Single ended amplier

[Kalee20]

Mark, I can see what you're on about. But, there's a language complication to muddy the water!

'Saturation,' when applied to bipolar transistors, means when the collector-emitter voltage is really low and the current flow is determined by the external circuit.

'Saturation,' when applied to FETs, means when the source voltage is fairly high and the current is independent of source voltage, being mainly determined by gate bias. FET's as amplifiers are almost always operated in saturation. It's almost opposite to BJT's... When the device is driven hard 'on', and the gate is driven well positive (MOSFET's), or zero (depletion JFET's), the device is said to be operating in the linear region.

'Saturation,' when applied to valves, is when the cathode emission is just sufficient to sustain the anode current, so that increasing the anode voltage or grid voltage makes no difference - the electrons are pulled away as fast as they're emitted. It's very rare to have a valve operating intentionally like this (the only case I can think of, is a white noise generating diode). Generally, a valve is saturated because the heater voltage is too low - or the valve is old and losing emission.

Having got that out of the way, what is relevant is, how low can the anode voltage go when driven hard on? In valve terminology this is referred to as 'bottomed' and is akin to the BJT being 'saturated.' And, in the absence of data, generally for a pentode or beam tetrode, 20-40 volts is a reasonable figure to assume (not as good as a bipolar transistor at all!).

Triodes, despite being simpler with only 3 electrodes, are actually more complicated as output valves. That's because they don't have the screen grid to help maintain the current at low anode voltages. So, they are much more difficult to turn hard on, and as output valves, they are less efficient. But this thread is about DA's amplifier using the trusty EL84, so let's not complicate things!

[Mark Hennessy]

You'll note that I put quotes around the word "saturation" because I know the situation is not so simple (!) with valves. The point of saying "saturation" in quotes was to draw an analogy with bipolar transistors, knowing that Andy has done quite a bit with transistor circuits.

Similar comments apply to FETs. I didn't use the term "saturation" in its literal sense - from the context, it's surely obvious that I'm referring to them being "saturated" in the sense of clipping the negative peaks of the output.

The rest of your post raises more questions for me, but I have to be somewhere else right now. More anon...

Cheers

[Mike Watterson]

The power FETS behave like resistors with a minimum on resistance, which can be low enough that there is no real "saturation" in the sense of valves or bipolar.

[Kalee20]

It's an unfortunate choice of words, really, 'saturation!'

Saturation in FETs refers to the region above the Id / Vd 'knee," the opposite of bipolar! And if the FET is driven hard on, as Mike Watterson says, it behaves like a low-value resistor and is (confusingly but, perhaps justifiably) referred to as the "linear region." If you operate the FET so that it enters the linear region, you get clipping on the negative output peaks!

But for valves, BJT's, or FET's, whatever we call it when the , we can't stop it being what it is. And when the thing starts clipping on the negative output peaks, we want to know the details! Bipolar junction transistors tend to clip with the thing dropping maybe a volt, maybe even just 50mV (less than the emitter-base drop!); FET's behave just like a low-value resistor, thus the voltage drop is small but load-current dependent; pentodes and beam tetrodes are even soggier but crudely, 20-40V is not unreasonable.

[Mark Hennessy]

The power FETS behave like resistors with a minimum on resistance, which can be low enough that there is no real "saturation" in the sense of valves or bipolar.

Indeed, I've already said this (emphasis added):

The one thing I'm not sure about - because of my lack of experience with valves - is just how "saturated" a valve can be. With bipolar transistors, transistors, it's pretty easy at a first and even a second look (though to do saturation properly gets surprisingly complex, but it's mercifully rare to need to go there). With MOS-FETs and JFETs it's even easier because the channel is basically resistive. But honestly, I'd be guessing when it comes to valves.

In that same post, I went on to say:

Obviously that, plus whatever is across the cathode resistor (assuming it's held still by a bypass cap) determines how low the anode can swing.

So it's completely clear that when I used the word "saturated" (in quotes), I was talking about the available output voltage swing from the anode (or collector, or drain). Specifically, how low it can go.

The exact mechanism inside the device is interesting in its own right, but is tangential to the point we were trying to get across (how to choose the standing DC voltage or current).

It's also worth saying that "saturation" is one of the terms used in the recording industry when over-driving something like an analogue tape machine or a "tube" stage for artistic effect.

So, we've answered the question about valves - 20-40V approx - and I certainly plan to read up on this for my own edification. But I'm left feeling guilty for asking the question because it's confused the thread, and the voltage isn't even relevant to the OP because he's using a transformer anyway! All I was trying to do was explain the thought-process behind setting DC conditions, using a resistive anode load for simplicity - which I hope has helped in the end.

So, by all means start a separate thread on this - I think it could be a worthwhile conversation. But can we please leave the subject alone in this thread? Remember - we're trying to help someone learn

[Diabolical Artificer]

No worries Mark, am still learning about transistors so all info is welcome. I find valves easier to understand than transistors, they use electrons only, non of these holes to complicate matters.

Anyhoo, rebuilt the amp yesterday and after a bit of fiddling, It works better. See pic of schematic. I'm going to stop with it here as I have a lot of stuff needs fixing, like solid state amps. I could improve it further by direct coupling and finding a better OPT, but to be honest I'm quite chuffed with it as is, it "sounds" quite good, not too bad for my second attempt at building one from scratch.

As you see on the schematic it says 1% ish distortion. this is because A) the Marconi isn't calibrated B) I'm not 100% certain I have the procedure right. I measured the frequency response which is pretty flat but not brilliant.

A few questions, we use RC filters to limit frequencies going into the amp, IE a high pass filter on the input etc, but where would we put the low pass filter to limit the amp to 15khz say? At the coupling stage between anode of ECC83/g1 of EL84? I'm also a bit unsure about how to work out the value of the cap that  couples g2 to earth. In some schematics I've seen this cap left off. Also do we really need the cathode bypass cap on the ECC83 if we use NFN?

Lastly I'm a bit hazy on how to find the R and C's values on the NFB. I used a 20 k pot, put a 1k sine in the amp, connected the OP to the distortion meter and scope and turned it till it till the distortion went down but not too much that we lost amplitude. I used a 390puff cap as that is what Mullard used in the 3-3. AFAIK though you put a 10k sq wave in and adjust C so you get flat horizontals.

Cheers, Andy.

[Kalee20]

Mark, apologies for derailed the thread somewhat with semantics and saturation!

DA, good you're quite chuffed with it, give yourself a pat on the back!

Ref queries... First, don't put the low-pass filter between the ECC83 and the EL84. Why? Because this is INSIDE the NFB loop. So, the NFB will fight to straighten out any top-cut. And secondly, top cut or bass cut adds phase shift, which can play havoc with stability. It's best to put OUTSIDE any NFB path. You could have a simple RC filter preceding the volume control. But excess HF going in tends to be less of a problem than excess LF going in, so it may be academic anyway...

Capacitor for g2 of the EL84? Just calculate it so that the voltage stays steady at the lowest frequency of interest, with a safety margin. The slope resistance at g2 is somewhat analogous to the ra of a triode, but that is unlikely to be specified. So, as a rough approximation, take it as Vg2/Ig2. And then make C equal to 1/10 of this at the lowest frequency of interest. Often, g2 is taken to HT+ so there isn't a problem; and sometimes there is a small value resistor in series which is low enough not to give significant AF voltage swing. It's there to prevent parasitic RF oscillations.

Bypass capacitor for the ECC83? You can leave it off, but if you do there are two significant feedback paths, one from the loudspeaker via the feedback resistor, and the other, more local current feedback within the ECC83 stage itself. Analysis would be horrendous, so I'd suggest put it in so you've got just the one feedback path.

Calculate the feedback resistor and capacitor? One way is as you've done it. Set the resistor to give you the reduction of gain (which is also the reduction in percentage distortion) that you can accept. And then play with the capacitor to give the best square-wave performance, with regard to overshoot, droop, and ringing, possible. It's not very mathematical, but it can give you a big insight into the behaviour and interaction of circuit elements.

[Diabolical Artificer]

Thanks Kalee.

Andy.

[Diabolical Artificer]

Re; calculating values for RC filters, when I built this amp I used a 330k R for the EL84's g1 off the top of my head. I thought I'd actually calculate it. AFAIK this R is there to keep g1 @ 0v and the datasheet usually has a g1 max in this case 300k, so I've bogged up here.Anyway, I think we find the value thus, we want Rg1 to be big as poss so we don't dissapate the power of the signal from our voltage amp, so lets say we use a 220k for Rg1, using 1/(2 x pi x f x Rg1) to find out the value of the coupling R from V1's anode, I get 1.4 etc ^ -8 for a roll off of 50hz (the OPT isn't going to go lower), which if I'm right means we use a 1n cap, right? This value seems very low, most amps I've seen use a 0.XXX or 0.0XXu.

So I got a 1000p/1n cap and a 220k R and connected my sig gen + to one end of the cap, with the R and scope probe to the other with the scope and sig gen ground/black leads connected to the other end of the R, I then swept up from 15hz to 50khz. There was attenuation below 50hz, but the amplitude remained low up to about 1khz and at 1k it was 0v, the signal then went up in amplitude as the frequency rose and remained more or less constant. I tried a few other experiments with different RC values and got similar results, at 1k the signal bottomed out at 0v.

Our RC filter/s is behaving more like a notch filter than a high pass filter.

Puzzled, Andy.

[Kalee20]

You think 1nF is low - I've seen 100pF!! Though it was with 2.2MΩ. (Vidor CN430 radio).

The calculation gives 1.4 etc 10^-8 which is 14nF actually, so 22nF would be reasonable.

This largely explains your measurements. 1nF and 220kΩ has a roll-off at 723Hz, so you will get progressively less transmission below that.

I don't know why you see 0V at 1kHz, but reckon it's to do with your set-up, maybe an earth loop creating false pickup which nulls out your actual signal.