Zen -> Cen -> Sen, evolution of a minimalistic IV Converter

[bcarso]

Originally Posted by EUVL

Here are all (15) JFETs left on Toshiba's product catalog.
All SMD, all N-channel, no high Yfs.


Patrick

So the 2145 looks like a dual 880. I note that the spread on transconductance is large, so the 4 mS in the table is probably a little pessimistic. However, it's clear Toshiba isn't making large area parts anymore, and in particular no P channel.

If one is going to consider parts for paralleling and either has low operating voltages in mind, or can cascode, it looks like the NXP BF862 is hard to beat. They do oscillate with little provocation though. Groner shows a circuit in which each has 220nH iirc in series with the gate, which avoids the thermal noise of a resistor.

[dirkwright]

Originally Posted by bcarso

Interesting. The datasheet is from 2007, but good on Mouser to carry it. A nice high breakdown voltage too.

When Toshiba discontinued the venerable 2SK389 dual they did have two SMD duals in two package styles that weren't the same chip as the 389, but looked o.k. What I was annoyed by: the package was only 5 pins, with the sources hardwired together Much less versatile. A six-pin package would have been so easy to do.

The LSK389 is still available.

[bcarso]

I got two six-packs of Ray-O-Vac "Platinum" NiMH AA cells in six-cell holders (they were out of stock on eight-cell ones, which I would have preferred). The open-circuit voltage for one series string of six was about 7.7V. Based on the change in voltage with various loads I deduced that the equivalent series resistance was about 550 milliohms. With excitation from a generator a capacitance term arises that reduces the voltage somewhat at frequencies around a kilohertz. And by around 20kHz the impedance begins to look inductive, of order 3.4uH (surprisingly high I thought). This is the result for one six-pack series string.

As far as noise, it tends to be extremely difficult to see due to magnetic field pickup in any typical lab environment. But by positioning two packs in series carefully, with a battery-powered preamp with a gain of 500 and bandlimiting with simple R-C highpass and lowpass, defining a BW of about 16Hz to 31kHz, I can't reliably see any increase in noise due to the batteries, whether open-circuit or with a 680 ohm load (hence about 22.6mA). The preamp noise referred to the input is about 700pV/root Hz.

One issue with such a battery supply is just getting decent contact to the cells, including the traditional clip and wires. I had to dissect and solder the connections for the clips to get anything sufficiently stable for the measurements.

From these results, admittedly with a very limited sample size, (a) I don't think battery noise is going to be a significant contribution to Sen noise, and (b) I wonder why one doesn't just center-tap the batteries and d.c.-couple to the output resistor. With standard Sen the d.c. offset will be reasonably small with good JFET matching, due to the fairly high output resistance (this accords with the good power supply rejection models earlier). A 1uF film cap across each series half tames (in sim at least) a high well-out-of band resonance, and there appears no need for anything more than that. However, it is slightly less convenient for turning the power supply on and off, since it will be necessary to lift both ends of the battery, and before doing that ground the output, to avoid horrendous transients propagating downstream if the switching isn't done precisely simultaneously.

For cascode-structure Sen the output impedances are very high, making the d.c. offset issue even less of a concern.

[EUVL]

Many thanks for publishing your test results. Useful info.

The use of 9V batteries was meant to start people off trying them in the first place.
For real applications I recommended AAA sized NiMH.
These can be obtained with solder tags.
That way you can easily make up a pack with no extra contact resistance of a battery holder.

I also use a simple L-C with highish internal R after the battery.
And the charger circuit I published use a relay to switch between PLAY and CHARGE.

So I think all the above mentioned have been solved ?


Patrick

[bcarso]

Solder tags would be a good way to go.

[EUVL]

Something like this.


Patrick

[CeeVee]

Originally Posted by EUVL

Something like this.


Patrick

56 AAA of those is what i'm looking for, not easy to find at a decent price...and genuine.

[EUVL]

ENELOOP AAA 1Z - Eneloop Micro, AAA, 1Z bei reichelt elektronik

Still cheaper than 4x 10VA R-cores, plus diode, regulators, ........


Patrick

[CeeVee]

many thanks Patrick,

that's way better price than i had seen up to now ...and half local price.

[bcarso]

Originally Posted by regal

So the proposal was Sen-ampstage with globlal NFB. Is this an opamp proposal? I 'm not sure what we debating. NFB won't work feeding to the input of the Sen because the current to voltage conversion hasn't occured unless you are somehow building an discrete opamp with it?

But your proposal reminds me of my favorite I/V prior to the Sen. A simple low resistance passive I/V followed by a Jfet phono preamp type stage which has NFB to itself (after the I/v conversion with the resistor.) So its a low 25-35db local feedback jfet amp for gain to allow the use of a smal passive I/V. Follow it by a nice filter/buffer, works nice.

But the Sen works better because it is simpler, less components, less power supply dependance, etc.

I can't graps how the Sen could be part of a GnFB loop. The Borbely design has that optimized the discrete opamp I/V, but its a high component design and doesn't sound quite as natural as the Sen in my opinion.

As promised, I've started a new thread with the most basic approach to reducing the distortion at the input of Sen for low source impedances, titled The InSense Current Conveyor. The sim results will be tested with a breadboard soon. It is tempting to jump immediately to a more complex cascoded design, but I'll try to restrain myself. It will also force me to reactivate the Ap, which has been down since a computer crash

Thanks again to Patrick for stimulating this work.

Brad