Here is a riveting topic for a new thread - fast low capacitance / high voltage diodes suitable for VAS anti-saturation clamping.
I need a >150V version of the 1N4148.
I'm pretty sure there is a 200V version out there but I can't for the life of me remember the part #.
Maybe someone can get this thread rolling by telling me what it is.
Cheers,
Glen
Onsemi's BAS21HT1 is what I'm using all along for such purposes. It's pretty good, although I'm told that some 1.5638MHz intermodulation products and a few hot headed recombining electron/holes pairs are clearly audible.
where's the 1.58Mhz junk coming from? the amp must be oscillating or picking up RF from a radio station
unclejed613 said:
It's probably coming from Sarcastia. Every audiophile knows that only Schottkys will do.
As to what to use for Baker clamps, I used to use the HER103, but I hear they've been discontinued.
Trr is slower than the 1N4148. the 4148 is 4ns whole the BAV21 is 50ns.. i've seen Baker clamps with 2 1N4148's in series (which reduces the capacitance to 2pf).
unclejed613 said:
Yeah, seems like you give up some on that in exchange for high-voltage capability. Same with the Rohm device that Ovidiu mentioned too.
Still, I'm really glad Glen started this thread, as I'm getting ready to order some parts. This thread made me look harder than I previously had. Mouser has the Rohm device for 5 cents each and the Fairchild device for 6 cents. Glad to see that these kinds of parts aren't in the dreaded "audiophile unobtanium" category 🙂.
andy_c said:
Don't be afraid to go for SMD parts. In particular if you etch your own PCBs, not having to drill hundreds of holes is a huge reward for the soldering effort.
Yes, there is a significant connection between breakdown voltage and Trr. It's possible to manufacture diodes with high breakdown voltage and small Trr, but it requires extra technological steps, hard to justify for a 5 cents device (out of which the chip should be 0.01 cents).
Two 1N4148 in series won't usually do; the reverse voltage spreads reverse proportional to the saturation current, which is not really well controlled in such devices, even if they are from the same tube.
I have lousy eyesight and am also somewhat of a klutz, so through-hole parts are a must for me 🙂. I have the boards made from one of those cheapo places that pool people's orders. They are about $20 each shipped in quantities of 5.
Damn.
It looks like any one of these suggested diodes will have to do - they all have pretty much identical specs - still not a match for the 1N4148's couple of pF junction capacitance, Trr of 4nS and low reverse leakage. 🙁
Cheers,
Glen
It looks like any one of these suggested diodes will have to do - they all have pretty much identical specs - still not a match for the 1N4148's couple of pF junction capacitance, Trr of 4nS and low reverse leakage. 🙁
Cheers,
Glen
so far, anyways.
buy now before they become scarce.
😉
and ditto, thanks glen for starting this thread. i am almost out of 1n4148 and wanted to know what other might be using for this purpose.
mlloyd1
buy now before they become scarce.
😉
andy_c said:
and ditto, thanks glen for starting this thread. i am almost out of 1n4148 and wanted to know what other might be using for this purpose.
mlloyd1
G.Kleinschmidt said:
Picky picky picky! 🙂 This thread turned out to be good news for me, as I found ones that were much better than what I had chosen a while back.
one company i worked for back in the 80's had the perfect answer for low leakage diodes, although the reverse voltage was only 40V. they used the B-C junctions of 2N2222 transistors as blocking diodes for the charging circuit of a capacitor "battery" that kept the memory settings in a stereo receiver. the leakage of 1N4148's was low enough that the cap would remain charged for about a week. the transistor junctions had such low leakage that the cap would remain charged for a month or more.
i suppose it would be fairly simple to find a 200V small signal transistor with a low Ccb. many of the detailed 1N4148 data sheets include a circuit for measuring Trr. if you have (or know somebody who has) a good pulse generator, you could test it for yourself (transistors aren't tested for the B-C Trr, since transistors are rarely used as diodes). generally transistors with higher voltage ratings have lower junction capacitances (not always) but since the chip area is larger than that of a diode, both the Trr and capacitance might tend to be higher.
i suppose it would be fairly simple to find a 200V small signal transistor with a low Ccb. many of the detailed 1N4148 data sheets include a circuit for measuring Trr. if you have (or know somebody who has) a good pulse generator, you could test it for yourself (transistors aren't tested for the B-C Trr, since transistors are rarely used as diodes). generally transistors with higher voltage ratings have lower junction capacitances (not always) but since the chip area is larger than that of a diode, both the Trr and capacitance might tend to be higher.
Well, I'm doing a SPICE model of the BAV21, and I'm running into a strange thing.
I grabbed a plot of the capacitance vs. reverse voltage from figure 8 of the datasheet and am fitting the data with CJO, M and VJ using the Excel solver. CJO comes out reasonable at 1.17 pF. VJ is reasonable, at 0.666. I would expect M to be in the range 0.3 to 0.5. But plugging any number in that range causes the simulated data to be pretty far off from the measured. Strangely, the best fit for M turns out to be a very low 0.0844. When using more typical M values from 0.3 to 0.5, no amount of messing with CJO and VJ gets the data anywhere near as close as it is with that weird 0.0844 value.
This has never happened to me before when fitting this kind of data. Ovidiu or anyone with knowledge of semiconductor processing, is this low number for M a consequence of some funky processing needed to make fast, high-voltage diodes? Maybe it's a measurement error? I'm stumped. I've attached the Excel spreadsheet, and you can see from the plot of measured and simulated capacitance that the sim just doesn't match the data when using common M values from 0.3 to 0.5.
I grabbed a plot of the capacitance vs. reverse voltage from figure 8 of the datasheet and am fitting the data with CJO, M and VJ using the Excel solver. CJO comes out reasonable at 1.17 pF. VJ is reasonable, at 0.666. I would expect M to be in the range 0.3 to 0.5. But plugging any number in that range causes the simulated data to be pretty far off from the measured. Strangely, the best fit for M turns out to be a very low 0.0844. When using more typical M values from 0.3 to 0.5, no amount of messing with CJO and VJ gets the data anywhere near as close as it is with that weird 0.0844 value.
This has never happened to me before when fitting this kind of data. Ovidiu or anyone with knowledge of semiconductor processing, is this low number for M a consequence of some funky processing needed to make fast, high-voltage diodes? Maybe it's a measurement error? I'm stumped. I've attached the Excel spreadsheet, and you can see from the plot of measured and simulated capacitance that the sim just doesn't match the data when using common M values from 0.3 to 0.5.
i don't think it's anything to worry about, since in a real circuit, Cj will be swamped by circuit board and lead capacitances, especially a Cj as low as what's in the data sheet.
andy_c said:
Andy,
It is to be expected high voltage diodes to have a smaller junction grading coefficient. Reason is the substrate that is lightly doped (so that the space charge region could extend and keep the electric field under the critical breakdown value). If the substrate is lightly doped, the junction is more abrupt (and "one side" asymmetric) and there's less reasons for any correction from the ideal junction C-V dependence. Don't know if M=0.084 is true for the BAV21 but certainly it's not absurd.
Here's though the BAV21 model coming with PSpice (which doesn't make it necessary correct). It still uses the default M, which to me is incorrect.
.model BAV21 D(Is=546.9n N=2.854 Rs=.1647 Ikf=.1486 Xti=3 Eg=1.11 Cjo=1.456p M=.3333 Vj=.5 Fc=.5 Bv=250 Ibv=100u Tt=72.13n)
Thanks. I'll keep it as-is for now. The M parameter rears its head again in the high-level injection equations, so I was concerned that a completely wrong M value might mess up the simulated currents in that region.
- Status
- Not open for further replies.