You have Forum experience enough to know that subjects wander allover the place, here and in any other thread 😉
That´s a complicated way to draw a plain vanilla transconductance amp 😉
Just sayin.
FWIW you can call many different points "chassis ground" ....be my guest.
No one seems to have posted a 'plain vanilla transconductance amp' circuit yet. eg emryBB's has Zo sorta 470R in parallel with a Zobel.
I would like one with at least 1k over 20-20k Hz.
If anyone is really interested in such a beast, this is Mills & Hawksford, which I quoted in my earlier post
https://www.aes.org/e-lib/browse.cfm?elib=6099
It also has an accurate summary of the advantages of current drive.
I was involved in some of their listening tests so can comment on the audible advantages .. but I won't 🙂 .. except to say the gains aren't worth their considerable extra complexity.
However, I AM interested in simpler circuits which might give Zo >1k over 20-20k Hz
I would like one with at least 1k over 20-20k Hz.
If anyone is really interested in such a beast, this is Mills & Hawksford, which I quoted in my earlier post
https://www.aes.org/e-lib/browse.cfm?elib=6099
It also has an accurate summary of the advantages of current drive.
I was involved in some of their listening tests so can comment on the audible advantages .. but I won't 🙂 .. except to say the gains aren't worth their considerable extra complexity.
However, I AM interested in simpler circuits which might give Zo >1k over 20-20k Hz
What benefits are you expecting from having such an ultra-high output impedance?
To reach 1kOhms at 20kHz your power amp needs to have a GBW of 20Mhz (simple single-pole Aol curve assumed).
5x the driver's impedance (or about 100Ohms) @20kHz is way enough and practical, IME.
It's for a different application entirely but I might want to repeat some of the DBLTs I did with Mills & Hawksford 🙂
Not sure about your GBW of 20MHz requirement. Have you got a specific circuit in mind? Please post if so.
BTW, IIRC, the Mills & Hawksford amp had output taken from the Collectors of the Output devices though that's not the only reason it has high Zo
I was using Marcel's formula, assuming an (already largish) 1 Ohm sense resistor:
So, with a more typical 0.22R sense resistor you'd need an amp with 100Mhz GBW product @ 20kHz, in other words, 74dB of open-loop gain @ 20kHz.
And that's without any stabilizing means to keep the chip amp happy at HF.
Improved Howland Current Pump is better suited to very high output impdance but needs careful trimming:
https://www.ti.com/lit/an/snoa474a/snoa474a.pdf
https://www.ti.com/lit/an/snoa474a/snoa474a.pdf
Don’t see any diference in output impedance howland current pump or simple grounded Rsense. I have done howland current pump amps. Still they are nice because the load is grounded so can be paralleld. I done this with lme49811 chip. So the performance is limited by this chip because Rsense is far from gnd and see high voltages. So composite with very fine opamps is not posible. So i went for ordinary Rsense grounded. That way performance can be achieved better. But anyway howland current pump is a nice solution in example with lm3886.
That's interesting. Since opamps are only 'interested' in the difference between their inputs, not what they are with respect to ground*, a non-grounded Rsense is pretty simple. So I don't understand your "So composite with very fine opamps is not posible" statement.
Jan
* within the power supply range of coourse.
Jan
* within the power supply range of coourse.
To measure the diference in example of between 60V and 61V which is 1V will be less accurate. Than to measure between 1V and 0V. I talk about what is max posible from electronics standpoint. It doesn’t mean that it would be audible with less acurate way, but i just like best posible way
This is incorrect. 1V between the opamp inputs is 1V between the opamp inputs. No difference.
Jan
Jan
Yes, but how you measure that 1V diference which is 60v above ground with precision opamp which is max 15v? Without loosing SNR and linearity?
You can (partially) bootstrap the master opamp supplies so the both input CM range is not violated and output voltage still is OK for CM range of the slave opamp.
The Howland has the advantage of a positive and a negative feedback branch and they are both independent of the sense resistor which allows a lot of tweaking. For example, a compensation across the negative leg feedback resistors can be two-pole C-R-C and this allows to tickle out some more output impedance... sketchy wrt to stability, of course.
The biggest improvement (in either topology) can be had from increasing the sense resistor in case one can tolerate the voltage drop and inefficiency.
Anyhow, IMHO aiming for 1k+ source impedance at 20kHz for a 4 or 8 ohms speaker is total overkill and will most definitely not change/improve sound over 100R or so... and if so, the change of circuit parameters to reach that high impedance may be dominant rather than true source impedance.
The Howland has the advantage of a positive and a negative feedback branch and they are both independent of the sense resistor which allows a lot of tweaking. For example, a compensation across the negative leg feedback resistors can be two-pole C-R-C and this allows to tickle out some more output impedance... sketchy wrt to stability, of course.
The biggest improvement (in either topology) can be had from increasing the sense resistor in case one can tolerate the voltage drop and inefficiency.
Anyhow, IMHO aiming for 1k+ source impedance at 20kHz for a 4 or 8 ohms speaker is total overkill and will most definitely not change/improve sound over 100R or so... and if so, the change of circuit parameters to reach that high impedance may be dominant rather than true source impedance.
if i make bootstraped measurement high up there i have to transport that measurement somehow down to 0V reference. Precision opamp should work at 0V reference , and precision current measurement should be done also at 0V reference. Any transporting of measurement from high to low voltage will degrade performance by the dividing ratio of transporting. The same effect as we want larger Rsense for precision. I really think there is no way around.
Back in the day I added a 0.1 ohm resistor in series with load, between power amp out and speaker Hot terminal, precisely to be able to ground one end of speaker and yet maintain high(ish) Z out.
I attenuated that common voltage 4:1 or 5:1 to turn rail to rail swing into something handled by sensor Op Amp inputs, its output became the NFB voltage.
It did work , with a little tweaking,
That said, I am lazy and minimalist so went back to the simple version, resistor between speaker "-" and ground.
I'm pretty sure there are ways but of course it adds cicruit complexity. As I said, keeping the master OpAmps inputs within CM range is the main idea but you have to keep an eye on the CM range of the slave OpAmp. Obviously, it is better to max out CM range on the slave as that one is inside the feedback loop.
Might I refer this discussion to my amplifier design, which achieves current output without any sensing resistor, and without too much complexity? See: https://robinet.co.uk/amplifier-design
John
John
Hello John,
thank you for the link. I had aquick look at your circuit. As far as i see it you use the Emiter Resistors as current sense, so there is current sensing. But you make your amplifier a voltage Amplifier by the feedback Resistors R14/R8. Without a closer look i would say it is a mixed feedback.
Bernd
thank you for the link. I had aquick look at your circuit. As far as i see it you use the Emiter Resistors as current sense, so there is current sensing. But you make your amplifier a voltage Amplifier by the feedback Resistors R14/R8. Without a closer look i would say it is a mixed feedback.
Bernd
Thank you, Bernd. You are right, of course. But these emitter resistors don't introduce any loss over and above those in a conventional amplifier, and in any case, they are of a very low value (0.1 Ohms). But yes, we can call it mixed feedback. And I should add, the load can be grounded, and bridge-working is very straightforward.
Last edited: