Use bleeder resistors in audio opamp design?

My stuff usually has extreme dynamic range requirements. Offset tweaking is a help, but it doesn't fix the weird bumpy noise floors caused by op amp filters in general. Any time you use feedback to change the frequency response of a circuit, you do things to the noise floor. That becomes a problem when you're applying a lot of gain in some frequency band, e.g. to turn a limp RC rolloff into a sharp filter. The noise gain can be pretty bad sometimes.

LC filters don't do this, and neither do good digital filters. So usually, analogue filters are for preserving dynamic range before digitizing.

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

capacitors.

I specifically said, "for moderate dynamic range jobs". It's horses for courses.

It isn't the topology I don't like, it's the idea of using 10 op amps in a built-up filter, with the usual issues. In an IC, where you have good matching and can do current-mode things and auto-zeroing to preserve dynamic range, I'm sure the limitations are quite different. My low frequency electronics design work to date has been exclusively applications circuits,(*) so that's where I'm coming from.

Back in the palmy days, I used your DC-accurate SC lowpass filter a few times, and liked it a lot. But even well-designed active filters have fairly serious dynamic range problems if you're trying anything very fancy. It's just a limitation of using feedback to turn an RC rolloff into something nice and sharp--it costs you noise gain.

Cheers

Phil Hobbs

(*) (Though I also do circuits in the infrared, which are much more like ICs.)

--
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

capacitors.

SC filters tend to be noisy. Switching spikes, PSRR, PSRR aliasing, and just generally bad noise levels.

John

e:

ote:

:

apacitors.

them in

r would

es on

aps in

ctros in a

series

nodes

fix it?

There

by

hen

be

ilt the

filter

if

ting

of

d

ow

is

an

nal

. p

onal

ss

four

rsus

an

gher

sign

n

e to

the

of

as

lume

u

t I see

s

and

ed

SNR

bly

lso

ork

l.net

or

od

Oversampled converters and DSP pretty much killed the SCF market.

:

te:

pacitors.

hem in

would

s on

ps in

tros in a

eries

odes

fix it?

There

y

en

be

lt the

filter

if

ing

of

w

is

n

al

nal

s

our

sus

n

her

ign

to

the

of

s

ume

I see

nd

d

SNR

ly

so

rk

.net

r d t

The matching in ICs is only really relevant for the caps. Charge injection pretty much makes op amp offset moot. SCF was technology that really had to go.

Incidentally, SK filters aren't really DC accurate. The op amp is still in the picture. There is a Fluke circuit done years ago that makes a truly DC accurate filter. LTC and Maxim did SCFs based on that scheme, though I don't know why Fluke never sued either company over patent infringements.

Note the LTC parts do not meet any classical filter shape as they state in their datasheets. I did analysis on the parts. It wasn't particularly hard to make them fit classical filter shapes (Bessel & Butterworth), but it took a root solver and root locus to design them optimally. Clearly something a LTC employee should have been able to do. I've often wondered how a company as good as LTC screwed this up, but could never get the inside story.

These were used in mission critical telemetry applications.

--
You can't fix stupid. You can't even put a Band-Aid? on it, because it's
Teflon coated.

capacitors.

in

in a

it?

Yeah right... Remove the caps and it's a unity gain buffer, offset null that and jam in the resistor for input bias, what's that? DC inaccurate?

Grant.

ote:

e:

te:

capacitors.

e them in

ter would

odes on

caps in

lectros in a

n series

e nodes

ll fix it?

g. There

m

g by

then

an be

built the

e

my filter

e

se if

eating

ue of

o

s.

add

Now

do is

r an

ignal

lf.

top

tional

less

f four

versus

can

er

higher

design

on

one to

of the

de of

was

volume

you

ilt I see

e

kes

, and

owed

in SNR

bably

'd

e

also

work

cal.net

for

in

good

w

ve

ff

ke

.net

l

e?

Across the resistor, it is DC accurate.

nd

the unit

put

t

Input referred noise is the name of the game. Apply gain first, then filter, and the filter noise is reduced by the front end gain.

capacitors.

in

in

in a

series

it?

There

Use zero-offset chopamps.

ftp://jjlarkin.lmi.net/Filter1.jpg

ftp://jjlarkin.lmi.net/Filter2.jpg

ftp://jjlarkin.lmi.net/FilterBoard.jpg

These mostly anti-alias for the real digital filters, in an FPGA.

John

But dynamic range is unchanged. All that happens then is that you clip sooner.

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

and

ince the unit

utput

r.

.net

t

Unless the filter is passive, you will always have the trade off of noise versus filtering. This is a given. But for a vast number of communications applications, the input is bounded, so you can select front end gain sufficient to cover the noise floor. I counter that unless your passive filter is tuned, the filter is basically not much of a filter. You simply can't get accurate enough passive components to produce a manufacturable product. Yeah some small production lot is possible by hand selection, but not a volume product.

Dynamic range is very precious, especially in optical measurements.

Usually you can maintain dynamic range without hand tweaking by using a gentle S-K filter (or, better, an LC) and doing the fancy stuff digitally.

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

...

Grrr, I don't care that much for DC accurate here, just as long as it saves some coupling caps. But I dislike some people's black/white view of the field, claiming they have the one and only true path, topology or method.

So pretty :)

Hey I've seen that circuit before! That's what my filters are, plus the other way too ;)

Yeah well I'm only making an active crossover, so surprised by the commentary a silly question raised, which is better fun the total abuse threads elsewhere on the group.

Anything I do upfront is going to be better than the passive Xover that's in my speakers now. Since I write that, some no doubt will jump onto the audiophoolary bandwagon. me? There's a question in my mind for decades I can answer, if I build this silly thing and see for myself.

Looking at some of the LTC opamps, they're quite cheap and very low offset voltage, much less than what I expected for non-chopper amps.

But I started this project when I bought couple tubes of opamps for half the single opamp price. Sourcing the caps was the hard part.

I think I'll be happy with a few mV overall offset, maybe trim out the DC on one of the opamp nulls in the signal chain. If it's not stable, block it with a cap. So many options... Interesting topics visited this thread, thanks to all contributors.

Grant.

Not me! My only claim is that S-K filters are sometimes useful. Others have stated that only amateurs use them.

We select parts and lay out boards for pretty!

John

Why are the edges of the daughterboards so rough?

Bad girls, bad girls, what you gon'a do?

--
You can't fix stupid. You can't even put a Band-Aid? on it, because it's
Teflon coated.

Beats me. It doesn't seen to affect the frequency response.

John

I doubt that it would. I'd just like to know so I don't do that.