ok I gotta go but I just happen to have a spare set of these jfets, readings are from pin 1 to pin 3 lsj74 40 ohms and pin 1 to pin 3 2sk170 38.8 ohms (source to drain readings), I think yours are o.k. in the ballpark I have put My jfets thru hell and back on My 2 F5's I built with no problems there pretty tough..... I'd recheck the bias proceedures.....
Last edited:
I wouldn't run or power up the output boards with no front end / input board hooked up to it, full voltage on the source and no voltage on the gate might blow the output mosfets, max Vgs difference between the source and gate is +/- 4 volts respectively, not sure but it might blow the mosfets maybe somebody else knows for sure..... however I have run the front end powered up with the output board not hooked up no problem.....
Last edited:
FE board Cascode transistors are fine. NPN tests 170 hfe and PNP tests 163. I'd accept any non-zero quantity within their range.
Time to test a P-channel output devices.
OMG.
After removing the first board, I found a 4-40 screw used to hold the boards on their standoffs shorting the Drain and Source on a P-channel output mosfet!!!! Sloppy....very sloppy.
So I unsoldered that mosfet and did the PNP diode test I found here just to make sure ( it must work, the short is a bypass...
How to Check a MOSFET Using a Digital Multimeter | Homemade Circuit Projects
It tests fine. Initial OL, then 0.3, just like a new one.
However, this does not address the issue with the funny biasing -- that I hope goes away with matched outputs.
Also, since the other channel threw the speaker protection relay, I suspect that something is horribly wrong there.... last time that happened *every* output mosfet was dead....
And the speaker protection needs to be fixed too, clearly with a more robust relay.
It is better than nothing for the F5T, but it won't stand the rail voltage. The Relay I used was part of the original BOM....it's good for 24 VDC. I had no idea.
Time to test a P-channel output devices.
OMG.
After removing the first board, I found a 4-40 screw used to hold the boards on their standoffs shorting the Drain and Source on a P-channel output mosfet!!!! Sloppy....very sloppy.
So I unsoldered that mosfet and did the PNP diode test I found here just to make sure ( it must work, the short is a bypass...
How to Check a MOSFET Using a Digital Multimeter | Homemade Circuit Projects
It tests fine. Initial OL, then 0.3, just like a new one.
However, this does not address the issue with the funny biasing -- that I hope goes away with matched outputs.
Also, since the other channel threw the speaker protection relay, I suspect that something is horribly wrong there.... last time that happened *every* output mosfet was dead....
And the speaker protection needs to be fixed too, clearly with a more robust relay.
It is better than nothing for the F5T, but it won't stand the rail voltage. The Relay I used was part of the original BOM....it's good for 24 VDC. I had no idea.
Horror show:
I've not touched it for days. It's been unpowered.
I removed the FE board from the broken channel, and check as the previous post. Removed, tested and resoldered the low powered resistors. All seemed well.
I verified that the other channel was not broken. It works fine. During this time, the power supply for the other channel was disconnected.
When I tried to re-install the FE board, I shorted the gate wiring to the "empty" power supply with the FE board, and two N channel outputs blew up.
I'd been working on the P-channel. The N-channel PS caps still had power after days of inactivity. Imagine my surprise.
Clearly, the drain resistors in the power supply are not doing their job - 2.2K at 3 Watts. Should be more than enough for a max 40 volt rail -- P = V^2/R, 1600/2200. Under a watt dissipation.
So, I replaced the outputs, checked that the output wiring to case was OL, and reinstalled the FE board and powered up.
No measureable flow across either of the N or P source resistors on the board I measured. DC offset was 35mV.
It would not budge when P1 and P2 were turned. Turned off the amp, the PS leds stayed on. Drained the PS manually with a 10ohm power resistor. NOTE: if there was a shorted output, the led in that channel would be off, and the DC offset would be rail voltage.
Checked the FE board Jfets. Seemed ok, but replaced anyway. Measured P1 P2. They were up around 400ohms and 700 ohms -- enough to drive the outputs.
Reinstalled the FE board to same effect, except DC offset now at 2 mV.
Checked Vgate voltage to gnd on the FE boards: 38 volts.
Pulled out the cascodes. Their HFE measures fine, at about 170 each.
The DC offset does not budge, no flow across the two source resistors.
The Speaker protection relay closes, showing continuity from output to binding post. If any flow anywhere, there would be a significant DC offset.
Horror show. I have no idea.
I've not touched it for days. It's been unpowered.
I removed the FE board from the broken channel, and check as the previous post. Removed, tested and resoldered the low powered resistors. All seemed well.
I verified that the other channel was not broken. It works fine. During this time, the power supply for the other channel was disconnected.
When I tried to re-install the FE board, I shorted the gate wiring to the "empty" power supply with the FE board, and two N channel outputs blew up.
I'd been working on the P-channel. The N-channel PS caps still had power after days of inactivity. Imagine my surprise.
Clearly, the drain resistors in the power supply are not doing their job - 2.2K at 3 Watts. Should be more than enough for a max 40 volt rail -- P = V^2/R, 1600/2200. Under a watt dissipation.
So, I replaced the outputs, checked that the output wiring to case was OL, and reinstalled the FE board and powered up.
No measureable flow across either of the N or P source resistors on the board I measured. DC offset was 35mV.
It would not budge when P1 and P2 were turned. Turned off the amp, the PS leds stayed on. Drained the PS manually with a 10ohm power resistor. NOTE: if there was a shorted output, the led in that channel would be off, and the DC offset would be rail voltage.
Checked the FE board Jfets. Seemed ok, but replaced anyway. Measured P1 P2. They were up around 400ohms and 700 ohms -- enough to drive the outputs.
Reinstalled the FE board to same effect, except DC offset now at 2 mV.
Checked Vgate voltage to gnd on the FE boards: 38 volts.
Pulled out the cascodes. Their HFE measures fine, at about 170 each.
The DC offset does not budge, no flow across the two source resistors.
The Speaker protection relay closes, showing continuity from output to binding post. If any flow anywhere, there would be a significant DC offset.
Horror show. I have no idea.
Last edited:
OK, so no real new news. It looks like resoldering the bleeders did the trick -- they are slow to drain fully, but should be safe. About 6min 30sec to get below 5 Volts without any load.
Please find an F5T .asc attached. The jfets are matched at 7.3 Idss. An LTspice sim with a +/- 15 V voltage source, and 1K resistors on the drains with Gate and Source shorted together to ground showed that perfectly.
If the cordell models are not supposed to be shown here, please delete. I believe I got them here.
Nonetheless, I don't know how to implement the thermistor ( I subbed a 4.7K resistor ) and I don't know how to deal with P1/P2. I set this up as P1=P2 = 0 ohms, so that there should be no current flowing across the source resistors.
However, there is.
Anyway, feel free to play with this. It looks like the published schematics need to have cascode transistors reversed. Once I changed it, it started to give reasonable values at the output when altering the resistor values I subbed for P1/P2.
I hope this gives a leg up to someone that wants to simulate it.
Please find an F5T .asc attached. The jfets are matched at 7.3 Idss. An LTspice sim with a +/- 15 V voltage source, and 1K resistors on the drains with Gate and Source shorted together to ground showed that perfectly.
If the cordell models are not supposed to be shown here, please delete. I believe I got them here.
Nonetheless, I don't know how to implement the thermistor ( I subbed a 4.7K resistor ) and I don't know how to deal with P1/P2. I set this up as P1=P2 = 0 ohms, so that there should be no current flowing across the source resistors.
However, there is.
Anyway, feel free to play with this. It looks like the published schematics need to have cascode transistors reversed. Once I changed it, it started to give reasonable values at the output when altering the resistor values I subbed for P1/P2.
I hope this gives a leg up to someone that wants to simulate it.
Last edited:
Actually, I think the attached file is fine. I think that version has the cascode transistors swapped and the resistors representing P1/P2 were chosen for little DC offset. Please check against schematic. P1/P2 are somewhere around 500 - 600 ohms.
I hope to use this to provide test voltages for the Front End board (Vgs) when there are no output boards connected -- and also when the boards are connected, but no load -- P1/P2 = 0. As well as when they are broken.
Note that you will have to set the source voltage as the reference to view Vgs for each polarity. If your boards are not measuring right, you *should* be able to short out or create open circuits to mimic the failure. At least I hope so.... beware of the diode bridge/thermistor or resistor at the audio circuit ground. It is not included here. Real life measurements to case ground will not be seen on the simulation, unless you add that bit of circuit yourself.
I hope to use this to provide test voltages for the Front End board (Vgs) when there are no output boards connected -- and also when the boards are connected, but no load -- P1/P2 = 0. As well as when they are broken.
Note that you will have to set the source voltage as the reference to view Vgs for each polarity. If your boards are not measuring right, you *should* be able to short out or create open circuits to mimic the failure. At least I hope so.... beware of the diode bridge/thermistor or resistor at the audio circuit ground. It is not included here. Real life measurements to case ground will not be seen on the simulation, unless you add that bit of circuit yourself.
Last edited:
corrected :
1: corrected circuit connections to output device bases.
2: used values for cascode as in my FE board
3; swapped Q7 and Q6 back to correct orientation.
Jfet operating voltages are pretty low. There is a Hum Breaking resistor installed ( R45 ). Changing this from zero to 1 changes the output DC offset near 10 mv. Changing R2 from 47.5K to zero is negligible.
Is there an optimal cascode voltage?
1: corrected circuit connections to output device bases.
2: used values for cascode as in my FE board
3; swapped Q7 and Q6 back to correct orientation.
Jfet operating voltages are pretty low. There is a Hum Breaking resistor installed ( R45 ). Changing this from zero to 1 changes the output DC offset near 10 mv. Changing R2 from 47.5K to zero is negligible.
Is there an optimal cascode voltage?
Last edited:
Got a little done today. Proceeding with caution.
DMM tests on the cascode BJTs show they are fine. Idss measurements on the Jfets show they are fine too, both around 7ma, using a 100 ohm resistor on the drain, and connecting Gate and Source together and running with a 8.9V battery. Good, the FE board is not eating Jfets.
LTspice says that the cascoded Jfets should see around 11-12V, even if the FE board is powered alone. Should probably short the ground and input when doing this.
Next steps: Measure Vdrain at cascoded Jfets, referenced to audio signal ground.
Question is: Should this be done with Jfets in circuit or out of circuit?
LTspice says the cascode transistors drop the voltage to about a third of rail voltage whether or not the Jfets are there.
I've put them back in, and don't want to take them out again.
If the cascode is functional, I think I will just try to bias again.... might have just stopped too early.
Will count how many turns from 0 to 5K on the trimmer -- 5K in parallel with 1K becomes 833 ohms at the test points. Verified.
DMM tests on the cascode BJTs show they are fine. Idss measurements on the Jfets show they are fine too, both around 7ma, using a 100 ohm resistor on the drain, and connecting Gate and Source together and running with a 8.9V battery. Good, the FE board is not eating Jfets.
LTspice says that the cascoded Jfets should see around 11-12V, even if the FE board is powered alone. Should probably short the ground and input when doing this.
Next steps: Measure Vdrain at cascoded Jfets, referenced to audio signal ground.
Question is: Should this be done with Jfets in circuit or out of circuit?
LTspice says the cascode transistors drop the voltage to about a third of rail voltage whether or not the Jfets are there.
I've put them back in, and don't want to take them out again.
If the cascode is functional, I think I will just try to bias again.... might have just stopped too early.
Will count how many turns from 0 to 5K on the trimmer -- 5K in parallel with 1K becomes 833 ohms at the test points. Verified.
Last edited:
Eric, you should be fine to test the FE board alone with the jfets installed. Don’t trust simulations when you remove parts in the sim. IMO you should verify that your power boards are ok before attempting to rebias. I think the optimal cascode voltage is in the 12-16volt range IIRC. Mine is set to 1/3 supply voltage which will be 15-15.5 volts typically.
Last edited:
Thanks Brian.
I've reflowed the solder for the bleeder resistors. Now, the bleeders work, but with their size and the size of the capacitor bank, they take over 6 minutes to discharge to under 5 V.
At this point, the bleeder is installed on the transformer side of the R in the CRC. It's probably better to install on the amp side. Maybe installing a second bleeder per rail on the amp side of the PS is a good idea, as it would cut the discharge time in half.
This amp was originally built 6 years ago and was my first amp build ever. I guess the bleeders on the Tea Bag boards took a bit more heat to make a solid connection than I could apply with a 25 watt dollar store soldering iron.
I'll measure the cascode voltage out of circuit and also measure the range of adjustment of both Vgates with respect to Vsrc.
There are still some wire locating issues I should clean up, which means possibly putting the main toroid on a raised floor. I am hesitating as that is a lot of work.
And there is still the DC relay/speaker protection issue as 24 VDC contacts got hit with near 40 VDC. Clearly, the speaker protection relay needs to be increased in DC switching capacity.
I've reflowed the solder for the bleeder resistors. Now, the bleeders work, but with their size and the size of the capacitor bank, they take over 6 minutes to discharge to under 5 V.
At this point, the bleeder is installed on the transformer side of the R in the CRC. It's probably better to install on the amp side. Maybe installing a second bleeder per rail on the amp side of the PS is a good idea, as it would cut the discharge time in half.
This amp was originally built 6 years ago and was my first amp build ever. I guess the bleeders on the Tea Bag boards took a bit more heat to make a solid connection than I could apply with a 25 watt dollar store soldering iron.
I'll measure the cascode voltage out of circuit and also measure the range of adjustment of both Vgates with respect to Vsrc.
There are still some wire locating issues I should clean up, which means possibly putting the main toroid on a raised floor. I am hesitating as that is a lot of work.
And there is still the DC relay/speaker protection issue as 24 VDC contacts got hit with near 40 VDC. Clearly, the speaker protection relay needs to be increased in DC switching capacity.
Last edited:
At last some progress!
Initial tests showed that the P channel cascode voltage was in the 36V range.
Checked the resistors making the cascode, by lifting one leg, measuring them to make sure, and reconnecting. Tested again. Cascode voltage still bad.
Removed the clip-on heatsinks from Q7 and the feedback resistors on the P channel and tried again.
The Q7 base voltage went to 12.85, and the emitter ( cascode ) to 12.28, in agreement with LT spice.
Woot!
Perhaps heatsinks were shorting out pins? Never happened before.....
Will verify N channel later -- not going to proceed further for a bit.
The overall build quality seems to be interfering with success. Eg. there is no reason for the heatsinks to be clip on. They should be screwed on. There is no reason for the feedback resistors to use heatsinks, when high power non-inductive wirewound are available.
Initial tests showed that the P channel cascode voltage was in the 36V range.
Checked the resistors making the cascode, by lifting one leg, measuring them to make sure, and reconnecting. Tested again. Cascode voltage still bad.
Removed the clip-on heatsinks from Q7 and the feedback resistors on the P channel and tried again.
The Q7 base voltage went to 12.85, and the emitter ( cascode ) to 12.28, in agreement with LT spice.
Woot!
Perhaps heatsinks were shorting out pins? Never happened before.....
Will verify N channel later -- not going to proceed further for a bit.
The overall build quality seems to be interfering with success. Eg. there is no reason for the heatsinks to be clip on. They should be screwed on. There is no reason for the feedback resistors to use heatsinks, when high power non-inductive wirewound are available.
JFETs are on the way. F-5T V1/2 TRANSISTOR KIT and PCBs waiting me at the post office. I ask because I will need help when I start building. I don't want to experiment with resistance values and solder-desolder several times.