EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: bd139 on October 03, 2017, 04:00:39 pm
-
Trying to reverse engineer the vertical amp in an oscilloscope my father built years ago which doesn't work. I think he ripped off a Heathkit scope for the design. There's a 1nF capacitor across the emitters on the two differential transistors in a design. Does anyone know what this is for?
I've simplified it a bit to include the current sink and pair only with default DC operating conditions.
The capacitor in question is the one marked 1nF. I'm assuming it shorts as the frequency increases therefore decreasing the effective 22R resistor value to a minimum of half but how does that affect the amp? Is it a form of frequency compensation?
(https://i.imgur.com/41wlqtO.jpg)
-
How does the gain change for differential and common mode inputs with and without the capacitor across? (Barlett's bisection theorem might be handy)
-
That's a good question. Did some analysis and that was a helpful pointer. Thanks I am enlightened now :)
Gdiff at DC is ~7.5 ignoring Re. As the frequency rises the Gdiff rises to 15. I'll sweep a signal from my DDS and measure the gain empirically as I think the other stuff in the compensation network after might fudge the gain a little bit.
Thanks again for your assistance.
Scribblings:
(https://i.imgur.com/oC908a1.jpg)
-
The capacitor helps to compensate for high frequency gain rolloff. Very roughly, the voltage gain of a common emitter stage is a function of the impedance at the collector divided by the impedance at the emitter. Reducing the impedance at the emitter thus increases the voltage gain and using a capacitor favors higher frequencies.
Many books on high-bandwidth discrete amplifier design also covered this technique. In particular, Tektronix had some books on 'scope design; Circuit Concepts I believe. If you can find the one on vertical amplifiers, it's well worth reading. I believe it's on the 'net somewhere. I remember finding this series in the high school library in the 70s and as I was building my own 'scope it really taught me a lot, some of which I actually still remember.
Cheers,
-
To add a little more subtlety (and/or confusion >:D ), consider also that the transistor's gain drops at high frequency.
If 2N3904 is used, then fT = 300MHz and hFE = 200 (ballpark). The gain begins to roll off at only 300/200 = 1.5MHz.
This has the effect of:
1. Reducing the input impedance of the amplifier stage;
2. because an impedance which falls as frequency rises is capacitive, the input has capacitive reactance as well.
This is in addition to device capacitance, which tends to be dominant in a circuit like this. For Ccb = 4pF and Rc = 330 ohms, we expect a load-referred cutoff of 1 / (2*pi*330*4p) = 120MHz (which wouldn't be bad for a discrete amplifier, eh?). At a gain of 7.5, we also expect a Miller effect capacitance of (7.5 + 1) * 4pF = 34pF, plus Cbe (~8pF) for a total 42pF input capacitance (which is different from the aforementioned phenomenon, and adds to it).
So for the base-circuit frequency response to match the collector-circuit response, you need a drive impedance at most 1 / (2*pi*120M*42p) = 31 ohms! This is hard to drive! You certainly won't get this bandwidth if the driving source is a higher impedance, like another copy of this stage (Ro = 330 ohms), which would get only 11MHz -- a much more believable number for a hand built oscilloscope!
By reducing the emitter impedance (with the 1nF capacitor in question), you're kind of kicking it while it's down: input impedance is increased with emitter degeneration, and by removing it, the input impedance drops that much faster. You end up with less bandwidth at a generous gain threshold (like -20dB), but more gain within a practical threshold (say -1 or -3dB). In effect, you've pushed down that extra gain, that wasn't being very useful anyway because it was in the stop band already, and concentrated it at the band edge, where it can be useful. Or to put it another way: the roll-off slope has been rotated, so that it is steeper, and drops more sharply from the passband, which therefore breaks at a somewhat higher frequency.
And since you have capacitance and resistance, you might ask: well, can we cancel out any of that capacitance with an inductor (peaking coil)? And indeed, that can be done as well, netting a modest -- but significant -- improvement in bandwidth: about twofold.
The back-and-forth between impedance, node capacitance, and compensation options like emitter C and peaking L, is the basic design process of wideband amplifiers. :)
Tim
-
Thanks both for the detailed reply. Much appreciated.
I’ve found a copy of the Tektronix concepts series book on that on w140.com. Will read. Thanks for the reference.
I think this has a bandwidth of about 15MHz so that makes sense. That’s not all that bad for a home brew scope! The actual transistors are BSX20 which are quite fast and they’re being driven by a BC109 emitter follower and that is driven from a 2n4416 FET follower with an unmarked transistor (grr) current sink. Might be able to push a bit more out of it. It has two later stages, one more amp and a pair of unmarked (grr some more) drivers on the 160v supply. I remember vaguely about peaking coils in some Heathkit books years ago. Didn’t pay much attention :). Will go back and read them.
However the CRT anode isn’t driven hard (it’s a 3BP1) so I’m not expecting anything remarkable in the brightness department. It’ll probably fade out around 5Mhz. I remember seeing it working when I was about 8 years old last.
The actual scope had a power supply failure due to a cruddy design (Zener + pass transistor + dry joint = phut!). Blew a few things up in it. Been slowly repairing and reverse engineering it for about a year now. Got the horizontal working and the power supply up so far. The whole thing is a rats nest of veroboard, random wires and dirt.
He was a good design engineer but a shitty production engineer :)
-
When I built my 'scope in the mid 70s, as a student I couldn't easily get transistors with a BVceo of 300 V with any speed for the deflection amp output stage so I used vacuum tubes. I believe I was able to get 1 MHz vertical bandwidth out of it. It also used a 3" CRT that was (and still is) older than me. Unfortunately, the phosphor is fragile and I damaged a 0.1" area in the center of the screen during the first few tests.
After college, I started to rebuild it but by then commercial 50 - 100MHz 'scopes became affordable so I never went past building a chassis and mounting the CRT in it. I was recently looking for transistors to repair some hp and Tek equipment from the 70s and went thru the parts I had collected for the 'scope. Turns out the 70s and 80s were the golden age for some of the transistors needed and it's really hard to get them today.
-
Yeah completely agree. The plate driver transistors in this are unmarked TO92 packages with an aluminium plate glued to the flat half. I got one of them off the plate and there is no sign of them ever being marked. Being who he was, he probably got a transformer out and a bag of reject transistors and just kept the ones that didn't explode when you put 160v across Vce.
Either way it's a fun project learning with discrete transistor circuits :)