Thanks, I'll check it out. One nice feature of the PMBFJ620 is that all 6 pins are brought out, which is useful for cascoding, source degeneration, etc. It is also symmetrical wrt to orientation, so it can be mounted in either direction. D-S are interchangeable, so PCB routing is simplified. The package is however an SOT363 with the low 190 mW Ptot limits, limiting the tail current in the LTP to a few mA with safe margins.
A dual BF862 in SOT363 would have been more useful as the low-noise, high-gain lower cascode device, and the Ptot would not have been a limiting factor when cascoded.
From the datasheet one notes that the principal heat paths are through the gate leads, like the 862. Although it's rather daunting, one can heatsink those leads by soldering wires which then conduct to something else nearby.
I got diverted to some other activities before completing all of the intended task, but not long ago paralleled a bunch of 862s by stacking them and soldering the gate leads to a piece of bus wire. This was a combined strategy for heat dissipation and lowest parasitic inductance. Preliminary experiments indicated that the resulting composite required much less than the typical snubbing impedance in gate leads for stability against VHF oscillation.
SIM using published specs -
Besides the issue of designing with published data, which is often suspect, the data is not kept up to date with changes in processing equipment, impurity improvments etc. The 2N5564 was first spec'ed as 50nv noise, but when I tested it for a preamp, decades ago, it was 5nv so I used it. It took many years for that published noise spec to get updated. Often the data from the original developer just keeps getting passed forward by all sources of the part. Decades old designs cannot be assumed to still measure the same today in all parameters. This only makes SIM work more questionable. Good, accurate models are required. But, a curve tracer, noise tester et al are available to get real numbers and spreads. Which is why I invested in SIM AND test equipment hardware. -RNMarsh
Besides the issue of designing with published data, which is often suspect, the data is not kept up to date with changes in processing equipment, impurity improvments etc. The 2N5564 was first spec'ed as 50nv noise, but when I tested it for a preamp, decades ago, it was 5nv so I used it. It took many years for that published noise spec to get updated. Often the data from the original developer just keeps getting passed forward by all sources of the part. Decades old designs cannot be assumed to still measure the same today in all parameters. This only makes SIM work more questionable. Good, accurate models are required. But, a curve tracer, noise tester et al are available to get real numbers and spreads. Which is why I invested in SIM AND test equipment hardware. -RNMarsh
Last edited:
A noise tester is a device that tests noise in specific devices, usually semiconductors, but resistors can be measured as well. A well known noise tester is the QuanTech, unfortunately it is not made anymore. However, I have one, Dick Marsh used to have one, and Audio1 has several. You might ask Audio1 for advice in this area.
Google Quan-Tech, noise. I guess the company no longer exists but their products are around used and for rental.
I showed several ways in my Linear Audio article to do it with a sound card and any FFT software. Below is data from NXP's site. The numbers at 100kHz (no 1/f) don't jive with the gm unless there are large extrinsic resistances.
kHz nV/vHz
0.01 11.06
0.120 8.82
1 5.95
10 2.84
100 2.06
The so called SPICE models are garbage from some MOSFET switch, what a mess.
Scott,
In your latest schematic post for resistors #21 and #22 is the reason you have two parallel resistors to get a 5 ohm value itself or is it for more power dissipation at this point? Am I to assume that these are two 1/4 watt resistors in parallel or is there some other reason to do that there? Just another dweeb question for you to answer.
Steven
In your latest schematic post for resistors #21 and #22 is the reason you have two parallel resistors to get a 5 ohm value itself or is it for more power dissipation at this point? Am I to assume that these are two 1/4 watt resistors in parallel or is there some other reason to do that there? Just another dweeb question for you to answer.
Steven
The PMBFJ620, I have 10 from Dimitri and will measure them and make a crude model.
I went to the parts bin and grabbed the closest to what I wanted. The doubling is not really necessary. The last schematic was just exemplary of exactly what I built, one need not get too picky about exact values since some folks might use 5% carbon film because that's what they have and others might go all 1%. Note that the two 226 Ohm resistors in parallel is deliberate in order to be exactly half of one. At 5% 240 or 270 would be fine here as would 249 at 1%.
A good point though is to be aware of the low power capacity of SMT and beef up where needed the output is a good place.
I had 25 of them from digikey since they are duals but they will probably
have a long shelf live here since I had to parallel even more of them than
in the BF862 case for acceptable voltage noise.
Correction: I only have 24 of them. I killed a fet with pliers, just to
watch its die.
Seems to be dual chip as was to be expected since they do not
share the gates, which are usually connected to substrate.
That makes them probably not much better in thermal
behaviour than 2 bf862 under a epoxy drop.
Gerhard
Last edited:
What's acceptible? If you need to parallel a lot of BF862's you are going sub-nano volt which is not particularly necessary on a line amp and as an input stage you now have to deal with the huge transconductance. From the data one pair of them equals/outdoes an OPA627 input pair.
It's too bad none of these seem to use a modern DI process where all three connections are isolated. As a side benefit the DI FET's can have negligible impact ionization. It was entertained here but you can't make a viable business argument.
I did better pulling 2SK170's out of the bag than the datasheet spec on the 2N5564.
Last edited:
Something about Toshiba's parts shows it can be done. Dimitri remarked that some out of bag singles were very very consistent. I also found few significant outliers in a large lot of 2SK364GR (same die as 2SK170).
We need to befriend a billionaire philanthropist
and persuade him to fund some worthwhile, if nonprofitable, parts development.
