My apologies.. I can be slightly unreliable.
Ideally you would want to synchronise the two circuits so they operate at the same frequency. It helps to stop them interfering with each other.
CT and RT are the timing components.
https://focus.ti.com/lit/ds/symlink/uc3842.pdf
Page 6)
There are equations in one of the other documents but in this case it is simple enough to use the graph from which it looks like setting CT to 1nF and RT to 22K will get you the switching frequency of about 100KHz.
The larger the core the lower its thermal impedance and therefore the greater the power it can dissipate for the same temperature rise above ambient....
I'l continue in a bit but for the moment I now have this on my 'spreadsheet',
From left to right,
The core type. Its effective area, Ae. The thermal resistance, Rth. This is in degrees centigrade per watt. Next is the amount of power it can dissipate if I wish to limit its temperature rise to 40C above ambient. The larger cores with their lower thermal resistance can dissipate more power for the same temperature rise. I split that available dissipation equally between the Core and the Windings, PCU/PCR. Next is the effective volume of the core Ve. Dividing PCR by Ve gives me Pv which is the specific power loss per unit volume that will result in that particular loss for each core.
The N87 relative core loss graph has that value, Pv, plotted on its Y axis.
The X axis is the switching frequency and a family of curves for various peak flux densities is given. As an example if I were to take the ETD29 the Pv Value is 134kW/M^3 which, and it involves a bit of guessed interpolation, would mean I have to limit the peak flux to 150mT. Just to confuse things that peak figure gets doubled to become 300mT.
Whilst the converter is operating in one quadrant of the B-H curve it is still allowed a peak-peak flux excursion as long as it does not exceed the saturation flux density for the material. Filling in the other numbers...
You will notice that you can operate the smaller cores at higher fluxes. You may remember from biology that the surface area to volume ratio determines how well something can get rid of its internally generated heat. The same applies here.
Some time in the past I selected a minimum input voltage of 250V to account for a single cycle line drop out. That will be our VIN. At this point the converter will be operating at the maximum duty cycle of 50% which will give Ton at 100KHz as being 5uS. So, now we can fill in the minimum required primary turns.
I am sorry this is taking some time and possibly time you do not have. Hopefully it, the transformer design, will be finished today.
Back in a bit.
Genome.
---------- Post added at 11:36 ---------- Previous post was at 11:25 ----------
Oh, R7 and R8.
Those are the components on the current sense pin. One is connected to the current sense resistor the other will be connected to the Ramp, RT/CT, pin to provide slope compensation. It is supposedly not necessary for this particular type of converter but does help to define the gain. The ILIM pin has a 1V threshold on it which with a 1R current sense resistor would give a 1A peak current limit. Slope compensation modifies that figure because it introduces a divider and it also injects the Ramp signal. As a result the peak current limit level is modified in this case going from 1A up to about 1.5A.
As mentioned this is 'current mode control'. The circuit controls the inductor current, turning it into a current source... That simplifies the output filter characteristic. As an LC section, ignoring the ESR, then above its corner frequency the voltage response would be second order with 180 degrees of phase shift. That can be accounted for but might make compensation more difficult. Current mode control 'hides' the inductor and make the response first order, a current source driving a capacitor, which is easier to deal with. There are other benefits as well.
Although it looks confusing having a current loop along with the suggestion that current is being controlled that loop is surrounded by the voltage feedback loop which ultimately controls the output voltage.
Genome.
---------- Post added at 13:14 ---------- Previous post was at 11:36 ----------
Now things start to fall to pieces..
From one of the data sheets this is a mechanical drawing of a typical bobbin,
From which we can extract certain dimensions.
Ww is the available winding width.
Wh is the available winding height.
In is the mean length of a turn.
An is the available winding area assuming no margins
An! is the available winding area with margins
Ww is reduced by 8mm. The requirement is for 'creepage and clearance' in order to comply with agency isolation requirements. This is an offline converter and you do not wish to expose the end user to mains voltages. 8mm works out as being margins of 4mm either side and, with appropriate insulation, the creepage becomes 8mm.
Putting that in the spreadsheet we get,
You will see that the smaller bobbins/cores lose quite a bit of the available winding area as a result of the requirement for those margins.
And then it gets worse, sort of. You may notice that I have rounded up the number of primary turns to the next highest even number. Depending on how things go that may or may not be required. Unfortunately when operating at high frequencies there are two/three factors that have to be taken into account.
The first is that with AC currents you experience something called 'skin effect' whereby the high frequency components are constrained to flow in the outer layer of your conductor. The AC impedance, and hence losses, of the wire is greater than it would be for DC. There is also something called 'proximity' effect' whereby within a specific area between overlapped windings the field energy is minimised... I can't say I fully understand it so I'll leave it up to Lloyd Dixon of Unitrode,
https://focus.ti.com/lit/ml/slup197/slup197.pdf
Otherwise you get into the realms of 'Dowell's Curves'. It seems Snelling was involved as well. Unfortunately I no longer have the document sent to me by the helpful people at Philips..
Anyway.
Wire Tables,
**broken link removed**
and a piece of software that will do some of the sums for you,
**broken link removed**
Unfortunately the link at the bottom is 'broken'. I'm surprised the copy of the site is still there. If you are feeling brave then the .zip file is here,
**broken link removed**
It's a Delphi(4, Pascal) program that implements, in part, the equations presented in,
Fortunately Dowell and Snelling plus lots of other people have worked all of this out. I've used a paper by J.Jongsma from the Central Application Laboratory, Philips Product Division Electronics Components and Materials, Eindhoven, The Netherlands.
Electronic Applications Bulletin, Vol 35, No 3, May 1978.
Minimum-Loss Transformer Windings for Ultrasonic Frequencies. Part 1 and Part 2.
It comes with the wire table database attached and deals with both skin and proximity effect. It is a bit 'old' but it works under Wine, my excuse I use Ubuntu. I don't know whether current versions of Windows will be happy with it but it was cobbled together under WIN95 so if it moans you might have to play with the 'compatibility mode' settings.
As Dixon suggests one of the methods for reducing proximity effect is to split the primary windings to sandwich those in the secondary which effectively halves the number of layers. It also has the benefit of reducing leakage inductance, that was the third one.
Time for a break....
Genome.