EMRFD Message Archive 1181
Message Date From Subject 1181 2007-11-23 14:31:45 Frank Floating grounds and ground return currents... I'm building a portable rig and while things are working, I'm trying
to puzzle out a few things.
I've got some op amp audio filters feeding on of those NE570 compandor
chips. I'm running the op amps on 12V and ground. A zener diode and
resistors have been used to create a 5.7V "rail", which is my
"floating ground" for these op amps that prefer to operate from +/-
So, when the op amp delivers output current, where's the return go? I
figure as long as the output current is positive, then the current
comes from the +12V, but if I were, for instance, to capacitivly
couple the op amp's output, then the negative-going output current
actually gets sourced through the op amp's negative power pin, which
is tied to my circuit ground.
These op amp audio filters are Sallen-Key styles, so there are
equalization components which normally would tie to "ground", but
instead I have them tied to that 5.7V rail...in some cases, these are
resistors, so they are setting the op amp's "center voltage" and
obviously the 5.7V rail is the right place.
Now, an op am has a PSRR of 40-60dB, so normally any noise on the
power supply is relatively moot...just doens't matter.
But now, I'm feeding the output of these op amps to the NE570
compandor chip. it has diff gain stage with very little
documentation, and it has internal op amps that are v-referenced to
positive power supply through resistors.
So, is my return current from my op-amp filters somehow still my 5.7V
line, or is it power supply/ground? It seems like the way this is all
working out, I've got to be religious about bonding the +12V, +5.7V
and zero volt rails with mongo capacitors, right?
Am I making too much out of this, or is it right to have this
1182 2007-11-23 18:02:34 Shawn Upton Re: Floating grounds and ground return currents... Ultimately, the return current all goes back to
"ground" or the 12V return. Afterall, it came in on
the 12V positive line, so therefore it has to go out
on the 12V negative line.
As for your differential input stage, it would be best
to tie undriven input to 5.7V. While 5.7V and
"ground" should both be AC grounds, I don't know if
your IC input is rail-to-rail input (that is, it may
not be happy being tied to the IC's negative rail).
If it is rail-to-rail, you could tie the undriven
input to ground, but any signal difference that shows
up (think other ground currents working into non-zero
resistances) will affect the IC and appear as another
signal. So the undriven input should go to 5.7V since
the signal is referenced to 5.7V (your new
ground)--especially if the IC inputs are not infinite
I'm not sure that return currents is the way to
analyze this; but rather where is the signal (at the
various stages) referenced.
1183 2007-11-23 23:45:01 jr_dakota Re: Floating grounds and ground return currents... I think you need better documentation on the NE570 first off .... try
As you are finding out using a floating ground is a real problem when
it comes to interfacing with other circuits that don't use the same
reference .... if you really feel you need a bipolar supply for the
opamps a better solution might be a charge pump voltage inverter ....
In your case where you only need to source 1 or 2 opamp chips you can
use a voltage inverter which can be made from a 555 timer (Which can
easily source 100mA) or you can get a dedicated voltage inverter IC
(Often called Charge Pumps).... You will lose a couple of volts so
with a 12 volt supply I'd shoot for +/- 8V which will still give you
more than enough headroom and use 78L08 and 79L08 regulators to clean
things up (This will also help isolate the audio circuit power supply
from the rest of power supply) ... I'd also consider building the
inverter in a PCB box or other small shielded enclosure because
although the inverter only runs at a few khz, it is a square wave
before filtering and regulation with plenty of harmonics in the LF
1184 2007-11-24 06:34:41 Shawn Upton Re: Floating grounds and ground return currents... If it were me, I'd skip the invertor. Probably a 10k
to a fixed voltage reference (I'm thinking something
like an LM4040 series, 5V, but a number of other
references should work), followed by a unity gain
op-amp with probably a 10uF tantalum and a 0.1uF mono
on the output. In my experience, 10+ microfarads
tends to be enough on general purpose op-amps to swamp
them into submission. Then one would have a
reasonably low impedance "ground" rail, without any
switching noise involved.
The shunt voltage reference is probably overkill (as
compared to say just a pair of 10k's voltage divider),
but if the +V line sags, then "ground" would stay at
the same potential as real ground (+V return that is).
1191 2007-11-25 04:16:44 jr_dakota Re: Floating grounds and ground return currents... I believe you are talking about biasing the + input to 1/2 Vcc as per
usual for a single supply op amp .... problem with that is the Sallen
Key circuits are hard to adapt to a single supply op amp hence he
tried to make a floating ground instead ... I tend to go with multiple
feedback filter circuits instead which are easy to bias to 1/2 Vcc
(but a lot harder to design than the Equal Value Sallen Key)
Another problem with using the 1/2 Vcc bias trick is when you have a
low level, low impedance circuit (Mic preamp, SDR backend)because your
biasing resistors become your dominant input noise source which is why
some people will take out a 4nv Noise Voltage op amp and replace it
with a (expensive) 1nV Noise Voltage op amp and can't tell a bit of
difference in noise because the biasing resistors set the minimum
input noise not the op amp itself, using the same basic circuit with a
bipolar supply and you then notice the difference in noise ....
Differential op amp circuits are another place single supply doesn't
work very well because the biasing resistors upset the balance of the
circuit and hence the inherent common mode rejecti
1192 2007-11-25 05:47:57 Shawn Upton Re: Floating grounds and ground return currents... Hmm, didn't think about resistor noise. One could use
small resistors and/or significant bypass capacitance
on the resistor divider; I'm not sure what the
difference in noise would be, comparing a 10k divider
to a shunt voltage reference. Some of those
references get expensive quickly; but then again, I
have seen some that are source/sink (LT1461 comes to
mind but that may be an obsolete part by now), meaning
no need to buffer the reference.
I just have to question how well one (at home) can
develop a SMPS to gain a bipolar supply on the same
board and not have some other noise related issue pop
1193 2007-11-25 11:42:40 jr_dakota Re: Floating grounds and ground return currents... If you input impedance is high enough, say 10K ohm common for line
level then the resistor noise becomes moot .... also once you get
through a stage of amplification and get the voltage levels up it
becomes moot also ... He probably satisfies both these criteria but in
his case it's the circuit topology that's the problem, since several
frequency determining components are tied to the non-inverting input
also tieing the bias resistors throws everything off ... It can be
compensated for but it's not very easy which is why I favor the
multiple feedback filter because the non-inverting input doesn't the
frequency determining components tied to it making it easy to bias
Charge Pumps are a bit different than your average SMPS, mainly it's
inductorless so you don't have the same EMI issues you do with an
inductor based switcher, if you run them high enough, say 50Khz or
better than filtering is easy (smaller caps) and it's above the audio
range .... The limitations compared to a inductor based SMPS are a
much lower current limit, and your output voltage tends to follow the
input voltage where a inductor SMPS circuit you can vary the input
voltage over a fairly wide range and keep the output voltage constant
... The output can also vary with current .... On the plus side
filtering and shielding requirements are less stringent
I'm using one in my SDR testbed, it's shielded in a PCB box with
feedthru caps and good filtering and bypassing ... Any noise would
readily show up in Power SDR (And it did at first) but once I shielded
and tweaked it the noise was better than 100 db down which is pretty
good even for a linear supply .... One of these days I'll get around
to better documenting the circuit I'm using since I'm going to need
another one so
1194 2007-11-25 22:56:08 Frank Re: Floating grounds and ground return currents... Yes, well, when I said "return" currents, I did mean "signal return".
The application is portable - small size and minimal parts are
important criteria and it is, in fact, a device with a mic preamp.
Therefore, there simply isn't space available for a switcher/charge
pump and requisite shielding. I had thought of that. I'm more than a
bit fanatical about noise, and as yet, have not encountered a
switching style supply of any kind that does not induce detectable
noise. I spent a bit of this past weekend fighting that in an Icom
IC-735; the swishing swooping background noises caused by that radio's
internal negative voltage supply are rather frustrating when you're
operating a DX contest with very weak antennas and need every bit of
noise floor the rig can deliver.
To reduce power supply noise issues, I developed a discrete mic input
stage, optimised for low IMD but validated in SPICE for PSRR. The
anticipated operating level into the Sallen-Key networks is roughly
300mV p-p and I expect resistor noise won't be an issue....but have
not undertaken the analysis of it. In the lowpass S-K sections, a
resistor does go from the non-inverting input to the 5.7V line, and I
used a series R/shunt C network from the line for noise isolation,
selecting the R value to yield less than 0.1VDC output offset at the
stage output, based on input bias currents of fairly low-grade op
amps. The S-K problem, as pointed out, is that even if your op amp
can take an input voltage down to negative rail, this would still bias
the ouput at negative rail, and you'd cut off all the negative-going
The highpass S-K networks have a DC connection from the previous
stage's output to the non-inverting input; they have a DC gain of
unity, so whatever the bias output voltage of the prior stage is, they
replicate, plus or minus bias offsets.
My only real concern had to do with the prospect of some of my later,
low-Z stages dumping "heavy" currents onto the supply/ground lines and
having those resultant induced "noise" voltages end up back at input
The NE570 is indeed a piece of work...but Philips had little choice,
as it is an implementation of an ancient Telco chip specification.
The fact that they've chosen internal op-amp biasing resistors that
tie to the op-amp's negative rail is unfortunate, but likely they had
no choice in the matter. I urge caution to anybody who chooses to use
this device or its cousins the 571/2/3. Make certain to download the
data sheets for the entire family and peruse the suggested circuits
carefully; I have not encountered a data sheet with more errata in it
than is found in this family's.
Anyway, I thank the list kindly for the thoughts; though I gained
nothing new, I consider the exercise akin to that of an engineering
design review: an opportunity to see if I've truly turned over all the
rocks in the quarry.
With the most gentlemenly of thanks,
1195 2007-11-27 00:14:25 jr_dakota Re: Floating grounds and ground return currents...