To choke or to CCS?!?

[chipwelder]

Related to the toroid thread

Long term plan 8 channels of 50 W GEM power... http://www.zen22142.zen.co.uk/Circuits/Audio/gem100.htm
There's two versions, a constant current source and a choke version.
I don't know why, but a friend of mine sez the choke will be better, I suspect it is beyond me to know why... maybe it is because I'm prolly going to ask him to bias or fiddle with the ccs until everything works out -  which I can't blame the oke for... So i'm open for peasant version discussion/suggestions...

Maybe related, but what does the size of the transformer have to do with anything but cahunas, weight and supply ripple?

[Ampdog]

Chip,

As you have probably noticed on that other thread, my main problem with a ccs in the role of a choke is with heat dissipation. The role of both here is to provide a high impedance to the power supply ripple i.e. at 100Hz, and as little resistance as possible at dc (the latter being a loss). A choke is an impedance which is normally a few k.ohm at 100Hz and with a dc resistance of some 20 - 80 ohm only, depending. As with a capacitor the (ac) current through a choke is ideally 90 degrees out of phase with the voltage across it, thus not dissipating heat except for the load current through the winding resistance. The ccs will have the same resistance at both ac and dc. Either route may be preferable depending on the magnitudes of current and ripple. (Chokes are expensive.)

The size of a transformer is primary related to the amount of 'magnetism' (magnetic flux) which is required to do its job, and the 'amount' of wire (i.e. copper) it needs. The number of primary turns are also related to the primary voltage. Then the thicker the wire, the lower the voltage drop and hence copper losses. For the kind of transformers likely to be encountered here, both core loss and wire loss are calculated to be around 3 - 5% each.

With a choke, the core might be big because enough 'free flux' has to be available after that 'locked' by the dc flowing through the windings. When enough current flows through the windings on a core to completely magnetise it (saturate it with flux), no extra 'free magnetism' will be available to be used to do it's job. The core size then has to be increased to give enough extra 'free' flux so that it can work as a choke. (There is usually a gap in the core to resist saturation - but the design is another story.)

[ludo]

Ampdog, maybe I'm missing something obvious here (that wouldn't be new ) But, the series resistance (80 Ohm) that limits the current to 450mA in that circuit if using a choke, will dissipate as much heat as the transistors & emitter resistors in a similar active CCS will? Is it not much the same then, the problem of getting rid of that heat in both cases?

I understand that the inductor would not dissipate heat if it had no series resistance, but then one would need an external resistance to set the current. And that resistor must dissipate.

So then is it not really a question of which is smaller and easier to get hold of and easier to mount on the heatsink along with the output stage?

I suppose the active CCS will run out of regulation sooner than the passive one, but only when the output node is getting perilously close to the +V rail. (The passive one is wonderfully impressive in this regard, Chipwelder. Maybe that's why your friend prefers it?) Also the choke will improve regulation as signal frequency on the output node increases. While the active CCS will gradually fall apart. But of course the higher the signal freq, the lower the level, generally.

There are two current sources there. The one from V- is active. If both are active, I'd think one can bargain on better tracking between the two, so more stable offset voltage as things heat up from a cold start. This may not amount to terribly much though.

Given the problem of calculating how to make that choke, I'd say the active CCS is a winner here. But maybe that's just the lazy approach.

[chipwelder]

Thanks for the peasant version Ampdog, I hope I'm getting it, we'll see below

That's OK I'm lazy too Ludo, however the getting the choke designed is not an issue...

So I'm guessing the main benefit of the choke is that it is less noisy in operation - or lets less noise through courtesy of the change in impedance vs. frequency (which is really helpful in keeping high frequency hash out of the output devices that "could" cause oscillations etc.) - the choke as a whole won't heat up as much, or shall we say by as many degC as the transistor... total heat remains the same, device temp not... so failure is not an option and stability is a given... but I may be wrong...

I then have to deduce that what makes sense for a 50W channel that operates sortof at Class A, a transformer larger than 200VA (i.e. if the heatsinks and devices will manage, a brief 200W into 2Ohm... - supply sag) would offer returns that starts to diminish in terms of less core saturation and possibly a quieter psu... so actually 300VA should be good, 500VA luxurious - if a little silly - not to mention costly...

[ludo]

Pottering around Communica today I looked at the transformers in the bins there. Something around 50VA isn't likely to weigh in under R100 (!)

The more I ponder the choke, the better it looks to me now. And the bigger the choke the better obviously. So probably your friend is right and I am just lazy. But the transistors will be way cheaper at just a couple of R. Especially if you have sizable heatsinks already.

If there is oscillation, more likely it wil be due to the lack of small base resistors on the output devices. Some don't like that.

How about active CCS for the low freq amps where it works better and chokes for the high freq ones?

About the transformer:

The hassle with sizing the power transformer, at least the way I see it, is "just" a matter of resistance of the secondary. (But of course it doesn't stay that simple.) Because the secondary charges the psu caps in huge pulses of current of fairly short duration, that resistance is a killer. It causes voltage drop. But in the interest of economy, limiting the current pulses by having the voltage drop there is actually a benefit. Now one has to figure out who's been economising where and for how good a reason: The secondary copper might be the most expensive material in the transformer. So that's a hint.

But there is also a limit to how much of a current pulse one can have without saturating the core. An amp-turns thing. Then it becomes fuse-popping territory. One can't have more flux than the core will support and one can't get more out than you put in. So you either pay for a bigger core, or live with the ripple when things get busy signal-wise. Or get psu caps like jam tins/many little ones (cheaper & better). Or up the frequency, but let's not go there.

Maybe stop where one can still afford it. What you save on the chokes might be better invested in a larger power transformer or just more psu caps. If you have a bad case of the DIY bug, which is a wonderful thing, then you can tell yourself that a large transformer is a good long term investment, eh? For 8 channels, 4 transformers now and the other 4 later when you've taken the Lotto or landed that big fish, is also an option?

How about smaller transformers for the high freq amps and larger for the low freq? T'Tech will hate you a little for the bits-n-pieces order and make you wait long, but it could work out better.

I hope that was useful, not just BS overload

[chipwelder]

No it is all useful... especially since it looks like R4k for the amount of trannies I supposedly need for 8 channels... will do with 6 channels now, but still R3k - eina... Wonder how much three 1kVA's will cost...
this is prolly why most people have one big fat transformer for two or more channels... BTW the trade-off given was 500VA with a choke 2kVA with CCS, but I'm sure there's a slight exaggeration there... Hence my questions here...

I think I just might do the ccs for lower freq.s however low is in this case up to about 2kHz. (I want to also build the 100w version for shits and giggles... probably borrowing parts and taking it to my buddy's system, which is more or less my reference - I am far from the reference however). BTW ludo, you must come and listen to the new x-o boards... it made a huge diffs.

I often wonder if that costly secondary isn't the differentiation of lundahl et al. audiophillic transformers vs. industiral types... I remember some guy writing about psu's and saying that you can measure the sag or idle current of the transformer and know how efficient and well designed it is... but I guess it is a case of buying a 300VA lundahl or a 500VA Industrially optimised transformer... works out to about the same...

[handsome]

Allow me to  toss in my five cents worth (dubious as it may be): with a choke input you will not have current pulses as choke inputs allow for more constant transformer loading that is your current waveform will approximate a sine wave versus a very steep sawtooth. this will relax transformer current requirements (eliminate ringing and lowering the ripple current rating of the capacitors) but increase the transformer's voltage requirement (higher secondary). Thus output from the choke will be 0.9x of your secondary voltage versus 1.4x in a traditional capacitor input scheme. However if you choose to use a  current source, you can seriously lower the amount of capacitance after the rectifier which will also lower the current pulses, as the (good) CCS will deliver a constant current regardless of the input waveform provided said waveform is always above the CCS drop out voltage. Personally i would do a choke input and a CCS......

[chipwelder]

Shucks am going to have to read that three times or so... bear with me...

[ludo]

I think Handsome is considering something else here. Never seen it in a SS amp, but hey, maybe it is a good idea.

The choke I was referring to is the current source thingie from the V+ rail to the signal output node of the amp. I was assuming straight transformer-->bridge-->caps for the power supply.

2KVA is insane. Not to denigrate the insane now, but I wouldn't consider that size for 50Watts. 300VA for two channels looks plenty crazy enough to me. Lots of caps.


Thanks for invite. Will PM.

[chipwelder]

Cool, well Ludo 50 into 8 Ohm, 100 into 4 ohm and 200 into 2 Ohm... not that my drivers would require that, but in a big speaker system they often do... I hear you though, I always thought as much as your output devices and heatsinks can handle into whatever ohms + a little for psu sag, under heavy going should be plenty... but now the Transformer gets more noisy as the going gets louder - if I unnerstand correctly... and that ends up in your input stage... you need lots of caps to keep it out, and lots of caps make the ESR/ESL higher, and that has a negative effect on the high frequency hash, and then you tune it with more bypass caps etc... which equals more components... So the bigger the better is only REALLLLY true for the transformer, except for your pocket, and it can possibly prevent lotsof complexity thingies later on... but nou ja, I know enough only to be dangerous...

[ludo]

If you parallel the caps, like in the PSU, their ESR and ESL get devided by the number of (similar) parallel devices. Especially in the PSU, one has opportunity for this and it can even save some pennies.  The main thing is of course that the capacitance increases. ESR is pretty tiny. I'm actually a bit sceptical about excessive bypass caps in parallel to electrolytics. They can actually cause lots of trouble some places.

Circuits that don't handle noisy supplies well, should get regulated supplies. Tall order to cater for 200W then.

[handsome]

the more capacitance you have after your rectifier the less time your caps have to charge up per mains cycle and therefore the sharper (shorter) the current pulse that charges them up as well as the larger the amount of current that flows. the incredibly huge pulses (they are) in combination with their very high frequency generate rf 'hash' that can get into your sensitive audio circuits and essentially demodulate down to audio frequencies where they will be heard. A choke input supply (like Musical Fidelity use) draws a constant current from the transformer eliminating any transient current behaviour in your psu but in solid state amps you need a very low resistance choke to avoid voltage drops.

Download the PSU designer from www.duncansamps.com, a brilliant program that allows you to design almost any psu, and look at the different current and voltage waveforms for choke and capacitor input supplies.

I realise now you are not talking about actual psu chokes but a choke as ccs. A ccs can be designed to perform better than a choke that is have a flatter impedance across the frequency range but it will get hot and uses parts. a choke is simpler.

Best of all of course, would be a ccs/choke feeding a shunt regulator......................mmmmmmm!

[Ampdog]

OH DEAR - - OOPS!!

I had it quite wrong! I have to confess I looked at the Graham Maynard circuit carefully only now. Why does the good fellow not draw his circuits as per normal convention? There was the power source away to the left on a tiny piece of real estate, etc. ...

I earlier presumed the 'choke' as per conventional power supply design, but it would appear that Graham has it somewhat in parallel with the output - at least not in a power supply filtering capacity. (I was slightly puzzled by a choke in an SS circuit; normally not used because of the prohibitive size then.)  My brief and incomplete previous arguments must be seen in that light. (I will come back to that in later post if I may; must get up this morning at 05:00.)

My previous vague numerics were somewhat off-normal. The ac action of a choke is proportional to the inductance (available L in the case of one carrying dc), which do not generate heat. In the case of a ccs, the 'internal impedance' is much higher than its literal V/I - a simple transistor can operate at Vce of say 10V and Ic of say 5mA, i.e. dissipation of 50mW, but its 'internal' or current impedance can be say 10K as a current source. But within the limits of its supply voltage and Vce. This is similar to a choke, although choke resistive loss can be as low as the design allows. My quoted value of up to 80 ohm was excessive; typical low current valve stuff. With my 100W amplifier, the power supply choke has 15 ohm resistance, that is 6W heat at 400mA (the component is quite large to be able to use thick enough wire). A ccs doing that job would dissipate more as it will need to be functional at the highest required voltage drop (at 100Hz peaks).

I will need to re-examine the Maynard circuit to see what exactly his choke is there for - it does not seem to appear as a power supply smoothing element in series with the +40V power supply. (Hope I do not need to come with another 'oops'.)

Just a further remark: It is correct that the higher the power supply (input or reservoir) capacitor the higher the peak charging current (and thus possibility of generation of 'switching' or rf pulses). But it must also be bourne in mind that charging time would be less because the discharge which needs to be replenished would be lower for a high capacitance. Furthermore the transformer series impedance would limit this current. i.e. it would reach a limit for ever higher capacitance. One must also keep in mind that the secondary series impedance act at 100 Hz (half a sine wave); it is higher than simply the dc winding resistance of the secondary, but the picture is also more complex as showed by a Spice analysis; lots of harmonics.

[chipwelder]

Eish... julle oaks raak nou seriously clever, i'll have to re-read this again and again after I finished this report...

[ludo]

Did some half-wit simulator wielding while Ampdog had a short(!) nap. Then we had a technical glitch with the forum. Kay, bless his heart, has fixed it by extending the confines of the playground for us. Now we have elbow room again. Thank you most kindly Sir!

My "non-artists" impression of the transformer (300VA 2 x 35V) is done with coupled inductors and some series resistances added. 4 Ohm primary and 0.3 Ohm in each secondary. Then a bridge, and then 4700uF caps. Another of the same but with 47000uF caps. I start the whole lot off at (+ and -)38VDC, wait for 1 second (the caps still charge up further) and then run some 3 Amp(peak) triangles off the caps at 100Hz for a bit. Then the triwave source shuts off (after running for 0.5 sec) and we can see how things recover. I also added the 500mA drains (CCSs) Chip would have in his amp over the caps.

Basically, fig 1 is more or less what one might expect to see in terms of voltage ripple on the caps. Only the positive rail shown. Negative rail does the same but upside down. Red trace is voltage on the 4700uF and green on the 47000uF.

Fig 2 is the current flowing in the secondary during the same sim. Red in the transformer with small caps, green in the transformer with big caps.

The charging pulses become impressively low (with small caps) in an amp that has low (say 100mA total) quiescent current for instance, at least while there is little/no signal. Fig 3 shows the currents with only 100mA quiescent in stead of 500mA coming off the caps, same colour scheme. Looks nicer when there's no signal, but not useful for this particular amp that does a kind-of class A imitation?

Sooo:

With big caps, there's a somewhat more consistent charging current (pulsed) in the secondary winding, and a much more consistent voltage on the caps. Ie, things are more equaly horrible with lots of capacitance.

Now which to prefer? Big and sudden voltage swings or constant troubles in the secondary (and the primary)? Eenie meenie miny... I like stable voltage on the caps.

Why?

'Cause the amp as it stands has limited power supply rejection. And I like to think the amp isn't as hard-wired to the transformer as it is to the caps.

On the other hand, the smaller caps might sound more funky. Due to the voltage step on the rail making it's way into the amp, at the beginning, and after a nice big musical wallop. So the wallop starts with an extra "plop" and the next mains cycle after the wallop then becomes a sort of point at the end of a sentence, as the voltage step from the recovering rail feeds into the amplifier circuit. (OK I exaggerate a bit but I'm not enthusiastic about that sort of thing.)

@Ampdog
How would one measure this Ampdog? With a single pulse longer than a mains cycle and looking at the settling behaviour on the amp output? Wish I had a 16 bit scope *with* bandwidth.

@Everybody clever
Any of the actually clever guys reading this ever made a proper model of a T'Tech 300VA toroid with a core that saturates and everything that opens and shuts? It's sort of what I half-tried but I could well be way far off. My voltages look a bit low for 2 x 35V secondaries. Care to share before I make more of a fool of myself? What should the coupling factor between the inductors be for a mains toroid in SPICE? 0.995 like I did here? 1 looks too good to be true.

@Handsome
Do you have a model number for that MF amp? Sounds like power factor correction (which is a fine form of good-neighbourliness. We shouldn't complain about the quality of the mains when we've been polluting it.) So it could be good, but I wonder about the regulation and stuff. Would like to see it.