Transcript ppt

Amplifier and Feedback Circuit Design
Philip Burrows
Queen Mary, University of London
On behalf of:
Colin Perry
Oxford University
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT1 (i)
needed amplifier to drive kicker at NLCTA
had to be fast relative to 170ns bunch train
wanted as much drive as could readily get
had to be done quickly and easily
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT1 (ii)
chose planar triode vacuum tubes as easiest way (type Y690A)
single output tube @ 5kV gave +/-8A into 50 ohms
two more tubes (ac coupled) to get enough gain
fast op-amp (THS3001) added later for more gain
tubes turned on for ~10us by IGBTs in cathode circuit
18A from 5kV when on = 90kW
but mean HV power at 10Hz = 10W
IT WORKED: risetime ~10ns, delay ~16ns
BUT: large (400 x 300 x 120mm)
poor linearity and pulse shape
inadequately engineered and probably unreliable
…and no-one keen on working on unit with 5kV supply and big
capacitors if it could be avoided.
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT2 (i)
repeat of FONT1 at NLCTA
aims:
modest increase in overall speed
better quality data
similar amplifier output and speed sufficient
requirements: better quality and usability
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT2 (i)
able to go solid-state because of:
the HV RF MOSFETs from Directed Energy Inc (now part of Ixys)
similar to usual switching MOSFETs
optimized for RF
RF type low-inductance package
chosen device: 500V max, usable to 25A
design of SLAC kicker
strips connected internally at one end
connect with very short (120mm) cables
drive with current source into an inductive load (200nH total)
peak drive voltage reduced to +/-100V for +/-8A
cf: +/-400V for +/-8A into 50 ohms
NB design of kicker gives only magnetic deflection, so reduced voltage
does not reduce sensitivity
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT2 (iii)
output stage:
differential (push-pull) pair of MOSFETs
a little source degeneration
centre tapped choke feeding 200V to drains
differential output drive to kicker
driver stage:
effective input C of FETs ~500pF
low Z (low R and low L) driver needed
10 high speed op-amps as buffers parallelled to
drive each gate
2 op-amps as phase-splitting gain stage to drive
buffers
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT2 (iv)
turning on output stage:
when ‘on’, draws 25A @ 200V = 5kW
FETs can’t take this continuously
so as in FONT1, turn on for ~10us
1W DC/DC converter enough for 200V supply
no heatsink needed for FETs
FETs normally biassed ‘off’ by drivers
turn on by 10us pulse into both sides of phase splitter
pulse takes gates of both FETs positive turning them on
possible because the op-amps can give 15V swing
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT2 (v)
IT WORKED:
current and timing much as FONT1 (+/-8A, 12ns risetime,
16ns delay)
good pulse shape, stable, easy to use
small (160 x 100 x 50mm)
2 were used in FONT2, into 2 kickers for increased deflection
PROBLEMS:
some instabilities at 200 to 300MHz
voltage gain very high when load impedance became
resonant
cured by adding RC stabilizing networks
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT3 (i)
a similar feedback experiment at ATF
much shorter bunch train meant reducing all delays
target for delay through kicker amplifier: 6ns
output drive sacrificed to do this with reasonable effort
design based on FONT2
same MOSFETs
differential drive to kicker
similar op-amp based driver
modified to increase speed and for different kicker
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT3 (ii)
kicker now driven more conventionally, matched to its cables
at small gaps (needed for sensitivity) ~50 ohms differential
matches a pair of 50 ohm coax in parallel to each side (except for
connector and feedthrough section)
kicker terminated in two pairs of 50 ohm cables, 10m long
200V to output is fed from far end (eliminating tapped inductor)
MOSFET stage has 25 ohm load resistors back-terminating the cables
these take half output current, leading to +/-4A to kicker
greatly improve stability
slightly improve delay
and anyway, FETs couldn’t safely stand voltage of full current swing into
kicker (+/-8A into 25 ohms = +/-200V)
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT3 (iii)
driver stage changed in detail:
now 16 OPA693 gain of 2 buffers to drive each gate
faster (700MHz vs 300MHz) & smaller (sot23-6 vs so-8)
OPA695 in phase splitter: also faster
both types have a ‘power-down’ pin
turned on along with output stage
simplifies power supply & avoids need for heatsink
turning on output stage:
changed because op-amps limited to +/-6V supplies
not enough swing to turn off FETs completely
drive capacitively coupled to gate
turn-on bias pulse fed to gate through resistor and inductor
output is turned on for 5us: conditions stable after 3us
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT3 (iv)
IT WORKS (or seems to, so far…):
small size: pcb 90 x 100mm (with front end 160 x 100 x 25mm)
good pulse shape
risetime 6.5ns, delay 6ns
A Note on Speed Issues
figures are with HF compensation in driver and output to boost speed
the FETs have a bit more source inductance than suggested by datasheet
this is the worst factor in limiting speed
a higher drive voltage would have been a better choice
a good solution would have used a pair of low voltage RF FETs as driver
a cascode configuration looked good and was considered: rejected as
needing too much development time
planar triodes still probably offer the highest performance, but would take
a much greater engineering effort
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
Front-end Feedback Circuits (i)
basic requirements:
additional gain for the low level BPM output
remotely controlled gain in the main loop
remotely controlled gain in the delay loop
sum the main and delay loop signals
disable the delay loop until just before the bunch train
remotely controlled delay round the delay loop is needed
FONT has always used a simple relay box switching cable
delays
will not be discussed further
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
Front-end Feedback Circuits (ii)
FONT1 and FONT2:
front end design was easy, based on AD835 multiplier chip:
a nice variable gain stage
includes summing function
250MHz bandwidth
switching the gain control input to zero disables the loop
FONT3:
much harder, to keep delay low
AD835 too slow (~3ns)
slightly exotic amplifiers needed
as built: total delay ~1ns
it is combined with the power amplifier board in a unit 160 x 100 x 25mm
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
Front-end Feedback Circuits (iii)
for FONT3 we have:
variable gain: pin diode attenuators
current controlled passive devices, delay ~0.1ns
gain stages: THS4302/THS4303 amplifiers
op-amps with internal feedback resistors
voltage gains x5/x10, closed loop bandwidth 2.5/1.8GHz
SiGe parts in QFN16 package
2 x10 stages in main loop path, 2 x5 in delay loop
loop disable: ADG901 CMOS RF switch
switches in ~3ns
bandwidth to ~3GHz
switching transient on output <0.5% peak signal level
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
Front-end Feedback Circuits (iv)
and also in FONT3:
variable HF boost in main loop path (with variable capacitance diode)
compensates for main output stage response
variable HF roll-off in delay loop path (with variable capacitance
diode)
matches response of delay loop to main loop via power amp
variable LF boost in main loop path (preset by solder links)
to correct for droop in pulse response
low pass filter on input
final part of BPM processing
provision for adding output from an AWG to input signal
to allow a non-zero position reference profile to be defined
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
Amplifier + Feedback Board
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05
FONT3 kicker system in total
front end and kicker amplifier (as described)
variable delay box (15ns range, in 1ns steps)
HV supply
12V to 200V DC/DC converter + 35uF plastic film output capacitors
support box (close to amplifer)
generate supply voltages (from 24V in)
timing to turn on amplifier and loop (from 2 NIM inputs)
test pulse and monitor outputs
interface to other control lines
control box (located remotely)
manual control through pots and switches, and supplies power
(was to be uC unit at the amp: dropped because of risk to schedule)
It all (with 50m control cable, 2nd amplifier unit, spares, tools, and
documentation) went in a suitcase without an excess baggage charge
Philip Burrows
Third Mini Workshop on Nano Project, KEK 30/5/05