JFETs as switch in 2-channel tube amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The supplies should hold up under all possible loading that can be presented.

The cap across the zener can't work as a reservoir cap because the zener clamps the peak voltage to 15v. As the AC cycle continues the voltage can only fall across the cap. If the cap were before the 150K then it would maintain a constant voltage (and hence constant current) across and through the zener. If you were to scope the 15 volt zeners I think you would see the supply isn't clean.

If you are simulating it then replace the zeners and their feed with a constant voltage supply (two 15 volt supplies) and get the circuit to work correctly from that :)
Unfortunately I don't have an oscilloscope, not yet anyway... :eek:
Anyway, I bought extra components for implementing what you suggested before, but first I tried replacing the resistors I was talking about to 2M2 and 33k, since it didnt require any drilling and rearrangement on the protoboard. I am attaching the pic with the voltages I measured.

However, the amp does not sound nice, there is still distortion present... :(

UPDATE:
I had a look at 2N5638s datasheet. In the 'Switching Characteristics' section it mentions Vgs=-12 V. Does it mean that I can/have to provide it with more negative Voltage for it to turn off?
 

Attachments

  • SwitchingcircuitVoltages.png
    SwitchingcircuitVoltages.png
    163 KB · Views: 108
Make sure that the zener stabilised supplies are OK under all switch conditions. I would have thought it better to rectify and smooth the HT first and then apply to the zener via a resistor or at least split the 150K witth a cap. You have to have -15 volts across the zener.
I missed to mention that. The V- rail supply was not stable with the 470k-6k8 pairs, but got stable with the 2M2-33k values!
 
Administrator
Joined 2007
Paid Member
UPDATE:
I had a look at 2N5638s datasheet. In the 'Switching Characteristics' section it mentions Vgs=-12 V. Does it mean that I can/have to provide it with more negative Voltage for it to turn off?

No, it means it will block peak signals up to the value of the Vgs voltage. Look at my link way back in post #11 :)

If the sound is distorted then the first thing to do is to simply bias the gates on via a 6K8 and 470K divider (as in your circuit) and make sure the sound is clean. Feed the divider from a clean 15 volt supply. Or just use a 9 volt battery and recalculate either of the two resistors.
 
Has anybody used HEF4066 to switch audio signals.

I have used several types of CMOS analog switches. They work great in a circuit where the total peak to peak voltage range of the analog signal applied to the switch is between Vdd and Vcc of the chip (usually 5 volts). Atempting to go outside this range activates the ESD clamp diodes on the input or output pins. This can cause distortion or chip death depending on the available drive current. They are useful in a solid state amp design, but don't work so well in between a pair of 12AX7's.

The cheap LDR isolator that I specified is rated for 60 volts. I have used it to well over 100 volts of signal in high impedance circuits without issues. It is also rated for 50 mW maximum dissipation. There are LDR isolators rated for 1/2 and 1 watt and a few hundred volts. These can be used for variable plate load resistors with a 12AX7.
 
I just found this thread. I see that there has been considerable discussion regarding a fairly complicated fet based channel switching scheme for a guitar amp.

I use simple LDR's. It is a Cadmium Selenide photoresistor and an LED inside a sealed package. LED on... resistance is about 100 ohms. LED off resistance is about 10 meg ohm. They work good for channel switching. The resistance curve is not linear but they are good enough that you can make a channel transition with a pot based foot pedal.

Soldano used LDR's for channel switching in some amps (SLO100). They have one extra stage in the overdrive channel so that the channels are out of phase at the summing junction. This way any leakage cancels.

Soldano used Vactec VTL5C1 which are expensive. I use these:

Hello Tubelab!

Yes, LDRs sound like they can handle the gain in a tube amp. The original design had LDRs but I chose to experiment with JfETs primarily due to their low resistances and their sturdiness in time, although i may be wrong here... I will definitely look into them! :)
 
No, it means it will block peak signals up to the value of the Vgs voltage. Look at my link way back in post #11 :)

If the sound is distorted then the first thing to do is to simply bias the gates on via a 6K8 and 470K divider (as in your circuit) and make sure the sound is clean. Feed the divider from a clean 15 volt supply. Or just use a 9 volt battery and recalculate either of the two resistors.

Mooly,
I was able to make some measurements while playing tonight. The amp sounds clean only in low gain, meaning I had to have the Volume pot on my guitar to 1-4 to get clean sound. Higher than that, there is distortion.

I then tied my DMM to each channel's JFET gates with crocodile clips and the voltage was swinging from 0.2 to -4 at the ON FET's gate and from -8 to -2 at the OFF FET's gate when i was playing, especially when strumming chords and the low strings(high output).

Btw, on sunday I'll put the extra caps/resistors to the power supplies, as you suggested, and test/measure voltages again.

Is a Source-Gate resistor relevant here? Can it be used to offset the signal voltage to Vc?
 
Last edited:
Administrator
Joined 2007
Paid Member
Try and go back on basics and confirm the conditions are correct.

1. The drain and source points in the circuit that all the FET's connect to should have zero volts DC present. If those conditions are met then there should be no need for any "offset" which would be very difficult.

2. Measure the gate voltage at Q4 and Q5 collectors because measuring on the gates via the 1Meg resistors will not be a true reading due to the small current the meter draws.

I still think there is a problem with the supplies :)

Also without a scope and seeing what the peak signal levels are, then it becomes more difficult to see where what is going on and where things are going wrong.
 
If you can't get the fet switching working I have a pretty simple relay switching desgn that I use is a couple amps that works pretty good. It uses two DPDT relays to switch between 2 channels and it has LED indicators to tell you what chanel you are on.
Hey Minion!

I am starting to think that Vpp levels are too high for the FETs and their gate supply levels, although I have to measure more and try one more topology... If all else fails I will definitely fall back to relays (although I'd like to try out LDRs too). Thank you very much, I appreciate it! :)
 
Try and go back on basics and confirm the conditions are correct.

1. The drain and source points in the circuit that all the FET's connect to should have zero volts DC present. If those conditions are met then there should be no need for any "offset" which would be very difficult.

2. Measure the gate voltage at Q4 and Q5 collectors because measuring on the gates via the 1Meg resistors will not be a true reading due to the small current the meter draws.

I still think there is a problem with the supplies :)

Also without a scope and seeing what the peak signal levels are, then it becomes more difficult to see where what is going on and where things are going wrong.
I nailed it!
I was curious this weekend to try out Kevin O'Connor's method which has the following differences:

1. Reversed biased 1N4148s on all FET gates.

2. +/-10 VDC for the Supplies, contrary to the -10/+0.2Vdc I was trying until now.

These 2 changes made the trick. And the experience gained is huge. i haven't really got the whole concept really well but it's a start!

I'll post the schematic when I get home. :)

Thank you Mooly! :)
 
Hi Guys

To use jfets as series switches requires that the control voltage for the gate exceed the signal amplitude being controlled. Since the signal is referenced to ground and is therefore bipolar, so must be the control circuit voltages.

For example, if the signal is 10V-peak, then the jfet control signal must be +/-12V or a higher. Using a unipolar gate voltage will cause distortion of half the signal.

The control circuit shown earlier in this thread is only good for shunt-connected jfets - not series ones. It is a simple matter to modify the circuit to provide this control range by adding NPNs tied to the -13V rail with their collector loads tied to +13V. The current here is negligible and can be drawn directly from any plate node that is well filtered.

For jfets used as series audio switches, the control lines to the gates should be ramped using a series 1M and 100pF to ground at the jefet gate-diode end. This will control thumps and provide a good transition from one channel to the other.

Generally, shunt switching is preferred whenever it can be used, as it is inherently quiet. The only downside is that the shunt element must work against a series resistance in the signal path. In a tube guitar pre, there are already series resistances, so this is not really a concern.

TUT has a 90-pg Switching Methods chapter. TUT7 (out next year) has another huge bit of info in the Automation chapter.

Have fun
Kevin O'Connor
londonpower.com
 
Hi Guys

To use jfets as series switches requires that the control voltage for the gate exceed the signal amplitude being controlled. Since the signal is referenced to ground and is therefore bipolar, so must be the control circuit voltages.

For example, if the signal is 10V-peak, then the jfet control signal must be +/-12V or a higher. Using a unipolar gate voltage will cause distortion of half the signal.

The control circuit shown earlier in this thread is only good for shunt-connected jfets - not series ones. It is a simple matter to modify the circuit to provide this control range by adding NPNs tied to the -13V rail with their collector loads tied to +13V. The current here is negligible and can be drawn directly from any plate node that is well filtered.

For jfets used as series audio switches, the control lines to the gates should be ramped using a series 1M and 100pF to ground at the jefet gate-diode end. This will control thumps and provide a good transition from one channel to the other.


Hello again Mr. O' Connor! :)

Glad to 'meet' you here! Your proposal works great. As you can see from the topic, I tried some other topologies that, despite the fact that they didn't work in my application, offered me lots of experience.

As for your books, I have 4 of them (TUT,TUT4, RSG and SPKR) and I totally recommend them!
Struth said:
Generally, shunt switching is preferred whenever it can be used, as it is inherently quiet. The only downside is that the shunt element must work against a series resistance in the signal path. In a tube guitar pre, there are already series resistances, so this is not really a concern.
Shunt switching is what I wanted to use in the beginning, but I was unsure of the exact topology (resistance values etc.) that it required. I will be testing it, however, in a future project
Mooly said:
Well done you


Yes it would be interesting to see the final arrangement that worked here.
Sorries for not having the time today, I will be posting it tomorrow with the voltages. Check my #36 post for Kevin's circuit, it's the one on the left.

Cheerz!
 
Last edited:
Ok guys, here it is.

Now a question emerges... what if I wanted to raise the gate voltage a bit more than +/-7.15? The thought behind this is to have more headroom...

Would it be wise to change the 1MΩ resistors with something smaller, say 680kΩ? Or is there another way to do this?

:scratch2:

Thanks!
 

Attachments

  • SwitchingcircuitVoltages.png
    SwitchingcircuitVoltages.png
    428.2 KB · Views: 56
Administrator
Joined 2007
Paid Member
:)

Easyone first (hopefully :)). The FET gate draws zero current at DC (if the biasing is correct and within limits) so the 1meg value shouldn't matter. 100k, 1meg or 100meg, the gate voltage should be the same although not measurable accurately without a FET input voltmeter (or old type valved VOM).

Hmmm... so the diodes seem to get the desired result. If it works it works :) although I see a "floating" gate now...
 
Administrator
Joined 2007
Paid Member
Yes, yet as I measured them while playing, the gate voltage changes according to the signal present in the channel! Those 1N4148 are really fast on this! :)

Btw, I really have to get a Fluke... I don't trust my $20 DMM anymore...

I can see how you say the gate changes according to the signal. The gate has virtually infinite impedance at DC so the voltage on the other pins can "transfer" via leakage and at AC via the junction capacitance of the FET.

Its one of those things where you need the working circuit in front of you to probe and test and see exactly what is going on.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.