Single ended amplier

[Mark Hennessy]

Yes, I've certainly looked at load lines for bipolars and JFETs and MOSFETs. They're in almost every text book. But, I really don't know why. Sure, I understand what they mean, but I've *never* used them during the design process. And just to be clear, I do a lot with transistors - mostly bipolar - at low, medium and high power levels. Most of what I do is analogue. But I get by without reference to load lines. Perhaps I'm missing something? I'd welcome concrete examples of something that load lines bring to the party during semiconductor circuit design - especially if there's no other way to do it.

That EL84 datasheet has a decent number of graphs! I'm really uncomfortable with the gm varying so much in valves, given that they are supposed to be transconductance amplifiers (where gm would be fixed in an ideal world). No wonder they distort like they do

Maybe I shouldn't worry, given that gm varies over a cycle with a bipolar transistor as it's directly proportional to Ic. We ignore that when thinking about small signals, but it's clear to see on the 'scope screen with large swings. One good reason to load with a current source! Plus, there is so much gm available that NFB does a splendid job of linearising it anyway.

The shape of gm vs. Vg - is that typical, or does it vary from device to device?

I note that the graph gives gm vs. Vg, and not the gm vs. Ia that I was after, but still interesting nonetheless.

At least the Ia vs. Vg curve is nice and reassuring

Andy (DA), Andy (radioman) makes a good point about measuring the inductance of the transformer primary. I'm wondering if shorted turns might be the problem? Do you have the means to measure inductance? I'm guessing that one of those Chinese component testers would probably do the job.

[Kalee20]

I've never used loadlines with semiconductors either.

Bipolar transistors do tend to spoil us, with their pretty predictable behaviour. And yes, gm is approximately 40x Ic where Ic is in mA, and gm is in mA/V... But this '40' constant is temperature sensitive, so perhaps it's not all good!

Loadlines with valves, I do most of the time, really useful!k

[pwdrive]

Mark wrote "The shape of gm vs. Vg - is that typical, or does it vary from device to device?"

A gm curve isn't that common in data sheets so far as I know but here's one for the 12AX7 plotted against anode current:

http://www.r-type.org/pdfs/12ax7.pdf

Lawrence.

[Mike Watterson]

Yes, I've certainly looked at load lines for bipolars and JFETs and MOSFETs. They're in almost every text book. But, I really don't know why. Sure, I understand what they mean, but I've *never* used them during the design process. And just to be clear, I do a lot with transistors - mostly bipolar - at low, medium and high power levels. Most of what I do is analogue. But I get by without reference to load lines. Perhaps I'm missing something?

They are not so much use on transformerless, totem pole or complementatry.  How many single ended transistor class A power amps have you done? That's about the only thing they'd be much use for!
Earlier stages you might have an emitter resistor to set bias current and thus easily choose a collector load resistor so that Vce is roughly 1/2 HT - (emitter resistor x Ic).
So I've never used load lines on graphs for transistors either. Also most transistors have a much greater spread than valves so the graphs are only a guide line. You design with localised DC NFB and assume a minimum DC gain.
Even on Valves I've only used them for the SE class A output stage. Earlier stages I make an estimate of what Ia I want and the grid bias for that (which likely only has cathode resistors on Indirect heated Cathodes, most of my valve designs are Battery direct filament Rod Pentodes anyway), then calculate the Anode load (if any). RF/Osc/IF stages the Anode & Screen currents set the HT decoupling resistor values for the desired voltage.

I found that most transistor books covering RF, IF, Oscillators and basic Audio are based on earlier valve texts. This is especially obvious on oscillators where the explanations in the valve text books are better. Hence the transistor collector load lines.

Transistor RF power amps might use 12V to 48V and are usually push pull transformer coupled. Ones for FM only might be class C or B. For SSB they may be Class AB or A. For digital they are often class A. I have some 1W wide band RF push-pull ICs which are class A and need transformers or bifilar chokes for DC load. That sort of stuff can sometimes need load lines.
I've no idea how to use "load lines" on totem pole or complementary Class B or Class AB transformerless push pull.
Digital Switches, Class C, Class B, Class D doesn't need load lines.

[pwdrive]

Below is a link for anyone trying to get to grips with valves and characteristics, triode characteristics start at article page number 52:

http://www.smcelectronics.com/DOWNLOADS/...THEORY.PDF

Lawrence.

[AlanBeckett]

We had our own 'Services Textbook of Radio' - several Volumes.
I can't find it on line but there are various Volumes for sale.

Alan

[Diabolical Artificer]

Lots of posts to reply to, i'll try and get it all in! Firstly the OPT I was originally using was indeed a "RS Standard Universal OPT - 217-567". I estimate it was made circa 70/80's. It has a "datasheet" which gives rough Z pri for various loads dependant on what pins you connect it to. I had it connected 1-3 pri/D-A sec. I did measure Vin/out and worked out a Z pri of 6700, I think. I presumed it was made for SE use but after what Andy Radioman said post #30, I notice it has a CT column, therefore it was intended for P-P use too. I'm no longer using it as the OP was rubbish maybe due to core saturation. I have a Marconi TF2700 , I will measure the L.

Re the ECC voltage amp - Kalee, well spotted, another bog up on my part, your right, I did a 2v input. Some of these old graph's are hard to use because of the scale, good idea Lawrence to measure using mm. One thing though, we don't need to be 100% accurate ( it's a valve amp not a precision measuring device) and can't really be as the graph represents an ideal valve not the one were using. I see your points, 47k would be better. What I'm puzzled by is that if we use a ECC83 we get 60v P-P out, which I thought would overdrive the OP valve causing distortion. Also wouldn't it mean that the volume knob would only go to "3" say? I thought to design a voltage amplifier stage we need an OP whatever the following valves g1 bias is + a bit extra for losses due to following g1 resistor and valve capacitance etc.

"The µ of a triode is a pretty reliable and constant characteristic. But mutiual conductance gm isn't. So if you use a really low value of anode load (like 1kΩ) the gain is going to be small and virtually dependent on gm, with the prospect of distortion as you run around the foot of any curve. Whereas if you use a very large value of anode load, like 100kΩ, or even better a constant-current load or a really high-value inductor, then the gain approaches µ which, being constant, implies low distortion. " There is far more than meets the eye to making a voltage amp stage than drawing a load line and choosing a bias point, I didn't know this Kalee. When doing it myself I've tried to chose a load and bias point so the operating point falls on the straight part of the grid line and avoid the curved part.

Post #29 - yes u, gm and dynamic anode resistance inter relate I think the word is. You can find out the dynamic anode resistance too by graphical analysis. There's no equivalent process for a transistor AFAIK. See Radio Designers Handbook- F Langford Smith, Chapter 2 Section 1 page 13. Thanks for the other link Lawrence.

Armed with all your advice I'll try and get time over the weekend to iron out the last wrinkles.

Thanks again, Andy.

[Mark Hennessy]

What I'm puzzled by is that if we use a ECC83 we get 60v P-P out, which I thought would overdrive the OP valve causing distortion.

Just a general point here about "gain structure".

If your triode has too much gain, then yes, it would be capable of "over-driving" the OP valve.

However, this doesn't matter because the OP valve will clip first when it runs out of headroom. And when the user hears clipping, they should turn down the volume.

The output stage should always clip first. If an earlier stage is clipping before the final output stage does, then something has gone wrong with the gain structure, and the amplifier will never be able to deliver its full output power.

Also wouldn't it mean that the volume knob would only go to "3" say?

That's more like it. In other words, the amplifier will have too much gain.

At which point, we use negative feedback in some form to reduce the gain to the desired amount - bearing in mind we need to have slightly more gain that we might think because some recordings are made at a lower level, and some sources are quieter than others - commercial hi-fi amplifiers usually have between 10 and 20dB of "gain in hand" for that reason. Which means that most commercial amplifiers are into clipping at about 12 o'clock on the volume control when playing a modern CD with high peak levels.

Negative feedback - stating the obvious, I know - will make a good amp better in every regard, providing you get it right. Some audiophiles think there is something wrong with NFB, but I assume we don't need to go there

[Kalee20]

I thought to design a voltage amplifier stage we need an OP whatever the following valves g1 bias is + a bit extra for losses due to following g1 resistor and valve capacitance etc.

Sometimes. If the output stage is a triode, probably yes. But not in this case.

What you need to do is to provide enough drive to the OP valve to swing it from double the standing Ia, to zero (so you get the same swing each side of the standing Ia). If it's a pentode, the anode current should be reasonably independent of anode voltage, so you don't need to take much notice of the load.

If the quiescent Ia is 40mA, say, you just need to check what the grid voltage would need to reduce to, to give 80mA, and make sure it's still negative. The EL84 might, for example, be biased at -7V for 40mA, and swing -2V to -12V (5V either side). There's no point in trying to take the grid 'up' above -2V because on the other half cycle you're already touching zero current, so about to hit serious distortion. But if your grid voltage for the double-quiescent, 80mA, happens to need to be positive, you're going to get distortion due to grid current. You'd then not be able to take advantage of the full swing in the other direction towards zero Ia, and so you need to rethink your operating point (less Ia, more bias). And then, you'd need to recalculate your optimum load (approx. 0.9 Va/Ia, as earlier posts).

If you use an ECC83, yes, you would have more gain than you need. You can 'lose' this gain with a bit of NFB, gaining a reduction in distortion, or by adding tone controls, or just by turning the volume knob down.


Mark Hennessey is dead right, the stage which goes into overdrive first should be the output stage. Otherwise you just aren't using all its capability. And of course you want to make sure you don't 'waste' its capability bu having it go into overdrive on frequencies you can't hear - so the frequency response curtailment should be done in earlier stages.

[Diabolical Artificer]

Ah, forgot about NFB Mark.

" provide enough drive to the OP valve to swing it from double the standing Ia, " not I sure I understand you here, if the OP valve is biased in Class A, how can the anode current double? I know the anode voltage swings higher than HT when current in the primary is 0, does current double too? If that is the case that means the PSU has to be capable of at least twice Ia.

Andy.