Ok, I played around with that first design and realized it wasn't very good at all. I played around some more in SwCAD before I had to go back to work, but since then I've come up with what I think could be a decent design. It's more complicated, but it is symmetric with a BJT driving stage and FETs at the output. If it works, I'd like to dub it the SLPUCD - symmetric, low-power UCD. When I have time tomorrow I'll work on figuring out some values for the driver stage and maybe try to tackle the feedback values too. I'm still a little confused as to what values would be a good starting point for the feedback components.
Please don't hesitate to post any tips, feedback, or suggestions about the circuit (whether you know what you're talking about or not!
).
I used the LT1818 opamp because it can source and sink up to 40mA of current, has a high slew rate, and can work off of split supply rails. If anyone knows of a good dedicated comparator that can handle a split supply, have a rail to rail that can source and sink current please let me know. Most of the LT ones advertise split supply and rail to rail, but the output only goes from ground to + supply... at least it does in SwCAD.
Almost forgot, these are the supply voltages:
V+=12V
V-=-12V
Vdd=5V
Vee=-5V
The 5V rails can easily be derived from the 12V rails or have their own isolated supply (the best option as the current draw should be very low).
Please don't hesitate to post any tips, feedback, or suggestions about the circuit (whether you know what you're talking about or not!
).I used the LT1818 opamp because it can source and sink up to 40mA of current, has a high slew rate, and can work off of split supply rails. If anyone knows of a good dedicated comparator that can handle a split supply, have a rail to rail that can source and sink current please let me know. Most of the LT ones advertise split supply and rail to rail, but the output only goes from ground to + supply... at least it does in SwCAD.
Almost forgot, these are the supply voltages:
V+=12V
V-=-12V
Vdd=5V
Vee=-5V
The 5V rails can easily be derived from the 12V rails or have their own isolated supply (the best option as the current draw should be very low).
Attachments
You both have valid points, thanks for the input. I have since read more into MOSFET driving schemes and have come up with a revised circuit. The driver part of the circuit simulates ok, but I can't seem to get the feedback values right to obtain a high switching frequency with the output filter (L=30uH, C=0.68uF) and load values (32ohm) I'm using. I'll post some part values later along with the SwCAD file so anyone can fiddle with it too.
Attachments
Low power right?
We are talking low power right? Why all the complication with the output? As suggested earlier a 4420 as the output stage would be more than adequate. This could be powered with split rails but may require a means of level shifting for the input signal. http://download.siliconexpert.com/pdfs/2005/08/10/semi_ap/1/mcl/mosfet driver/mic4420.pdf
The only thing that bothers me would be if the amp didn’t oscillate properly the output would be full rail and would destroy any phones connected to it. Since we are talking high freq at the output a small cap could be used to couple the output into the filter, 1uf film? This would give the DC protection we need and also insure no offset.
This implementation would be good for several watts of power so could be tested with small speakers. Various input schemes could be tried including the suggested lm319. 2 chips and a few small parts implementation looks very good and would be very friendly to battery portability.
Roger
We are talking low power right? Why all the complication with the output? As suggested earlier a 4420 as the output stage would be more than adequate. This could be powered with split rails but may require a means of level shifting for the input signal. http://download.siliconexpert.com/pdfs/2005/08/10/semi_ap/1/mcl/mosfet driver/mic4420.pdf
The only thing that bothers me would be if the amp didn’t oscillate properly the output would be full rail and would destroy any phones connected to it. Since we are talking high freq at the output a small cap could be used to couple the output into the filter, 1uf film? This would give the DC protection we need and also insure no offset.
This implementation would be good for several watts of power so could be tested with small speakers. Various input schemes could be tried including the suggested lm319. 2 chips and a few small parts implementation looks very good and would be very friendly to battery portability.
Roger
Sorry, but it's not just a high frequency. It's a high frequency modulated by a much lower frequency. The cap will need to be as large as it would in a conventional amp (A, AB, B).
BRWX.
What's your goal for this amplifier. Why class-d, what advantages do you think it has for a headphone amplifier.
Tim__x said:
The idea is to see if a class-d amp built for headphones has all the detail and fidelity that the class is known for with speakers.
My guess is that at these low powers class-d will not sound as good as a good, simple class-a or a/b amp.
But let's see if Brian can prove me wrong!
BWRX said:
I built a UCD180-AD amp specifically to power my AKG K1000 headphones after having such good luck with my Audiodigit 2020 amp. The Audiodigit was powered by SLA batteries. It was absolutely silent and had amazing detail.
The UCD is better. It has a much more relaxed upper end, way better bass, and there is no comparison in complex choral or orchestral passages. Of course the UCD costs 4 times what the Tripath 2020 did to build and has 180 watts instead of 8.
I tried my Sennheiser HD650's with both amps and was disappointed but mainly because the K1000s are so much better sounding to begin with.
AKG K1000s with the UCD are the best sound I have gotten so far as a primarily headphone user. My front end is "PC Audio" (Foobar and iTunes) to Scott Nixon USB DAC to Sowter TVA to Hypex UCD180-AD.
I have the T-amp for sale on EBAY right now and am keeping the UCD.
Hope this helps...
Cap question?
Tim,
At the output pin we would have a PW modulated full frequency signal that only would require a small cap. After the filter we would have mostly audio and this would require the large cap. The only problem is this cap must be quite a bit larger than the following filter cap or we would have problems with interfering with the filtering and more importantly the feedback.
Roger
Tim__x said:
Tim,
At the output pin we would have a PW modulated full frequency signal that only would require a small cap. After the filter we would have mostly audio and this would require the large cap. The only problem is this cap must be quite a bit larger than the following filter cap or we would have problems with interfering with the filtering and more importantly the feedback.
Roger
First of all, you shouldn't put the cap before the output filter with post-filter feedback because you'll lose all DC feedback.
Secondly, no a small cap won't work. The pre-filter output of a class D amplifier contains large low frequency components. If it didn't, ampliverters would be the easiest thing in the world to design.
I quite agree. My questions were rhetorical.
P.S. About the cap, think of it this way; if you remove the filter cap the amplifier will still work (if clocked or pre-filter) just with a first order filter, right? Now add the series cap, being in series, the order doesn't matter, switch the inductor and the cap so it goes inductor, cap, load. Still think you can use a small cap?
Secondly, no a small cap won't work. The pre-filter output of a class D amplifier contains large low frequency components. If it didn't, ampliverters would be the easiest thing in the world to design.
I quite agree. My questions were rhetorical.
P.S. About the cap, think of it this way; if you remove the filter cap the amplifier will still work (if clocked or pre-filter) just with a first order filter, right? Now add the series cap, being in series, the order doesn't matter, switch the inductor and the cap so it goes inductor, cap, load. Still think you can use a small cap?
I think it's near rediculous to simply assume a low power class d can't sound excellent. Is your sole basis for this assumption simply because you've never heard one, or because Tripath (hardly a robust design) was missapplied to it?
Pretty well any hearing aid these days is a class d amplifier, while they may not be high end, I really doubt it's nothing but switching noise they're listening to in them.
It just takes a few good design choices, and I wouldn't be so quick to lay doubt on the technology as a whole for low power useage until it's been explored in depth.
If you really think it may have problems, please enumerate your areas of concerns so that they may be investigated and if required addressed at the design level as required, or at least debated.
This whole thing reminds me of a few years ago when class a/b guys used to say that class d is only good for subwoofer use, look at em flocking now.
Pretty well any hearing aid these days is a class d amplifier, while they may not be high end, I really doubt it's nothing but switching noise they're listening to in them.
It just takes a few good design choices, and I wouldn't be so quick to lay doubt on the technology as a whole for low power useage until it's been explored in depth.
If you really think it may have problems, please enumerate your areas of concerns so that they may be investigated and if required addressed at the design level as required, or at least debated.
This whole thing reminds me of a few years ago when class a/b guys used to say that class d is only good for subwoofer use, look at em flocking now.
I have no doubt that class D amplifiers can sound excellent, and that at higher powers they can sound better than their class A/AB counterparts. I just don't think they can (I'm quite sure they can't) beat a class AB headphone amp. It's possible to make headphone amplifers with bandwidth into the Mhz and distortions of all kinds in the -140db region, class D can be good, but not that good (yet).
classd4sure said:
Who said that?
Hey classd4sure, I got nothing against Class-D amps, I build them!
I'm just thinking in practical terms, that's all. It seems to me that the advantages of Class-D really come into play when power goes up. I see the main Class-D advantage as not asking the power transistors to work in the typical linear mode. It's duty cycle that varies in the prefilter signal, not voltage. As we all know.
But when we need an amp to drive a load that requires very little power, mostly just milliwatts, it becomes a practical matter of power handling. It isn't too hard to design and build a Class A amp to supply that much power and do a damn good job of it. Given the same cost and complexity, Class-D may have a very hard time competing - at these low powers.
So I am not writing off the project at all, just stating that the competition is going to be stiff. I have posted my concerns about noise floor and other issues in previous posts.
OK gentlemen, thank you for those replies, it helped me understand your take on the matter.
Tim your point low power class d being competitive on such a high performance scale is a valide one. Class d thus far does excel in high power applications, but they also suffer to some extent when it comes to distortion at those higher powers. Alot of that is due to todays current component technology I think.
So it seems to me that at such lower power levels of concern, one could potentially push these components much further, and perhaps enough to more than compete with a class a/b counterpart at least.
"But when we need an amp to drive a load that requires very little power, mostly just milliwatts, it becomes a practical matter of power handling. It isn't too hard to design and build a Class A amp to supply that much power and do a damn good job of it. Given the same cost and complexity, Class-D may have a very hard time competing - at these low powers."
I'd agree with that as well. I think the noise floor issues and such are simply a matter of having a decent layout, this in practice would require more time, cost and complexity than a class A counterpart. So where's the advantage then right?
Assuming that at best we'd hopefully at least equal the sound quality of class A in this application, the only motivation for going to the extra trouble of class d would have to be increased battery life. By the way, does anyone know what the usual Ipod has for an amp?? I guess there's probably at least some small market for this.
Maybe a little off topic, maybe not, but I wonder how well anything powered by a lithium based battery would take to the pumping effect.
Anyway, I think a number of designs of the simplest possible nature should be evaluated for this purpose, the low power gives alot of design freedom we normally would not find, and greater care towards layout should be the area of concentration.
Cheers
Tim your point low power class d being competitive on such a high performance scale is a valide one. Class d thus far does excel in high power applications, but they also suffer to some extent when it comes to distortion at those higher powers. Alot of that is due to todays current component technology I think.
So it seems to me that at such lower power levels of concern, one could potentially push these components much further, and perhaps enough to more than compete with a class a/b counterpart at least.
"But when we need an amp to drive a load that requires very little power, mostly just milliwatts, it becomes a practical matter of power handling. It isn't too hard to design and build a Class A amp to supply that much power and do a damn good job of it. Given the same cost and complexity, Class-D may have a very hard time competing - at these low powers."
I'd agree with that as well. I think the noise floor issues and such are simply a matter of having a decent layout, this in practice would require more time, cost and complexity than a class A counterpart. So where's the advantage then right?
Assuming that at best we'd hopefully at least equal the sound quality of class A in this application, the only motivation for going to the extra trouble of class d would have to be increased battery life. By the way, does anyone know what the usual Ipod has for an amp?? I guess there's probably at least some small market for this.
Maybe a little off topic, maybe not, but I wonder how well anything powered by a lithium based battery would take to the pumping effect.
Anyway, I think a number of designs of the simplest possible nature should be evaluated for this purpose, the low power gives alot of design freedom we normally would not find, and greater care towards layout should be the area of concentration.
Cheers
Those were my thoughts as well.classd4sure said:
Exactly. To quote a smart man "Everything should be made as simple as possible, but not simpler." At low powers you could use complementary devices and other techniques, but that doesn't mean you can't do just as well with normal devices or layouts running well within their limits. A higher switching frequency was one thing I originally had in mind for the design and that means fast turn on and turn off of the output devices. I don't know exatly what could be achieved with a very simple circuit but I'm sure a slightly more complex scheme would still outperform something simple at low power. For example the latest schematic I posted has some provisions for quicker turn on and faster turn off of the output FETs.classd4sure said:
The thing about class d is that a lot of transient current capability is required for the best performance - even at low power levels. If you want simple you can just use a comparator or opamp with a lot of output current capability to feed the LC filter but you can surely do better by using some FETs and BJTs for the output and driving stages, respectively.
I expected a lot of cynicism but that doesn't really matter. I like this project because it should be rather cheap and simple and is a great way to become more familiar with the intricacies of class d (for myself and certainly anyone else who wishes to follow along).
Hi Brian,
Hence my thoughts on using small devices instead of the usual high power mosfets.
Remember the razer. That will better allow you to push the speed while keeping transient current draw as small as possible.
What you could use is a comparator with emitter followers, AC coupled to complimentary mosfet drivers.
But then there's aspects such as source follower output stage that is worth exploring dependant upon where you find the greatest problem areas are, such as transient current or freewheel current or deadtime ....?
Regards,
Chris
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.