| TroelsM |
Hi!
Having build an amp similar to the "amp30" at this page: http://listen.to/AUDIOEXPERIMENT I got curious and started looking for at way to use the very simple approach with higher power, only N-channel Fets and without ANY parts that are not common.
The result is shown in this post and there a few sim's are in the next post's.
It is only a model for simulation-purpose and therefore there is no bypass-caps and there is no input filter. The goal is to build this amp with normal leaded components and see if works. The best results be with SMD, but I want to keep it simple and see if a good PCB-design cant save som off the "damage" done by the leaded components.
The Fet's are to be IRFP264 as these seems to be very common and I have a lot of them. I dont have a PSpice model for them. Anyone?
R9, R10, L1, R14, L3 and R17 simulate the PCB. - Or something like it..
U1A and U1F drive the BD's and at the moment of switching they have to deliver a high peak-current to the BD bases. Therefore The inverters should probably be made up of 3-4 inverters in parralel.
Deadtime is taken care of by the RCD-networks: D1, R5, C2 and D5, R19, C5.
All comments are velcome. My Questions (for now...) are:
1) What do you think of the driver-curcuit? I want to use discrete componets to make the the amp cheap and easy to build without any "special" parts.
2) I would like to add feedback after the filter but I hav'nt found a way around the phase-shift-problems?
3) If this amp is to be build as a bridge then only one modulator should be used, but then feedback is only taken from one part of the bridge. I have used this approach before and it works, but some sort of feedback from both sides of the brigde would be better.
4) The LM311 does a level-shift from +/-12V to -48/-60V.. is this ok?
5) Q2, R2 and R8 does the levelshift to the high-side-driver. I'm told that this part off the amp is critical and any better solutions are welcome.
6) If the output is ringing there should be added snubbers. What about diodes across each of the IRFP264's? I have some RHRG30120 (30A, 1.2kV, "Hyperfast"), but BYW29 might work?
That all for now. Sims and comments will be posted soon.
TroelsM |
|
|
| TroelsM |
First: a better Png-version of the Schematic:
TroelsM
Note: I remember something about getting a better PSRR if there were a resistor from each power-rail to the feedback-point...? |
|
|
| TroelsM |
Seems to be fast and clean, but of course that's gonna get alot worse in the real world..
TroelsM |
|
|
| TroelsM |
Here's the first set of sim-result's:
The amp dos'nt osc at startup but it starts after ½ a period when the signal goes negative. I don't think this is a problem.
At high pos. voltage at the output the freq drops very much. Hav'nt figured out why. It's very late 4.08 in the morning). I´m not gonna work on that now.
TroelsM |
|
|
| classd4sure |
Hi,
Excellent effort.
My only comments at this time are, remove or lower R22 and R17 if you really want the current to flow to the BD's.
There's a few projects around which answer alot of your questions, dig up the "reference design" thread it's too full of good info to ignore.
You're drivers will hold a small amp together, should be fine for the experiments you're doing, that doesnt' mean that they won't need all kinds of tweaking though, parallel as many inverters as required to get the job done.
3. correct.
4. I highly doubt it, check the data sheet look into AC coupling or level shifting properly. Was smart to ask though, simulators aren't all telling.
5. It is. Opto or magnetic coupling are options, it'll depend if your current solution is OK or not.
6. Diodes across the fets aren't mean to cure ringing, they're meant to prevent the body diode from becomming saturated which will then cause some wicked shoot through/dissipation/failure. You really need the right diode for the job, schottky's.
Keep at it, this is great stuff.
Chris |
|
|
| TroelsM |
Hi. Spanks for the fast answer.
It's getting too late here, but one thing: Can you suggest any improvements to the drivers? If its not gonna ruin the whole keep-it-simple-and-cheap-thing, then any ideas would be welcome.
TroelsM |
|
|
| tlf9999 |
is this really that simple? it looks pretty complicated to me.
if you want to stay all discrete, you may want to look into TA-N88, a sony digital amp from years ago - it actually looks quite similar to your design.
a few questions - I am learning Class D myself as well:
a) why not use a gate driver, like ir2011 or the lm chip I mentioned in another thread? what is the advantage of having all discrete gate drivers?
b) why use the comparator for level shifting? wouldn't a simple transistor do that just as well? |
|
|
| classd4sure |
| quote: | Originally posted by TroelsM
Hi. Spanks for the fast answer.
It's getting too late here, but one thing: Can you suggest any improvements to the drivers? If its not gonna ruin the whole keep-it-simple-and-cheap-thing, then any ideas would be welcome.
TroelsM |
Nah I'd say start with that and see how it works. Simplest is best to start with and that seems fairly simple.
When you do it with discretes like that you're going to learn alot so keep at it. When you want better there's already circuits on here that will teach you even more, and meet all your criteria. Lucky you :)
Regards,
Chris |
|
|
| JohnW |
I've heard alot about the Sony TA-N88 Class D amplifier - does anybody have the circuit of this beast?
Thanks in advance,
John |
|
|
| phase_accurate |
I don't own the TA-88N schematic I only know its block diagram (after which I developed my first class-d amps). And I might find out the patent number again.
I also once used BD135/6 as drivers that were themselves driven by 4049 CMOS drivers.
This topology was actually quite fast but I started to experiment with the fist IR2110 as soon as it came out.
Specialised driver ICs are indeed simpler to use but the ubiquituous BD transistors are more easily obtained of course.
Regards
Charles |
|
|
| Pierre |
I am not sure at all about the level-shifter implemented inside the LM311...
Please have a look at another thread that is very active lately : "how to fix this sch". I proposed a simple topology for level-shift, gate drive and bootstrap. It hasn't been tested but should work better or worse. It is in fact the individual stages are very similar to your proposal, but it is simpler, and it uses a level-shifter based on a transistor after the PWM modulator.
Have you implemented that sch in real life?
Best regards,
Pierre |
|
|
| TroelsM |
I'm not going for an discrete solution but a cheap and easy one. The inverters are dirt-cheap and do a nice job at "cleaning" the signals.
The schematic hav'nt been implemented yet, but one of these days it will be.
The levelshifter from the LM311 will be fixed soon.
I'll have a look at some of the other threads (again) and return with the updates to the schematic
TroelsM |
|
|
| classd4sure |
| quote: | Originally posted by TroelsM
Here's the first set of sim-result's:
The amp dos'nt osc at startup but it starts after ½ a period when the signal goes negative. I don't think this is a problem.
At high pos. voltage at the output the freq drops very much. Hav'nt figured out why. It's very late 4.08 in the morning). I´m not gonna work on that now.
TroelsM |
You're right, it isn't a problem..... in simulation :)
I think you'll find it will be in reality. All it takes is a simple circuit to precharge it, and a slightly more complex one to provide a delay at turn on to ensure it has the time to fully charge.
Some level of frequency modulation is natural in most self oscillating amps but it shouldnt' drop too much. That's why it's good to switch at 500kHz instead of 320kHz. Then you dont' really care all that much if it drops down even 150kHz.
There may be ways of limiting it with tweaking feedback as well. I couldn't tell you how though. Maybe Charles can.
Regards,
Chris |
|
|
| classd4sure |
| Sorry, what's on page 5 exactly? |
|
|
| TroelsM |
Pierre: If this is the sch youre referring to (http://www.diyaudio.com/forums/atta...tamp=1129882065). Then I don't get it... have you simulated it? The level at the base, at top BD139/BD140's wont ever be higher than approx 5V and it should be min 56v+12v = 68V.
Next point: At the moment of switching the BD's have a very low current gain and the base-curent will be pretty high. In your sch the basedrive wont ever be very high becurse of the 1k resistors.
Thats why I use the 40106-inverters. They give a clean switch and a high "current-gain" ast high Freq.
Please do coment.
TroelsM |
|
|
| Pierre |
You are right! There is a mistake in that sch. Top of R2 should go to the bootstrap voltage, not to +12V.
No, I haven't simulated it, and I am afraid I don't have much time to do it now, if someone is willing to do it for me... |
|
|
| TroelsM |
If I get the time i might simulate it later, but I still think that the current to the bases is way too low to get at fast switch.
TroelsM |
|
|
| Pierre |
| You may be right about base currents also. I only posted the sch as a proposal. Adding a couple of TC427 driver ICs instead of the BD's should work much better, anyway. |
|
|
| nitrate |
Hello all,
Interesting thread, can't wait to see how it developes. What frequency are you planning on running this amp at? I can't see the totem pole driver arrangement going too fast, also why the multiple dead time? One on the fet gate, one in the logic?
You all seem to be using the same schematic/sim software. Which is it?
Keep up the good work!
Mad.P |
|
|
| classd4sure |
| quote: | Originally posted by classd4sure
Sorry, what's on page 5 exactly? |
Nevermind, it's the Sony amp you mentioned, thanks for posting it. It's very simple. |
|
|
| TroelsM |
The RD's on the gates are pretty standard. The "Xtra" deadtime is a lot easy'ere to trim to get very short deadtime.
Totem-pole? Please explain?
In the sim the freq is approx 250kz, but that is a function of the delay in the inverters and therefore the freq in the real-world sch could be a lot different.
I´m using PSpice.
TroelsM |
|
|
| nitrate |
| quote: | | Totem-pole? Please explain? |
Hi,
The totem pole i'm referring to is Q1/Q3 and Q4/Q5. They are arranged in a configuration known as 'totem-pole'. It is a standard way of driving at low frequencys and works well but i've never seen one working above 100KHz, the low switching times and crow bar current will cause havock. Talking of speed if you want more you will find changing the gates for TTL instead of Cmos will speed things up, the waveform of TTL gates are much sharper as well. Unfortunatly you must use them at a max of 5V, this complicates things a little.
Have fun,
Mad.P |
|
|
| TroelsM |
Aha. Got it. Why should it not work above 100Hz? I have used it in an amp running at 500Khz and the waveforms were pretty decent.
Any links/Pdf's on the problems you talk about? I would like to learn more about it. After sll its a somewhat important part of the Sch..
TroelsM |
|
|
| Kenshin |
1.You can't use the same 40106 chip for high side and low side...
2.Simple design, probably I would begin to use N channel since this.
What's the actual ton/toff of high side?
3.why so many filters in the japanese amp TA-N88 ? |
|
|
| phase_accurate |
| quote: | | 3.why so many filters in the japanese amp TA-N88 ? |
Years ago everybody thought that even the tiniest bit of ripple had to be "ironed out" in order to reach audio nirvana.
Only within the last 10 years there was a paradigm change in that reasonable amaounts of ripple aren't regarded as disturbing as a filter of high order itself.
Regards
Charles |
|
|
| nitrate |
Greet's
I don't know if anyone has ever written a paper about the pro's and con's of totem pole drivers, I'm sure somewere somebody has LOL. I'm just going off my own experience in trying to use them in these kinds of amps. Alos i have not seen any practical designs that use them, pity really as they are so simple:bawling:
Mad.P |
|
|
| TroelsM |
Nitrate: What kind of problems are you referring to? If you could explain which transistors you have used and what the problems were maybe we could find and/or solve some of the problems?
I'm guessing that a lot of discrete drivers have failed because of very poor PCB-layouts with dozen of nH trace-inductance and slow transistors.
Kenshin: No, you're absolutely right. I´m going to use a separate 40106 for the highside. The Sch is only a sim-model. Some of the inverters might not be used and I hav'nt figured out how many 40106 I'll need in total.
I would like some more input on the post-filter-feedback but maybe I'll build the powerstage and examine the switching of the driver and Mosfet. If those parts are working well then I guess that a LOT of people will be very interested in the design.
TroelsM |
|
|
| TroelsM |
I Have siulated the performance of the totempole-driver but i cant find any serious faults with it. I tried to ad a lot of parasitic inductance but it only made things a wee bit worse.
-But anyhow i decided to try another driver-curcuit in the simulator. Two different schematics are added to this post. The top-schem is a little faster at turn-on because off the "kick/boost" from the C1, R2-connection around Q2. The timeconstant is so small that C1/R2 are only conducting the first 100nS of the switching action but the give a good kick to turn Q1 on faster.
C1/R2 does the "AC" switching and Q2 does the DC-switching...?
Both Curcuits seems to work fine, but they will be harder to implement with leaded components as there are more componets in the critical loop driving the gate.
Comments? Ideas?
Sim-results will be in the next post.
TroelsM |
|
|
| TroelsM |
Here's the sim-results. The sim's are done without any parasitic/PCB-inductance.
M1 are clearly turned on faster and the loss (area under curve) is therefore smaller. Not a big difference but it's there.
The inverters in the sch might have to be made up off 2-4 inverters in parralel.
TroelsM
EDIT: just tried the totempole-driver from my original-schematic. It gives the same switch-times and losses as the best of the two "new ones". The Totem-pole is, as i mentioned before, a lot simpler to implement though.
I would like to try both in the real world, but I´m not shure when I'll get near a scope' again..
TroelsM |
|
|
| JohnW |
TLF9999,
Thanks for posting the Schematic of the Sony,
John |
|
|
|
