LM399 - running the heater off a dirty, dirty boost converter!

[robert.rozee]

i'm thinking around an idea: could one run just the heater of an LM399 off a +/- boost converter without significant degrading performance of the zener portion?

the converter i'm thinking of using is this one:
 https://www.ebay.com/itm/314255439007
a simple XL6007 based boost converter with a crude capacitor + 2x diode to provide the -ve rail. input would be from a single 18670 LiIon cell (3.7v). the majority of the total power usage would therefore be coming via the dirty XL6007.

the supply for the zener side (and associated LT1001 opamp) would be created by a ICL7660 Switched-Capacitor Voltage Converters driving a 4- or 5-stage capacitive voltage multiplier from the CAP+ pin (pin 2). the current requirements here would be extremely low, just a few mA.

my plan is for a portable 'artifact', with the zener voltage presented directly to the output terminals. charging of the 18650 cell would be from a socket accepting USB5V.


any thoughts?


cheers,
rob   :-)

[Kleinstein]

The output side of the boost converter is not necessary bad. I would definitely add one more stage of filtering with an inductor and capacitor though. Another point to look at it that the converter is OK with the rather high current drawn during heat up.

The LM399 does not really need a negative supply. It is perfectly fine with just a positive, like 16 V maybe 12 V for the heater and the ~ 0.5-1 mA for the zener.
An important point is getting the ground / neg side connections right. So battery ground, ouput ground, heater ground, OP-amp ground and Zener ground are kind of different branches, though connected at some points.
There is a little effect of the heater supply voltage - but the SMPS should offer regulation.

I would be more worried with the ICL7660 - they produce a awful lot of current spikes and are not nice to the supply side. Chances are it is worse than a magnetic boost converter.
I would more like use the same boot converter as for the heater with some fitler and maybe an LDO for some 10 V or so. In principle the OP-amps would not even need an LDO.

For battery operation I would not use LT1001 / OP07 any more. OPA202 or OPA207 are comparable to better noise and much lower supply current.

[KT88]

It is possible to use switched mode rgulators in such applications. But I won‘t rely on any ebay stuff and roll my own. It would require some effort to get it right though. A Cuk converter could be the better option as it provides an inherently cleaner output.

[iMo]

Any such converter creates ripples which are difficult to filter. The ripple could be easily 50-100mVpp (or more). You would need LC type filters to attenuate them. I would not use a switcher with the 399, except the burn-in. The heater and zener are two blocks of electronics, coupled internally via a reverse biased diode, thus feeding pulses into the heater may mess with the zener too..

[robert.rozee]

Kleinstein: i've seen in various places mention of the LM399 producing slightly better performance when the heater is driven from a split supply, hence the choice. it is easy enough to achieve (addition of 2 diodes, 2 capacitors), and ensures that both heater terminals are kept well clear of ground - ie, there is no heater ground and the diode coupling heater to zener will be held well and truly reverse biased.

H+ and H- will be connected via inductor/capacitor filter. boneDragon sent me a PM with a schematic of his own reference design, using 10mH series, 220uF//10uF to ground as filters. this seems a reasonable starting point. i don't see the boost converter having any major problems with initial current draw while heating up, worst case the -ve rail will slump for a few seconds.

since i'm using the opamp just to supply the zener, and taking two wires directly from the zener to the (approx 7v) output terminals, there will be very little possibility of getting the grounding wrong.

if an LDO performs better than an ICL7660 then the LDO is a far better/simpler solution. however, the '7660 would be used in a non-standard configuration, basically as just a single hi/low switch, so current spikes may not be such an issue. or, i could just use a pair of 9v primary batteries to supply the opamp/zener, as the current draw here is so low.


imo: measuring the ripple will be an interesting problem, my scope only goes down to 500uV/div. i do have a 34401A, which may show something, but not quite the right tool for the job! my LM399 has been burning in for approx 2 years now.


cheers,
rob   :-)

[Kleinstein]

The temperature regulation works a little better with a higher voltage. There should be no big different between +-15 V and a single 30 V. The advanatge with a single supply is that it is easy to make sure that the H- is not significant more positive than the zener neg side. If a dual supply is used, especially from a relatively weak converter it would be good to have a schottky diode between H- and GND to prevent H- to get too positive.
The charge pumped negative supply may not be very stable - so I would keep the option to use just the positive supply and use GND directly for H-. The filters should than be only on the positive side, maybe with the exception of a common mode choke for the USB supply.

If designing a PCB (and not using prototype board) I would include space for an RC damper as needed for the ADR1399 reference type, just in case.

The ripple from the SMPs tends to be not that small - at least not before the filter. So a scope with 5 mV/div would still be OK to see something.

[robert.rozee]

the boost converters arrived from over the seas, so i put something together to test the ripple produced. a 1k resistor simulates the LM399's heater, while an LM317 provides a regulated 12.2v - a 270 ohm resistor from Vout to the adjust pins draws around 4mA which should be sufficient for testing.

the filters and an LM317L were assembled on a dodgy plug-in prototyping board, and so quite a bit of jiggling was required to get things operating with even moderate reliability. with a soldered assembly i'd expect somewhat better results. i have some LM1117-adj LDO regulators on order to replace the LM317L.

here is the (partial) schematic:



and here is the resulting ripple (from top to bottom +15v, -15v, +12.2v:




cheers,
rob   :-)