I suspect I am going to smack myself in the face when someone provides the answer, but for now I am stuck. I want to simulate in LTspice the load an audio amplifier places on an unregulated PSU. For +ve signal current is drawn from the +ve side of the PSU and and for -ve signal values current is drawn from the -ve rail. I can't just place a current load across the rails as the current is returned to GND and I can't just place a sinusoidal current load from each rail to GND because only the positive (neg) part of the cycle is drawn from the +ve rail (-ve rail). 😱
Build a phony baloney output stage consisting of a pullup transistor (NPN? NchMOS?), a pulldown transistor (PNP? PchMOS?) and two ideal voltage sources, one in series with each base or gate, which emulate the "Bias Spreader". Fiddlefart around in .DC analysis, changing the voltage sources values, until you get the desired amount of DC bias current in the output transistors. Now switch to .TRANsient analysis and apply your music waveform to the bias spreader. Done.
Attachments
I was thinking it ought to be considerably simpler than modelling an amp. I merely want to model the sinusoidal current waveform 'split' between the rails (if that makes sense) but can't see how to split it. I'm not even so much worried about the DC bias draw as it likely rounding error for what I am currently looking at but that could be added. I'm most focused on checking cap ripple, RC resistor power dissipation etc.
EDIT: I've just seen Mark's post. Seems like a good 'halfway house'. Thanks. I will take a look.
EDIT: I've just seen Mark's post. Seems like a good 'halfway house'. Thanks. I will take a look.
And a behavioral model approach
If you're really look to concentrate on the power supply, you can use a behavioral model of the power amp current drain. Note that the model becomes a bit unrealistic if you make Vin large enough so that max(Vin)*R1>V1, but otherwise, it is a reasonable representation of the load current taken by a class B amp.
If you're really look to concentrate on the power supply, you can use a behavioral model of the power amp current drain. Note that the model becomes a bit unrealistic if you make Vin large enough so that max(Vin)*R1>V1, but otherwise, it is a reasonable representation of the load current taken by a class B amp.
Attachments
Good! The post#5 circuit uses 4 components (V3, B1, B2, R1 ... R2 is not strictly necessary) while the post#4 circuit uses 5 components (Vin, V1, V2, Q1, Q2, LOAD).
If you inserted the sinusoidal stimulus directly into the B1 and B2 equations, you could get it down to 3 components. Maybe somebody will discover how to get it down to 2??
If you inserted the sinusoidal stimulus directly into the B1 and B2 equations, you could get it down to 3 components. Maybe somebody will discover how to get it down to 2??
Last edited:
Hey Mooly, you don't happen to have one for a USB-powered DAC, do you? If you do I'd be interested 🙂
By the way, many thanks for that thread on using LTSpice, super helpful.
> checking cap ripple, RC resistor power dissipation etc.
Assuming ample rail caps: These are many-cycle loads. Just hang resistors on the power supply. A totem-pole output stage driving Sine into 8 ohms is down-to about 48 Ohm equivalent (two 24 Ohm for split supply).
Resistor dissipation will not get critical in a half an audio cycle.
Ripple voltage will be different if that rail is on or off, but your main question is the worst-case ripple.
Cap ripple current may be an issue. Full resistor load gives the worst-case. Actually current is higher on the on side than the off side, but again it won't blow-up in a half an audio cycle.
(If you fear 0.1Hz waves, use 12 Ohms one-rail for sine, 8r for square.)
In speech/music apps, signal induced ripple current will NOT be an issue. The average is far-far below max. The constant 100/120Hz ripple is the main thing. The way we now over-over-size main caps, this is usually not a problem. (Beware if you cleverly use very small-volume caps.)
A side-effect of this approach is that you can built the supply and bench-test, compare prediction to actual result, expose any typos or false assumptions.
Assuming ample rail caps: These are many-cycle loads. Just hang resistors on the power supply. A totem-pole output stage driving Sine into 8 ohms is down-to about 48 Ohm equivalent (two 24 Ohm for split supply).
Resistor dissipation will not get critical in a half an audio cycle.
Ripple voltage will be different if that rail is on or off, but your main question is the worst-case ripple.
Cap ripple current may be an issue. Full resistor load gives the worst-case. Actually current is higher on the on side than the off side, but again it won't blow-up in a half an audio cycle.
(If you fear 0.1Hz waves, use 12 Ohms one-rail for sine, 8r for square.)
In speech/music apps, signal induced ripple current will NOT be an issue. The average is far-far below max. The constant 100/120Hz ripple is the main thing. The way we now over-over-size main caps, this is usually not a problem. (Beware if you cleverly use very small-volume caps.)
A side-effect of this approach is that you can built the supply and bench-test, compare prediction to actual result, expose any typos or false assumptions.
Hi,
besides a approprately chosen resistor as load, a current source should suffice.
Either the simple "current" model from the LT-libraries of which You define its behaviour (also to find under "Load"), or the "g" model, a voltage controlled current source.
With "bi" You can design a arbitrary current source.
I´d probabely use a resistor that defines the idle current plus a paralleled current source for the .ac and .tran sims.
jauu
Calvin
besides a approprately chosen resistor as load, a current source should suffice.
Either the simple "current" model from the LT-libraries of which You define its behaviour (also to find under "Load"), or the "g" model, a voltage controlled current source.
With "bi" You can design a arbitrary current source.
I´d probabely use a resistor that defines the idle current plus a paralleled current source for the .ac and .tran sims.
jauu
Calvin
I think the attached works well - following Mark's suggestion. Now the challenge of finding caps with sufficient ripple rating. (Or perhaps in the example posted I need to shift the 0.47 ohm resistors one cap to the right.)
PRR/Dr Nick, the problem I had with a simple resistive load is that the current drawn is basically DC (as in a Class A operation). The load is massively overstated. Unless I am mistaken.
djoffe/Calvin, thanks for the suggestion regarding using a behavioural current source. I will study djoffe's example as I am unfamiliar with using this LTspice feature. (A simple current source did not work - see post 1.)
PS: Mooly, none of your links in post 1 of your tutorial (which I stumbled upon when searching for a solution to this) work if a user has set a number of posts per page which isn't the default. A common problem with this forum's software.
PRR/Dr Nick, the problem I had with a simple resistive load is that the current drawn is basically DC (as in a Class A operation). The load is massively overstated. Unless I am mistaken.
djoffe/Calvin, thanks for the suggestion regarding using a behavioural current source. I will study djoffe's example as I am unfamiliar with using this LTspice feature. (A simple current source did not work - see post 1.)
PS: Mooly, none of your links in post 1 of your tutorial (which I stumbled upon when searching for a solution to this) work if a user has set a number of posts per page which isn't the default. A common problem with this forum's software.
Attachments
Last edited:
@SGK
Take an look on LT4320, it is ideal bridge controller (replace Gretz with MOS-FETs).
With such device you can get few Volts more on output with same transformer and also get improvements regarding power dissipation and current ripple waveform trough rectifier/MOS-FET.
Take an look on LT4320, it is ideal bridge controller (replace Gretz with MOS-FETs).
With such device you can get few Volts more on output with same transformer and also get improvements regarding power dissipation and current ripple waveform trough rectifier/MOS-FET.
Mark, I read that to mean decoupling capacitors on the amp PCB itself (output stage) i.e. decouple to a "PWR_GND" and wire from amp PCB PWR_GND to star point. Also connect PSU PWR_GND to star point with separate wire. Signal GND kept separate from PWR_GND and wired separately to the star point.
The use of decoupling on the PSU itself was something I debated briefly with diyAudio member astx. I had feebly suggested that decoupling on the PSU did not help much. He argued you can never have too much decoupling (C5 and C6 in post 12) and that the 1uF caps were there to deliver peak current and short circuit HF ringing from outside the PSU. We also discussed the purpose of the Zobel something I had not seen in such PSUs before but have since. Are you suggesting any decoupling on the PSU should be separated yet again?
(PS: astx had designed these elements into his PSU also because he wanted the PSU to be suitable for powering Class D amps)
The use of decoupling on the PSU itself was something I debated briefly with diyAudio member astx. I had feebly suggested that decoupling on the PSU did not help much. He argued you can never have too much decoupling (C5 and C6 in post 12) and that the 1uF caps were there to deliver peak current and short circuit HF ringing from outside the PSU. We also discussed the purpose of the Zobel something I had not seen in such PSUs before but have since. Are you suggesting any decoupling on the PSU should be separated yet again?
(PS: astx had designed these elements into his PSU also because he wanted the PSU to be suitable for powering Class D amps)
Last edited:
Interesting product. Given the currents that might be involved it would require MOSFETs with very low Rds(on).
Thanks. I'm just looking at that but can't see an easy answer. Normally I take the page number out for a link. Just trying this for curiosity. One link in each pair has the -5 page ID.
This should go to post #85
http://www.diyaudio.com/forums/soft...spice-iv-beginner-advanced-5.html#post4047521
and so should this:
http://www.diyaudio.com/forums/soft...ltspice-iv-beginner-advanced.html#post4047521
and this one should
and so to this
Edit... those seem to work here and yet replacing a link in post #1 of the LT thread as a test doesn't work correctly. We know the forum software is creaking and groaning, one reason a move to a new platform has been proposed by Jason.
🙁 No answer for that at the moment.
Attachments
If the amplifier is on a different PCB than the power supply, there should be more than one "ground" wires between the amp PCB and the star ground. Have a look at Figure 45 (page 41) of the Akitika GT-101 assembly manual. It's a well thought out star-on-star arrangement. GT-101 is also a single-supply amplifier: VCC and GND but no VEE. Don't let that distract you.
They all took me to post 85.
Post 1 item 2 link:
http://www.diyaudio.com/forums/soft...ng-ltspice-beginner-advanced.html#post4025431
doesn't take me anywhere except post 1. Right click on relevant post number, copy link, paste and remove the page reference should work.
Post 1 item 2 link:
http://www.diyaudio.com/forums/soft...ng-ltspice-beginner-advanced.html#post4025431
doesn't take me anywhere except post 1. Right click on relevant post number, copy link, paste and remove the page reference should work.
- Status
- Not open for further replies.