Everyone gets confused. That's the max open circuit voltage OUT of the PV array, these are speced as MOSFET drivers so their VT would have to be less than that.
In this application they look like a current to current converter of 1/1000, 1mA of LED current makes 1ua of PV current. Into 1 Meg this makes a Volt, you are free to flip the polarity because the array is galvanicly isolated. The graph is on the data sheet, there is a max speced by heat in the LED. I'll post a complete circuit today, got too late last night.
That's great. They are sold as switches so yes I got confused. I thought they were on off devices.
Thanks,
They stack fine but 8V would pinch-off almost any FET I use. I made my own 65V versions on a project chip.
339A stock Osc ckt -
In case anyone is wondering what happened to ideas about paralleling and Rsource trimming --- I did the parallel using same jFET from another spare unit... used it backwards and forwards. The Rsource trim does find a null but it isnt any deeper than the stock 339A osc with a single jFET. HP had found the sweet-spot with this method. So onward and upward with SW's suggestion. For those following along, I've attached the osc section of the schematic. Thx-RNMarsh
In case anyone is wondering what happened to ideas about paralleling and Rsource trimming --- I did the parallel using same jFET from another spare unit... used it backwards and forwards. The Rsource trim does find a null but it isnt any deeper than the stock 339A osc with a single jFET. HP had found the sweet-spot with this method. So onward and upward with SW's suggestion. For those following along, I've attached the osc section of the schematic. Thx-RNMarsh
Attachments
Last edited:
Hi Demian,
The guys using these for output protection replacing relays with solid state relays say 100ms to turn off a mosfet. A Jfet has a much lower capacitance.
The PV's can be paralleled for speed up. The manufactures of these PV's make them for SMP applications.
I don't think this should be a problem at least for the 339a osc. Consider the 60uF cap in series with the 2k resistor on the jfet feedback leg.
The circuit Scott proposed replaces the feedback arrangement.
We will see.
Cheers,
Try it at 10KHz. Found a much better null there than at 1KHz.
I was getting 100dB-110dB on the 2nd at 10KHz and now I get -120dB or there about.
The amplitude control has to be trimmed as well as the Rsource trim.
The two trims have to be worked alternately to get the best null. If you think you have best try going a little further with the trims. Some times the 2nd rises and fall again much deeper.
I have tried it with and without paralleling a few times.
Adding a degeneration R to source, and yes I did cut the trace, had an opposite effect to what was proposed.
I do think we have hit bottom with oscillator. I'm experimenting with larger caps in the AGC filtering and integrator. Too big a cap here kills the oscillator. My target is THD20+N -120db.
A THSH might do better for the AGC. Jfets seem to enjoy a really fast AGC.
Get the ripple down.
Cheers,
Last edited:
As a point of reference for this pro-ject: The "Victor" 1KHz fixed frequency oscillator measures in at .000125% THD+N. That's about -118-119dB
Compare that with my forum #1215 test data on my modded 339A. Thx-RNMarsh
[I finally did the output buffer change -- LME49710 -- only got 2-3dB better. But it adds up once the big drops in THD are done.]
Compare that with my forum #1215 test data on my modded 339A. Thx-RNMarsh
[I finally did the output buffer change -- LME49710 -- only got 2-3dB better. But it adds up once the big drops in THD are done.]
Last edited:
Thanks Rick.
I have some new trick up my sleeve that dramatically reduces the distortion of the oscillator.
But I need to get a handle on all of it before sharing the magic.
Stay tuned.
Cheers,
Here's a quick summary, I found that series connected JFET's with gates all connected together can dramatically improve THD. I didn't show this but you get the idea. The two 100k's could include a pot to trim seconds. The second plot shows a 100 Ohm sense resistor for the LED and 100K resistor on the PV array this gives 1000 to 1 or 1 to 1 on the output. The speed looks like it should not be a problem.
Attachments
@Richard, David and any others interested: Here are two plots I made of my basically unmodified 339A oscillator output (orig. opaqmps, etc, but I added a trim to the feedback resistor on the JFET to help with 2nd H -- this tirm hasn't been touched in almost two years).
The oscillator was at full output of 6.3VRMS at 1kHz, feeding a 30k wirewound pot to adjust level, then through the Active Twin-T filter to drop the fundamental by 40dB and on to the ADCs -- so the oscillator was running essentially unloaded. I've found that the distortion goes up by about 2-3dB when I use the internal level pot and switched attenuator to set the output to 1VRMS.
One plot is the EMU 0204 ADC and ARTA software, the other is with the QA400 and b0.97 software. The plots look very similar, and show the very low distortion of the HP design -- and both have about the same THD calculated manually, although the SW numbers differ. I offer this as a reference for comparison to any improvements to the oscillator -- hope they help -- note that ARTA calculates the THD+N about 20dB higher than the THD alone.
The oscillator was at full output of 6.3VRMS at 1kHz, feeding a 30k wirewound pot to adjust level, then through the Active Twin-T filter to drop the fundamental by 40dB and on to the ADCs -- so the oscillator was running essentially unloaded. I've found that the distortion goes up by about 2-3dB when I use the internal level pot and switched attenuator to set the output to 1VRMS.
One plot is the EMU 0204 ADC and ARTA software, the other is with the QA400 and b0.97 software. The plots look very similar, and show the very low distortion of the HP design -- and both have about the same THD calculated manually, although the SW numbers differ. I offer this as a reference for comparison to any improvements to the oscillator -- hope they help -- note that ARTA calculates the THD+N about 20dB higher than the THD alone.
Attachments
@Richard -- The EMU drivers work for me in XP SP3; I think there is a driver package that runs in Win7, though Creative may call it a beta. ARTA then runs the EMU from its device setup menu. Like David, I'm beta testing the QA400. The EMU and QA both having rising noise floors (due to noise-shaping used) above 30kHz, so their dynamic range goes away fast. BTW, I have an HP 239 which is very similar, and it is equally good. But if you're going to be driving low-Z loads, then the LT1468 seems to be the real deal.
Have you pushed the limit of FFT size and averaging or computed the noise floor in volts/root-Hz to see how much you can get. You should be able to compute the expected value with SPICE. The same issues apply voltage noise, noise gain, current noise, and the impedance of the gain network.