Class D amp using TL494

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
snip Heinz :
I think it will work BUT only if you have c o n s t a n t powersupplys.
If not, you need a 2. divider from ps-midpoint to get a new reference!
snip

Hi ledmania,

Now I see your supplyvoltage is not
2X 25V but it`s 2X50V. this is even worse without:

"a 2. divider from ps-midpoint to get a new reference!"

in your circuit (post1) you have without inputsignal constant 25V at the topside of the speaker, OK!

Which voltage is on the bottom side?

When I interpret the complete circuit (You don`t draw it!) is on the bottom side of the speaker the mid-point of the
powersupply (- 50V / + 50V).

-50V is also the ground for your 2V reference.

In this case the speaker will see a offset of 25 V!


Regards
Heinz!
 
powerbecker said:
snip Heinz :
I think it will work BUT only if you have c o n s t a n t powersupplys.
If not, you need a 2. divider from ps-midpoint to get a new reference!
snip

Hi ledmania,

Now I see your supplyvoltage is not
2X 25V but it`s 2X50V. this is even worse without:

"a 2. divider from ps-midpoint to get a new reference!"

Hi Hienz,

For easy viewing of the power supply, just imagine it that you have a two pcs. of battery with 25V each connected in serries, then imagine that you put your voltmeter across the batts with red probe on the + side, and black probe on the - side...Of coarse you will get +50V right? But if you change the polarity of the probe across the batts, you'll get -50V....How about putting the black probe at the -50V then the red probe is put at the center points of the batts, here you'll get +25V..This is exactly what my power supply is.
But I can't figure out what is that mid point reference that you want me to include for:rolleyes:


powerbecker said:
in your circuit (post1) you have without inputsignal constant 25V at the topside of the speaker, OK!

Which voltage is on the bottom side?

When I interpret the complete circuit (You don`t draw it!) is on the bottom side of the speaker the mid-point of the
powersupply (- 50V / + 50V).

-50V is also the ground for your 2V reference.

In this case the speaker will see a offset of 25 V!


Regards
Heinz!

Hienz, There will be no offset of 25V that will occur across the speaker because the speaker is referenced to the mid point of the supply that we call neutral.

Regards,
ledmania
 
Ledmania,
Now I think I understand you: you were using a wrong and confusing naming for the rails in the schematics: if you put -50V, everybody understand -50V referenced to GND, and +50 means +50V referenced to GND so we all assume you have +/-50V supplies.

So following your explanation, you have +/-25V supplies. Then it is true that you have 2V offset in the NFB network as you wished, but the problem of the dependency on the supply rails voltage is still there, and believe me: supplies for amplifiers are not so estable at all. Even if you plug it in different homes, you will get different voltages and that will create an important offset at the speaker.
 
snip Heinz :
I think it will work BUT only if you have c o n s t a n t powersupplys.
If not, you need a 2. divider from ps-midpoint to get a new reference!

snip Pierre
but the problem of the dependency on the supply rails voltage is still there, and believe me: supplies for amplifiers are not so estable at all. Even if you plug it in different homes, you will get different voltages and that will create an important offset at the speaker.

Hi ledmania

a picture says more than 1000 words (when it is complete and correct).
I believe Pierre and I think right.
The problem (You see it?) was solved with a new reference.
For clarity I "draw" this :

Regards
Heinz!
 

Attachments

  • better1.jpg
    better1.jpg
    66.9 KB · Views: 3,181
Assuming ±25 volt rails and the speaker ground and the chassis ground (center tap of the transformer) are at the mid-point (0 volts)...

Why not run the 494 with its ground pin (7) at 0 volts? Then you could use a good quality op-amp (5532 driven from ±15 volts) to drive pin 3 of the 494 (with proper protection for the input). Then the audio input and feedback would operate essentially normally with no references other than ground (0v). The 494 could drive an opto-coupler to drive the 2010.

I don't know if you've tried this circuit yet but you may need to take feedback from the IC side of the inductor.

Also, the 5532 probably won't work for over-current protection because it can't swing it's output close enough to it's negative rail input. An LM358 would work, I believe.
 
Pierre said:
Ledmania,
Now I think I understand you: you were using a wrong and confusing naming for the rails in the schematics: if you put -50V, everybody understand -50V referenced to GND, and +50 means +50V referenced to GND so we all assume you have +/-50V supplies.

LOL!!:D..At last, you understant it..Well, thats nice :)
But if you have experience in repairing TV/VHS using japanese or
American diagrams then my sch. would be "simple" to you.
I finally realized my mistake in over estimating the comprehention level of the viewers of the sch and should removed that anoying -50V, then should exaggerate the appearance of that pathetic arrow neutral that brought lots of controversy:clown:
Oh well..


.
Pierre said:
So following your explanation, you have +/-25V supplies. Then it is true that you have 2V offset in the NFB network as you wished, but the problem of the dependency on the supply rails voltage is still there, and believe me: supplies for amplifiers are not so estable at all. Even if you plug it in different homes, you will get different voltages and that will create an important offset at the speaker.


Nice point pierre, In fact I anticipated that problem before I design the circuit because of the nature of the current involved on it.So, this beast will be same as to those class D amps that will surely drain your wallet in putting lots of filter caps on PS.

On the other hand, for those who want to try in building this amp with tight budget, then you can tweak the circuit very easy by substituting low power fets on the o/p then crankdown the PS to 24V. So you will get here now 12V at the neutral. Nothing to change on the circuit except recalculating the current sense resistor and R15 and R14 again with 2V Vdrop.

You will also notice that the current sense on the 4 parallel resistors at the source of the fet was calculated with a tripping point of 10amps which is maybe too high for the beloved fet to endure. Based on the data sheet it has 16amps of max current, so maybe I have to reduce it to something like 8? or maybe 5 if you love your fet very much:clown:.

I think the current sense is reliable enough to shutdown the fet driver in safe mode wether the short circuit is accidental or "intentional" on the o/p side.

The setting of the current sense resistors is an arbitrary choice for the builders of this amps depending on their target output power. the calculation is simply ohms law.

Regards
ledmania
 
Perry Babin said:
Assuming ±25 volt rails and the speaker ground and the chassis ground (center tap of the transformer) are at the mid-point (0 volts)...

Why not run the 494 with its ground pin (7) at 0 volts? Then you could use a good quality op-amp (5532 driven from ±15 volts) to drive pin 3 of the 494 (with proper protection for the input). Then the audio input and feedback would operate essentially normally with no references other than ground (0v). The 494 could drive an opto-coupler to drive the 2010.

I don't know if you've tried this circuit yet but you may need to take feedback from the IC side of the inductor.

Also, the 5532 probably won't work for over-current protection because it can't swing it's output close enough to it's negative rail input. An LM358 would work, I believe.

Perry,
Your suggestion on paragraph 2 of your post will simply jeopardize the 700uA pulldown current at pin 3 of Tl494. Please go back to your PDF on that part..It says the opamp is "only active high" and rely only on that 700uA as pulldown.
Also the use of ne5532's o/p to drive pin 3 with +/-15V will surely hurt pin 3 b'coz the opamp was powered only by 5Vref(pin 14).
And the use of optocoupler, well its another story.

Your comment regarding the use of ne5532 as the current sense amp, and you are concern that it will not work because of the low sink capabilty? well, please look at the PDF of IR2011 and notice their i/p. You will see that it has "schmitt trigger" i/p with sharp hysterises response. this means that even ne5532 cant switch digitally, then schmitt trigger will take care of it..

Regards
ledmania
 
Hi guys,

The sch that you have seen on my 1st post, then another correction was made with the 2nd and 3rd sch, regarding the NE5532. You will notice that the original supply was 12V. But then was change to 5Vref to match the supply of Tl494 error amp supply. I think ne5532 will not work on that supply level. So, that opamp must be substituted with another opamp qualified to run @ 5V.

Regards,
ledmania
 
powerbecker said:
Hi ledmania

a picture says more than 1000 words (when it is complete and correct).
I believe Pierre and I think right.
The problem (You see it?) was solved with a new reference.
For clarity I "draw" this :

Regards
Heinz!

Hi Heinz,

LOL!!!:D. I think I have to congratulate you for this! Your proposed new reference makes a lot of sense to me!.......Lets say the PS varies,then of coarse the 2V on that node will also vary therefore the + side i/p of the opamp will copy it!!! Bingo!!

Good job Heinz!!!
I will add your proposal into the final schematic...
Regards
ledmania
 
Hi guys,

Here is the complete sch of the circuit that was added with a new reference node as suggested by Heinz.(thanks Heinz)

I have no time to look for qualified opamp that will run beter @ 5V to replace NE5532 as over current protector. So, you still see it on the sch.

LO and behold.....
An externally hosted image should be here but it was not working when we last tested it.


Regards
ledmania
 
That looks better.
Perry.
Using optocouplers to drive the IR2010 would be very slow and will also ruin the main aim of this design: simplicity with decent performance.

About the feedback network, I still think that the feedback should go to the negative input of the error opamp, and adjust the output polarities (swap them) if necessary. Error amps in Class-D always use the negative input for feedback and input. Have you checked this? How do you calculate the gain of the whole circuit as it is now?

Do you have the intention of building it in real world soon and test its performance, Ledmania?
 
I know of two commercially successful designs that use optocouplers. One uses them to drive the outputs. The other uses them to drive the driver IC. Both designs have been in use for years. They run at ~90kHz. Unless he's running at a significantly higher frequency, the optocouplers would be fast enough.

Of course, you're correct in noting that it will increase complexity.
 
Thank`s ledmania!

snip Pierre
About the feedback network, I still think that the feedback should go to the negative input of the error opamp, and adjust the output polarities (swap them) if necessary. Error amps in Class-D always use the negative input for feedback and input. Have you checked this?

I "feel" the same.
Also I doubt if You will get sufficiently bandwidth when feedback includes the outputfilter. May be you better should take FB before the filter.
For speakerimpedances much higher then filter-Z you should add a R/C-damper in parallel to the speaker.
How high is the switching frequency?

Regards
Heinz!
 
Pierre said:
That looks better.
Perry.
Using optocouplers to drive the IR2010 would be very slow and will also ruin the main aim of this design: simplicity with decent performance.

About the feedback network, I still think that the feedback should go to the negative input of the error opamp, and adjust the output polarities (swap them) if necessary. Error amps in Class-D always use the negative input for feedback and input. Have you checked this? How do you calculate the gain of the whole circuit as it is now?

Do you have the intention of building it in real world soon and test its performance, Ledmania?


Thanks for liking it, Pierre.

On the 2nd paragraph of you post, well,, when I am designing this circuit about a week ago, I also decided to put that feedback on the negative side of the i/p ..But when I realized that there are several "strings" of phase invertion within the loop, I simply changed my mind and favors that "eye brow raising connection".
Its really funny at first sight, but if you follow the strings one by one, then you will smile and say,, yes, "theres a reason to do that".

About your question regarding the gain??LOL!!:D now I'm still calculating it LOL!!:D ...As far as my initial calculation, the 1st gain of the opamp is based on conventional opamp gain calculation which is.........Vgain=Rfb/Rin =10...so, the sensitivity is 200mv and the o/p is 2V ..Peak to peak of this voltage varries the o/p to +/- 50%duty cycle giving maximum swing to +/- 25V...Well, theoretically speaking, and neglecting switch losses and duty cycle imperpection, then this is not the practical way because we are dealing with pulsed signal o/p at the moment...But on the lenear side "after the filter" then I think I'll have to analyze very carefully because I am avoiding "assasination" by some intelectuals by giving wrong math.LOL!LOL!!LOL!!..:D :D

ledmania said:


Pierre,
Yes you are looking at my post #1..
Please pay attention to the detail of the sch and contemplate on it carefully because there are some flaw on it but I'm not going to exposed it yet because I'm waiting for more adept guys that will notice it.
I think that my kirchhoffs voltage law know how is just working fine so the voltage drops on each identified nodes are just ok.

Regards
ledmania

When I'm done in designing the circuit, and finally hit the button "send" to class D forum,,, Oh! my! I've noticed that flaw on the input of the opamp which will cause some adept guys to react.
I'm waiting for someone to suggest something on that part and "lo and behold" an adept guy from Berlin hit the bulls eye!! Bravo! Heinz!! Bravo!!..

Regards
ledmania
 
Ledmania,
If you inject feedback through the - input you will get, globally, an inverting amplifier, whose gain is Rfb/Rin (where Rin is the resistor in series with the input source). The only thing to revise then is the polarity of the PWM signals, but that supposes no added cost or complexity.

About taking feedback from the speaker output, I don't recommend that. Having made a lot of prototypes with both pre and post filter feedback, the added complexity of the compensation network, reduced phase margin, etc, of the post-filter version doesn't worth the pain IMHO. Have in mind that the filter will produce a 180º phase shift that you must be able to compensate, and believe me, that's not easily done properly.
For this application, you won't get better performance by taking feedback after the filter. Believe me: take it from the mosfets output. At most, you will need a low pass RC network to filter out some of the square wave before passing it to the error amplifier.

Best regards,
Pierre
 
Pierre said:
Ledmania,
If you inject feedback through the - input you will get, globally, an inverting amplifier, whose gain is Rfb/Rin (where Rin is the resistor in series with the input source). The only thing to revise then is the polarity of the PWM signals, but that supposes no added cost or complexity.


Pierre,
You are correct on your statement above.
I am thinking to revise the circuit and use the - i/p instead, then I will put down the hi side inverter to lo side to compensate for the change in phase invertion, then transfer the Rfb and Rin to - i/p then presto!!!.

The original circuit is indeed a genuine negative feedback loop amp that also obey on a standard Vgain equation of Rfb/Rin and is intended to complement on your "qouted conclusion" on the 1st page of my post.. But I also realized that it is intended also for students or beginers alike that perhaps those beginers may bring cofusion to that "seemingly" wrong looking i/p... .
But now, I think I'll have to follow their hearts content;)
And also to yours..;)


Pierre said:
About taking feedback from the speaker output, I don't recommend that. Having made a lot of prototypes with both pre and post filter feedback, the added complexity of the compensation network, reduced phase margin, etc, of the post-filter version doesn't worth the pain IMHO. Have in mind that the filter will produce a 180º phase shift that you must be able to compensate, and believe me, that's not easily done properly.
For this application, you won't get better performance by taking feedback after the filter. Believe me: take it from the mosfets output. At most, you will need a low pass RC network to filter out some of the square wave before passing it to the error amplifier.

Best regards,
Pierre

Pierre,
Everybody are entitled to raise their own opinion, but it so happen that you and me have different opinion on this part. I dont like to argue with you on this part and I will leave the decision to the person who want to assemble a class D amp using my circuit.
All I can say is that my intuition and IQ tells me to put the NFB "after the filter". Please take a look at the new schematic and notice that there are two leg of voltage divider that sandwiched the speaker...
this two leg is composed of R14 and R15 on the left and R20 and R21 on the right.."Note" that the 2V on this two nodes always"seeks" into equilibrum due to NFB to make sure that there is always "0" offset voltage accross the speaker...Can you imagine if we transfer the left leg before the filter and will sandwich that 200Khz noise there???..Please think of it:)
If that cute little opamp knows how to complain then it will shout "Pierre!!!! give me a break!!!!"...LOL!!LOL!!:D :D :D


Regards
ledmania
 
ledmania said:



Pierre,
You are correct on your statement above.
I am thinking to revise the circuit and use the - i/p instead, then I will put down the hi side inverter to lo side to compensate for the change in phase invertion, then transfer the Rfb and Rin to - i/p then presto!!!.

The original circuit is indeed a genuine negative feedback loop amp that also obey on a standard Vgain equation of Rfb/Rin and is intended to complement on your "qouted conclusion" on the 1st page of my post.. But I also realized that it is intended also for students or beginers alike that perhaps those beginers may bring cofusion to that "seemingly" wrong looking i/p... .
But now, I think I'll have to follow their hearts content;)
And also to yours..;)

Regards
ledmania


So, Here is the final revised schematic that we are talking about....

Lo and behold!!...............

An externally hosted image should be here but it was not working when we last tested it.
 
If you put a low pass RC filter after the voltage divider you shouldn't have any problems while taking feedback after the output IMHO.
But anyway it is easy to move to one form of feedback to another, so both possibilities can be explored.
Compensating an after filter feedback is not an easy task, be prepared to review your feedback and estability knowledges as I did one year ago when I tried to do so (I succeeded by the way, but I don't think it worths the effort, anyway).
 
Pierre said:
If you put a low pass RC filter after the voltage divider you shouldn't have any problems while taking feedback after the output IMHO.
But anyway it is easy to move to one form of feedback to another, so both possibilities can be explored.
Compensating an after filter feedback is not an easy task, be prepared to review your feedback and estability knowledges as I did one year ago when I tried to do so (I succeeded by the way, but I don't think it worths the effort, anyway).


I also want to add, that putting the NFB before the filter will oblige
you to put several strings of RC lo pass filter to get better noise reduction in which the RC filter have low Q giving shallow roll off. While comparing to LC filter which is proven to have much higher Q if thicker wire is used will give sharper knee of roll off, and of coarse it is already free and available already across the speaker.

By the way, the 2V on that two leg of the devider is not really 2V as seen by the two i/p's of the opamp but the voltage difference is the actual voltage as seen across it. This is in the millivolt region, typicaly less than 15mv. So, mixing very small noise accross this region will also amplify by the entire amp.

Well of coarse you are correct in exploring the right location of that NFB. Only the real builder of this beast can really prove "which is which"

Regards
ledmania
 
Hello Ledmania & others,

Have you guys considered the drawbacks occuring with BUS PUMPING when this type of non Zero Referenced Feedback type is used....This type would Certainly generate lot of DC error at the output During Bus Pumping and when the amp is outputing low frequency BASS signals.....

Consider it, during Bus -Pumping....

K a n w a r
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.