F5 dc offset issue

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

PRR

Member
Joined 2003
Paid Member
Weird issue with my F5. One channel if input is shorted 0mv dc offset. If input is not shorted 180mv offset. If preamp is connected (cap coupled preamp, no dc) offset is 60mv. Any ideas?

You mean like this F5?

FET voltage error would not depend on source resistance.

But it is acting like there is 0.75uA flowing at the input. The reduction with preamp suggests the preamp has a 100K resistor after its cap, a very likely setup.

(However we would expect the short-input to still show some 18mV output offset due to the 4.7k.)

The 2SK170 gate current at this high-ish channel voltage should be 10 to 50 nA at 25 deg C. More than 10X lower than you seem to have. The gate current will rise with temperature, but unless it is awful hot it should not be that huge. Also there "could" be some compensation from the 2SJ74's opposite gate current.

As a +/-23V Power amp, I would not be too over-concerned about 60mV DC on the speaker. This is 0.2% voltage and <0.001% power of what the speaker can obviously stand as audio.

But it does not seem quite right. Maybe someone who knows these JFETs better will know if 0.75uA gate current is reasonable.
 

Attachments

  • F5-gates.gif
    F5-gates.gif
    17.8 KB · Views: 465
You mean like this F5?

FET voltage error would not depend on source resistance.

But it is acting like there is 0.75uA flowing at the input. The reduction with preamp suggests the preamp has a 100K resistor after its cap, a very likely setup.

(However we would expect the short-input to still show some 18mV output offset due to the 4.7k.)

The 2SK170 gate current at this high-ish channel voltage should be 10 to 50 nA at 25 deg C. More than 10X lower than you seem to have. The gate current will rise with temperature, but unless it is awful hot it should not be that huge. Also there "could" be some compensation from the 2SJ74's opposite gate current.

As a +/-23V Power amp, I would not be too over-concerned about 60mV DC on the speaker. This is 0.2% voltage and <0.001% power of what the speaker can obviously stand as audio.

But it does not seem quite right. Maybe someone who knows these JFETs better will know if 0.75uA gate current is reasonable.

Yes, mine is an f5 turbo, same amp with 30v rails and 2 pairs mosfets.

Preamp has 47k to ground after cap.

Iam not so worried about the 60mv offset but Iam worries about losing a jfet and having a massive dc spike...guess I will need to do more investigating.
 
The one and only
Joined 2001
Paid Member
Don't rule out oscillation. When you cascode those Jfets you enter new
territory on stability, and a different (low) source impedance may be
creating a parasitic that shifts the operating point.

Also, if you are cascoding the Jfets, perhaps try putting some resistance
between the Drain of the Jfets and the Emiiters/Sources of the cascode
transistors, like 100 ohms or so.
 
Thanks Nelson. Thanks for the amp too!

I have no cascode currently with 30ish volt rails. The jfets have a small heatsink on them and do not get especially warm.

Full disclosure: I do "hot swap" preamps all the time. May have caused some damage?

Only one channel is exhibiting this behavior and honestly, I was unaware of it until I noticed different preamps were making the amp show different DC offset even though the preamps are cap coupled. Amp sounds and behaves fine.

So it could be oscillation and not leaky jfets? Guess I better do some more exploration before I replace the jfets.
 
Don't rule out oscillation. When you cascode those Jfets you enter new
territory on stability, and a different (low) source impedance may be
creating a parasitic that shifts the operating point.

Also, if you are cascoding the Jfets, perhaps try putting some resistance
between the Drain of the Jfets and the Emiiters/Sources of the cascode
transistors, like 100 ohms or so.
Hi Nelson,
I did not cascode the Jfets because I wanted to keep to the original design which you had in v2. I also used the recommended PSU voltage. Will this be a concern for long term amp stability?
 

PRR

Member
Joined 2003
Paid Member
Question: F5T rails are ~35V, it will definitely exceed the 25V. Did I miss something in my setup?

I can't know what Mr Pass was thinking, but in hindsight we see the JFET offers a small simple distortion signature, two together reduces that, and the plan I snipped says 23V which is higher than we might use for phono or RF preamp but is safely within the 25V rating.

The "25V" rating is because in MOST JFET work there is no reason to go higher. Most JFETs go into small-signal stages where a dozen Volts is ample for the work they do. Mostly when a JFET is wanted the HIGH input impedance (low Gate current) is a key factor, and Gate current does increase at high voltage, so most designs keep Vds down to a dozen Volts.

JFET gates are just junctions like any transistor or chip. There is a historically popular "36V process", which is why so many opamp chips are rated +/-18V MAX. 99 out of 100 will survive 40V for years, though maybe not decades; and of course that uncertainty is why we are told "36V".

But JFETs are sold on "low-low gate current". While they may not break-down at 36V, the gate current rises to a point that "looks bad" on the datasheet. That's probably why the 25V limit.

We now know that the plan I snipped says 23V, and you are running the same general plan (without cascode) at 35V. And not breaking-down, but acting somewhat strange. This is not a shock. While it has worked and may work for many more years, you are "off label" as the pharmacy people say. There is apparently a cascoded version which addresses this?
 
I can't know what Mr Pass was thinking, but in hindsight we see the JFET offers a small simple distortion signature, two together reduces that, and the plan I snipped says 23V which is higher than we might use for phono or RF preamp but is safely within the 25V rating.

The "25V" rating is because in MOST JFET work there is no reason to go higher. Most JFETs go into small-signal stages where a dozen Volts is ample for the work they do. Mostly when a JFET is wanted the HIGH input impedance (low Gate current) is a key factor, and Gate current does increase at high voltage, so most designs keep Vds down to a dozen Volts.

JFET gates are just junctions like any transistor or chip. There is a historically popular "36V process", which is why so many opamp chips are rated +/-18V MAX. 99 out of 100 will survive 40V for years, though maybe not decades; and of course that uncertainty is why we are told "36V".

But JFETs are sold on "low-low gate current". While they may not break-down at 36V, the gate current rises to a point that "looks bad" on the datasheet. That's probably why the 25V limit.

We now know that the plan I snipped says 23V, and you are running the same general plan (without cascode) at 35V. And not breaking-down, but acting somewhat strange. This is not a shock. While it has worked and may work for many more years, you are "off label" as the pharmacy people say. There is apparently a cascoded version which addresses this?
Interesting perspective. I thought the cascode is to drive more current and isn't anything to do with the voltage. I could be wrong.
 
> JFET gates are just junctions like any transistor or chip. There is a historically popular "36V process", which is why so many opamp chips are rated +/-18V MAX. 99 out of 100 will survive 40V for years, though maybe not decades; and of course that uncertainty is why we are told "36V".

May I ask if you do know, or if this is your speculation ?
Have you perhaps measured the actual breaksown voltage of the GS or GD junction of a 2SJ74 ?
Or can you perhaps explain the different Vgds specifications of 2SJ74 (25V), 2SK170V (40V), 2SK246/2SJ103 (50V), if they are all from the "historically popular 36V process" ?

Just for interest, there are also JFETs with much lower Vgds, such as the beloved BF862 (20V).
2SK932, 2SK3557 are even lower at 15V, even though the latter is almost identical to BF862 in spec.
Many of the datasheets will have a graph of gate current against voltage.
One can easily see at what voltage they begin to increase significantly.

Vgds (the gate-drain breakdown voltage) is specified at that when the gate current goes beyond µA level.
Some manufacturers specify that at 1µA (J111), some at 10µA (2SK3557), some at 100µA (most Toshiba's).
And normal gate current is ~100pA level.


Cheers,
Patrick
 
> I thought the cascode is to drive more current and isn't anything to do with the voltage.

Cascode (as applied to the drain of a JFET) is applied to fix the drain voltage of the device below the cascode.
The motivation may be to protect the to-be-cascoded device's drain from excessive voltage.
It may also be to shield the drain from drain voltage variations.
In any case it definitely has to do with voltage.

References :
PassDiy
JFETs: The New Frontier, Part 1 | audioXpress
The Tube CAD Journal, Hybrid Cascodes


Patrick
 
> I thought the cascode is to drive more current and isn't anything to do with the voltage.

Cascode (as applied to the drain of a JFET) is applied to fix the drain voltage of the device below the cascode.
The motivation may be to protect the to-be-cascoded device's drain from excessive voltage.
It may also be to shield the drain from drain voltage variations.
In any case it definitely has to do with voltage.

References :
PassDiy
JFETs: The New Frontier, Part 1 | audioXpress
The Tube CAD Journal, Hybrid Cascodes


Patrick
I am enlightened now! Thanks Patrick.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.