Paradise Builders

Just finished measuring the Idss of Q90 and Q99.

Q90 J310 Idss = 35 mA
Q99 2SK170 Idss = 6.7 mA

These seem to be in the right ball park. The problem must be elsewhere.
I don't know how you measured the 35mA Idss.
Continuous 35mA @ 10Vds is very likely to damage your jFET.
Even for only a second or two to let you see the voltage on a meter.

Lowering the voltage to half value, say around 5Vds will not give you a true reading of Idss.

Could you describe your non damaging Idss method for high Idss devices?
 
Member
Joined 2004
Paid Member
I don't know how you measured the 35mA Idss.
Continuous 35mA @ 10Vds is very likely to damage your jFET.
Even for only a second or two to let you see the voltage on a meter.

Lowering the voltage to half value, say around 5Vds will not give you a true reading of Idss.

Could you describe your non damaging Idss method for high Idss devices?

Thank you for responding :).

Sounds like I should not have done this, but I measured Idss for a second or or so using a 10 V lab supply and a Fluke DMM in series with the drain (mA setting). Gate and source tied together. I could see the current drift down as the device is warming up. The Idss of some devices started in the low 30s, others around 40 mA. I used the latter. You are right that this is flirting with the maximum power dissipation of the device.
 
Member
Joined 2004
Paid Member
Check the Vp and the Id at a fixed known value of Vgs. Then plot those two values on the Vgs vs Id curve for the device and see if the slope is similar to the datasheet curves.
Hi Andrew, I just finished measuring three J310s and they seem to match the spec. The plot for one of the devices appears below. Should this meet the need for the paradise buffer?
 

Attachments

  • Capture d’écran 2014-12-30 à 3.37.05 PM.png
    Capture d’écran 2014-12-30 à 3.37.05 PM.png
    30.4 KB · Views: 281
  • Capture d’écran 2014-12-30 à 3.09.47 PM.png
    Capture d’écran 2014-12-30 à 3.09.47 PM.png
    55.5 KB · Views: 260
Member
Joined 2004
Paid Member
A question for the gurus… How much of a mismatch between the positive and negative sides of the circuit does it take to overwhelm the servo?

I measure a mismatch of .25 V between the Base of Q18 = 14.65 V versus the Base of Q7 = -14.90 V.

Likewise, I measure a mismatch of .21 V between the VcascH = 14.18 V versus the VcascL = -13.97 V.

Are these too large?

I tried adjusting Vplus or Vminus to reduce these, but reducing one mismatch has the opposite effect on the other.

Unless mistaken, VcascL is set by D1 and Q98, and VcascH by D2 and Q97, so I could rebuild these with new parts for better symmetry.

Now if I need to improve the symmetry between the positive and negative mirrors, how can I do this? Is there a convenient spot to add a trimmer? Alternatively, would it be silly to start trimming the 120 ohm resistors on the emitters of the bipolars?

Many thanks in advance for your help! Regards - Pierre
 
The Idss of the inside jfet (Q99) is 6.8 mA.

I measured 41 mV across the 5 ohm resistor R12 in the buffer, so know that there is 8.2 mA of current flowing through it. It looks like the buffer is working?
!

the 6.8mA Idss device will pass less than 6.8mA when inside a cascode that give <<10Vds on the active (inside) device.
expect ~80%. i.e. around 5mA to 6mA.
Look at your plot, find 5.5mA. Read off the Vgs for that current. (about 2.1Vgs to 2.25Vgs)
That becomes the Vds for your active device.
This must be greater than 2times the Vp of the active device. i.e. your active device must have a Vp <1.1V

You then test the Vds and the Id of the cascode to see if the predicted values actually occur.
Then build up a matching set for the other channel.
 
Last edited:
Can I presume the cascoded active device is at the output?
There are no voltages marked there !

I think your -17.74Vdc is placed in the wrong position.

The upper cascode seems to have only 2.63Vdc across it.
That is far lower than 2times the Vp of the cascode device.

You must adjust your bias to get an output node voltage approximately half the supply voltage. i.e. near 0Vdc.
 
Member
Joined 2004
Paid Member
!

the 6.8mA Idss device will pass less than 6.8mA when inside a cascode that give <<10Vds on the active (inside) device.
expect ~80%. i.e. around 5mA to 6mA.
Look at your plot, find 5.5mA. Read off the Vgs for that current. (about 2.1Vgs to 2.25Vgs)
That becomes the Vds for your active device.
This must be greater than 2times the Vp of the active device. i.e. your active device must have a Vp <1.1V

You then test the Vds and the Id of the cascode to see if the predicted values actually occur.
Then build up a matching set for the other channel.

! indeed… I removed and double checked the Idss of Q99 to measure 6.5 mA (not 6.8 mA). The current flowing through R12 and R15 is 8 mA. There is a slight POSITIVE Vgs on Q99 (.1V or so) therefore it could be that some current is coming from the gate...
 
Member
Joined 2004
Paid Member
You must adjust your bias to get an output node voltage approximately half the supply voltage. i.e. near 0Vdc.

Thank you Andrew. Yes I understand this.

Note that Vriaa is at 15.34 V instead of -1.5 V… And I am trying to find out where the root cause is: Is it is because of asymmetry in the mirrors, asymmetry between VcascL and VcascH, the servo not working properly, or the buffer not biasing normally?

I guess it would be straightforward to isolate the buffer by lifting and grounding the gate pin of Q99 and grounding the bottom leg of R48.
 
Hi,

the JFETs and resistors of the buffer in the pdf schematic seem wrong.
Q91 must be like Q9 a 2SK170 here.
The value of the source resistor R15 should be the same as that of R12.
If that isn´t the reason for the offset, could You then pull the DC-Servo OPAmp and connect Pin6 temporarily to signal gnd.
Measure all the testpoints and report.

The two JFETs Q97 and Q98 are connected as CCSs and they are degenerated by their source resistors quite heavy.
So the currents through Q97/D2 and Q98/D1 are very low (~0.5mA), maybe too low for the Zeners to regulate cleanly.
In a first step (after putting the OPAmp back in the socket) You may try to parallel a say 10k resistor to R10.

This increases the current through Q98 and should give a slight increase in Vcascl and a more stable zener voltage.
See what happens to Vasch and Vascl.
If the offset improves I´d suggest to replace the 1k source resistors R10 and R11 with a lower value for the resistors (I´d start with ~500R, meaning You could parallel a second 1k to the existing 1k).

jauu
Calvin