+-
to
m :ohms
ohms
output.
Phil, I guess I don't understand the problem. Are you saying that an ordinary resistor or attenuator at room temp that will reduce the gain by the required amount adds too much noise?
If not, then why not use a digitally controlled RF attenuator with FET switches that work in RF front ends at 50 (or 75) Ohms. If this adds too much noise, then I don't see how ANYTHING can do what you need unless it is cooled below room temp??
Mark
- o s s
tput.
That's an interesting part Bill. Have you (or anyone) ever tried it for AGC in an oscillator circuit?
George H.
Is the DAC going to be trimmed once in a while, or is it going to be trimmed constantly as temperature changes, or is it going to scurry around to follow some signal level?
John
in
There are various methods for that, but it has to be very quick. The key is to correct for temperature shifts caused by large signals, which can happen fast enough to cause noticeable distortion, and that means kilohertz control bandwidths. That's pretty hard to do in a thermal control loop, so I'm hoping to use sort of a feedforward scheme, measuring the temperatures and dorking the gain slightly to make everything work as the naive user expects.
The motto for this line of gizmos is "Be Done, Now." Ideally, that means that even if you bang on it pretty hard, it stays fast-quiet-linear-boringly-predictable, as thought the actual circuit problems didn't exist.
Target is a worst-case error of 0.1% of the photocurrent. That way, you never have to worry about it--you can just get on with your measurement. We'll see how close we can get.
The thought has crossed my mind a number of times. We just have to get Nico to pony up with that beer he owes us. ;)
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs (atsign) electrooptical (period) net http://electrooptical.net
No, if I could get a 100-ohm digital pot with good performance, I'd be done. The 1k ohm ones have about 60 ohms worth of really drifty wiper resistance, basically due to putting 8 2->1 CMOS muxes in series, so they don't get there, at least not with decent resolution.
It has to work at baseband with really low 1/f noise.
As my Dad used to say when he gave me a job to do, "If it were easy, I'd have done it myself."
I may resuscitate an old idea and use a spare transistor on the die as a heater and temperature sensor, but I'd really rather not if I can avoid it.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs (atsign) electrooptical (period) net http://electrooptical.net
Dissipation due to large signal excursions can cause pretty significant distortion, I think. The die is roughly 43 mils square by 600 um thick, which makes 0.0009 cc, or just about exactly 2 mg.
At 700 J/kg/K, it takes about 1.4 mJ to make a 1 K excursion, so if the dissipation changes by a few milliwatts, we can get several-percent changes in collector current in well under a second.
So I'd want to have kilohertz sorts of control bandwidths if possible.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs (atsign) electrooptical (period) net http://electrooptical.net
+-
to
m :hms
hms
output.
No. When I worked at Cambridge Instruments I worked on an old machine with a tunable input bandwidth, when the adjustable resistances were CdS opto-resistors, adjusted by an adjacent LED, and that worked - though production had to rely on experienced technicians to set the thing up in the first place, by bending the leads on the LEDs.
It wasn't something that we re-engineered into the newer machines - bandwidth adjustment in the later models tended to be done by using a small MOSFET to connected one of a capacitor to ground (or not).
-- Bill Sloman, Nijmegen
Phil Hobbs a écrit :
Make it dual path. Have your series 5R resistor for low noise, then sum a say up to +/-20 gain controled path with say another 100R resistor. That way the "gain control path" noise and distortion is attenuated by a
26dB factor and won't annoy any more.-- Thanks, Fred.
Ouch. The thermal transient response of the transistor will depend on more than just the silicon die. There's the mounting paddle and the epoxy, too. It will be messy.
That's as intelligent a comment I can make before I've had my big Saturday bowl of latte.
John
Have you looked at digitally controlled and voltage controlled RF attenuators like these
You might be able to wire one up in an unusual way to do what you need (i.e. not use it directly as a Zo = 50 Ohm attenuator) Mark
Sure, as in the OTAs we've been discussing. It turns out that finding a quiet enough one is hard...the LT1228 has way too much output noise.
I suppose I might be able to find a nice quiet CFA that'll give enough gain at the required bandwidth...but with +-8% range, we get at least
1/12 of the noise of the amplifier and its resistors.I'm looking at ways of doing three-way thermal balancing on-chip. The cutest one is like this:
+15V 0-----*------*-----*-----* | | | | | | | | --- | | | LED \ / R 200 R 200 R 200 V R R R --- R R R | | | | (5 mA each) | | | | | / / / | K K K *----|-----|-----| 100 pF | |\ |\ |\ | | | | *---CCCC-----* | | | | 1M | | R | | | | |\ | R | | *---RRRR--*----|-\ | R | | | | \ | R | | | | >---*--* | | | | | / | | |/ |/ |/ +6 0------|+/ | | | | | |/ | *----|-----|-----| | | |> |> |> | | | | | 750 | | | | | | | *-----*-----*--------RRRR-------------* GND \ / `-----y-------'One NPN from each array
The op amp forces the dissipation in the right-hand array transistor to be 5 mA * 6.6V = 33 mW. The other two can dissipate anywhere from almost nothing to about 5 mA * 14.6V = 73 mW.
When it's in range, the two left-hand sections adjust their power dissipation to keep their V_BEs exactly the same as the RH transistor's. For ideal transistors, with zero or matched R_EE', equal emitter areas, and infinite Early voltage, this forces them to all be at the same temperature. It's a cute form of integrating feedback loop--any nonzero error causes the collector voltage to ramp until it is exactly in balance.
The R_EE' and emitter areas can be measured and then taken out by fixed offsets on the bases, but the Early effect sets a limit on the DC loop gain, unfortunately, so it will need an auxiliary slow control loop to bring the three close to balance. Still, this approach is about as quick as it gets.
Another approach is to do it all digitally, by alternately measuring the V_BE of each transistor and driving it to whatever dissipation is required.
That has the disadvantage of putting both thermal and electrical transients right where I don't want them, so I'm kind of leaning towards bandaging the all-analog approach above, by putting in an auxiliary heater.
If I could find a good gain control method, I'd still prefer that. Maybe a combination of a slow external heater loop plus a reduced-range DPOT gain control....
Instrument design sure is fun.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs (atsign) electrooptical (period) net http://electrooptical.net
That's an interesting idea too. I didn't think of Hittite because of JL's experience with some of their 'dc coupled' amplifiers that turned out not to work properly below 100 MHz, but maybe these switched things would be better.
I could do something similar with a really low resistance R-2R network and some 2N7002s, as someone already suggested. Linearity might be an issue, I'm not sure.
I'm trying to get 0.1% worst case nonlinearity, because that's the level at which people stop worrying about it. "Be Done, Now" is the idea.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs (atsign) electrooptical (period) net http://electrooptical.net
+-
... has to go between a
If it's a low voltage signal, can you use a jFET in the voltage- controlled-resistance region to shunt to ground?
I missed this the first time round. That's an interesting idea--I'll have to do the math to see if I can get enough range.
Thanks
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs (atsign) electrooptical (period) net http://electrooptical.net
It's a possibility, but it's hard to know what the JFET is actually doing at any given moment. Something that allows verification would be much more comforting.
Thanks
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs (atsign) electrooptical (period) net http://electrooptical.net
ut +-
as to
ohm
so:
ohms
ohms
ng
e output.
A colleague will be doing an adjustable gain thing. He'll be interested. I've seen the CdS-led opto things, a bit messy.
George H.
+-
to
hms
hms
utput.
.net
t- Hide quoted text -
I don't know about digital pots. You say there's always 80 ohms in the wiper? (At least that explains why there are no 100 ohm ones.)
With two pots, there's all sorts of different ways to hook them up. You might leave one as you first posted and add the second as a rheostat to the wiper from one side or the other.
Or maybe the rheostat in the upper arm of the potentiometer?
Now if you had three pots.....
George H.
Hide quoted text -
You're still putting the switches in series; wiper resistance adds.
Same.
What I've done is to use analog switches and resistors in parallel. "Wiper resistance" in the 1ohm range can be had and higher power dissipations are possible (my real need). Monotonicity can be an issue but that wasn't a serious problem in my application.
Except that one of ^-----^ these paths (or both) will saturate
...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
Remember: Once you go over the hill, you pick up speed
ut +-
as to
ohm
so:
0 ohms 0 ohmsng
e output.
cal.net
. l.net-Hide quoted text -
I'm not sure exactly what Phil's trying to do. But I was picturing one pot as a coarse adustment. And then the rheostat as fine adjustment on the 'fat' leg of the pot. Then the wiper resistance is not going to matter, it's always in series with several hundred ohms. (Oh, he'd have to tweak his 10 ohm resistors to keep the fat leg on one side all the time.)
George H.
iper
e aElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.