I found that changes in the MOSFET source resistors (by paralleling in values from 2.7R to 10R) did not have any material effect on the THD%.
Of course, this could be one of those dog food problems (that's what we called Linear Programming back in the day...)
I set up a Rube Goldberg adjustment system for the JFET source resistors with a pair of 10-turn pots, in series with 100R -- (The JFET source resistors are 12R//68R = 10.2R on each leg). Even a fraction of an ohm change (one click of the pot) can significantly change the THD% -- at least when you're working in the thousandths of a per cent. There seems to be a sweet spot where the 3rd starts to dominate the second.
Of course, this could be one of those dog food problems (that's what we called Linear Programming back in the day...)
I set up a Rube Goldberg adjustment system for the JFET source resistors with a pair of 10-turn pots, in series with 100R -- (The JFET source resistors are 12R//68R = 10.2R on each leg). Even a fraction of an ohm change (one click of the pot) can significantly change the THD% -- at least when you're working in the thousandths of a per cent. There seems to be a sweet spot where the 3rd starts to dominate the second.
KSTR said:
Hi Klaus,
I was intrigued by circuit you proposed in post #1288. If you ground-reference the load you'll be surprised by result (see attachment), but you might have to change Rb and R14. It's selbstverständlich that B1 and B2 are opposite in phase (they are just strangely drawn in your sch.)
This circuit simulates very good and it will be easy to match JFETs and MOSFETs (them being of same type), so real life circuit shouldn't be far from simulated version. I plan to build one as soon as I complete the PSU so I'll let you know about results.
Attachments
sousmielie said:Hi, I'm having trouble sourcing j-fets, all I can get that I think could maybe work are: 2SJ103L1 and 2SK246. Any comment?
Thank you
They have less gm, but they will work all right (although not as good as K170/J74).
Build it with JFETs you got and enjoy it - people have reported great results with even less appropriate devices
Thanks, juma,juma said:
but actually I only have copied/redrawn circlomanen's circuit from that other thread, changing the bias conditions and the cap coupled gain resistors with that single, direct one. With or without those GND resistors from the load the perfomance doesn't change in "perfect spice world" (exept for the additional 200R of load), in reality it'll probably perform better with them because of less sensivity to side effects... and probably better with a blocking cap in series with the gain resistor, too... we'll see, and I wish you good luck, hoping no serious issues show up that have been overlooked so far.
- Klaus
> in reality it'll probably perform better with them because of less sensivity to side effects...
The few circlotron circuits I have built, my experience is that you do need the ground reference, and 100R is about the highest value you should use for a power amp. Without them, you still get the differential output, but they might go out of symmetry wrt ground by quite a bit, especially with fast signals like 10kHz square wave.
> and probably better with a blocking cap in series with the gain resistor, too...
I do not spice, so I might be wrong.
But as far as I can see, the reason for having the bootstrap cap (2200uF) is to increase open loop gain, as the bias of the JFET will be reduced to something like 2mA. Without the bootstrap cap, the drain resistor of the JFETs would be something like 700ohm using F5 devices and resistor values elsewhere. But otherwise I cannot see why it would not work. 2SK170 and the like needs at least 4-6mA to operate in their low noise regime. The outputs might have a common mode voltage of a few hundred mV wrt Gnd, but since it is common mode, who cares. The elegance of the F5 is a simple circuit with very high bandwidth but no caps. There are also quite a bit of discussion about the sonic effect of the bootstrap in source (emitter) feedback circuits in Geoff's JLH Class A wbsite.
http://www.tcaas.btinternet.co.uk/jlhupdate.htm
under the section "removal of the feedback cap" & Fig. 2.
Had I not been still waiting for Fairchild MOSFETs, I would have built one without caps as per Circlomanen's circuit in post #1286. But maybe someone who spices regularly would care to do a quick sim, just to prove me wrong.
Patrick
The few circlotron circuits I have built, my experience is that you do need the ground reference, and 100R is about the highest value you should use for a power amp. Without them, you still get the differential output, but they might go out of symmetry wrt ground by quite a bit, especially with fast signals like 10kHz square wave.
> and probably better with a blocking cap in series with the gain resistor, too...
I do not spice, so I might be wrong.
But as far as I can see, the reason for having the bootstrap cap (2200uF) is to increase open loop gain, as the bias of the JFET will be reduced to something like 2mA. Without the bootstrap cap, the drain resistor of the JFETs would be something like 700ohm using F5 devices and resistor values elsewhere. But otherwise I cannot see why it would not work. 2SK170 and the like needs at least 4-6mA to operate in their low noise regime. The outputs might have a common mode voltage of a few hundred mV wrt Gnd, but since it is common mode, who cares. The elegance of the F5 is a simple circuit with very high bandwidth but no caps. There are also quite a bit of discussion about the sonic effect of the bootstrap in source (emitter) feedback circuits in Geoff's JLH Class A wbsite.
http://www.tcaas.btinternet.co.uk/jlhupdate.htm
under the section "removal of the feedback cap" & Fig. 2.
Had I not been still waiting for Fairchild MOSFETs, I would have built one without caps as per Circlomanen's circuit in post #1286. But maybe someone who spices regularly would care to do a quick sim, just to prove me wrong.
Patrick
juma said:
They have less gm, but they will work all right (although not as good as K170/J74).
Build it with JFETs you got and enjoy it - people have reported great results with even less appropriate devices
Thanks, I'll give it a go then. My supplier told me (about 10 minutes ago) they can get me the other JFET's but it's going to take a while if I don't place a big order... I'm gonna draw up a PCB as we speak. This is going to be my first PASS Amp.
sousmielie said:
Thanks, I'll give it a go then. My supplier told me (about 10 minutes ago) they can get me the other JFET's but it's going to take a while if I don't place a big order... I'm gonna draw up a PCB as we speak. This is going to be my first PASS Amp.
Don't be lazy
, read the whole thread and you will find a lot of informations, other people's experiences and PCB layout proposals. Reading the thread can actually save you time.
EUVL said:....
Had I not been still waiting for Fairchild MOSFETs, I would have built one without caps as per Circlomanen's circuit in post #1286. But maybe someone who spices regularly would care to do a quick sim, just to prove me wrong.
Patrick
I don't have Fairchild P MOSFETs, but I have matched J74s so a polarity reversion will do the trick (see attachment). This one simulates so good that even if reality becomes an order of magnitude worse, it will still be a great amp
Attachments
sousmielie said:
Thanks, I'll give it a go then. My supplier told me (about 10 minutes ago) they can get me the other JFET's but it's going to take a while if I don't place a big order... I'm gonna draw up a PCB as we speak. This is going to be my first PASS Amp.
Genade!! "Sousmielie"?? Wat `n naam... he-he!!
Who is your supplier?
Cheers
D
GlidingDutchman said:
Genade!! "Sousmielie"?? Wat `n naam... he-he!!
Who is your supplier?
Cheers
D
ja wat, het dit in die ou dae by RAU gekry...
I get my components from mantech electronics in jhb and RSE (just around the block). I might just be a little lazy (and in the mood for some diy) today, but you're right should start at p1... 
juma said:
I don't have Fairchild P MOSFETs, but I have matched J74s so a polarity reversion will do the trick (see attachment). This one simulates so good that even if reality becomes an order of magnitude worse, it will still be a great amp
How about a hybrid with four feedback loops?
sousmielie said:
ja wat, het dit in die ou dae by RAU gekry...
Okay....
Jacco - "SP", tja, voorwaar een besckrijwing!!
Sousmielie - Try Electronics 123 as well as Communica in Midrand as well as Roleen Electronics in Boksburg.
D
