• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Tube input and mosfet output

Status
Not open for further replies.
Hello, i have been playing around with this combination - just for the fun of it - sweet tube sound and power - is it worth building ?

Do you have any comments ?

Kind regards
 

Attachments

  • tubemos.jpg
    tubemos.jpg
    35.5 KB · Views: 845
Hi Stormo,
I don't think that will perform well. The impedance faced by the driver tube is too low. I don't understand why the 220K resistor is there either, it forms a voltage divider and lowers the impedance still more. You need to swing 70 V peak to drive this. A 10uF coupling cap will be expensive and can be reduced if you get the impedance up to a reasonable level.

My feeling is that you should avoid parallel tube sections. Possibly use one 12AX7 section for each channel and a follower to drive the output stage. A classic choice for that would be a 12AT7, others will have an opinion on that. Your output stage needs work too. Your drivers are operating in class B, could be argued class C. There will be crossover distortion from the driver stage. Add gate stopper resistors to the output FET's.

It will make noise as it sits. I don't know that you would like the noise.

-Chris
 
I can't comment on the mosfet schematic.

However generally speaking :
The 12ax7 is bound to struggle driving the input capacitance of the mosfets.

If possible you should try to use something like an ECC99/6N6P/5687 or similar.

I know a guy who is helping his buddy build something like that. Maybe he could give you a couple of pointers. He posts onder 6h5c. http://www.diyaudio.com/forums/member.php?s=&action=getinfo&userid=14574

Regards,
Bas
 
I agree with Anatech here...
12AT7 would give you approximately the right gain with littel or no feedback, and a lot less output impedance.
You can make things better by converting the first NPN into a follower configuration, terminated with a bootstrap current source (essentially, change places of BJT and bootstrap). This will increase your output stage's input impedance a LOT. You need to bias the base of this BJT to approx. 0V plus one diode drop, easily done with a pair of resistors and trim pot (although there are better ways, you may want to try using a Baxandal pair instead of a sole BJT).
The bias for the output stage is insufficient. It needs two diodes to compensate for two BE junctions, but in order to make this thermally stable, you need to include some emitter resistors. In particular, split your uotput BJT/MOSFET CFPs into lower and upper, then add two resistors in series (about 0.1-0.15 ohm should do), connect output to center point. The two diodes should be carefully chosen (do not use 1N4148 and the like as their forward drop is too high) and in good thermal contact with the driver BJTs. You may even use diode conencted BJTs of thes ame kind as the drivers, and mount them one on the other (keep in mind you need to isolate the metal tabs frome ach other!).
Your output MOSFETs are not chosen too well, I'm afraid. For starters, IRF9540 and 640 make a better pair. 540 and 9640 have well over 2x difference in gm! IRFI parts are not suitable for outputs although their isolated case may be very attractive - the thermal resistance to heatsink is too high for this application.
 
Hi
Several problems on first glance, but an initial word.

Valves are voltage drive, transistors are current drive, try not to mix and match.

Move the 220K resistor from valve anodes to ground to inbetween the 10uF cap & 1K and ground, you are dragging the anode down.

Biasing the first transistor with resistors from the + and - rails, not good, you cannot cancel the noise or ripple, find altermative method that is clean and free from nasties.

Cascode output stages are very unstable in terms of biasing and are prone to oscillate, you can use a single pair of mosfets to drive a speaker far more effectively and remember the first statement about drive.

IRF540 and 9640, bad mix, the RDS-on is totally wrong, match the fets for the same RDS or close to, also make sure the VDS is high enough.

I would be more inclined to use a mosfet to drive the output stage rather than transistors, simple and more effective, three mosfets and a single valve and you have more than enough power drive + matching the fets and you have 40W easy.

Lastly, HT, you do not need this much, 200V maximum maybe even less than this, you only need to swing 10V rms clean with an output stage working properly.

With all amplifier designs, the simpler it is the more transparent.

Have fun

Chris
 
Hi Chris,
Actually, there are a few hybrid amps out there using bipolars and tube front ends. They work fine. Most of the distortion and problems come from the tube section.

Bipolar outputs are more linear and have less distortion than mosfet outputs. I don't know if this applies to lateral mosfets. It is not very hard to make the current amp section a high enough impedance for the tube to drive.

If the gain stage = 1, then you need about 50V peak of drive (for +-25 V supplies). The high voltage on the tube section allows a more linear curve to be followed.

-Chris
 
chrisfound said:
IRF540 and 9640, bad mix, the RDS-on is totally wrong, match the fets for the same RDS or close to, also make sure the VDS is high enough.

Definitely a bad mix but Rdson is not t he parameter you should be looking at, although it is correlated with the one you do need to look at.
Although CFP makes the circuit less susceptible to it, the parameter you need to match is gm (Id/Vgs relationship), just take care to match by curves, not by the typical spec number as this is often taken under different conditions for different MOSFET types, so you cannot compare directly.
Matched gm (also often marked S) also, to some extent, means matched Rdson and even total Cin, but will give you very different maximum Vds, typically 2x higher for the N-type device. Not particulairly important, as long as it's well over the span between the power rails.
I do agree that CFP tends to be problematic, be especially careful with inductive 'Re' (putting this in quotes as it's reallt the emitter degeneration resistor for the equivalent transistor formed by the BJT and MOSFET CFP combo). Mosfet Cgs and lower gm makes it a bit less susceptible to oscillation, but the BJT collector resistor has to be chosen carefully.
 
Hello everyone and thanks for your comments 🙂

Some changes have been made :

* Still parallel 12ax7 tube input
* old 220 K resistor to ground from anode is removed
* 1.st bjt is now a follower at 8.5 mA, still biased from the rail
* a Vbe multiplier has been added, holding Mosfets at 120 mA
* driver stage removed
* gate stoppers added for stability
* emittor resistors added for stability

Next step is building 🙂)
 

Attachments

  • tubemos.jpg
    tubemos.jpg
    30 KB · Views: 433
Stormo said:
Hello everyone and thanks for your comments 🙂
Some changes have been made :
* Still parallel 12ax7 tube input


At some point you may want to split up part of the cathode resistor into two separate resistors, one for each section of the tube, and adjust them for equal current. Not extremely important so you can leave this for later.

* old 220 K resistor to ground from anode is removed
* 1.st bjt is now a follower at 8.5 mA, still biased from the rail

Try splitting the 3k resistor from forst BJT emitter into 2x1.5k, and conenct the midpoint to the output via a large-ish cap (220uF or so, preferably bypassed by a small foil cap). This will gibe you a bootstrap 'current source' for the follower to work into.
Without it, you will have a rather uneven slew rate for positive and negative going signals, also, this problem will worsen the larger the input swing.
Also, your biassing for the follower will inject hum, ripple, and a portion of the fullwave rectified output back into the input (keep in mind you are driving it with a high impedance sopurce!). You can fix this in a fairly simple manner: disconnect midpoint of 248k and 570k resistor from the base of the follower BJT, then reduce the value of these resistors tenfold or so. You may want to sever their conenction and put a trimmer pot inbetween. Then, connect the wiper of the trimmer pot, or the conenction of the resistors if you have not insterted the trimmer, to ground via a capacitor at least the same value as your coupling cap, and also, to the base of the BVJT via a 150k resistor. If you do not insert the trimmer, your resistor values will need some adjustment (248k will need to be lower or 570k higher) in order to get zero bias. The input impedance of the follower at DC will remain largely unchanged, and will be 150k for AC. The advantage is that now your bias is referenced to AC ground where there should be no hum.

* a Vbe multiplier has been added, holding Mosfets at 120 mA

Your BJT may overcompensate some with temp rise, but for testing this will not be a serious issue. There are several easy ways to cure this.

[/quote]
* driver stage removed
* gate stoppers added for stability
* emittor resistors added for stability
[/quote]

You mean source resistors. These will not increase stability (in fact, if they are signifficantly inductive you may run into serious trouble), because the BJT multiplier used for biasing already overcompensates with temperature, with the resistors it will overcompensate even more. You also lose gm with the resistors and increase the output impedance (which will be highish to begin with). In either case, no big dieal, we are not talking anything drastic here, but you will have several things to play with and tweak 🙂

Good luck building, be sure to write how it went!
 
Status
Not open for further replies.