2C34 amp.

[tri-comp]

Originally Posted by Diabolical Artificer

Thanks, I have that datasheet, as you see no small signal gain only hFE. So how does one compute hfe for another Ic?
A.

What's wrong with the lower picture on page 3 of the MJE340 ON-Semi datasheet ?
To me it looks like DC Current Gain (hFE) can be read all the way down to Ic = 1mA

[Diabolical Artificer]

hFE Big F Big E is different to hfe small f small e isn't it? The former is DC current gain the latter is AC signal gain, no?

A.

[tri-comp]

But you're running the CCS at a constant (DC) current of 3mA. (+/- whatever, if it's found to have an impact on the THD figure)
You're not going to worry about the transistors AC-performance in a CCS ...or did I get this wrong ?

[Radio Wrangler]

Normally upper case is taken to mean the large signal current gain (which can be somewhat nonlinear) and lower case means a small signal value.

Hfe (either case) is neithre AC nor DC specifically. It's normally flat over frequency up to a point and then it falls (roughly a first-order fall) to unity gain at a frequency called ft where it passes through a gain of 1.

It's like the gm of a valve. You can plot curves of static DC performance and measure the slope of Ia over Vg and get the mutual conductance. You can stick the valve in a valve tester and it puts a small AC voltage component on the dialed-in grid bias voltage, and it measures the resulting AC component of the anode current. So the Gm is there at DC and AC.

David

[Diabolical Artificer]

Where I got confused T is that the formula I had and the one you posted called for hfe. I presumed inserting hFE would influence the result. A bit like if a cake recipe calls for Syrian dates how do I know if Turkish dates will do? Not knowing the finer points of transistor cooking I tend to stick to the recipe literally in order to save the time an extended cooking coarse would take. I shall gen up on the subject so I know the difference.

Thanks David, I was looking through Art of last night and had to look hard to find hfe. In valve circuits no distinction is made between large/small signal. So is this just a case of semantics and hFE/hfe the same or does hfe have a different parameters?

Back to the amp, I'm going to put it on the back burner for a while as I've not got the cash to buy OPT's (again) that this amp could do with. the ones I have , have a too low Z ratio. I was hoping to be able to use the ones I have and find a mains tfmr from my stash. Building this amp to completion at this time would mean winding a mains tfmr and buying a couple of OPT's.

Before I put it away though I will get my head round computing Z out for CCS's and modify the CCS I'm using.

Thanks for your help/comments, Andy.

[Radio Wrangler]

Originally Posted by tri-comp

What's wrong with the lower picture on page 3 of the MJE340 ON-Semi datasheet ?
To me it looks like DC Current Gain (hFE) can be read all the way down to Ic = 1mA

It does look that way, doesn't it? but it's only a plot of typical performance. It gives you an indication of how hfe varies with collector current, but it isn't a guaranteed specification. What ON sets as limits is the hfe at 50mA/10V/25C in the spec table IE anywhere between a minimum of 30 and an unspecified maximum.

So rescale that graph so the 50mA 25C point is at hfe=30 and it then shows the lowest gain device they will ship. Considering that the upper spec is theoretically infinite isn't terribly helpful.

So a properly designed circuit should work correctly and to its spec with ant transistor with gain equal to this minimum plot, or greater.

If you design around typical figures, your circuit will work provided you get a typical transistor to put in it. A transistor from anywhere else in ON's production spread might or might not work.

Transistor parameters tend to vary more between bathes than between individuals within a batch. So even if you bought a box of transistors, if the first device out the box had a gain outside the range your circuit could handle, then you may likely find all the rest are similar.

A good design has the range of each parameter which it can accept matched to the range of that parameter the maker will ship.

This principle holds true for valves or transistors or nuts and bolts.

The spreads though are far wider for transistors than they ever were for valves. Good transistor circuits can be designed, but they involve more work than valve designs.

David

[tri-comp]

Hi David, thank you for 'splaining.

In this circuit the CCS is there to establish a high Z at the CF-cathode.
How high isn't that important as in any case, no matter what hfe you'll find on your MJE340's, it is bound to be extremely much better than using a simple resistor in the range 10KOhm ~ 47KOhm which was commonly used in the past.

My calculation shows a CCS Z-out of 12MOhm but even if that off by as much as x10, plus or minus, it's still a vast improvement over the simple resistor.

Using a single-transistor CCS as Andy did, isn't much of an improvement over the resistor if any at all. Especially not when using a power-device as the rather significant internal capacities are bound to degrade the frequency response of the coupling.

[Diabolical Artificer]

Thanks as always David. Been reading up on hfe/hFE and CCS's as used in valve amps. Firstly it seems hfe has fallen out of favour and beta is used mostly as hfe/hFE is confusing. Damn right. Writter's like Morgan Jones and other valve guru's are vague on the subject of CCS's which doesn't help.

So from what I've gleaned the Zout formula simply put is - Zout = Re x beta upper Q x beta lower Q, no need to worry about hoe. Also like a lot of valve circuits it seems tranny circuits are ish too. EG beta is 30ish for a MJE340 and 100ish which means my CCS had a Z out of 19k ish, but with an extra Q would be 1.9Mish. A lot of what you were going on about T now makes sense, as does the mention in a lot of tomes that any circuit that relies on beta is flawed.

Yesterday I put another tranny on my CCS - b*gger wouldn't work. After much head scratching I found I'd put the zener/Vref in the wrong way round, this meant Vref was 0.6v instead of 6.2v. Which also means putting an emitter R in is pointless, no wonder I had to turn the preset to near zero ohms for it to work.

As I used these unmodified in the 2C34 amp, it's also no wonder results were less than optimum. It means my CF's were running on fumes , analogous to wimps getting sand kicked in their faces instead of being mighty muscle bound jocks : )

I've took the amp apart now, but will re-visit it at some point.

A.

[bikerhifinut]

I have to admire your tenacity Andy.
I would have taken the easy way out ages ago! I am very risk averse as you know.
I usually stick to known topologies and circuits and do no more than refine them with closer tolerance better performing modern passive components.
Mostly I think I get a very acceptable result.

Anyway high Fives mate you push the envelope and that has got to be appreciated.

A.

[Radio Wrangler]

hoe gives the output conductance of the transistor acting alone.

In a constant current source, there is a resistor in the emitter path and this provides degeneration (== feedback) and that feedback modifies both the input and output impedances.

Once you start to see how the collector current as a component of the emitter current creates voltage across the emitter resistor, and thereby opposes the voltage drive to the base, those transistor amplifiers touted as having no feedback start to become hilarious.

Bikerhifinut, in the nineteen eaghties the west started to copy the Japanese quality circles, where manufacturing processes and products were gently perturbed and those changes found to improve the product were retained. Repetition steadily optimised the product. It was subtle and effective, but had one big limitation. It peaked the performance, but it could only take you to the peak of the hill whose slopes you started from. It could not take you to a different and taller peak if you had to cross a valley to get to it.

Similarly, fine-tuning an existing design may miss opportunities that larger scale changes might yield.

Learning to be an effective designer isn't easy. It's no where near as hard as most people think it is! It also opens doors you'd not even think were there.

David

[Diabolical Artificer]

As I understand it David, hoe is minimal in this CCS, so can be left out.

FB, I have trouble visualising this concept; I think of FB as being introduced by some sort of loop, a bit like running out the front door of a house, and going round the back into the back door. So how does the voltage or current get "back", in this case, is it going down through the resistor and back into the base? The same applies be it through or because of a cathode bypass cap. the FB "path" in a cathode/emitter follower also alludes me at present.

"collector current as a component of the emitter current " Surely Ic and Ie are one and the same thing? As current flows from collector to emitter. Sometimes understanding what's happening in a Q or valve come to that, is like trying to follow individual kids in a playground of 100 pupils. I,Z etc are over the shop. I'm getting there, been taking the Art of Electronics to bed ( last of the swingers eh) and light is creeping under the door of perception.

Andy, bless you. We just have different approaches, things really bug me if I can't understand what's happening be it in a circuit or mechanical device or in a human's head. Your approach sounds more relaxing; we're just coming at the same thing from different angle's, neither is the right or wrong way.

Andy.

[Radio Wrangler]

Ie = Ic + Ib or else Mr Kirchoff gets upset. In a transistor with plenty of gain, Ic is much the biggest part of Ie, but it's good to remember it's not all of it.

You apply a voltage(wrt ground) to the base of a transistor. The emitter is grounded through a resistor. There is a power supply to the collector.

The transistor turns on and pulls up the top end of the resistor until the emitter sits about 0.7v below the applied base voltage. Ohms law now gives you the emitter current. The base current, taken from whatever is driving it, is 1/hfe of the emitter current. The collector current is pretty much equal to the emitter current, or Ie-Ib if you feel picky.

To a first approximation, whatever drives the base sees the emitter resistor times hfe.

If a random perturbation raises the emitter voltage a little for a moment, the transistor sees less Vbe and reduces its current to compensate for the perturbation. If the emitter voltage is perturbed downwards, then the transistor fights it with more current. This is the feedback integral to an emitter follower. The collector current gets taken along for the ride.

If we have a load in the collector and vary it, hre causes the Vbe for a given current to vary. this will affect the current in the current source, so the current source no longer looks like an ideal infinite-impedance source to the collector load.

If the emitter resistor was zero, this change in vbe seen as a fraction of vbe would scale the current.

But with a significant resistor size, the change in vbe has to be seen as a fraction of the total emitter voltage plus vbe. the change in current is therefore far less

There IS feedback, but the loop is of zero area it's because the emitter resistor is both a part of the loop supplying voltage to the base, and of the loop carrying power supply current through collector load, collector, emitter, emitter resistor, and back to the power supply.

The base loop is the input driving current to the transistor
The collector loop is the output of the transistor

The hfe gain of the transistor couples these loops in the forwards direction

hre couples them a little in the reverse direction

The emitter resistor couples them strongly in the reverse direction, and just a little bit in the forwards direction.

For formal analysis a hybrid-Pi model with all four H-parameters gives a set of simultaneous equations that need solving. This is OK on degree courses, done manually. A better analysis needs to model variation of transistor parameters. Spice does this, but most people using it are unaware of this.

David

[julie_m]

It's been awhile since I actually got my hands properly dirty with the actual maths, but isn't Ie = Ic * (1 + 1 / hFE) by definition? The base current also has to flow out of the emitter.

[Radio Wrangler]

I'd kept it conventional by taking currents in the expected directions as positive.

However, if you consider the current of an electrode as the value going inwards, then to be consistent then all three should be the same and emitter current becomes negative

So Ie = -Ic - Ib

Also if you want to do an analysis as part of a mesh network, then consistency says you have to pick one direction of circulating currents, say clockwise, and then apply it to all meshes, so you'd analyse your transistor circuit in terms of current into base and out of emitter as part of one mesh, then you'd do current into emitter and out of collector as part of the next mesh.

There are many conventions which can be used for current flows. There are also many other models of transistors and sets of different forms of parameters.

H parameters are a *******ised set that aren't consistent within themselves as a mixture of resistances and conductances.

David

[Radio Wrangler]

All assuming NPN, of course...

David

[julie_m]

It depends which sign convention you are using.

Some textbooks liked to have all the currents at a circuit node adding up to zero, so you could tell by the sign which way it was flowing. Others liked to keep the sum of currents flowing into a node equal to the sum of currents flowing out of that node, all written positive. That way, a negative answer meant you had guessed a direction wrongly.

There is no real difference mathematically, since you effectively just change the sign when moving something from LHS to RHS of an = sign or vice versa anyway; but one way is usually easier to understand than the other, and which is which is a person-to-person variable.

As I understand it, h parameters are a convenience for getting a voltage and a current out for a voltage and a current in, and mean not all the numbers in the matrix are dimensionally compatible -- there's a resistance, a conductance and two dimensionless constants.