2SJ313DocO Spice model
Hey all,
try and see how this model works for Your simulations please?
No guaranties and warranty for cautionary reasons of course...:
*$
.MODEL 2SJ313DocO NMOS(
+ LEVEL=3
+ L=2.0000E-6
+ W=1.3000
+ KP=1.0030E-6
+ RS=20.000E-3
+ RD=20.000E-3
+ VTO=-1.8000
+ RDS=1.0000E6
+ TOX=2.0000E-6
+ CGSO=90.000E-12
+ CGDO=45.000E-12
+ CBD=200E-12
+ MJ=.96725
+ PB=3
+ RG=10
+ RB=1.0000E-3
+ GAMMA=0
+ KAPPA=0)
Best of luck...
DocO
Hey all,
try and see how this model works for Your simulations please?
No guaranties and warranty for cautionary reasons of course...:
*$
.MODEL 2SJ313DocO NMOS(
+ LEVEL=3
+ L=2.0000E-6
+ W=1.3000
+ KP=1.0030E-6
+ RS=20.000E-3
+ RD=20.000E-3
+ VTO=-1.8000
+ RDS=1.0000E6
+ TOX=2.0000E-6
+ CGSO=90.000E-12
+ CGDO=45.000E-12
+ CBD=200E-12
+ MJ=.96725
+ PB=3
+ RG=10
+ RB=1.0000E-3
+ GAMMA=0
+ KAPPA=0)
Best of luck...
DocO
Must say that Erno Borbely's designs is a totally different cup of tea... being fully differential complementary designs..as opposed to this single ended minimalistic design-concept...
This here is different as the design consist of so few parts...the nature of the parts seem to play an even bigger role...
I think there is much to learn from Vladimir's take on a simple Jfet/mosfet preamp,, where he uses Jfet for input and some special HF-mosfets for the second stage...The VAS stage in our amplifer here...
Also I question the selection of the SK170 for the input..would rather use the BF862...which has twice the gain..and lower noise at lower currents.
The key to this design seems to find a good simple way to control the offset when making the Amplifier DC-coupled....Though I really see no specific point in this....!!
If the cap in the feedback really plays a significant role.. we may ask Swordfish...(do you like to fish by the way) to try to bypass the cap with some film-caps and listen to the sonic difference...
This here is different as the design consist of so few parts...the nature of the parts seem to play an even bigger role...
I think there is much to learn from Vladimir's take on a simple Jfet/mosfet preamp,, where he uses Jfet for input and some special HF-mosfets for the second stage...The VAS stage in our amplifer here...
Also I question the selection of the SK170 for the input..would rather use the BF862...which has twice the gain..and lower noise at lower currents.
The key to this design seems to find a good simple way to control the offset when making the Amplifier DC-coupled....Though I really see no specific point in this....!!
If the cap in the feedback really plays a significant role.. we may ask Swordfish...(do you like to fish by the way) to try to bypass the cap with some film-caps and listen to the sonic difference...
I have an idea that is simple enough for me for a servo offset trimmer - but I'm not attached to very few components. For me, enough components to make the very best sound is the right amount - I will draw it out and post later.
Mmmm, bypassed caps can sound worse than just caps but whatever you do to them they can never sound as clean and clear as a wire - it seems that not many strive for this - me, I do strive for this. For me it's what Hi Fi is all about.
The idea of bypass was just to see the influence...to get a picture of how much the cap matters.. I mean we do use caps in the circuit to add stability by reducing bandwidth...in the bootstrap (if we use that) and in the feedback we use it to reduce gain to a minimum at DC...
The pure approach would for sure be the DC...but at what cost..??
The pure approach would for sure be the DC...but at what cost..??
2sj313 model
Sorry chaps,
I use micro-cap 9 and model worked there.
Maybe you need to edit some syntax for Multisim?
Best of luck
DocO
Sorry chaps,
I use micro-cap 9 and model worked there.
Maybe you need to edit some syntax for Multisim?
Best of luck
DocO
Hi Lineup,
Nice circuit!
One thing - when you power two CCS from a common reference, one, generally the lower current CCS, should have a largish base stopper, around 1K5. This decouples the two bases and prevents interactions which might otherwise affect stability.
SWF,
Congratulations on your success with the new AC coupled circuit (ie the one with a 1000uF fb shunt cap). This is a terrific achievement.
Would you like to show 'n tell?
Mikelm,
I use a single transistor servo, along with a green led, a trimpot, and a 15V zener, and it works very well in one of my commercial lines. Could you PM me your email?]
Cheers,
Hugh
Nice circuit!
One thing - when you power two CCS from a common reference, one, generally the lower current CCS, should have a largish base stopper, around 1K5. This decouples the two bases and prevents interactions which might otherwise affect stability.
SWF,
Congratulations on your success with the new AC coupled circuit (ie the one with a 1000uF fb shunt cap). This is a terrific achievement.
Would you like to show 'n tell?
Mikelm,
I use a single transistor servo, along with a green led, a trimpot, and a 15V zener, and it works very well in one of my commercial lines. Could you PM me your email?]
Cheers,
Hugh
Hmm, yes the bf862 looks quite good. I will order some. Getting more gain in the input stage really helps, hence my support for using a MOSFET in this position.
I hate the name swordfishy. I signed up in a hurry one day meaning to ask one question and never thought I'd use it for so long. I wish I could change it. I just happened to have a fishing magazine in front of me which is why I chose it. Swordfish was taken so I added the "y". Call me whatever you like.
I will try bypassing the feedback cap tonight and let you know the results.
SWF,
This tells me that with a heatsink, the offset drift would be even better, a good result, and the reason I felt AC coupling was necessary, albeit without a servo to preserve the simple elegance of your circuit.
Could I post a complete LTSpice analysis using 2SK170/IRF9610? I have just completed it and the H2 is at -90dB, an excellent result, with almost 90% of the distortion in H2 and H3, all musical. This is with Iq set at 500mA, BTW.
Cheers,
Hugh
This tells me that with a heatsink, the offset drift would be even better, a good result, and the reason I felt AC coupling was necessary, albeit without a servo to preserve the simple elegance of your circuit.
Could I post a complete LTSpice analysis using 2SK170/IRF9610? I have just completed it and the H2 is at -90dB, an excellent result, with almost 90% of the distortion in H2 and H3, all musical. This is with Iq set at 500mA, BTW.
Cheers,
Hugh
Please do post it - for the benefit of others, I already have the model right here!
I don't mind the devices running warm to be honest. Even at 20ma they're warm but by no means hot. Only slightly above ambient. They come up to temperature very quickly with no heatsink.
FETZILLA UNDER SPICE
Here is the Spice analysis of FETZILLA. I have reduced the jfet stage current to 4mA for reasons of dissipation, increased rail voltage to 36 for 50W of output power into 8R, and amended gain very slightly to 21.6 (26.7dB, close to the THX standard of 29dB).
This is a VERY good amplifier, hats off to Lineup, SWF, and all others who contributed.
Cheers,
Hugh
Here is the Spice analysis of FETZILLA. I have reduced the jfet stage current to 4mA for reasons of dissipation, increased rail voltage to 36 for 50W of output power into 8R, and amended gain very slightly to 21.6 (26.7dB, close to the THX standard of 29dB).
This is a VERY good amplifier, hats off to Lineup, SWF, and all others who contributed.
Cheers,
Hugh
Last edited:
Very nice Hugh! These nice results are what I'm enjoying at the moment!
The only things I have to add are:
1) I am using less input degeneration, which I think is crucial to making the input jfet perform. How did you arrive at the figure you're using? I think a switch to the bf862 miib mentioned would work wonders for overall performance.
2) I have managed to get by with only 200R gate stoppers on the output fets, and none at all on the vas. I have managed to get away with this twice now, so hopefully not a fluke. The jfet drain and 9610 gate are tied pin to pin.
3) I am using only 1uf across the bias spreader as I recall that in simulation it actually made things worse at higher frequencies and powers. Are you able to test this theory? I have some 5uf poly caps I can try. Might be worth a go.
Looks good though huh?!
The only things I have to add are:
1) I am using less input degeneration, which I think is crucial to making the input jfet perform. How did you arrive at the figure you're using? I think a switch to the bf862 miib mentioned would work wonders for overall performance.
2) I have managed to get by with only 200R gate stoppers on the output fets, and none at all on the vas. I have managed to get away with this twice now, so hopefully not a fluke. The jfet drain and 9610 gate are tied pin to pin.
3) I am using only 1uf across the bias spreader as I recall that in simulation it actually made things worse at higher frequencies and powers. Are you able to test this theory? I have some 5uf poly caps I can try. Might be worth a go.
Looks good though huh?!
Hi SWF,
Yes, good results.
You have to watch those jfets. Even the K170 is only rated to 40V, and with 36V rails it's pulling Vds of just under 30V, you wouldn't want to go higher. And if you are forced to use a cascode, then I'd be tempted to go bipolar input stage, OR, use the DN2530.
I did the OLG calculation on LTSpice. For an output of 28.2Vpp, the differential across the input gate and Q1 source is 0.93mV. This gives an OLG of 28,200/0.93, which is 89.6dB. Since gain is fixed at (680+33)/33, 21.6 or 26.7dB, we have a loop gain (at 1KHz) of close on 63dB, which is actually pretty high. I was surprised at this; it's about lineball for a BJT design, and would give high DF and flat frequency response.
In short, given that 33R rather than 20R on the fb shunt allows use of a smaller blocking cap, improves overload capacity and the loop gain is high anyway, I would be reluctant to deviate from this from a technical POV, but if it transpires that it actually sounds better, then why not?
I will leave provision on the board for a gate stopper on the VAS; this will permit a variety of devices to be tested, which is a practical feature for a forum design.
I use a Black Gate NP 22uF rated to 6.3VW as bias generator cap, and have enjoyed very good results. We could use a smaller one, but at the very high current of this VAS, 20mA, it needs more rather than less to keep the bias voltage constant. I will have a look at the model, see what happens when it's varied.
Yep, damn good design, this one is really worth running with!
Cheers,
Hugh
Yes, good results.
You have to watch those jfets. Even the K170 is only rated to 40V, and with 36V rails it's pulling Vds of just under 30V, you wouldn't want to go higher. And if you are forced to use a cascode, then I'd be tempted to go bipolar input stage, OR, use the DN2530.
I did the OLG calculation on LTSpice. For an output of 28.2Vpp, the differential across the input gate and Q1 source is 0.93mV. This gives an OLG of 28,200/0.93, which is 89.6dB. Since gain is fixed at (680+33)/33, 21.6 or 26.7dB, we have a loop gain (at 1KHz) of close on 63dB, which is actually pretty high. I was surprised at this; it's about lineball for a BJT design, and would give high DF and flat frequency response.
In short, given that 33R rather than 20R on the fb shunt allows use of a smaller blocking cap, improves overload capacity and the loop gain is high anyway, I would be reluctant to deviate from this from a technical POV, but if it transpires that it actually sounds better, then why not?
I will leave provision on the board for a gate stopper on the VAS; this will permit a variety of devices to be tested, which is a practical feature for a forum design.
I use a Black Gate NP 22uF rated to 6.3VW as bias generator cap, and have enjoyed very good results. We could use a smaller one, but at the very high current of this VAS, 20mA, it needs more rather than less to keep the bias voltage constant. I will have a look at the model, see what happens when it's varied.
Yep, damn good design, this one is really worth running with!
Cheers,
Hugh
That all sounds good!
You mention a BJT input. I have tried this, and it sounded rather good, though maybe a little more clinical than the fet.
I would be very interested to see what design you come up with using a bjt input and fet vas - if you can be bothered simulating it. I pretty much dropped one into the original circuit and it worked quite well, however at the time I was using the bjt vas. It definitely had potential though.
You mention a BJT input. I have tried this, and it sounded rather good, though maybe a little more clinical than the fet.
I would be very interested to see what design you come up with using a bjt input and fet vas - if you can be bothered simulating it. I pretty much dropped one into the original circuit and it worked quite well, however at the time I was using the bjt vas. It definitely had potential though.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.