I have been trying to build the CCS according to the schematic found here:
diytube.com • View topic - EZ Sink: Depletion-mode MOSFETs and the LM334
LM334 is configured for 7mA.
First attempt, I built the circuit, connected it to the 6SN7 plate, applied 250V B+ to the drain of the DN2540...and poof. No more MOSFET. Basically have an expensive resistor.
Thinking there was some issue with the way current is applied from the cathode warming up, I subbed the tube for a 1000 ohm resistor as a load. Ammeter shoots up to about 15-20mA...and no more MOSFET.
To check that I have the LM334 configured correctly, I remove the MOSFET from the equation. I hook up the LM334 to one of my low voltage bench supplies, using it to source current to the same 1k resistor, and it behaves as intended. I get a constant 7mA from the max voltage of the power supply down to the dropout voltage of the device.
I pop in another DN2540 (source connected to the V+ of the LM334, gate tied to the V-, and consequently, the load) in circuit, and the MOSFET fails once again.
I've been testing the DN2540s cascoded as a CCS for about a week prior and outside of one oops, they had been performing admirably in the exact same role. My goal with the LM334 was to be able to have a repeatable current reference vs the variability I had run into with using a current set resistor (I was just using a trimmer to get the current set) based on whatever the Vgs was of the particular 2540 I had plugged in at the time.
I'm about to pull what is left of my hair out over this. I'm down to a single good FET now and really don't care to destroy it...however it is managing to happen. Any ideas?
diytube.com • View topic - EZ Sink: Depletion-mode MOSFETs and the LM334
LM334 is configured for 7mA.
First attempt, I built the circuit, connected it to the 6SN7 plate, applied 250V B+ to the drain of the DN2540...and poof. No more MOSFET. Basically have an expensive resistor.
Thinking there was some issue with the way current is applied from the cathode warming up, I subbed the tube for a 1000 ohm resistor as a load. Ammeter shoots up to about 15-20mA...and no more MOSFET.
To check that I have the LM334 configured correctly, I remove the MOSFET from the equation. I hook up the LM334 to one of my low voltage bench supplies, using it to source current to the same 1k resistor, and it behaves as intended. I get a constant 7mA from the max voltage of the power supply down to the dropout voltage of the device.
I pop in another DN2540 (source connected to the V+ of the LM334, gate tied to the V-, and consequently, the load) in circuit, and the MOSFET fails once again.
I've been testing the DN2540s cascoded as a CCS for about a week prior and outside of one oops, they had been performing admirably in the exact same role. My goal with the LM334 was to be able to have a repeatable current reference vs the variability I had run into with using a current set resistor (I was just using a trimmer to get the current set) based on whatever the Vgs was of the particular 2540 I had plugged in at the time.
I'm about to pull what is left of my hair out over this. I'm down to a single good FET now and really don't care to destroy it...however it is managing to happen. Any ideas?
The schematic in the link has a fatal error, which will almost certainly destroy the FETs:
It has no protection for the FET's gate oxide. Even a high-voltage FET like the DN2540 needs protecting, because the fragile oxide layer will be ruptured with even a momentary voltage of 20V or more across gate-source, in either direction (look at the Maximum Ratings table, page 1 of its data sheet). The gate is high-impedance too, to make it easier to destroy!
It's easily fixed: for a depletion FET, put a zener diode across gate-to-source. The voltage must not exceed 20V even under pulse conditions, so I recommend a value less then 10V. In fact the DN2540 is in cutoff somewhat below 2V, so even a 2.7V zener will work. Connect cathode to source, for depletion FETs in this circuit.
There is another error in the text: with an active load in the source, one cannot omit the gate stopper of a FET, even if connected to ground. 100Ω is ideal, don't increase it too far from that with high-voltage applications; and in any case don't get the Drain supply too near to the 400V rating, even for a moment. The FET can be destroyed by brief D-S transients, too. Watch out for start-up and shut-down peaks.
It has no protection for the FET's gate oxide. Even a high-voltage FET like the DN2540 needs protecting, because the fragile oxide layer will be ruptured with even a momentary voltage of 20V or more across gate-source, in either direction (look at the Maximum Ratings table, page 1 of its data sheet). The gate is high-impedance too, to make it easier to destroy!
It's easily fixed: for a depletion FET, put a zener diode across gate-to-source. The voltage must not exceed 20V even under pulse conditions, so I recommend a value less then 10V. In fact the DN2540 is in cutoff somewhat below 2V, so even a 2.7V zener will work. Connect cathode to source, for depletion FETs in this circuit.
There is another error in the text: with an active load in the source, one cannot omit the gate stopper of a FET, even if connected to ground. 100Ω is ideal, don't increase it too far from that with high-voltage applications; and in any case don't get the Drain supply too near to the 400V rating, even for a moment. The FET can be destroyed by brief D-S transients, too. Watch out for start-up and shut-down peaks.
Last edited:
Rod is right int hat zeners are a belts and braces approach (and cheap.. so why not) but still, loads of people use the likes of dn2540 without them with no issues (I have 4 ccs in operation, at higher voltage and current than yours, and they have been happy for years.
Some basic questions -
Are you isolating them properly?
What is the schematic, something seems odd to me, 100% sure you have the right pinouts?
where did you buy the supertex - some people report dodgy chinese copies from the ebay / aliexpress.
Some basic questions -
Are you isolating them properly?
What is the schematic, something seems odd to me, 100% sure you have the right pinouts?
where did you buy the supertex - some people report dodgy chinese copies from the ebay / aliexpress.
Last edited:
The DN2540 (in CCS "upper" position) tend to destroy at power on (especially over 300V B+), even with protective zeners.
Several years ago I lost a few bunch of them, so in this position I use exclusively -about- 1kV FETs.
Several years ago I lost a few bunch of them, so in this position I use exclusively -about- 1kV FETs.
> loads of people use the likes of dn2540 without them with no issues
Usually, they have a low-value resistor between gate and source, to make a current-source. In such a case it is much more difficult to find >20V peaks, and the circuit typically survives without protection.
But with a current-source IC in the source lead, we have an arrangement with a purposely high dynamic impedance. It's a completely different situation, and G→S protection is essential. I'm not surprised that it does not survive even one power-ON event without it.
The zener will also protect the LM334 from overvoltage damage, so the value of the addition is doubled.
The DN2540 is not avalanche-energy rated for the D-S voltage rating (as switching FETs often are); so vanishingly-short peak-voltage pulses at 400V+ will destroy it. Stray inductance around the wiring could be enough to blow them, at switching events, if the DC voltage is already high. A 10Ω resistor plus stacked-construction film capacitor (1-5µF MKP) in the drain lead will increase robustness - but the capacitor's wiring loop around drain-to-gate must be low-inductance: for example strips of copper-foil for the ground-run.
Usually, they have a low-value resistor between gate and source, to make a current-source. In such a case it is much more difficult to find >20V peaks, and the circuit typically survives without protection.
But with a current-source IC in the source lead, we have an arrangement with a purposely high dynamic impedance. It's a completely different situation, and G→S protection is essential. I'm not surprised that it does not survive even one power-ON event without it.
The zener will also protect the LM334 from overvoltage damage, so the value of the addition is doubled.
The DN2540 is not avalanche-energy rated for the D-S voltage rating (as switching FETs often are); so vanishingly-short peak-voltage pulses at 400V+ will destroy it. Stray inductance around the wiring could be enough to blow them, at switching events, if the DC voltage is already high. A 10Ω resistor plus stacked-construction film capacitor (1-5µF MKP) in the drain lead will increase robustness - but the capacitor's wiring loop around drain-to-gate must be low-inductance: for example strips of copper-foil for the ground-run.
They came from Mouser, so no worries there. I don’t buy anything but resistors and maybe the odd jellybean transistor off of eBay or aliexpress.
I double checked the pin outs after the first one went. I’ve got a bad memory for that kind of stuff anyway, so I usually scribble down the pin outs of any device with more than 2 terminals. My own sanity check.
Thank you for all of the suggestions.
What you suggest is probably spot on, this is all on a breadboard, so I have a ton of leads everywhere, plus the capacitance/inductances between the strips in the breadboard itself. What really had me confused was the device failing even when I stepped the voltages down to less than 10v. Looking at the data sheet I thought I would have been well within the safe region of the device at that point.
A gate stopper was my next step, but given that I was down to my last device, I didn’t want to be completely SOL until I could order some more so I thought I’d seek some help first. I’ll probably put in another order (I’ll just get 50 of them this time!) and add protection for the gate and drain when I try it again.
LM334 current is a function of temperature (227uV/C) and it's noise is likely higher than a resistor. See the datasheet.
Last edited:
Asking for my understanding...
The data sheet says that the noise level should be acceptable in a single stage transistor voltage amplifier with a gain up to about 2000. Since I’m only using it up to a gain of roughly 20 (before feedback) then shouldn’t it be negligible anyway?
How would it compare to the noise of the cascode of 2 DN2540s?
The DN2540 has excellent low noise credentials. Walt Jung measured approximately 1.1nV/rt.Hz into a 100Ω ac-load. So the current noise is ca. (1.1nV/rt.Hz)/100 = ca. 11pA/rt.Hz. This is a measured figure, not a modelled or data sheet claim. It is unexpectedly good for a MOSFET; these usually suffer from lots of excess flicker-noise.
At 5mA, the LM334 gives a typical figure of 250pA/rt.Hz (100Hz and up). Quite a bit more, but it depends where it is used in your amplifier chain, and what impedance it works into.
At 5mA, the LM334 gives a typical figure of 250pA/rt.Hz (100Hz and up). Quite a bit more, but it depends where it is used in your amplifier chain, and what impedance it works into.
It will be the 1st amplification stage, direct coupled (probably) into a LTP. I was going to go with a cathodyne for simplicity sake, but I apparently won’t have the headroom for anything but a negligible amount of feedback.
I assume that the effective load would be the output impedance of this stage in parallel with the the following one, then use ohms law to determine the noise voltage?
I assume that the effective load would be the output impedance of this stage in parallel with the the following one, then use ohms law to determine the noise voltage?
I destroyed some because the tip of my soldering iron was not grounded: electrostatic discharge killed them!
> I assume that the effective load would be the output impedance of this stage in parallel with the the following one, then use ohms law to determine the noise voltage?
Yes, that will translate the data sheet spec into V/rt.Hz. You can extrapolate the typical values curves to suit the current you plan to run, and use them to take account of frequency.
Since you are nearly there with the build, do an early noise test! I would run the CCS into a single resistor, similar to the real load, and check the rms noise for real. Then you can calculate where it will get to at the speaker end, where S/N can be considered (e.g. at 1 Watt output, rms comparing rms noise).
Yes, that will translate the data sheet spec into V/rt.Hz. You can extrapolate the typical values curves to suit the current you plan to run, and use them to take account of frequency.
Since you are nearly there with the build, do an early noise test! I would run the CCS into a single resistor, similar to the real load, and check the rms noise for real. Then you can calculate where it will get to at the speaker end, where S/N can be considered (e.g. at 1 Watt output, rms comparing rms noise).
Yes, grounded soldering irons are important! And even with low voltage circuits and closely-spaced leads, like the standard TO-220 format, using no-clean solder is worth the trouble. And/Or carefully cleaning off all the flux after everything has cooled down. The flux from older types of solder wire can be incredibly current-leaky.
I never thought about the need for a grounded soldering iron. I wondered why mine had a 3 prong plug...
How likely is it that components get damaged but don't completely die so the end product still suffers.
How likely is it that components get damaged but don't completely die so the end product still suffers.
unfortunately this works the other way round, too
you can easily charge your body by walking over a carpet made of synthetic materials,; if you now pick up a mosfet part for soldering, this charge (up to 100pf charged to several kV) kills the mosfet gate *because* your iron is grounded ...
when I worked in the industry long time ago they exclusively used those Weller irons which had a safety transformer and 12V heaters with just 2 wires and deliberately no safety earth connection; instead all workers handling mosfet components in both lab and production had to wear wrist straps which were earthed through several series connected megohm resistors to prevent the person and the components being charged; you could get fired if you didn't wear your strap ...
such things happen, I had mosfets with internal gate protection develop many microamps gate leakage due to handling (charged body effect) but still working as switches; presumably the internal diodes did their job and prevented gate oxide rupture but got leaky themselves in doing so ...