output devices are spared from voltage gain - laterals act as source followers only. Voltage gain is left to k2013/j313 which are more adequate for this task and they don't have to provide much current - driving two pairs of laterals is an easy task.
What else?
1. I'd say acceptable
2. I used a PSU from another amp (+/-25V) so there is a bit more swing possible than with F5 but roughly, that's it.
3. If biased to same current, yes
Using two pairs of output devices allows heavier biasing and driving lower Z loads deeper into class A but increased heat is the price to be paid for that.
Thanks for the answers, juma.
Can I rephrase to see if I understood this? (Sorry if it's obvious to everyone else...) F5 in standard form has 1.3A current through a single pair of output devices, so when you say "same current" this would mean about 0.65A through *each* output device, since they come in two pairs in your schematic. In other words, "same current" means same total current, and therefore about the same total amount of heat. It follows that each device is operating farther from its limiting values, so you could bias "more heavily" if you wanted to, but obviously this means it gets hotter...
So, did I get it?
Cheers
Nigel
Can I rephrase to see if I understood this? (Sorry if it's obvious to everyone else...) F5 in standard form has 1.3A current through a single pair of output devices, so when you say "same current" this would mean about 0.65A through *each* output device, since they come in two pairs in your schematic. In other words, "same current" means same total current, and therefore about the same total amount of heat. It follows that each device is operating farther from its limiting values, so you could bias "more heavily" if you wanted to, but obviously this means it gets hotter...
So, did I get it?
Cheers
Nigel
I tried using Hitachi/Renesas lateral MOSFETs in F5 but they lack transconductance and can not provide sufficient gain as drain followers without harming the sound.
Since they perform great as source followers it's obvious that another gain stage is needed between input JFETs and output laterals. ....
Schematic resembles Buzquito but feedback network is F5-like.....Here are the schematic and PCB layout (both open for discussion
I had a "similar" idea almost 2 years ago...
http://www.diyaudio.com/forums/pass-labs/123640-f4-type-amp-but-gain.html#post1518771
Yes you did - "same current" for the amp as a whole
Without source resistord you'll have current hogging regardless of Id - that's how many good lateral MOSFET died...
Only insanely tight matching might get you rid of source resistors.
Yes, I remember that. I don't like the sound of JFETs with such a low Id. Second stage (with BJTs) is also not at all to my liking.I had a "similar" idea almost 2 years ago...
http://www.diyaudio.com/forums/pass-labs/123640-f4-type-amp-but-gain.html#post1518771
How did that project ended?
Fab-Four Amp
Hi
I have increased the jfet Id to 5 times more than original (5 to 6 ma from what I remember). For the second stage I have used a BJT enhanced cascode.
It ended like this:
fab-fourpoweramp (fabaudio)
Hi
I have increased the jfet Id to 5 times more than original (5 to 6 ma from what I remember). For the second stage I have used a BJT enhanced cascode.
It ended like this:
fab-fourpoweramp (fabaudio)
Matching costs a lot, but I believe it is needed for sound quality, irrespective to current equalization. If laterals are so much different that die without source resistors, than sound will be so-so.
Nice
Matching is always nice, but it's not the thing that will per se make your amp sound great or bad.
Lateral MOSFETs have their distinctive properties but diyng from too much current is not what makes them exclusive - BJTs and HEXFETs have Pd max limit too.
Actually, you are saying that output devices which use source resistors can never sound good, right ?
No, I only say that paralleled laterals tend to diminish an effect of dismatch between them, opposed to HEXFETs and BJTs, that tend to rise even a minor dismatch. Therefore HEXFETs and BJTs ALWAYs need source/emitter resistors, while laterals usually don't need them, except for probably a case of terribly dismatched pair. But in the last case sound will also suffer.
Not in my experience (2sk135/2sj50).
What laterals did you use to obtain such results?