fast track/hold

[Ban]

For my digital amplifier project I need a very fast(<50ns acquisition)
Track/Hold amplifier. The accuracy should be better 0.1%. I could not find
anything that is still in production and affordable.

--
ciao Ban
Bordighera, Italy

 

[John Larkin]

On Wed, 31 Mar 2004 06:25:10 GMT, "Ban" <bansuri@web.de> wrote:

For my digital amplifier project I need a very fast(<50ns acquisition)
Track/Hold amplifier. The accuracy should be better 0.1%. I could not find
anything that is still in production and affordable.

All the really good track/hold circuits are buried inside the front
ends of A/D chips.

John

 

[Ban]

John Larkin wrote:

On Wed, 31 Mar 2004 06:25:10 GMT, "Ban" <bansuri@web.de> wrote:

For my digital amplifier project I need a very fast(<50ns
acquisition) Track/Hold amplifier. The accuracy should be better
0.1%. I could not find anything that is still in production and
affordable.

All the really good track/hold circuits are buried inside the front
ends of A/D chips.

John

That is my feeling as well, because many T/H chips like the BurrBrown SHC605
are obsolete now, but any high-speed A/D-converter has some S/H built in,
that matches its performance. Unfortunately I want to make an analog-working
sigma-delta front-end for a pulse-width modulated digital amp and do not
want a quantisation. I could however use a fast 10 or more bit A/D and drive
a D/A with it. But this should really be an ultimate possibility.

--
ciao Ban
Bordighera, Italy

 

[Tim Wescott]

Ban wrote:

John Larkin wrote:

On Wed, 31 Mar 2004 06:25:10 GMT, "Ban" <bansuri@web.de> wrote:


For my digital amplifier project I need a very fast(<50ns
acquisition) Track/Hold amplifier. The accuracy should be better
0.1%. I could not find anything that is still in production and
affordable.

All the really good track/hold circuits are buried inside the front
ends of A/D chips.

John


That is my feeling as well, because many T/H chips like the BurrBrown SHC605
are obsolete now, but any high-speed A/D-converter has some S/H built in,
that matches its performance. Unfortunately I want to make an analog-working
sigma-delta front-end for a pulse-width modulated digital amp and do not
want a quantisation. I could however use a fast 10 or more bit A/D and drive
a D/A with it. But this should really be an ultimate possibility.

Just off the top of my head, have you looked at the class-D amplifier
chips that are available these days? Would one of them do the trick?

You could also find a parallel-out ADC, clock it, and just look at the
MSB. It's a bit of overkill, but you'd get the effect of the thing's
fast S/H amp.

Or you could get one of the analog switches that are out there and roll
your own S/H with a high-speed Op-amp.

For that matter, how about a high-speed comparator, and latch the
output? I assume that you're going S/H -> comparator -> driver anyway,
why not go buffer -> comparator -> latch -> driver? You'll have
metastability issues, but you were going to have those anyway.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

 

[John Larkin]

On Wed, 31 Mar 2004 17:48:51 GMT, "Ban" <bansuri@web.de> wrote:

John Larkin wrote:
On Wed, 31 Mar 2004 06:25:10 GMT, "Ban" <bansuri@web.de> wrote:

For my digital amplifier project I need a very fast(<50ns
acquisition) Track/Hold amplifier. The accuracy should be better
0.1%. I could not find anything that is still in production and
affordable.

All the really good track/hold circuits are buried inside the front
ends of A/D chips.

John

That is my feeling as well, because many T/H chips like the BurrBrown SHC605
are obsolete now, but any high-speed A/D-converter has some S/H built in,
that matches its performance. Unfortunately I want to make an analog-working
sigma-delta front-end for a pulse-width modulated digital amp and do not
want a quantisation. I could however use a fast 10 or more bit A/D and drive
a D/A with it. But this should really be an ultimate possibility.

Does your sigma-delta need a s/h? Most s-d quantizers dump the signal
directly into an integrator. A s/h will likely trash your signal:noise
ratio by folding high frequency noise back into the signal range.

John

 

[John Woodgate]

I read in sci.electronics.design that John Larkin <jjlarkin@highlandSNIP
techTHISnologyPLEASE.com> wrote (in <oorl601oif0f01mrea1bes29tficbkh84p@
4ax.com&gt about 'fast track/hold', on Wed, 31 Mar 2004:

All the really good track/hold circuits are buried inside the front
ends of A/D chips.

Do the A/Ds have poor brakes or do the T/Hs have no stop lights?
--
Regards, John Woodgate, OOO - Own Opinions Only.
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[Ban]

Tim Wescott wrote:

Ban wrote:

John Larkin wrote:

On Wed, 31 Mar 2004 06:25:10 GMT, "Ban" <bansuri@web.de> wrote:


For my digital amplifier project I need a very fast(<50ns
acquisition) Track/Hold amplifier. The accuracy should be better
0.1%. I could not find anything that is still in production and
affordable.


Just off the top of my head, have you looked at the class-D amplifier
chips that are available these days? Would one of them do the trick?

Yes and no

You could also find a parallel-out ADC, clock it, and just look at the
MSB. It's a bit of overkill, but you'd get the effect of the thing's
fast S/H amp.

I need an analog voltage to compare it with a triangular wave to find the

length of the pulse. I want it to be constant as long as the period of the
triangular wave(1.2us). The aquisition time should be as short as possible,
not more than 60ns, because this is in the dead-time where the output goes
to zero.

Or you could get one of the analog switches that are out there and
roll your own S/H with a high-speed Op-amp.

I'm afraid that might be the case, I do not have the required test equipment
either, unfortunately. FET switches are too slow and have high pedestal
errors with that 100p sampling capacitor. Maybe a hot carrier diode bridge
as a switch. My triangle generator uses

Fast opamps are no problem anymore now, THS4503. (I hope)

For that matter, how about a high-speed comparator, and latch the
output? I assume that you're going S/H -> comparator -> driver
anyway, why not go buffer -> comparator -> latch -> driver? You'll
have metastability issues, but you were going to have those anyway.

Tim Wescott

I'm still waiting for nice suggestions which are much appreciated. Thanks so
far Tim and John.

I have a parallel all-digital version in work, the problem is there to find
a fast synchroneous up/down-counter(80MHz) and magnitude comparator.

--
ciao Ban
Bordighera, Italy

 

[Tim Wescott]

John Woodgate wrote:

I read in sci.electronics.design that John Larkin <jjlarkin@highlandSNIP
techTHISnologyPLEASE.com> wrote (in <oorl601oif0f01mrea1bes29tficbkh84p@
4ax.com&gt about 'fast track/hold', on Wed, 31 Mar 2004:


All the really good track/hold circuits are buried inside the front
ends of A/D chips.


Do the A/Ds have poor brakes or do the T/Hs have no stop lights?

LOL

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

 

[Tim Wescott]

Ban wrote:

Tim Wescott wrote:

Ban wrote:


John Larkin wrote:


On Wed, 31 Mar 2004 06:25:10 GMT, "Ban" <bansuri@web.de> wrote:



For my digital amplifier project I need a very fast(<50ns
acquisition) Track/Hold amplifier. The accuracy should be better
0.1%. I could not find anything that is still in production and
affordable.

Just off the top of my head, have you looked at the class-D amplifier
chips that are available these days? Would one of them do the trick?


Yes and no


You could also find a parallel-out ADC, clock it, and just look at the
MSB. It's a bit of overkill, but you'd get the effect of the thing's
fast S/H amp.


I need an analog voltage to compare it with a triangular wave to find the
length of the pulse. I want it to be constant as long as the period of the
triangular wave(1.2us). The aquisition time should be as short as possible,
not more than 60ns, because this is in the dead-time where the output goes
to zero.


Or you could get one of the analog switches that are out there and
roll your own S/H with a high-speed Op-amp.



I'm afraid that might be the case, I do not have the required test equipment
either, unfortunately. FET switches are too slow and have high pedestal
errors with that 100p sampling capacitor. Maybe a hot carrier diode bridge
as a switch. My triangle generator uses

Fast opamps are no problem anymore now, THS4503. (I hope)


For that matter, how about a high-speed comparator, and latch the
output? I assume that you're going S/H -> comparator -> driver
anyway, why not go buffer -> comparator -> latch -> driver? You'll
have metastability issues, but you were going to have those anyway.


Tim Wescott

I'm still waiting for nice suggestions which are much appreciated. Thanks so
far Tim and John.

I have a parallel all-digital version in work, the problem is there to find
a fast synchroneous up/down-counter(80MHz) and magnitude comparator.

Oh, I know the answer to that: Inside a TI TMS320F2810 DSP chip.
150MHz, way-versatile PWM generation (including dead-time for
shoot-through), _and_ they throw in a DSP core in the bargain.

Or you could use an FPGA. Getting 80MHz out of a Xilinx Spartan part
should be straight forward, if not a cake walk.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

 

[Fred Bartoli]

"John Woodgate" <jmw@jmwa.demon.contraspam.yuk> a écrit dans le message
news: zDIp2BEe1waAFwMn@jmwa.demon.co.uk...

I read in sci.electronics.design that John Larkin <jjlarkin@highlandSNIP
techTHISnologyPLEASE.com> wrote (in <oorl601oif0f01mrea1bes29tficbkh84p@
4ax.com&gt about 'fast track/hold', on Wed, 31 Mar 2004:

All the really good track/hold circuits are buried inside the front
ends of A/D chips.

Do the A/Ds have poor brakes or do the T/Hs have no stop lights?
--

Excellent one John !


Thanks,
Fred.

 

[mike]

Ban wrote:

Tim Wescott wrote:

Ban wrote:


John Larkin wrote:


On Wed, 31 Mar 2004 06:25:10 GMT, "Ban" <bansuri@web.de> wrote:



For my digital amplifier project I need a very fast(<50ns
acquisition) Track/Hold amplifier. The accuracy should be better
0.1%. I could not find anything that is still in production and
affordable.

Just off the top of my head, have you looked at the class-D amplifier
chips that are available these days? Would one of them do the trick?


Yes and no


You could also find a parallel-out ADC, clock it, and just look at the
MSB. It's a bit of overkill, but you'd get the effect of the thing's
fast S/H amp.


I need an analog voltage to compare it with a triangular wave to find the
length of the pulse. I want it to be constant as long as the period of the
triangular wave(1.2us). The aquisition time should be as short as possible,
not more than 60ns, because this is in the dead-time where the output goes
to zero.


Or you could get one of the analog switches that are out there and
roll your own S/H with a high-speed Op-amp.



I'm afraid that might be the case, I do not have the required test equipment
either, unfortunately. FET switches are too slow and have high pedestal
errors with that 100p sampling capacitor. Maybe a hot carrier diode bridge
as a switch. My triangle generator uses

Fast opamps are no problem anymore now, THS4503. (I hope)


For that matter, how about a high-speed comparator, and latch the
output? I assume that you're going S/H -> comparator -> driver
anyway, why not go buffer -> comparator -> latch -> driver? You'll
have metastability issues, but you were going to have those anyway.


Tim Wescott

I'm still waiting for nice suggestions which are much appreciated. Thanks so
far Tim and John.

I have a parallel all-digital version in work, the problem is there to find
a fast synchroneous up/down-counter(80MHz) and magnitude comparator.

I'm kinda confused.
According to the back of my envelope, you need 7 time constants to get
..1% accuracy. If you want to do it in 50nS, you need a 7 ns time
constant. 100pF cap means driving from 70 ohms or less. Then there's
blowby. 0.1pF across your switch will use up all 0.1% of accuracy.
Then there are switching transients.
Sounds to me like a kinda hard problem.
mike

--
Return address is VALID.
Bunch of stuff For Sale and Wanted at the link below.
Toshiba & Compaq LiIon Batteries, Test Equipment
Honda CB-125S $800 in PDX
Yaesu FTV901R Transverter, 30pS pulser
Tektronix Concept Books, spot welding head...
http://www.geocities.com/SiliconValley/Monitor/4710/

 

[Ban]

mike wrote:

Or you could get one of the analog switches that are out there and
roll your own S/H with a high-speed Op-amp.



I'm afraid that might be the case, I do not have the required test
equipment either, unfortunately. FET switches are too slow and have
high pedestal errors with that 100p sampling capacitor. Maybe a hot
carrier diode bridge as a switch. My triangle generator uses

Fast opamps are no problem anymore now, THS4503. (I hope)


I'm kinda confused.
According to the back of my envelope, you need 7 time constants to get
.1% accuracy. If you want to do it in 50nS, you need a 7 ns time
constant. 100pF cap means driving from 70 ohms or less. Then there's
blowby. 0.1pF across your switch will use up all 0.1% of accuracy.
Then there are switching transients.
Sounds to me like a kinda hard problem.
mike

Mike,
you are absolutely right, that is why I thought the diode bridge would do.
With 5mA you get below 10 Ohms switch resistance and a 0.025pF*2 should be
possible when you drive the upper and lower bridge connections with the
output of the Hold-capacitor buffer through a diode and with a good layout.
This would also reduce the errors to a fixed offset, which gets compensated
for by the delta-sigma feedback from the output (everything is dc-coupled).

--
ciao Ban
Bordighera, Italy

 

[Ban]

Ban wrote:

I'm kinda confused.
According to the back of my envelope, you need 7 time constants to
get .1% accuracy. If you want to do it in 50nS, you need a 7 ns time
constant. 100pF cap means driving from 70 ohms or less. Then
there's blowby. 0.1pF across your switch will use up all 0.1% of
accuracy. Then there are switching transients.
Sounds to me like a kinda hard problem.
mike

Mike,
you are absolutely right, that is why I thought the diode bridge
would do. With 5mA you get below 10 Ohms switch resistance and a
0.025pF*2 should be possible when you drive the upper and lower
bridge connections with the output of the Hold-capacitor buffer
through a diode and with a good layout. This would also reduce the
errors to a fixed offset, which gets compensated for by the
delta-sigma feedback from the output (everything is dc-coupled).

Here is a wonderful app. note, highly recommended for high speed design.
http://www-s.ti.com/sc/psheets/sbaa045/sbaa045.pdf
--
ciao Ban
Bordighera, Italy

 

[Ian Buckner]

"Ban" <bansuri@web.de> wrote in message
newshNac.165098$O31.6863287@news4.tin.it...

mike wrote:
Or you could get one of the analog switches that are out there
and
roll your own S/H with a high-speed Op-amp.



I'm afraid that might be the case, I do not have the required
test
equipment either, unfortunately. FET switches are too slow and
have
high pedestal errors with that 100p sampling capacitor. Maybe a
hot
carrier diode bridge as a switch. My triangle generator uses

Fast opamps are no problem anymore now, THS4503. (I hope)


I'm kinda confused.
According to the back of my envelope, you need 7 time constants to
get
.1% accuracy. If you want to do it in 50nS, you need a 7 ns time
constant. 100pF cap means driving from 70 ohms or less. Then
there's
blowby. 0.1pF across your switch will use up all 0.1% of
accuracy.
Then there are switching transients.
Sounds to me like a kinda hard problem.
mike

Mike,
you are absolutely right, that is why I thought the diode bridge
would do.
With 5mA you get below 10 Ohms switch resistance and a 0.025pF*2
should be
possible when you drive the upper and lower bridge connections with
the
output of the Hold-capacitor buffer through a diode and with a good
layout.
This would also reduce the errors to a fixed offset, which gets
compensated
for by the delta-sigma feedback from the output (everything is
dc-coupled).

--
ciao Ban
Bordighera, Italy


Have you looked at the Peregrine Semiconductor PE4140?

http://www.peregrine-semi.com/prd_mixer_pe4140.html

Regards
Ian

 

[mike]

Ban wrote:

mike wrote:

Or you could get one of the analog switches that are out there and
roll your own S/H with a high-speed Op-amp.



I'm afraid that might be the case, I do not have the required test
equipment either, unfortunately. FET switches are too slow and have
high pedestal errors with that 100p sampling capacitor. Maybe a hot
carrier diode bridge as a switch. My triangle generator uses

Fast opamps are no problem anymore now, THS4503. (I hope)


I'm kinda confused.
According to the back of my envelope, you need 7 time constants to get
.1% accuracy. If you want to do it in 50nS, you need a 7 ns time
constant. 100pF cap means driving from 70 ohms or less. Then there's
blowby. 0.1pF across your switch will use up all 0.1% of accuracy.
Then there are switching transients.
Sounds to me like a kinda hard problem.
mike


Mike,
you are absolutely right, that is why I thought the diode bridge would do.
With 5mA you get below 10 Ohms switch resistance and a 0.025pF*2

OK...I'm still having a hard time with 0.025pF with discrete parts on a
circuit board...but I don't have to make it work in production. ;-)

should be

possible when you drive the upper and lower bridge connections with the
output of the Hold-capacitor buffer through a diode and with a good layout.
This would also reduce the errors to a fixed offset, which gets compensated
for by the delta-sigma feedback from the output (everything is dc-coupled).

I think this sez you don't really need anywhere near .1% accuracy for a
single sample. That makes it a LOT easier.

The Tektronix S6 sampling head uses a six diode traveling wave gate
structure. It's basically a track/hold system. All six diodes are in
series and turned on. To sample, you turn off the diodes. Whatever was
on the middle section of transmission line is the sample. Excellent
blowby characteristics, mostly attributable to the extra two diodes,
but has low efficiency. Have to add lots of gain and let the feedback
fix up the errors on the next sample.
mike



--
Return address is VALID.
Bunch of stuff For Sale and Wanted at the link below.
Toshiba & Compaq LiIon Batteries, Test Equipment
Honda CB-125S $800 in PDX
Yaesu FTV901R Transverter, 30pS pulser
Tektronix Concept Books, spot welding head...
http://www.geocities.com/SiliconValley/Monitor/4710/

 

[Bill Sloman]

"Ban" <bansuri@web.de> wrote in message news:<RfEac.128752$z23.5425945@news3.tin.it>...

Tim Wescott wrote:
Ban wrote:

John Larkin wrote:

On Wed, 31 Mar 2004 06:25:10 GMT, "Ban" <bansuri@web.de> wrote:

<snip>

I'm afraid that might be the case, I do not have the required test equipment
either, unfortunately. FET switches are too slow and have high pedestal
errors with that 100p sampling capacitor. Maybe a hot carrier diode bridge
as a switch. My triangle generator uses

Schottky diode bridges can be pretty fast. Driving them can be a
swine. I used a transformer for the job once (in a 10MHz system). I've
sometimes wondered about a transmission line transformer winding, but
tend to wake up screaming a little later.

Fast opamps are no problem anymore now, THS4503. (I hope)

For that matter, how about a high-speed comparator, and latch the
output? I assume that you're going S/H -> comparator -> driver
anyway, why not go buffer -> comparator -> latch -> driver? You'll
have metastability issues, but you were going to have those anyway.

Tim Wescott

I'm still waiting for nice suggestions which are much appreciated. Thanks so
far Tim and John.

I have a parallel all-digital version in work, the problem is there to find
a fast synchroneous up/down-counter(80MHz) and magnitude comparator.

The 80MHz up-down counter is no problem - the Motorola MC100E136

http://www.onsemi.com/pub/Collateral/MC10E136-D.PDF

can be stacked up to form counters of arbitrary length good for count
rates up to several hundred MHz.

For the magnitude comparator look at the MC100E166

http://www.onsemi.com/pub/Collateral/MC10E166-D.PDF

As has already been pointed out, modern fast programmable logic
devices should be able to manage 80MHz without too much trouble. Post
a query on comp.arch.fpga or just search - this used to be a
frequently asked question.

----------
Bill Sloman, Nijmegen

 

[Ban]

Phil Hobbs wrote:

Bill Sloman wrote:

I'm afraid that might be the case, I do not have the required test
equipment either, unfortunately. FET switches are too slow and have
high pedestal errors with that 100p sampling capacitor. Maybe a hot
carrier diode bridge as a switch. My triangle generator uses


Schottky diode bridges can be pretty fast. Driving them can be a
swine. I used a transformer for the job once (in a 10MHz system).
I've
sometimes wondered about a transmission line transformer winding, but
tend to wake up screaming a little later.


A dual-gate FET can do a great job of sampling. Back in the palmy
days, I used to use 3N201s for this, but a small dual-gate GaAs FET
should work great. Because of the built-in cascode, the gate-drain
capacitance is extremely small, and they will switch in a nanosecond
without difficulty. Linearity is reasonable if not stellar.

Cheers,

Phil Hobbs

THX a lot Bill and Phil, I had a look at these ECL-beasts, and want to avoid
them. They just are out of my budget and eat up current, need also fast
comparators to interface to CMOS.
I also decided to drop the analog approach, too complicated and unstable.
I'll post my result when it is ready.
I will work out a design based on 74LV series and when it functions, I might
port it into an array.
--
ciao Ban
Bordighera, Italy

 

[Bill Sloman]

"Ban" <bansuri@web.de> wrote in message news:<XOhbc.10671$rM4.347875@news4.tin.it>...

Phil Hobbs wrote:
Bill Sloman wrote:

I'm afraid that might be the case, I do not have the required test
equipment either, unfortunately. FET switches are too slow and have
high pedestal errors with that 100p sampling capacitor. Maybe a hot
carrier diode bridge as a switch. My triangle generator uses


Schottky diode bridges can be pretty fast. Driving them can be a
swine. I used a transformer for the job once (in a 10MHz system).
I've
sometimes wondered about a transmission line transformer winding, but
tend to wake up screaming a little later.


A dual-gate FET can do a great job of sampling. Back in the palmy
days, I used to use 3N201s for this, but a small dual-gate GaAs FET
should work great. Because of the built-in cascode, the gate-drain
capacitance is extremely small, and they will switch in a nanosecond
without difficulty. Linearity is reasonable if not stellar.

Cheers,

Phil Hobbs

THX a lot Bill and Phil, I had a look at these ECL-beasts, and want to avoid
them. They just are out of my budget and eat up current, need also fast
comparators to interface to CMOS.

ECL-to-CMOS conversion isn't expensive. I tended to do it with a
long-tailed pair of transistors.

The great advantage of ECL for fast analog work is that - as
current-steering logic - it is very mch quieter than current switching
logic like TTL and CMOS. Very fast comparators and A/D converters have
ECL-compatible outputs, and fast D/A converters have ECL-compatible
inputs because TTL and CMOS in the immediate vicinity of fast analog
creates too much interference.

I also decided to drop the analog approach, too complicated and unstable.
I'll post my result when it is ready.
I will work out a design based on 74LV series and when it functions, I might
port it into an array.

Burying all the fast logic in an array does have the advantage that
the switched currents are being switched between tracks that are very
close together, and the capacitances being charged and discharged are
relatively small - make the inputs and outputs differential and the
array can work almost as quietly as ECL.

------
Bill Sloman, Nijmegen