Any experience with negative impedance?

[Adrian Tuddenham]

Question: since you've established the goal is characterising an existing
speaker and its acoustic environment, shouldn't the test be performed with
period equipment? Or, given the availability and known limitations of
that, then a modern equivalent designed to emulate key features?

If they're going for historical accuracy, I expect the original
installation contended with the same wire length already, plus a tube amp,
evidently one without the benefit of NFB by your description (typical
triode amps achieve a damping factor of about 3, not counting the
transmission line of course). You should be adding resistance, quite a
bit by modern standards.

I'm sure you've already belabored all this to the client, so I'd just be
curious to know the reasons behind it.

The performance of the loudspeaker is my client's main interest. We can
investigate whether it changes with driving impedance at a later date,
but extablishing the starting point, uncluttered by other variables, has
to be the first move.

Although we know the circuits of two out of the three original
amplifiers (different amplifiers were used at different times), we are
missing key data such as output transformer ratios and construction.
With triode output valves and no overall feedback, the transformer ratio
is going to have a big influence on the output impedance and general
performance of the amplifier.

 

[Adrian Tuddenham]

On Mon, 28 Jan 2013 10:38:23 +0000,


If you are running 3-wire T&E, you can do remote sense, analogous to a
3-wire RTD. Signal, return, sense. The sense wire gives you half of
the total wire drop, so gain up by 2 and add that to the outgoing high
side signal.

The return wire *is* the sense shunt.

Something like this?

http://www.poppyrecords.co.uk/other/images/Circuit2.gif

 

[[email protected]]

I really don't understand the need to monitor the end point of the
transmission line? If you were to loop back a pair of wires only to
be used to control feed back grain in the amp, it would not make any
difference, the amp circuit would simply increase the gain to force the
end point to match the desired set point at the source.

Simply put, if you just turn up the volume, you'll end up with the
same results with out all the bull crap, also, doing that feed back like
that is only going to introduce more problems.

The impedance of the loudspeaker varies over the audio range, so the
current will not be constant. This means that the volt-drop will vary
with frequency and cannot be compensated by just turning up the gain.[/QUOTE]

Are you trying to null the speaker or the wiring?

 

[Adrian Tuddenham]

An other thing, from experience: You are driving a high gain opamp
differentially from an unknown impedance source (input connector).

Maybe an input transformer and a normal transistor input stage would be
better. That stage should be designed so its maximum AC output can never
send the TDA power amp into clipping. The idea is that with an 'unknown'
impedance at the input you cannot guarantee the input circuit input will
never oscillate if the input is part of the feedback circuit of the opamp.
If you use transformers you need to shield against hum pickup (rotation,
mu-metal).

I am trying to keep the price down, otherwise I would have preferred to
ues an input transformer. This equipment is intended to be used with
very short input connections, so balanced line isn't necessary. The
sources will probably be a computer and a professional grade radio
tuner.

I can't see anything in my input circuit which would lead me to expect
oscillation with any source impedance from fully inductive to fully
capacitive. Could you explain why you think this is possible?

The amplifier circuit has now been upgraded:

http://www.poppyrecords.co.uk/other/images/Circuit2.gif

I do not know this TDA7295, I have used the TDA7294, it is a very stable
MOS amp that does not need the usual RC filter to ground at the input.
http://panteltje.com/panteltje/amplifier/index.html mm was year 2000, time
flies, amp still working I drive heavy inductive loads via a step up
transformer like my cryo-cooler with it...

I have downloaded your data sheet and cannot see any difference between
the TDA7294 and the TDA7295. I'm sure there must be a difference, but I
cannot see it.

 

[Adrian Tuddenham]

Are you trying to null the speaker or the wiring?

Only the wiring, but the volt-drop of the wiring depends on the
impedance of the loudspeaker, which varies with frequency. I have to
find a way of continuously monitoring the volt drop* and adding it to
the amplifier signal.

*Either by actual measurement or by eztrapolation from some other
related parameter.

 

[[email protected]]

Are you trying to null the speaker or the wiring?

Only the wiring, but the volt-drop of the wiring depends on the
impedance of the loudspeaker, which varies with frequency. I have to
find a way of continuously monitoring the volt drop* and adding it to
the amplifier signal.[/QUOTE]

Can you measure the speaker/wiring? How about a DSP with the
(complex?) inverse transfer function of the wiring, before the power
amp?

*Either by actual measurement or by eztrapolation from some other
related parameter.

Exactly. Do it in the digital domain. It won't be perfect but it
might get you close enough.

 

[Tim Williams]

The performance of the loudspeaker is my client's main interest. We can
investigate whether it changes with driving impedance at a later date,
but extablishing the starting point, uncluttered by other variables, has
to be the first move.

Ok, so it's actually an open variable. Excellent!

I dare suggest that you define the system as loudspeaker *and* wires.
Trying to drive it with a low or negative resistance is certainly not in
the spirit of the original design, and the speaker was designed to operate
best on period equipment.

Remember, modern speakers, which prefer high DFs, were only developed
because modern amplifiers also achieve high DFs. DF is no guarantee of
sound quality, and a speaker can be designed to favor very high, moderate
or very low DF (there are speakers out there designed for constant-current
amps -- and apparently they also work very well).

Although we know the circuits of two out of the three original
amplifiers (different amplifiers were used at different times), we are
missing key data such as output transformer ratios and construction.
With triode output valves and no overall feedback, the transformer ratio
is going to have a big influence on the output impedance and general
performance of the amplifier.

Yeah, but they wouldn't have used any ratio outside of 2 to 5 times plate
resistance, which is the maximum power rule-of-thumb. Occasionally,
audiophiles stray in the upward direction, to keep distortion lower, but
this severely cuts into power output.

Tim

 

[legg]

On Mon, 28 Jan 2013 10:38:23 +0000,

The requirement is to design a specialist one-off audio amplifier to
drive a loudspeaker at the end of a long cable for experimental
purposes. The voice coil is nominally 15 ohms with a pure resistance of
10 ohms. The loop resistance of the cable will be somewhere between 2
and 4 ohms. It would be desirable to have a damping factor of 10 or
better, so I need to reduce the effect of the cable resistance in some
way.

I have considered the use of line transformers or a four-wire feedback
system, but neither of them is practical in this particular case. As
this amplifier will form part of a fixed installation and will only ever
be used to drive the one loudspeaker, I am now looking into the
possibility of giving the amplifier 3 ohms of negative output impedance.

The simple solution is to use an off-the-shelf unit for the power
amplifier, the TDA7295 should be more than adequate. With a 0.5 ohm
resistor in the return loudspeaker wire, I can measure the output
current and derive a suitable voltage for adding to the input signal.

The practical situation generates at least two problems which the simple
theory does not take into account:

1) The capacitance between the conductors of the long loudspeaker
circuit will give the effect of a falling loop impedance as the
frequency increases, leading to ultrasonic instability.

2) The long unscreened run of loudspeaker wire, which might be in close
proximity to mains cables and dimmable lighting circuits, may pick up
interference by capacitive coupling, which will then be injected into
the feedback point and amplified.

Both of these might be overcome by a capacitor across the 0.5 ohm
resistor, the value being chose to only take effect above the highest
wanted audio frequency.

The problem is that I have to construct this equipment without access to
the loudspeaker or wiring, and it then has to be certain of working
correctly the first time it is installed. For this reason I would be
very grateful if anyone has had any practical experience of anything
similar and could let me know of the snags they have encountered and the
solutions which worked.

Don't do it and say you did.

RL

 

[Adrian Tuddenham]

Yes, but I'm still concerned about cable inductance. The three conductors
will have (mumble mumble) tricky mutual inductances, so a purely ohmic
correction may get wonky at high frequencies. As might a simple
negative-resistance amplifier.

How long is the cable run?

Somewhere between 40 and 100 metres.

If the inductance is significant, it should only be the compensation
which suffers, not the stability.

Above 10 Kc/s it is most unlikely that there would be any significant
audio output from this unit and the voice coil impedance would be a lot
higher than the nominal 15 ohms, so I am not particularly concerned
about failure of the cable compensation at higher frequencies.

 

[Adrian Tuddenham]

Ok, so it's actually an open variable. Excellent!

I dare suggest that you define the system as loudspeaker *and* wires.

This is not an ancient audiophile project, this is an attempt to measure
one thing at a time. That is the way to do objective research: measure
each parameter with as little interference as possible from all the
others.

 

[[email protected]]

On Tue, 29 Jan 2013 19:50:50 +0000,



Can you measure the speaker/wiring? How about a DSP with the
(complex?) inverse transfer function of the wiring, before the power
amp?


Exactly. Do it in the digital domain. It won't be perfect but it
might get you close enough.

Thanks to John Larkin's suggestion to use the spare conductor as a
sensing wire, I can do it with one op-amp, two resistors and a
capacitor:[/QUOTE]

I thought you'd already discounted additional (sense) wires. Didn't
you object to the Kelvin connection idea?

http://www.poppyrecords.co.uk/other/images/Circuit2.gif

- for analogue problems use analogue solutions.

You have far more control in the digital domain. I thought you had
already rejected the obvious feedback solutions.

 

[Adrian Tuddenham]

Thanks to John Larkin's suggestion to use the spare conductor as a
sensing wire, I can do it with one op-amp, two resistors and a
capacitor:

I thought you'd already discounted additional (sense) wires. Didn't you
object to the Kelvin connection idea?[/QUOTE]

There were objections to using a full four-wire sensing system, however,
the three-wire system is relatively easy to impliment and should be
adequate.

You have far more control in the digital domain. I thought you had
already rejected the obvious feedback solutions.

I have one month to get this built and working. I can't afford to mess
about experimenting with digital 'solutions' and any use of digital will
immediately involve myself and the client in a long bureaucratic
rigmarole of EMC testing.

 

[Adrian Tuddenham]

As long as it doesn't oscillate.

I had some US style 14/2 Romex handy, so I measured a meter of it.

https://dl.dropbox.com/u/53724080/SED/Romex_LC.JPG

Thanks for that info. It looks as though I don't have much to worry
about there.

As someone predicted, inductance is minimum if you use the paralleled
pair against the bare ground wire. In the 3-wire remote sense config,
there is probably a best permutation. This could be, and should be,
Spiced.

Is your T&E similar geometry to Romex? Will it be in a
conduit? That would > change things a little.
http://en.wikipedia.org/wiki/File:Romex_cable.agr.jpg

That looks very similar. T&E doesn't have the paper insulation and the
colours are different, but the geometry is almost identical. I was
proposing to use 1.5 sq.mm load conductor size (the earth is slightly
smaller) which has a resistance of 12.10 milliohms per metre of single
conductor. If the inductance doesn't prove to be a problem, I would
prefer to use the load cables to feed the voice coil and the earth wire
as the sensing wire - mainly because that is how the installer will wire
it up, no matter how many times I instruct him to the contrary.

As the client will be needing a drum of cable, I will buy it and make
the necessary tests and measurements before sending it on to be
installed on the premises. That way I shall have the opportunity to
sort out any problems 'in house' before arriving on site.

How come you guys can say Kc/s and we have to use KHz? Not fair.

I can claim to be old enough to have always used it.

 

[Fred Abse]

The signal level at that point is nominally 0dBu.

For the benefit of us non-audio guys, what are dBu referred to?

 

[Adrian Tuddenham]

For the benefit of us non-audio guys, what are dBu referred to?

0dBu is the voltage which would be called "0dBm" if it were terminated
in 600 ohms i.e. 0.775v rms (...because dBm can only refer to power
into 600 ohms).

 

[Adrian Tuddenham]

Well, how about a combination of those?
And do not forget the usual mains plugged into it.
Series resistors, limiting diodes?

That is planned, but not drawn yet.

What I expected to see was some relay that only connects the speaker after the
output was found to be OK.



I don't know either, been 14 years, it is a good chip, the TDA7294.

As to 'as cheap as possible'? (was that what you mentioned?) seems to
conflict with 'research project'. Component price usually is not a point
with one of designs, it is the hours.

I tend to get involved in projects where the hours are less important
than getting the job done to my satisfaction - it's called "being self
employed".

Other important thing: I do not see any RF filters in the input.

Very good point, I shall have to add those when I am satisfied that the
circuit does what is required of it. (This drawing is only a prototype
at this stage.)

I had to decouple each and every input including the mains to my mixer in
Amsterdam, as the electric tram drivers radio came right through (AM
detection in input circuit). We had that happen in a hall a couple of
times. Really is a distraction...

I am surprised that you didn't get noises from the traction current too.
If I remember correctly, Amsterdam trams have a carbon strip on their
current colectors, so that probably helps to reduce contact noise. It
is much easier to fit on-board filters to a tram than to a trolleybus,
because the tram has a well-earthed chassis.

The old UK trams used a brass wheel for current collection and had no
suppressors. On rainy days, people at the outer end of a tram route
would switch on their wireless sets to listen for a tram approaching, so
as to avoid standing a long time in the rain waiting for it.

I see you stabilized the + and - 30 to the power amp, loop bandwidth big
enough? Those are slow transistors... Or is it just over-voltage
protection against high mains? Sometimes a simple emitter follower with
cap and zener to ground can function as C multiplier and give extra
filtering over a wide band, and limit voltage at the same time.

The real reason was to give me some control over the peak power of the
amplifier. If I find it is capable of delivering dangerous transients
to the loudspeaker, I can change the zener diodes and bring down the
power rails to a lower value. There are reservoir capacitors on the
pins of the power amplifier chip (as recommended in the data sheet), so
the speed of the regulating device is not important.

 

[Fred Abse]

0dBu is the voltage which would be called "0dBm" if it were terminated
in 600 ohms i.e. 0.775v rms (...because dBm can only refer to power
into 600 ohms).

The dBm I use are relative 1mW in 50 ohms.

 

[Adrian Tuddenham]

No, but in the TV studio in Hilversum they had strange earth currents in
the video cables every time a train went past it. There is a special train
stop across from the studio complex.

A few thousand amps D.C. in conductors about 4 metres apart can do
wonderful things to nearby CRT television monitors and pickup tubes.
The return conductors rails should not be in multiple contact with
earth, but faults can occur and remain unnoticed until someone complains
that they are getting several volts difference along an 'earthed' wire
(or the foundations of a steel-framed building have just been eaten
through).

When working in Amsterdam next to the railway central station I had 400 Hz
or so pickup from the trains there.

That was more likely to be signalling currents than power.

 

[Adrian Tuddenham]

The dBm I use are relative 1mW in 50 ohms.

I presume you are working with RF. Is there an RF equivalent of dBu,
based on voltage without reference to impedance?

 

[[email protected]]

On Wed, 30 Jan 2013 12:34:47 +0000,


I thought you'd already discounted additional (sense) wires. Didn't you
object to the Kelvin connection idea?

There were objections to using a full four-wire sensing system, however,
the three-wire system is relatively easy to impliment and should be
adequate.

You have far more control in the digital domain. I thought you had
already rejected the obvious feedback solutions.

I have one month to get this built and working. I can't afford to mess
about experimenting with digital 'solutions' and any use of digital will
immediately involve myself and the client in a long bureaucratic
rigmarole of EMC testing.[/QUOTE]

Just put a "CE" sticker on it and be done with it. ;-)