Solid state phase splitter

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Any experience with that? Suggestion on a good schematic?
I have a "source-o-dyne" phase splitter made with an LND150 MOSFET in a small push-pull 6AK6 amp I built a few years ago. It works very well, exactly like a valve-based cathodyne, but even better.

(A true valve cathodyne has some nasty characteristics when overdriven unless you take several precautions. There is a section on this in Merlin Blencowe's valve preamps for guitar book.)

IMO, a real valve is wasted in a cathodyne. With so much negative feedback, the cathodyne produces too little harmonic distortion to hear. It sounds sterile-clean, just like a transistor. So why not just use a transistor (MOSFET) instead, and save the triode for somewhere else in the circuit, where it can actually contribute to the sound?

A generic schematic and LTSpice simulation is attached. The exact MOSFET you use doesn't matter much, as long as it's rated for a sufficiently high voltage. I just picked something built into LTSpice. The back-to-back 10 volt zener diodes are necessary to protect the MOSFET gate.

Incidentally, minimum B+ is at least four times the largest peak signal you want to drive your output tubes with. Put another way, if you use, say, 300 volts B+ on the MOSFET, it can spit out two phase-reversed signals of up to almost 75 volts peak (150 volts peak-to-peak) each.

That's without any load on the source-o-dyne outputs. Once you hook them up to the EL84 control grids, there will be some loading from the grid bias resistor, and heavier loading from grid current once the EL84s start to overdrive, so the maximum available signal swing will decrease.


-Gnobuddy
 

Attachments

  • N_Ch_Enhancement_Source_O_Dyne_2018-06-20_18-56-12.png
    N_Ch_Enhancement_Source_O_Dyne_2018-06-20_18-56-12.png
    39.4 KB · Views: 623
  • N_Ch_Enhancement_Source_O_Dyne_2018-06-20_19-06-10.png
    N_Ch_Enhancement_Source_O_Dyne_2018-06-20_19-06-10.png
    53.1 KB · Views: 590
IMO, a real valve is wasted in a cathodyne. With so much negative feedback, the cathodyne produces too little harmonic distortion to hear. It sounds sterile-clean, just like a transistor. So why not just use a transistor (MOSFET) instead, and save the triode for somewhere else in the circuit, where it can actually contribute to the sound?

-Gnobuddy

That's exactly what I am thinking!

Especially in the case of projects that are small in terms of component number and output power, it will be useful to save space in the chassis (or use a smaller one), and also require less power from the power transformer which has to power one filament less.

And what about using the case of a phase splitter based on a differential pair cathode biased? Using an equivalent mosfet circuit will it influence the sound?
 
For the sourc-o-dyne I will make sure to not feed them with more than a 1/2 B+ peak to peak signal if you want to avoid SS clipping, the preceeding AX7 might swing 2/3 B+ and a tone stack could have no attenuation above 3-4khz with treble dimed...also put a 100k grid stopper (EL84 will stand if cathode bias) on power valve, especialy the cathode driven one to avoid the "nipple" effect.
the differential one has been done by R.G.Keen. High Gain Phase Inverter - MOSFET Follies :cool:
hard to avoid clipping this one!
 
We can now easily find 500V MOSFETs, so we can drop-in a MOSFET and run it on the same 300V rail as the tube we are replacing.
This is undoubtedly true, but in such a case, then go for a full SS amplifier. Personal taste only: I dislike mixing SS and tubes. Different technologies, different behavior, different requirements. Not for me.

In the other hand, a MOSFET with 300V per se doesn't warranty that you will get clean 16Vpp from them.
 
That's exactly what I am thinking!
I'm with you. If I ever build a tremolo oscillator into an amp, I'll use a MOSFET for that, too. And I think MOSFETs make complete sense for the send and receive electronics in an FX loop as well.

And what about using the case of a phase splitter based on a differential pair cathode biased? Using an equivalent mosfet circuit will it influence the sound?
A MOSFET differential-pair PI is a no-go, unfortunately. The reasons are not immediately obvious, so I'll start at the beginning:

A proper (triode valve) diff. pair, unlike a cathodyne, does have voltage gain (say 25x - 30x to each output). It also overdrives gradually as the signal level goes up, and this provides enough distortion to be quite audible, at least if the amp is designed so the output stage hasn't gone into hard clipping long before the phase splitter starts to overdrive.

To me, the long-tailed-pair valve phase splitter has a sort of "growl" to it, probably from significant amounts of third-harmonic distortion. It is a push-pull amplifier, after all: that means even harmonic distortion is cancelled out, and only odd harmonics are generated (assuming perfectly balanced MOSFETs.)

So, to replace a proper valve diff amp PI, our hypothetical MOSFET differential amp phase-splitter has to do two things: it has to provide voltage gain like a valve diff amp PI, and it has to overdrive and "growl" like a valve diff amp PI does.

It's easy enough to wire up a pair of MOSFETs in the same differential amp configuration (though self-bias is a no-no). But we immediately have a problem: the voltage gain will be way higher, maybe ten times higher, than a valve diff amp. This is because modern MOSFETs have enormous transconductance compared to a half-12AX7 or similar triode.

Okay, we can fix the too-high voltage gain by putting an unbypassed source resistor betwen each MOSFET and the shared "tail" resistor. But the fix introduces another problem - now you have a sterile-clean PI that provides no audible distortion at all. It will sound like a cathodyne, not like a long-tailed pair.

Maybe there is some clever way around these issues. But at the least, it's not a case where a simple substitution (MOSFET for triode valve) will work well.

Incidentally: FETs and self-bias are a very bad idea. It doesn't work at all for enhancement-type FETs, and it doesn't work reliably for depletion-mode FETS either, because of the wide parameter spreads all semiconductor devices suffer from.

This is why, for instance, the circuit I posted for a MOSFET "source-o-dyne" doesn't use self-bias. Instead, a pair of resistors across the power supply biases the MOSFET gate to one-quarter B+. The source of the FET will settle to within a few volts of that, and with a few hundred volts B+, we can just ignore those few volts, and say the source will be at one-quarter B+. Now chuck in equal value source and drain resistors chosen for reasonable current flow and reasonable power dissipation in the MOSFET, and you're done.


-Gnobuddy
 
<snip>

...the differential one has been done by R.G.Keen. High Gain Phase Inverter - MOSFET Follies :cool:
Thanks for the link! I note that RG Keen experienced exactly the issues I just described in my last post. He didn't mention that adding local NFB to lower the MOSFET PI gain also linearizes away any trace of audible distortion, but perhaps he was okay with an audibly transparent PI, rather than one that sounds like one made with a pair of triodes.


-Gnobuddy
 
This is undoubtedly true, but in such a case, then go for a full SS amplifier. Personal taste only: I dislike mixing SS and tubes.
Understand and agree about the personal taste part.

My personal taste goes the other way to yours. I have struggled for a long time to try and get good guitar tone with an entirely SS amplifier, and have never got anything better than just acceptable. So pure SS guitar amps don't thrill me.

In my opinion, it is exactly the differences between MOSFETs and valves that makes it so interesting to mix and match them:

  1. Valves are great at making distortion that sounds good with guitars; MOSFETs aren't.
  2. High-ra preamp valves (like a half-12AX7) are bad at driving any kind of difficult load; MOSFETs are really good at this, so they make excellent buffers for triodes.
  3. Triodes struggle to generate a voltage gain of 60 times (36 dB); MOSFETS are excellent at producing clean voltage gain for FX loop recovery, active tone controls, tremolo oscillators.
Because of these differences, to me, it makes a lot of sense to use valves as musical distortion-generators, and MOSFETs where we need audibly transparent buffers or perfectly clean voltage gain.


-Gnobuddy
 
Don't want to mention the name of the amp and manufacturer.
YouTube
Works fine if you want a PI with no audible distortion.

You could warm up the cleans elsewhere with real valves, but it sounds as if that particular amp designer didn't bother. I'm guessing the target demographic for that model doesn't care too much about clean tones, and is okay with the thin steely sound.

IMO, this is what good clean tone sounds like (start listening at 30 seconds into the clip if in a hurry): Fender '65 Twin Reverb Guitar Amp Clean Tone Product Demo - YouTube

I have yet to hear that sort of "valvey" or "tubey" cleans from any solid-state amp, though the Boss Katana 50 and 100 are getting pretty close. Not straight MOSFETS in analog circuits, though, but rather, hundreds of thousands of lines of computer code running on a DSP chip.


-Gnobuddy
 
I am pretty sure the PI of the twin is in the clean zone of its range of operation in the clip. I'm pretty familiar with the Twin sound as the first four years of my musical endeavors had my cousin playing through one. I was just showing the LTP circuit with Mosfets as it seems to provide gain, was not commenting on the sound as I have not tried it. If it remains as clean as the 100% NFB splitter as described on the previous page it might be an option to someone also.
 
If it remains as clean as the 100% NFB splitter as described on the previous page it might be an option to someone also.
Agreed, that thought has crossed my mind as well. It would work as a cathodyne substitute that does not suffer from "nipple distortion" like a real cathodyne or source-o-dyne.

The attached image shows the problem. The zener diode mimicks what happens when grid current starts to flow in an overdriven output valve.

In this example, I used a 10V zener, so if Vout1 exceeds 10 volts peak, the zener starts to draw current. This acts like bypassing the FET source resistor, which bumps up the voltage gain to the FET drain. This in turn shows up as "nipple distortion" on the second PI output, Vout2 (blue trace.)

The MOSFET long-tailed-pair you showed will not suffer from this problem.


-Gnobuddy
 

Attachments

  • nipple_distortion_001.png
    nipple_distortion_001.png
    52.4 KB · Views: 233
A 47k/100k 1/4w resistor is as cheap as efficient to reduce the mA hungriness of the following power tube...MOSFET are nice in this application with the usual countermeasure...be happy and do not hit the rail ;-)
 

Attachments

  • Crowhurst remedy.JPG
    Crowhurst remedy.JPG
    121.7 KB · Views: 205
  • Crowhurst1958 nipples.JPG
    Crowhurst1958 nipples.JPG
    49.2 KB · Views: 244
I am pretty sure the PI of the twin is in the clean zone of its range of operation in the clip. I'm pretty familiar with the Twin sound as the first four years of my musical endeavors had my cousin playing through one. I was just showing the LTP circuit with Mosfets as it seems to provide gain, was not commenting on the sound as I have not tried it. If it remains as clean as the 100% NFB splitter as described on the previous page it might be an option to someone also.

I see totally the point for Fender style clean amps. What do you think about the mosfet phase splitter that you are showing, for driving a power stage in an high gain metal amp? Do you think the phase splitter is also used to color the sound, or everything comes from the preamp in this style of amps?

Anyway, I own the amplifier that has that phase splitter (still not naming brands), and they say that "a well designed mosfet PI has the exact same performance as a tube PI, therefore there is no need to use an extra valve".

I'm tend to not trust exactly what they mainly because they call their amps "all tube", but they are as hybrid as any valvestate.

The approach I would like to use is to switch to solid state only where it doesn't really matter sonically, i.e. the rectifier is one, and I was wondering if it can be the same for the case of the PI in a heavy metal scenario...
 
A 47k/100k 1/4w resistor is as cheap as efficient to reduce the mA hungriness of the following power tube...
I agree that large grid-stoppers are one of the keys to managing cathodyne distortion.

But as the attached image shows, there is still distortion even with a 47k grid stopper. I've used a typical Fender value of 56k for the cathodyne's cathode and anode resistors. When the output valve starts to flow grid current, that effectively puts the 47k grid stopper in parallel with the 56k cathode resistor.

That drops the effective cathode resistance to 25.55k, bumping up the voltage gain at the anode from nearly unity (1.0) to nearly 2.2 times.

Because the cathode resistor (or source resistor, in this case) is so large, IMO it is virtually impossible to make the output valve grid stopper so big that there is negligible "nipple distortion". If we make that grid stopper too big, there might be unwanted treble roll off within the guitar frequency band.

The long-tailed-pair circuit doesn't suffer from this problem, and if we build it with MOSFETs, it's also cheap and compact.


-Gnobuddy
 

Attachments

  • nipple_distortion_w_grid_stopper.png
    nipple_distortion_w_grid_stopper.png
    63 KB · Views: 139
Yes you're right, it's better to lower the Rsource/Rstopper factor. the LND150 is supposed to stand 3 or 4 ma at a "wrong" bias voltage of 100v (here we design at 1/4 B+ around 80v), 33k R source&drain should be safe. with a 100k stopper it's hard to show the nipple.

If we assume that total input capacitance of an EL84 pentode mode is about 35pf including stray, the 100k stopper give you less than .2db attenuation at 8khz and the 220k+drain Zout will reach -1db at a frequency the guitar speaker & guitarist ear (the golden one, next to the splash cymbal) will ignore.

Now we should take care of the total grid load resistance which is 300k max/grid bias and 1M/self bias for the EL84 so 220k+100K is not that bad.

the drawback might be harder grid current clipping from a higher source impedance...should we consider using a 2x115v-2x24v small stepdown toroid load in a LTPI?

The long tail pair could also suffer from spike effect according to MerlinB, if anode/drain voltage of the inverting device is clamped due to grid current then it is more a cathode/source follower and pass all the signal to the non-inverting device instead of 1/2...or something like that. The stopper may be less than with the cathodyne , you have to sim or measure it to be sure.
 

Attachments

  • spike1.jpg
    spike1.jpg
    200.3 KB · Views: 134
  • freq.JPG
    freq.JPG
    123.3 KB · Views: 113
I see totally the point for Fender style clean amps. <snip>
No idea on the circuit performance, I do not have the time I would like to play with electronics even though I have the parts to try the circuit. It is the playing/listening/comparing that takes all the time. Hopefully not too many years longer. As far as the amps are concerned, they seem to sound well enough on youtube (with its limitations) but I would like to hear a good recording of mildly distorted playing.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.