Transistor tester

Just tried, scrolled the frequency through in very fine steps. Couln't notice any oddities.

EMI is something you hardly have an influence over. There's typically half a dozen manufacturers and that's all there is.

Shielding isn't really expensive. Now that the Bay Bridge is built the Chinese probably won't hog all that steel anymore :-)

st

Unless you get your own inductors wound at one of the smaller winding shops . Joerg insists that they no long exist, but I found one last year in the Net herlands to wind a handful of prototype coils for me - onto standard former s, which I'd (mostly) bought from Farnell (a broad-line distributor) to be used with RM cores that I'd (mostly) bought from Farnell.

If Farnell is still stocking the stuff some people are still desiging their own and getting them wound in smallish volumes.

Unless you are designing your own coils.

And there's always mu-metal, which isn't cheap, is a swine to fabricate, an d has to be heat-treated after you've fabricated it, but has a much higher permeability.

Hey! My rigol has a digital filter with adjustable 'corners'. I wonder if it's 'before' the FFT function? Then I could get rid of all the 'stuff' folding down.. That would make the FFT more useful. (Well the rigol fft has other issues too, it seems to have reflection, folding type spectra at the high frequency end... like DC is mixed with the highest frequency.)

George H.

Yeah, but the point is you might as well use a linear for a one-off project.

Le Sun, 07 Jul 2013 11:51:30 -0400, Phil Hobbs a écrit:

Beads and series RC snubbers?

Say you allow 1ms settling time per point, and set your R resistance to

For 1V step the snubber generates a 20mA current step and you need 24 time constants to get it down to say 1pA. With 25 time constants the error is 300fA, not considering DA of course.

For the fixture use SOT23 or whatever gold plated footprints with a spring loaded lever to press the DUT on it's landing area. Emitter "pad" is a largish copper pour stitched to the GND plane. Base and collector are accessed through 50R Tlines connections snubbed right at the DUT pads.

Thanks for the nice schematic. On the second sheet you've got some switched cap circuit that looks to be taking the difference of the emitter votlage. (switched by the hi-gate and lo-gate) Are the hi and lo gates also switching the current somewhere? I couldn't see it.

And following that same signal path the next amp stage has an offset that comes from the temperature reference of the Ref02 5 volt reference.. that's a neat trick!

George H.

Right, that's what I've almost always done in the past. In this case I needed a negative supply, and I'm trying to get more comfortable with switchers for other purposes.

The main point of using switchers in this design was to find out what the pickup was like when using good quality toroids and a chip with slowish switching edges. Not too bad, it turns out, even though it's built in a smallish box.

Cheers

Phil Hobbs

The currents are switched square-wave fashion by the NOR-gate flipflop, with the flying caps connected to the outputs. HIGATE and LOGATE are the sampling pulses, which connect the flying caps to the buffered VBE and the lowpasses to the base current TIA. Using a 4017 for sequencing is one of those classical 1970s tricks that are still useful when you want to avoid using a micro. (Or when you want to generate a sine wave, of course.) ;)

The REF02 output is nominally PTAT, but with no guarantees. Still it's close enough to keep the fairly narrow offset adjustment within range. The next version, if any, will have a temperature controlled test fixture. There's also something wrong with the beta error output for PNPs--it has an offset of about +200 mV even with a MAT03, which it shouldn't. (MAT02s and MAT03s are sort of my standard blameless Ebers-Moll transistors, at least below ~20 uA I_C.)

Cheers

Phil Hobbs

As you always say, trying to get comfortable with switchers is good medicine. Eventually you will come into situations that I have been in a lot, where anything linear is totally prohibitive because of power dissipation and size.

The same goes for PWM drive of bigger loads, class-D, and so on. As long as one develops a thorough understanding of all the analog things going on inside those circuits it is much less voodoo than people think.

OK I tried this, massive fail! I then realized that it's a silly idea. Once I digitize the signal all the frequencies above Nyquist's are folded down, and there's nothing to be done.

George H.

The high end scope I use (R&S) has various modes for this. All of them maintain the top sample rate (20Gsps). The much maligned "high resolution" mode does signal averaging, greatly improving vertical resolution and reducing noise. There are peak detection modes that capture the highest and lowest values during a time interval. And various DSP modes that allow you to set the bandwidth, again reducing noise and increasing vertical resolution.