T
Tim Williams
- Jan 1, 1970
- 0
I decided to break out the breadboard and build a circuit reminiscient of
my Heathkit's deflection amp. This is, in order: a bipolar diff amp with
CCS tail and resistive collector load; a PNP gain stage (with emitter
degeneration), then the output, which in this case is cascode. To save
time and hassle, I built exactly one half, which means instead of a diff
pair, I grounded the input's emitter and biased its base with the function
generator and a resistor. Fair enough. I used 2N3904's for the NPNs and a
2N4403 for the PNP, being that I lack 3906's. Supply is +/- 9V, and bias
for the first stage around 5mA, second 15mA and cascode, 25mA or so.
Cascode is supplied from +150V through 2.6 kohms collector load.
So with this setup I went about testing some transistors in my parts box.
I know I have some of those chroma/video amplifier transistor guys in my
parts bin, several of them housemarked.
I found that the TO-202 things (Toshiba, pinout EBC, suprisingly) work
nicely, getting up to 2MHz or so off the bat (no emitter compensation
anywhere). That's about what the 2SC1569's do as well. I tried switching
transistors (like from switching supplies), which sucked (0.5-1MHz), I
tried other transistors, I even tried one that started oscillating at low
bias current. (In other words, high Vce --> avalanche. When biased "off",
touching the collector with a finger caused the amplitude to increase and
the falling edge to sharpen. I may have to see what fall time this
transistor produces, given a more substantial capacitance to work on. Heh,
and maybe see if the C-B junction snaps ala Larkin. ;-) )
I even tried a MOSFET, which was interesting: more voltage gain, somewhat
different bias voltage, but generally slower, probably in no small part
because I grabbed a 600V, 1 ohm device.
I tried a TO-126 transistor, 2SD986 I think it is. This was among the best
performers. I looked up the datasheet only to realize that it's a fucking
darlington! How can I get better performance from a darlington!? That ain
't right...
Finally, I explored my stash of faster transistors. I worked my way
through some TO-126 types, one rated as 400MHz fT. Pretty nice. Vcbo is
right on the bleeding edge though, like 180V. Still, I went for the last
stretch: BFQ225, a 1GHz, purportedly 100Vcbo unit! And, lo and behold,
there it worked! With emitter capacitors added (and various and sundry
ferrite beads and bypass caps to keep the breadboard from oscillating in
the hundreds of MHz..), I got out to about 20MHz -6dB.
This exploration smacks of Larkin... damn the ratings, just hook it up and
see if it burns!
I've got to wonder if there's possible damage from running over ratings
though. These aren't tubes...
On the theoretical side of things, I have a question: fT is defined for
current, correct? (h_fe = 1, an AC current parameter.) So, then the
maximum voltage gain (with respect to frequency) essentially depends on Ccb
and h_fe, right? And of course, you can get more if you restrain Vce
(cascode or emitter coupling), but that doesn't do any good when I want
absurdly high Vout here. So, how close to fT can you get while still
swinging a lot of collector voltage? What kind of voltage gain can be
generated?
How do they, like, Tek themselves for instance, do it in scopes? Well,
when they made good analog stuff anyway... The, what, 2465 is it, does 4
channels out to 500MHz or something like that, no sampling of any sort
needed. Now, I know part of that is tube design, but they still need to
make voltage. How much, how, and with what?
To get more than 20MHz at more than 20Vp-p I'm starting to have delusions
of vacuum tubes, but then, I keep trying to reassure myself that they have
the same problems (like Cgp), and that I already tried carefully to get the
best rise and fall times in that class D tube circuit...
Tim
my Heathkit's deflection amp. This is, in order: a bipolar diff amp with
CCS tail and resistive collector load; a PNP gain stage (with emitter
degeneration), then the output, which in this case is cascode. To save
time and hassle, I built exactly one half, which means instead of a diff
pair, I grounded the input's emitter and biased its base with the function
generator and a resistor. Fair enough. I used 2N3904's for the NPNs and a
2N4403 for the PNP, being that I lack 3906's. Supply is +/- 9V, and bias
for the first stage around 5mA, second 15mA and cascode, 25mA or so.
Cascode is supplied from +150V through 2.6 kohms collector load.
So with this setup I went about testing some transistors in my parts box.
I know I have some of those chroma/video amplifier transistor guys in my
parts bin, several of them housemarked.
I found that the TO-202 things (Toshiba, pinout EBC, suprisingly) work
nicely, getting up to 2MHz or so off the bat (no emitter compensation
anywhere). That's about what the 2SC1569's do as well. I tried switching
transistors (like from switching supplies), which sucked (0.5-1MHz), I
tried other transistors, I even tried one that started oscillating at low
bias current. (In other words, high Vce --> avalanche. When biased "off",
touching the collector with a finger caused the amplitude to increase and
the falling edge to sharpen. I may have to see what fall time this
transistor produces, given a more substantial capacitance to work on. Heh,
and maybe see if the C-B junction snaps ala Larkin. ;-) )
I even tried a MOSFET, which was interesting: more voltage gain, somewhat
different bias voltage, but generally slower, probably in no small part
because I grabbed a 600V, 1 ohm device.
I tried a TO-126 transistor, 2SD986 I think it is. This was among the best
performers. I looked up the datasheet only to realize that it's a fucking
darlington! How can I get better performance from a darlington!? That ain
't right...
Finally, I explored my stash of faster transistors. I worked my way
through some TO-126 types, one rated as 400MHz fT. Pretty nice. Vcbo is
right on the bleeding edge though, like 180V. Still, I went for the last
stretch: BFQ225, a 1GHz, purportedly 100Vcbo unit! And, lo and behold,
there it worked! With emitter capacitors added (and various and sundry
ferrite beads and bypass caps to keep the breadboard from oscillating in
the hundreds of MHz..), I got out to about 20MHz -6dB.
This exploration smacks of Larkin... damn the ratings, just hook it up and
see if it burns!
I've got to wonder if there's possible damage from running over ratings
though. These aren't tubes...
On the theoretical side of things, I have a question: fT is defined for
current, correct? (h_fe = 1, an AC current parameter.) So, then the
maximum voltage gain (with respect to frequency) essentially depends on Ccb
and h_fe, right? And of course, you can get more if you restrain Vce
(cascode or emitter coupling), but that doesn't do any good when I want
absurdly high Vout here. So, how close to fT can you get while still
swinging a lot of collector voltage? What kind of voltage gain can be
generated?
How do they, like, Tek themselves for instance, do it in scopes? Well,
when they made good analog stuff anyway... The, what, 2465 is it, does 4
channels out to 500MHz or something like that, no sampling of any sort
needed. Now, I know part of that is tube design, but they still need to
make voltage. How much, how, and with what?
To get more than 20MHz at more than 20Vp-p I'm starting to have delusions
of vacuum tubes, but then, I keep trying to reassure myself that they have
the same problems (like Cgp), and that I already tried carefully to get the
best rise and fall times in that class D tube circuit...
Tim