ripple rejection

[uneumann]

I'm just sharing something that may be of interest. In the recent FET noise thread , a cascode circuit emerged as an input stage capable of lower noise and equal or greater gain than a classic triode (12ax7) input stage. In pursuing the FET circuit further, it has another benefit that may be useful. Turns out that the FET cascode can be extended with a current mirror to achieve a very high power supply ripple rejection of almost 70db. This performance depends on the high drain impedance of the cascode circuit, so a similar current mirror doesn't do much for a triode or pentode stage. The circuit below shows three current mirror arrangements (many mirror circuits are possible) and they all perform equally well so the choice of mirror circuit seems non critical.
Current mirrors are common in opamps and SS power amps. I'm unaware of any used in guitar preamps, but I wouldn't be too surprised if it has. Anyone have a pointer to one?

A power supply with 120Hz ripple added (upper right) and a 1k/10uf RC filter that produces ~1v peak ripple (0 db) to power the 3 FET circuits. A standard triode input stage powered with 1v peak ripple would have horrible hum. The triode stage (on the left) therefore has an additional 10K/100uf filter. This additional filter is what is needed to make the triode stage hum comparable to the FET stages powered by the full 1v peak ripple.



Some waveforms illustrate what's going on. The PS ripple at C4 is ~1v peak (below). This powers the three FET cascode stages.



With all stage inputs grounded, the PS ripple visible at R17 (the leftmost cascode circuit output) is ~0.35mv peak (waveform below). This is ~ 69 db down from the PS rail ripple.



A PSRR curve of the R17 signal shows ripple rejection over varied freq's (below). Since the 120Hz PS ripple level was set to at 0db, the output levels are absolute stage PSRR values. The dip at 120 HZ is interesting. It is set by adjusting the value of C7 (and similar in other stages). I haven't fully grasped why there's a dip there, but it seems related to the parasitic cap's of the two FET stages and the fact that the rail ripple feeds the upper FET gate. Some "magic" level and phase shift of ripple at that Q13 gate leads to the dip. A 470nf C7 value moves the dip to 60Hz, which is what voltage-doubler and tripler supplies produce and is harder to filter with rail caps.



The output of the tube stage is ~0.4mv peak at R41 (shown below) - similar to the cascode output above. However, the additional RC filter of 10K (R42) and 100uf (C16) is needed to produce this ripple level from the tube stage. The RC filter reduces the PS ripple at C16 to ~1.3mv peak. The triode stage has a ripple rejection ratio of only about 10 db.



The source-follower FETs (Q6 and Q9) are not strictly needed, but the output impedance of the mirrors is high, so I added them to see if they impact things. Other than buffering the output, they seem benign.

OK - so does it matter? After all, you can simply add bigger RC filters to a supply to get any level of hum reduction you need. Better rejection means that fewer/smaller/more reliable filter caps can be used. Whatever filtering you have, it can't hurt to lower ripple by 60 db. Perhaps small (~1uf) mylar/poly caps are sufficient filtering in the preamp. It's another option for circuit designers to explore.

As for tone, there is no measure for that. I do know that I've tested the rightmost FET circuit and it behaves as the simulation predicts to the best of my measurement capabilities. It also sounds very clean, as the simulation predicts. That means that "tube color" will have to be added in later stages - but now the signal is 5-10v peak (stage gain is ~70-100 depending on values) with lots of headroom and very low hiss and hum. As a physically small circuit (no bottle or sockets), it can be located right at the input jack, with minimal lead length/dress issues.

[R.G.]

I'm very fond of current mirrors in general SS work. Use them all the time. You're right, using current mirrors is one way to get good power supply nose rejection.

For the case of triodes in a guitar amp, you can do it another way too good advantage. It's pretty simple to use a source-follower "capacitor multiplier" regulator per stage with a MOSFET to get a lot of ripple reduction. The low currents per stage make this entirely feasible, and using one regulator per stage with the low current and modest voltage drop can eliminate the need for heat sinking.

[Mike Sulzer]

I like the current mirrors because they work down to dc.

Why 150K for all the resistors up at the power supply level? If I understand the operation of the circuit, (and please forgive me if I have it wrong) the available voltage swing in your output sections is limited because of those large resistors. This means that the output section clips before you have used all of the available current swing in the JFET below. This is of no consequence if the input signal is small, since it will not clip, but the output of some guitar pickups can be quite large, using nearly all or even all the available range of a 12AX7, for example. Depending on the choice of the JFET, its transconductance could be larger, but probably not much smaller than that of a 12AX7. Thus it is at least as important to be able to use the full current swing.

[uneumann]

Why 150K for all the resistors up at the power supply level? If I understand the operation of the circuit, (and please forgive me if I have it wrong) the available voltage swing in your output sections is limited because of those large resistors. This means that the output section clips before you have used all of the available current swing in the JFET below. This is of no consequence if the input signal is small, since it will not clip, but the output of some guitar pickups can be quite large, using nearly all or even all the available range of a 12AX7, for example. Depending on the choice of the JFET, its transconductance could be larger, but probably not much smaller than that of a 12AX7. Thus it is at least as important to be able to use the full current swing.

Agreed. The headroom in the example is limited to about 80v peak. While that's plenty for SC players - it may clip with hot PUs or pedals. Change some values (lower R20, R13 and increase R15) to add gain and headroom. The circuit is just showing the concept. With a PS of ~400v there's no problem getting more headroom. Many 12ax7 input stages run about 150-160v plates, so realistically, you only get about 80-100v peak headroom there as well.

I picked high values for R20 and R13 to linearize the mirrors and to accentuate the dip in PSRR. The dip depends on the signal amplitude on the Q13 drain. Somehow that's coupling to the gate and source of Q13 to cancel ripple currents through R20. When the R20 ripple current is low, so is the R13 mirror ripple current. If you optimize for headroom and forgo or reduce the dip, you still get about 64db rejection. It's a design choice - optimize for your needs. In my application, a voltage tripler produces the ~400v supply - with cap multipliers , but 60Hz rejection still matters and I'm setting C7 for a dip at 60Hz. A high value for R20/R13 also keeps Q13 from overheating.

[Gnobuddy]

Current mirrors are common in opamps and SS power amps. I'm unaware of any used in guitar preamps, but I wouldn't be too surprised if it has.

A few years ago I read Merlin Blencowe's preamp book, and started thinking about how the load resistance affects the nonlinearity of a triode's transfer function. The attached pics tell the tale: as load resistance falls from infinity to zero, you get progressively more and more nonlinear (harmonic) distortion.

My interest is in guitar amps, where triode non-linearity is a desirable thing. So I thought about ways to create a near-zero load impedance, and two ways immediately came to mind. The first is to use a cascode topology, and the second is to use a current mirror as the anode load. The attached current-mirro schematic dates from around 2013, which was several computers, several residences, and one country away from my current living circumstances. The schematic shows a variable anode load for the triode, with the current mirror responding only to the resulting anode current (but not to the anode voltage). The idea is to have some sort of "dial a nonlinearity" control.

The circuit needs development. As shown, DC bias conditions and output headroom vary dramatically with as the anode load is changed. I never built the current-mirror circuit, but I think it is worth further exploration at some point.

-Gnobuddy

[uneumann]

The circuit needs development. As shown, DC bias conditions and output headroom vary dramatically with as the anode load is changed. I never built the current-mirror circuit, but I think it is worth further exploration at some point.
-Gnobuddy

Seems worth some consideration...
Interesting to see what ideas emerge when you leave familiar territory.

[Gnobuddy]

Interesting to see what ideas emerge when you leave familiar territory.

Agreed! I have thought for some time that there may be room for interesting new possibilities to emerge in hybrid guitar amp circuits, where you mix and match the most desirable characteristics of valves and semiconductors.

The traditional common-cathode triode gain stage has no power supply ripple rejection to speak of. But then I realized that another piece of relatively recent electronics technology - 270 uF, 330V "flash capacitors" that cost a buck each from surplus outlets - can provide 60 dB (!!!) of ripple filtering at 60 Hz when used with a 10k resistor in the B+ line!

-Gnobuddy

[Enzo]

On the other hand, by the time you get to the third, fourth, fifth B+ node, there is very little ripple anyway. Hum in the front stages tends to come from other sources.

[Gnobuddy]

On the other hand, by the time you get to the third, fourth, fifth B+ node, there is very little ripple anyway.

Agreed, cascaded filters are very effective. And they were a perfect solution at a time when only small-capacitance electrolytic caps were available.

In my case, I was experimenting with small-signal pentodes in preamps, and was running them off a separate (lower) B+ rail to keep them from having excessive gain. Nothing else was running off that lower voltage rail, so there was no good reason to have multiple B+ nodes. That's when I realized that a single filter with a contemporary large-value flash capacitor did a fine job all by itself. Today, it's a cheap, effective, alternate approach to have in one's design toolbox.

-Gnobuddy

[R.G.]

One question occurs to me at first reading. I've always thought that current mirrors had quite high outptut impedances, not very low ones. Seems to me that a current mirror load for a triode would be much the same as a current source load - which it seems to me it is.

What am I missing?

[uneumann]

One question occurs to me at first reading. I've always thought that current mirrors had quite high outptut impedances, not very low ones. Seems to me that a current mirror load for a triode would be much the same as a current source load - which it seems to me it is.

What am I missing?

You're right - they do have HiZ outputs. That's why I used a source follower to buffer the output. The HiZ output is in the output leg of the mirror. The input leg is exactly opposite - it's a LoZ load - unless loaded intentionally by a resistor. The input leg is better thought of as a current sink. The input side usually has a BE junction (with collector-base shorted) as a load for the driving current. That's a very lowZ load. The reflected current matches the sink current, but it's usually a collector sourcing the load current so that output is HiZ.

[MarkusBass]

Would you care to post the simulation from the first post (including required models)? I wanted to use similar circuit in one of my new builds.

Mark

[darkfenriz]

RC filter on the standard common cathode gain stage has another feature: even with mild overdrive of the stage it will produce preamp stage sag (if cold biased) or anti-sag (if warm biased).
Those DC fluctuations will propagate though the stages (and coupling caps) changing the operating points of the following stages.

In contrast your circuits are piece of decent engineering job of pursuing some characteristincs of triode stage while getting rid of other characteristics - it seems like with current mirrors you'll preserve the transconductance curve and with anode loading you may also preserve anode characteritics. But why?

With fets and cascodes you're moving towards perfect gain stage = buffer with gain.

[uneumann]

Would you care to post the simulation from the first post (including required models)? I wanted to use similar circuit in one of my new builds.
Mark

OK - here it is... This should help. Tweak for your needs.

[uneumann]

RC filter on the standard common cathode gain stage has another feature: even with mild overdrive of the stage it will produce preamp stage sag (if cold biased) or anti-sag (if warm biased).
Those DC fluctuations will propagate though the stages (and coupling caps) changing the operating points of the following stages.

In contrast your circuits are piece of decent engineering job of pursuing some characteristincs of triode stage while getting rid of other characteristics - it seems like with current mirrors you'll preserve the transconductance curve and with anode loading you may also preserve anode characteritics. But why?

With fets and cascodes you're moving towards perfect gain stage = buffer with gain.

Well, OK, point taken - but why not? What's wrong with a perfect gain stage? As pointed out above, strong local FB means there is little coloring of the signal by the devices (FETs, transistors, etc) themselves. The stage just gives you a nice big clean and quiet version of the input you're then able to EQ and sag and harmonic-ize more easily without noise creeping in. Most triode input stages fed by most PUs try to achieve exactly the same thing. Of course, if you OD the input, differences can arise. I'm trying to keep the input clean and let the later stages handle all of the coloring and OD character. It's just a design decision. There is no right/wrong way here. Different system behaviors result from each decision - but they can all lead to good guitar amps. The huge variation of well-liked amps out there attest to the fact that variations work and in fact are embraced by many. I prefer to push a bit beyond the predictable tried and true way to do things. You never know what you find out there until you try.