?

Please ignore, I see a problem with what I did.

?

?

Thank you, Tony.

?

Here’s the problem I encountered with the unmodified LM317A_TRANS while hoping to use the DC current source I1 to create a programmable current source. See ProgCurrSource.asc.

?

The drop across Rc leaves only 12.5mV across Rs, leading to a 10mA current through RL. There are no AC currents flowing through either ADJ or through Rc, and there is no DC current through ADJ.

?

The PSRR at the OUT2 terminal below 100Hz is a measly 25dB. At OUT1, the same PSRR is 55dB.

?

With LTspice 17.1.14, simulating with .ac dec 100 10 10000 yields the following errors, although it also obviously produces the above results:

?

Questionable use of curly braces in "b§e_abmgate yint 0 v={if(v(a)>{{vthresh}},{{vdd}},{{vss}})}"

??? Error: undefined symbol in: "if([v](a)>((vthresh)),((vdd)),((vss)))"

Questionable use of curly braces in "b§e_abmgate yint 0 v={if(v(a)>{{vthresh}},{{vdd}},{{vss}})}"

??? Error: undefined symbol in: "if([v](a)>((vthresh)),((vdd)),((vss)))"

u2:_u1_u12:dd1: Emission coefficient, N=2.6437e-312, too small, limited to 0.1

u2:_u1_u11:dd1: Emission coefficient, N=2.6437e-312, too small, limited to 0.1

u1:_u1_u12:dd1: Emission coefficient, N=2.6437e-312, too small, limited to 0.1

u1:_u1_u11:dd1: Emission coefficient, N=2.6437e-312, too small, limited to 0.1

Direct Newton iteration for .op point succeeded.

?

?

?

?

?

There are at least two unencrypted models I saw today: LM317_TRANS (No note on PSRR) and a more extensive model LM317A_TRANS, which states:

?

* Model Usage Notes:

*

* A. Features have been modelled

*????????? 1. Startup Response

*????????? 2. Peak Current Limit

*????????? 3. Dropout Voltage vs Output Current?

*????????? 4. Line & Load Transinet (sic) Response

*

* B. Features haven't been modelled

*????????? 1. Quiescent Current vs Input Voltage

*????????? 2. Temperature dependent characteristics

*????????? 3. PSRR Response

?

If the more extensive model says no to PSRR, I don’t hold out hope for the less extensive one.

The notes in the LM317A model are nonsense. PSRR is modelled, although it doesn't quite stack up with the datasheet, which states 68dB (typ) - the model delivers -55dB. For comparison, for the LM317, the datasheet figure is 57dB and the model delivered 62dB. The model behaviour overall is better than for the LM317, which doesn't model the output capacitor effect on PSRR. In the LM317A, it is modelled. The LM317A PSRR figure for when a 10uF capacitor is hung on the ADJ pin can't be achieved.

What doesn't seem to be modelled is the effect of the load current on PSRR. Maybe this is what the notes were supposed to cover.

The two transistor level models don't model current limiting correctly - it kicks in at ~800mA with some foldback, whereas the two TI behavioural models limit at ~2.2A, but with no foldback. None of them will have limiting modelled fully, because it is determined by Tj in the actual device. The LM317A behavioural model doesn't implement temperature-dependent limiting at all, whereas the LM317 version does, but the analysis explodes at 75?C.

Conclusion: all the models are flawed, but the TI LM317A model is the most reliable. I modified it so the inherent PSRR was 68dB, as per the TI datasheet Figure 10, and it showed the additional 10uF capacitor on the ADJ pin lowers this to 85dB, although Figure 10 suggests it should be closer to 82dB. These are close enough.

I didn't test the other characteristics.

--
Regards,
Tony