JFET listening impressions - you may be surprised.
I have spent the morning listening to the JFET front end with dfferent types of music and now feel I can make some solid observations. First up, it is not all good.
The midrange clarity has definitely improved. More air around voices and the amp just makes everything sound easy, like it's lazily following the source with no real effort required to do so. It is a nice sound.
However, and this is a big factor, the wonderful bass performance from before has disappeared. Completely gone. It is possible that the old version was exaggerating the bass and that this one is more faithfully following the level of bass in the recording, but I'm not sure I like it. The difference between the MOSFET and JFET front end is night and day.
I thought maybe things would improve if I made the feedback capacitor larger, so I doubled it to 440uF. No difference.
This circuit is identical to the last one except I have use polypropylene capacitors this time rather than polyester. I have a hard time believing that the capacitors could make this much of a difference to the bass performance so I'm going to blame the JFET.
The VAS transistors are from the same tube and the output fets are the same ones I used last time.
The difference is so obvious that were I to make a final version to put in a box, I would be leaning towards the MOSFET version.
So, any ideas why this is so? I badly wanted the JFET to be better, and in some ways it is, but that oh so important bass performance just isn't there and this is very disappointing.
Is it possible that the circuit really needs a device with some decent transconductance in the front end to perform? The MOSFET's transconductance is something like 6x higher.
I have spent the morning listening to the JFET front end with dfferent types of music and now feel I can make some solid observations. First up, it is not all good.
The midrange clarity has definitely improved. More air around voices and the amp just makes everything sound easy, like it's lazily following the source with no real effort required to do so. It is a nice sound.
However, and this is a big factor, the wonderful bass performance from before has disappeared. Completely gone. It is possible that the old version was exaggerating the bass and that this one is more faithfully following the level of bass in the recording, but I'm not sure I like it. The difference between the MOSFET and JFET front end is night and day.
I thought maybe things would improve if I made the feedback capacitor larger, so I doubled it to 440uF. No difference.
This circuit is identical to the last one except I have use polypropylene capacitors this time rather than polyester. I have a hard time believing that the capacitors could make this much of a difference to the bass performance so I'm going to blame the JFET.
The VAS transistors are from the same tube and the output fets are the same ones I used last time.
The difference is so obvious that were I to make a final version to put in a box, I would be leaning towards the MOSFET version.
So, any ideas why this is so? I badly wanted the JFET to be better, and in some ways it is, but that oh so important bass performance just isn't there and this is very disappointing.
Is it possible that the circuit really needs a device with some decent transconductance in the front end to perform? The MOSFET's transconductance is something like 6x higher.
Last edited:
swordfishy,
Just some thoughts:
I think that jfet sounds best when biased at more than 70% of their IDss so they say...
Also, I wonder if paralleling jfet (same IDss) would increase the transconductance...
Fab
SWF,
Yes, I assume you use a 1K gate stopper as well, my bad.
Fab, I think you are right.
A mosfet has typically 5-10 times the transconductance of the jfet, which is measured in tens of millisiemens.
This will affect loop gain, greatly decreasing it. It should also make the amp easier to stabilise of course. However, low loop gain will increase output impedance and seriously impact the damping factor, a big factor in rendition of bass.
Jfets don't have a lot of power handling either. There are some impressive Supertex vertical N mosfets, the DN2530, which cost about a dollar from Mouser and which can be run at higher current. These are typically 0.3S, or 300 mmho, so much more loop gain is possible with these. I will see if I can obtain a Spice model from an engineer friend.
However, there are other factors to the bass. It might not all be DF.....
Hugh
Yes, I assume you use a 1K gate stopper as well, my bad.
Fab, I think you are right.
A mosfet has typically 5-10 times the transconductance of the jfet, which is measured in tens of millisiemens.
This will affect loop gain, greatly decreasing it. It should also make the amp easier to stabilise of course. However, low loop gain will increase output impedance and seriously impact the damping factor, a big factor in rendition of bass.
Jfets don't have a lot of power handling either. There are some impressive Supertex vertical N mosfets, the DN2530, which cost about a dollar from Mouser and which can be run at higher current. These are typically 0.3S, or 300 mmho, so much more loop gain is possible with these. I will see if I can obtain a Spice model from an engineer friend.
However, there are other factors to the bass. It might not all be DF.....
Hugh
Last edited:
All, danspy especially. I have told you a lie and I apologise. I always thought the source of the input fet was at nearly zero volts and thus for using a jfet you would need a negative gate voltage. This is not the case. Depending on drain resistance the source is always at a positive voltage of at least a few volts, and therefore even with a JFET you need a positive bias voltage. Therefore the bias arrangement I posted in post 336 will be fine.
The good news about this is that the layout for a jfet or MOSFET version is identical making pcb construction easy. My bias voltage is current somewhere around 5v, needing 3 LEDs to get enough voltage. Maybe a single Zener would do better here Hugh?
Anyway, onto the bass side of things.
Hugh, glad to know my theory about needing more transconductance in the input stage wasn't completely stupid! I must be learning something.
I have improved it slightly. As you guys suggested I have increased the jfet current from 1-2mA to 5mA or so by dropping the drain resistor to 250R. I also added the input filter and added more PS bypassing. While the bass is still a little weak, the overall sound is improved. More separation and transparency. Very good.
However, even the bass issue is not too bad. While there is definitely much less than the MOSFET version, there is still quite a lot compared to other amps I have heard. So overall it's in pretty good shape now.
Hugh, perhaps dropping the source resistor of thie input stage from 100R to 20R or something might help the OLG and improve damping factor?
FAB, paralleling input devices might work, but it would be nice if we could avoid it I think. Do you think that sound wise paralled jets would still be better than a single MOSFET?
The good news about this is that the layout for a jfet or MOSFET version is identical making pcb construction easy. My bias voltage is current somewhere around 5v, needing 3 LEDs to get enough voltage. Maybe a single Zener would do better here Hugh?
Anyway, onto the bass side of things.
Hugh, glad to know my theory about needing more transconductance in the input stage wasn't completely stupid! I must be learning something.
I have improved it slightly. As you guys suggested I have increased the jfet current from 1-2mA to 5mA or so by dropping the drain resistor to 250R. I also added the input filter and added more PS bypassing. While the bass is still a little weak, the overall sound is improved. More separation and transparency. Very good.
However, even the bass issue is not too bad. While there is definitely much less than the MOSFET version, there is still quite a lot compared to other amps I have heard. So overall it's in pretty good shape now.
Hugh, perhaps dropping the source resistor of thie input stage from 100R to 20R or something might help the OLG and improve damping factor?
FAB, paralleling input devices might work, but it would be nice if we could avoid it I think. Do you think that sound wise paralled jets would still be better than a single MOSFET?
Hi Hugh.
That would be me. Then i am sentenced to forced labor, until i by my heart can say i will not use more than ten active devices in an amp.
Does this include the currentsources...
- Sonny
Last edited:
SWF,
I've tried your model but can't get it to work using Notepad and LTSpice, any tips?
The entire 1st stage source current passes through the series fb resistor, therefore to keep voltage drop across this resistor to the minimum, a low value should be used. This mandates an even lower value for the shunt fb resistor, forcing use of a very large shunt capacitor. If we assume that 22R is about as low as you can go on the shunt resistor, then the 6dB corner frequency would be reached when the Xc of this cap is at 22R. That corresponds to 3.3Hz for a 2200uF cap, which is surely close enough for government work.
If we now set a gain of 18 for the amplifier, the series feedback resistor becomes 375R. 5mA passing through this resistor would give a voltage drop of 1V875, which means the source will be around 1V9 ABOVE the output, which we set to 0mV by adjusting the bias on the gate of the input device. The 2,200uF feedback cap will thus be polarised with just under 2V, should be rated to 5V or 6.3V, and should be protected with a 4V7 zener to prevent damage should the output stage blow.
Hope this explains the fb network clearly,
Cheers,
Hugh
I've tried your model but can't get it to work using Notepad and LTSpice, any tips?
The entire 1st stage source current passes through the series fb resistor, therefore to keep voltage drop across this resistor to the minimum, a low value should be used. This mandates an even lower value for the shunt fb resistor, forcing use of a very large shunt capacitor. If we assume that 22R is about as low as you can go on the shunt resistor, then the 6dB corner frequency would be reached when the Xc of this cap is at 22R. That corresponds to 3.3Hz for a 2200uF cap, which is surely close enough for government work.
If we now set a gain of 18 for the amplifier, the series feedback resistor becomes 375R. 5mA passing through this resistor would give a voltage drop of 1V875, which means the source will be around 1V9 ABOVE the output, which we set to 0mV by adjusting the bias on the gate of the input device. The 2,200uF feedback cap will thus be polarised with just under 2V, should be rated to 5V or 6.3V, and should be protected with a 4V7 zener to prevent damage should the output stage blow.
Hope this explains the fb network clearly,
Cheers,
Hugh
This is why I recommend at least 5mA for a 2SK170BL.
John Curl always says that JFETs runs better with high current.
Hugh.
Supertex has got SPICE models at their website.
Supertex - SPICE Models
swordfishy
I would like to see a schematic of that circuit you use.
You said you would post a circuit.
There are many factors that could make you perceive less BASS.
For example if the upper of the frequencies are more dominant, bright
then the BASS, eventhough not changed, would seem less.
Ideally you should setup 2 circuits. One with MOSFET input and one with JFET.
And then you can switch to make a better A-B compare.
Fcrawley
There is still experimenting with different circuits.
But I am sure someone will make a PCB in the end.
Hi all,
last night I came across this thread and read it in a rush. Thanks to all contributers, it was fun and instructive!
I have a few Q's, though.
Greg, which are your main speakers? Which is your daily poweramp?
All, I never activley biased a 2sk170. The datasheet makes clear that this FET at or near IDSS exhibits a very low Vgs. I am not sure that you are in a good operational mode if you bias the gate to 0mV if the source sits at nearly 2V, but maybe I'm missing something?
Or perhaps we learn here that a mosfet input stage could sound better than pure data and common sense would imply?
Nice project!
Rüdiger
last night I came across this thread and read it in a rush. Thanks to all contributers, it was fun and instructive!
I have a few Q's, though.
Greg, which are your main speakers? Which is your daily poweramp?
All, I never activley biased a 2sk170. The datasheet makes clear that this FET at or near IDSS exhibits a very low Vgs. I am not sure that you are in a good operational mode if you bias the gate to 0mV if the source sits at nearly 2V, but maybe I'm missing something?
Or perhaps we learn here that a mosfet input stage could sound better than pure data and common sense would imply?
Nice project!
Rüdiger
Oops, I thought Greg used a 2sk170 here. With a high Id it would sit aprox. 50mV below the source when left self biased.
Rüdiger
EDIT: Has someone done a LT-Spice sim and would share it here?
Last edited:
lineup, onvinyl, mikelm,
Maybe a FET VAS will improve things...but we are going to a lot of effort to improve a circuit to make a jfet work...when it was already so good without it
The circuit I said was at 2 volts was the one with a 500R drain resistor (1-2mA fet current), and 1k/100R feedback network. I think if you simulate that circuit you will see that my measuements were absolutely correct.
Moving on, my current circuit has a 250R drain resistor and 200R/20R feedback network (about 4mA current). Bias voltage is at very close to zero volts now (a few hundred mV).
onvinyl,
To answer you question, which I assume you're asking to see whether my opinion means anything Well, I don't have a Krell and B&Ws, that's for sure, and I don't have golden ears either.
The speakers I do most of my mod to mod testing with are a small set of Jamo bookshelves and when I think it's worth the effort I drag in my Soavo 1 Floorstanders. These are not well known, but they are high end Yamaha, and on par with their famous efforts from the 1970s. They have a flat frequency response and sound very bland with most amps - hence my looking for a bit of low order distortion.
Here is the AVHUB summary for them:
My daily amplifier is a meticulously constructed set of Peter Daniel Gainclone PCBs, which sound better than they have a right to. I also have a Cambridge Audio Azur 840W which is their flagship multi $k product designed by none other than Douglas Self. It is currently on EBAY as it is so sterile it could disinfect a hospital. Amazing figures but no soul whatsoever, and definitely not a good match to my already bland speakers. I hate it.
I have also built and owned a Sansui 500 Valve Amp from the 1960s, Musical Fidelity A1, JLH class A, aleph 3, Aleph 30 and several zero feedback designs of my own design as well as a heap of high end consumer audio - pioneer, sony, etc, etc.
So now you have an idea of where my opinions come from.
BTW, I think you might be on to something with your comment about the mosfet input...
As this is all getting so confusing I will post a new schematic later tonight so we all know where things are at!
Maybe a FET VAS will improve things...but we are going to a lot of effort to improve a circuit to make a jfet work...when it was already so good without it
The circuit I said was at 2 volts was the one with a 500R drain resistor (1-2mA fet current), and 1k/100R feedback network. I think if you simulate that circuit you will see that my measuements were absolutely correct.
Moving on, my current circuit has a 250R drain resistor and 200R/20R feedback network (about 4mA current). Bias voltage is at very close to zero volts now (a few hundred mV).
onvinyl,
To answer you question, which I assume you're asking to see whether my opinion means anything
Well, I don't have a Krell and B&Ws, that's for sure, and I don't have golden ears either. The speakers I do most of my mod to mod testing with are a small set of Jamo bookshelves and when I think it's worth the effort I drag in my Soavo 1 Floorstanders. These are not well known, but they are high end Yamaha, and on par with their famous efforts from the 1970s. They have a flat frequency response and sound very bland with most amps - hence my looking for a bit of low order distortion.
Here is the AVHUB summary for them:
My daily amplifier is a meticulously constructed set of Peter Daniel Gainclone PCBs, which sound better than they have a right to. I also have a Cambridge Audio Azur 840W which is their flagship multi $k product designed by none other than Douglas Self. It is currently on EBAY as it is so sterile it could disinfect a hospital. Amazing figures but no soul whatsoever, and definitely not a good match to my already bland speakers. I hate it.
I have also built and owned a Sansui 500 Valve Amp from the 1960s, Musical Fidelity A1, JLH class A, aleph 3, Aleph 30 and several zero feedback designs of my own design as well as a heap of high end consumer audio - pioneer, sony, etc, etc.
So now you have an idea of where my opinions come from.
BTW, I think you might be on to something with your comment about the mosfet input...
As this is all getting so confusing I will post a new schematic later tonight so we all know where things are at!
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.