Constant Current Diodes in Valve Circuits

[GussyLoveridge]

I've not seen constant current diodes used in valve circuits - at least any that I've seen personally.

Could we use them in the cathode circuit for common cathode stages?

I'm guessing there is some reason we don't right?

 

[Rob Flinn]

I have seen circuits that use a current source. i.e a transistors with some resistors.

 

[GussyLoveridge]

Oh yes, I've certainly seen those as well!

 

[abbey road d enfer]

I've not seen constant current diodes used in valve circuits - at least any that I've seen personally.

Could we use them in the cathode circuit for common cathode stages?

I'm guessing there is some reason we don't right?

What advantage do you expect?

 

[GussyLoveridge]

I don't think I know enough about it yet to infer any advantage. Only that through my readings I've seen mentions of constant current sinks and sources (both valve and solid state) being used in various ways often to flatten the load line and reduce distortion (I think?) but I've not seen a constant current diode used.

It's just a curiosity.

 

[abbey road d enfer]

A CCS in the cathode would result in constant current in the anode, which yields zero output. Now decoupling it with a capacitor may, or may not give interesting results.
However, using it as anode load may increase the output headroom, just like a parafeed. That would be if the CC diode can withstand the voltage.
I guess it would work also in a cath follower.

 

[merlin]

A constant-current diode is basically a JFET with a built in resistor, and you do see those used in valve circuits occasionally. The applications are a bit limited though, since the voltage ratings are not high. You could use one as the tail in a differential pair.

 

[thor.zmt]

I've not seen constant current diodes used in valve circuits - at least any that I've seen personally.

I have. A very expensive Hai Enten Brand from Japan even used them as Anode load, the 100V Version.

Could we use them in the cathode circuit for common cathode stages?

Sure, it will nail down the Anode voltage even with varying tube characteristics, a bypass capacitor (in series with a resistor if we want to degeneration) will be needed

I'm guessing there is some reason we don't right?

They are relatively rare and they do not like high voltages that well. That limits utility in Tube circuits.

Given that an active anode load with a TO-220 High voltage FET is easily arranged and can handle high voltage much better, they are more common.

One little trick is to replace the Cathode Resistor of the upper tube in a SRPP with CRD. Instant hybrid Mu-Follower.

Thor

 

[Cqwet Dbdfte]

Use the DN2540 depletion mode MOSFETs, cascoded. Handles 400V.
No CLD will do that
As a plate load on a triode you get the mu gain.
Alternatively a really high voltage enhancement mode MOSFET controlled by an LM431 type, fed ~ 1mA from a LND150.
Did that 40 years ago, for a 500Vpp driver.

 

[Rusan]

Use the DN2540 depletion mode MOSFETs, cascoded. Handles 400V.

I'll second that! For triode plate loads, I've used a DN2450 cascoded with both another DN2450 and with a LM334, and the IXYS 10M45 as well. The DN2450/LM334 gives about 150M impedance at 50Hz, 12M at 1kHz and 2M at 20kHz. To my ears, it does some really nice sounding things when sitting atop a 6922. Full mu and a flat loadline, without resorting to that pesky, old-fashioned 10kV PSU and 2kW plate resistor...

With the CCS-loaded 6922, THD is extremely low and nearly pure 2nd harmonic up until pushed hard, where some fairly significant 3rd harmonic content enters. However, as one would expect, at the onset of clipping the distortion skyrockets all at once. I've also experimented with 80H inductors as 6922 plate loads, which aren't as low distortion as a MOSFET CCS due to lower overall impedance, but they do allow it to swing up to the full B+ voltage with (relatively) low distortion.

On a side note, I've used the ZVN0545A and IFR820 high voltage MOSFETs as source followers after 12AX7s and 12AT7s, and they seem to work quite well. Clean, quiet, no filament supply, and gobs of headroom.

 

[Cqwet Dbdfte]

If you use a FET in a current source for a triode you can take out the signal at the FET's source rather than the plate, as the load current will not add to programmed load current to tube.
Or you can look at it as a plate has high output impedance while a FET source has low output impedance.
A resistance loaded triode has varying current which will make large signals asymmetric.
Using HV MOSFETs for followers makes perfect sense as no voltage gain is extracted and all the gain is used to buffer the signal.
6922, 6DJ8 etc are very nice tubes. Very linear with constant current, I use 10mA with LED bias.

 

[Wavelength]

GL,
Yes using them for a cathode follower is a good idea. The gain will reach 1 and have a pretty big swing before they fall out of regulation. I have used them for buffers in pedals and output circuits when the tube is biased from a gain stage. You have to watch the voltage max and min but all good.
If your using this with the grid referenced to ground you might be better off with a large choke. I do that a lot for direct coupling single ended tube amps as it can get ~20% more power in Class A and with some tubes A2.

 

[Rusan]

I do that a lot for direct coupling single ended tube amps as it can get ~20% more power in Class A and with some tubes A2.

I once built a direct-coupled SE amp using a choke-loaded 6922/6DJ8 driving a triode-strapped EL34, with the 6922's power supply node being the top of the EL34's cathode bias resistor, thus creating a self-biasing DC feedback loop. It was somewhat trendy back 20 years ago, and was called the "direct reactance drive" topology.

It had no instabilities that I could find, and sounded unbelievably sweet. Trioded EL34s have quite low 3rd harmonic and extremely low higher orders, but very high 2nd harmonic. I used a 5K reflected load instead of the more usual 2.5K-4K to maximize overall linearity, at the cost of a fraction of a watt output. Unfortunately, I never got to put it on my buddy's HD analyzer to see precisely what was going on. I just know it sounded killer. I never tried the topology with a big beam tube or directly heated triode, but I'll bet it packed plenty enough muscle to drive a 6550 or 300B into A2.

 

[Cqwet Dbdfte]

The ear (reportedly, 30%?) masks 2nd harmonics, as the inner ear generates its own, but 2H really upsets %THD meters.
This is very unfortunate, as audio technology development has been feeding off these meters readings since they began.
Push-pull amps nicely suppress even order harmonics in the output transformer, besides phase splitter also has this tendency.
We can still live with all kinds of push-pull harmonics as long as it is inaudible/tolerable. With 50W of tube idle power and ~1W loads my 4x6550amps would rarely leave class A region.
SE tube amps is a great idea, but their low power makes speaker choices problematic. Would be nice to have the space for some 105dBspl/1W/1m corner horns.

 

[thor.zmt]

Use the DN2540 depletion mode MOSFETs, cascoded. Handles 400V.
No CLD will do that
As a plate load on a triode you get the mu gain.

Or use 1/2 5687 and put the J-Fet CCS/CRD into the cathode.

It also offers a Mu-Follower output from the 5687 Cathode. I preferred this over all solid state options.

The probable reason is that all the larger high voltage FET's have a lot of parasitic capacitance. And nonlinear capacitance at that. A real tube with a J-Fet CCS/CRD is much less variable and much lower capacitance.

https://www.bartola.co.uk/valves/tag/mosfet-follower/

One example where I have used this was a 5842 / 5687 Mu-Follower DC coupled to a 300B Grid in a "stacked supply" Amplifier. Another occasion was with two halves of a 5687 forming the amplifier, fixed grid bias on the lower tube. The 5687 has enough bias voltage to get a J-Fet CCS/CRD into the flat part of the V/I curve, a 12AU7 would as well, a 12AX7 on the other hand, no.

Thor

 

[Cqwet Dbdfte]

@thor.zmt The high perveance 5687 (9H) would be more popular if it had a 12AU7 type (9A) pinout. A 300% higher heater current may also be a obstacle. Not for the tube rollers. I have quite a few spare 5687's.

I like the Bartola DN2540 circuit (that's the one you are referring to?), but the first FET has the input at 0V, and the output is still AC-coupled. The split PSU adds complication, maybe not better than an input cap?

Using current sinks in cathode followers may have some application, I prefer a resistor, a chassis mounted one may be needed as it worked out better in my DC coupled 211 driver. I could not see any improvements in my circuit with a current sink. I have not tried a MOSFET for it either, using 6BL7 paralleled.

The DN2540 does not really fit into the "Large HV MOSFET" category, and what happens to the bootstrapped input capacitance in a follower?

Thanks!

 

[thor.zmt]

@thor.zmt The high perveance 5687 (9H) would be more popular if it had a 12AU7 type (9A) pinout.

I would hate that. As is getting NOS 5687 is still afforable.

The SinoSov 6N6/6N30 can make a standin - with a 6DJ8/6922 pinout.

I like the Bartola DN2540 circuit (that's the one you are referring to?), but the first FET has the input at 0V, and the output is still AC-coupled. The split PSU adds complication, maybe not better than an input cap?

I referred to Bartola's site for his notes on the slew rate limiting and the excess HF distortion all those large die FET's produce.

I personally use the STN1HNK60 (SOT223 SMD, there are TO-92 and TO-220 version with the same die), with Crrs 3.8pF and Ciss 156pF is a better choice than most, but it still looses to most tubes.

Using current sinks in cathode followers may have some application, I prefer a resistor, a chassis mounted one may be needed as it worked out better in my DC coupled 211 driver. I could not see any improvements in my circuit with a current sink.

Well, if (say) we have -12V (heater) and +120V (anode) available, connecting a CRD will allows around +20dBu signal swing, without loading the cathode. Do that with a resistor.

And as CCS it is 1 part, compared to 4 for a ring of 2 and it is 2-terminal.

A application I really like is shown here (tube part):

The DN2540 does not really fit into the "Large HV MOSFET" category, and what happens to the bootstrapped input capacitance in a follower?

Crrs is not that low. And yes, Ciss is bootstrapped, but depletion FET's have fairly low transconductance compared to enhancement types, but where does the current to charge and discharge it come from?

Are you sure you read the articles at the Bartola Site? Start here:

https://www.bartola.co.uk/valves/2016/01/09/slew-rate-part-iv-and-source-followers/

His latest anode load gyrators use DN2540 to bootstrap a J-Fet at the bottom... Kind'a (yes, I know it's different) like I do with a Tube instead of DN2540.

Thor