Please post a circuit diagram of your scheme.
I've attempted 'simple' current amps but even a simple spec. like >100R output over 20Hz - 20kHz ends up much too complicated.
There's loadsa very naive 'current amps' around 😡
I added a very general diagram in the thread "Common Emitter OPS?" that you also contributed too (?). Simply imagine Zs in that diagram as the series combination of the (ground connected) current sensing resistor and the loudspeaker.
One point of note is that I operate the amplifiers at effectively unity gain (wrt the sensing resistor output) which is substantially below that of a conventional power amplifier: Stability margins are therefore not marginal if adopting a conventional topology! The output resistance is considerably more than 100R, however.
I am also considering documenting my work so that I can publish it in a separate thread. That may take some time...
Wouldn't cutting the other drivers be a lot more complicated than just boosting and limiting?
Thermal compression affects different frequencies differently, as you can see in B&O's investigation of thermal compression compensation. It seems to be much easier to boost, and then ultimately limit to protect the woofer than to try to reduce the levels accordingly for the other drivers.
In my system, the huge power only happens for very short time periods. It's like having a much higher crest factor because of the ~20dB or so bass boost. While the drivers will still suffer a bit, it wouldn't be too bad for 99% of the time.
In my system I do all the DSP in software on a single board computer. So once I get the info on Re in there I can do whatever I want with it: boost, cut, etc. Reducing power to match the instantaneous efficiency of one driver to another seems more prudent than increasing power, and that's why I would use that approach (if I did it at all). I could do this to all the drivers in the multiway system, import the data for each, and then take action to balance things out in the crossover (WRT gain and frequency response. That's the beauty of working with software - you can easily invent and implement new things. It wouldn't be trivial, but it is definitely doable. Note that this is not using any current drive at all, it is just compensating for VC heating.
Do you mean
with OutputStage.pdf?
It's not clear in OuputStage.pdf if there any feedback and if so, where it is taken from.
I would certainly appreciate a rough schematic of a working prototype by PM.
As Soundbloke wrote here, Zs is the combined impedance of the sense resistor (at the ground-side leg) + the speaker. The signal from the sense resistor is used to form a global feedback for the opamp driving the gates while the output stage itself is a 'railshifter' transimpedance config with local voltage feedback to stabilize and balance it.
Last edited:
Thank you so much for answering so perfectly 🙂 Although I have never heard the term 'railshifter' before and I hope it doesn't direct people to think there is no gain in the stage - or at least no equivalent of gain if an equivalent voltage output is imagined.
How about a pic? worth a thousand words bla bla. Otherwise everything is as clear as mud 😡
Some idea of resistor values would be good too.
I have said previously in this thread that I will try to document my own work on this topology. The explanation given by "KTSR" is clear, however.
I have a "working" two-pole compensated implementation where the open loop rolls off at 40kHz and therefore maintains the output impedance over the audio bandwidth and falls thereafter. A first order compensation scheme requires earlier rolling off but still maintains a more than adequate output impedance to be labelled a current source over the audio bandwidth.
I have settled on 0.22-0.47R as the current sensing resistors using single MOSFETs - not that it tells you much.
I have 2 applications.
One is yet another implementation of my Powered Integrated Super Sub tech. The working conditions, speakers bla bla are well specified and I can ring whatever changes necessary to allow for different Ro and push that to unimportant frequency bands still within the audio bandwidth.
The 2nd is for a measurement amp .. eg to investigate the THD reducing properties of current drive. For this, I need a 'high' Ro over 20Hz - 20kHz. I thought this would be trivial at first but its not without a lot of extra complexity ... eg Mills & Hawkesford AES E-Library Distortion Reduction in Moving-Coil Loudspeaker Systems Using Current-Drive Technology
I'd be happy with >200R 20Hz to 20kHz.
If you have a working circuit, I'd appreciate even a rough schematic ... with all (most?) resistor values of course.
Last edited:
Soundbloke cares about HF compensation, and that is indeed the problem: Electronical current drive works by feeding back the load, in other words its feed-back loop includes the load. In order to do pretty perfect current drive, we need to feed back the load by a great extent. Any stray reactance of the load becomes a problem.
In the circuit i linked to earlier i decoupled the load using a pretty large series inductance, so i can use long speakers cables. Downside were, i cannot obtain current dominating voltage drive, but i do neither need nor want that dominance anyway.
If you want to do that, you need a smaller series inductance, short cables and an HF-compensated electro-magneto-acoustic transducer. The transducer becomes part of the system, forget about plug-n-play.
In the circuit i linked to earlier i decoupled the load using a pretty large series inductance, so i can use long speakers cables. Downside were, i cannot obtain current dominating voltage drive, but i do neither need nor want that dominance anyway.
If you want to do that, you need a smaller series inductance, short cables and an HF-compensated electro-magneto-acoustic transducer. The transducer becomes part of the system, forget about plug-n-play.
Last edited:
I'm not sure which circuit you linked too. Could you please link to it again for a senile old man? 🙁
Is this something you use on a daily basis ... or was it just an experiment?
Here for your convenience. I only simulated this one. Once upon a time i did current drive with a 10 Watts amplifier, a cheap 20cm fullrange driver and a sealed box, for subwoofer use. This raised Qtc from 0.8 to 2, but even without amplitude equalization it sounded good, because clang was much reduced.
Not particularly I don't - only when chasing ultimate and largely unnecessary performance measures.
It is only a problem in the way in which you have attempted it. My approach uses a transconductance amplifier/output stage and works inherently by avoiding feedback back anything to do with the load impedance. That is the basic premise of providing a high output impedance.
(This is only the case for current drive, however. For the motional feedback also discussed in this thread it is obviously not the case).
I am not sure when building from scratch why you would chose to use a conventional voltage amplifier as the basis? Or even why a load stabilising network would be required? A simple, resistively loaded transconductance amplifier seems such an obvious starting point.
Hi, the start of the URL, everything before diyaudio.com/..., is broken. I think, you can fix that for yourselves. Soundbloke, what is simple about a transconductance amplifier? There are small-signal transconductance ICs, and they have hi clang. To me, MOSFETs are disorder, while FETs such as BF245 are nice to have.
What is "clang"? I presume it is not a scientific term of any relevance?
As to MOSFETs, they have issues because like all engineering they represent a compromise. That compromise is minimal in this application and a 100mA FET is hardly relevant either.
Clang is harmonic distortion.
MOS in MOSFET abbreviates a manufacturing process but not an electronical characteristicum. The difference of FETs to bipolar transistor boils down to the control electrode being isolated from the set electrodes. Whereas small-signal FETs are handy for dealing with hi-impedance sources, MOSFETs confuse me. Most logic nowadays is done in MOSFETs, and i do not understand it.
MOS in MOSFET abbreviates a manufacturing process but not an electronical characteristicum. The difference of FETs to bipolar transistor boils down to the control electrode being isolated from the set electrodes. Whereas small-signal FETs are handy for dealing with hi-impedance sources, MOSFETs confuse me. Most logic nowadays is done in MOSFETs, and i do not understand it.
Last edited:
- Status
- Not open for further replies.