I did not realize that 330pF and 22R was a Zobel to level the impedance between the two amp legs. How does that work? Anyhow, I hope me adding a 330pF and 10R to ground on each leg before the L doesn't mess things up. It seems to work as it sounds good and the harshness that was present before is gone.
What is the switching freq (where in the datasheet and is it selectable like 3116?).
I did an A/B comparison again last night with my current TPA3116D2 setup running same mods and I still think the TDA7492 sounds more dynamic with crisper and punchier bass. Hard to believe that such a cheap amp (with worse HD figures) is challenging my favorite TPA3116D2 Ybdz that I have used all this time. It seems that the TDA7492 is adding about 1.5dB on the highs above 2kHz compared to the 3116 which was more neutral. Not sure how real this is but I applied an EQ shelf to cut out and it sounds better and less fatiguing. The bass seems to be slightly higher in HD by maybe 4 to 5 dB more - where more power is being drawn.
Switching frequency on page 10 of data sheet. 400k Max, think typical was 310k. Didn't read as though selectable, but mentioned both internal and external oscillator variants. It means they recommend output LP filter frequency of 27k.
Its a shame the version I am viewing doesn't include a parts breakdown referring to the circuit schematic as the revision of the 3116 data sheet did.
Its a shame the version I am viewing doesn't include a parts breakdown referring to the circuit schematic as the revision of the 3116 data sheet did.
Well, if one compares the distortion of the amp to our speakers, I do not think we have to worry too much about it. Talking about amp distortion, many tube amps are way worse but still sound fine.
My 2 cents.
Been messing around with the 3116 board tonight by adding the pre-LC zobel as with the 7492. Got a good hour of comparing, albeit low volume because kids in bed.
Here's what I had set up....
3116 board running off 20V 5A laptop brick, signal fed from a headphone socket of Samsung galaxy 3. The amp end the headphone wires feeds through 2+ and 2- wires each having a 470nf nichicon passing to the pins of the input caps underneath, basically bypassing the stock input caps. The board is still a pretty stock YJ black/blue. Only mods so far are heatsink removed, DC decoupling caps are now Lelon 470uF RZW, and the 1nF smd/t replaced with a leaded Philips version. Both channels wired to 8inch full range woofers direct without xo. These are paper cone woofers from a 1978 Advance VII speaker, which I recently refoamed. Just to note the speaker wires were about 8ft cheapo diy store stuff.
What I did was to add the zobel to 1 channel only, enabling a side by side comparison without swapping and changing.
As I said, this wasn't to talk about the 3116, but more what I found with the mod.
All round the zobel improved sound. I am not sure it made the bass deeper, or louder, but it definitely became more precise. I could clearly here the strike of the bass without it sounding punching, whereas the other channel lost the strike a little to the deep bass. The bass in voices even became clearer, which to some would sound as if the bass was lost to highs, but the bass was still there as deep as ever, just less muddled together. With the highs, I didn't notice excessive highs. Only managed to listen to voices on WiFi internet radio during chat segments, and yes I noticed a tiny bit of sibilence coming through, but this was a radio station talk show, I could also clearly hear intake of breaths as well.
Music wise, I listened to jazz, classical, dance music and heavy metal. All sounded cleaner, clearer, deeper, higher, and had better separation of the different types of sounds/instruments all round. One of my downloads I like to test with is the instrumental of the Clean Bandit tune, think its the one with Jess Stone in vocal version, those in UK would recognise it as the Sainsburys TV advert music. The difference was very subtle, but when I placed ear about 12inches from cone I could clearly hear these experiences....remember I wasn't listening at a particularly high volume at all. Unfortunately, I wasn't able to crank up the volume to get a feel as to whether the soundstage would have shifted slightly towards the pre-LC zobel channel. Hopefully I'll get the chance with speakers outside on a sunny day this week.
I don't believe I was hearing an inaccurate representation of so many different types of music, nor of the human voices discussing the bands musical inspirations. Everything became more open, less muddled, sounds more separated.
All in all for my final system design the mod fits in nicely as the 1inch tweeter which will accompany the woofer on an OB is going to be 24-35db down by the time Its xo slope hits vocal ranges.
Just my two penneth on the effect of the pre-LC zobel as on the TDA7492 schematic when used as a mod on the TPA3116.
Here's what I had set up....
3116 board running off 20V 5A laptop brick, signal fed from a headphone socket of Samsung galaxy 3. The amp end the headphone wires feeds through 2+ and 2- wires each having a 470nf nichicon passing to the pins of the input caps underneath, basically bypassing the stock input caps. The board is still a pretty stock YJ black/blue. Only mods so far are heatsink removed, DC decoupling caps are now Lelon 470uF RZW, and the 1nF smd/t replaced with a leaded Philips version. Both channels wired to 8inch full range woofers direct without xo. These are paper cone woofers from a 1978 Advance VII speaker, which I recently refoamed. Just to note the speaker wires were about 8ft cheapo diy store stuff.
What I did was to add the zobel to 1 channel only, enabling a side by side comparison without swapping and changing.
As I said, this wasn't to talk about the 3116, but more what I found with the mod.
All round the zobel improved sound. I am not sure it made the bass deeper, or louder, but it definitely became more precise. I could clearly here the strike of the bass without it sounding punching, whereas the other channel lost the strike a little to the deep bass. The bass in voices even became clearer, which to some would sound as if the bass was lost to highs, but the bass was still there as deep as ever, just less muddled together. With the highs, I didn't notice excessive highs. Only managed to listen to voices on WiFi internet radio during chat segments, and yes I noticed a tiny bit of sibilence coming through, but this was a radio station talk show, I could also clearly hear intake of breaths as well.
Music wise, I listened to jazz, classical, dance music and heavy metal. All sounded cleaner, clearer, deeper, higher, and had better separation of the different types of sounds/instruments all round. One of my downloads I like to test with is the instrumental of the Clean Bandit tune, think its the one with Jess Stone in vocal version, those in UK would recognise it as the Sainsburys TV advert music. The difference was very subtle, but when I placed ear about 12inches from cone I could clearly hear these experiences....remember I wasn't listening at a particularly high volume at all. Unfortunately, I wasn't able to crank up the volume to get a feel as to whether the soundstage would have shifted slightly towards the pre-LC zobel channel. Hopefully I'll get the chance with speakers outside on a sunny day this week.
I don't believe I was hearing an inaccurate representation of so many different types of music, nor of the human voices discussing the bands musical inspirations. Everything became more open, less muddled, sounds more separated.
All in all for my final system design the mod fits in nicely as the 1inch tweeter which will accompany the woofer on an OB is going to be 24-35db down by the time Its xo slope hits vocal ranges.
Just my two penneth on the effect of the pre-LC zobel as on the TDA7492 schematic when used as a mod on the TPA3116.
Xrk suggested I post photo of mod mock up. Will do tomorrow. For those interested in pre-LC Zobel i.e. Boucherot cell, here's a few links to info.......
Zobel network - Wikipedia, the free encyclopedia
Boucherot cell - Wikipedia, the free encyclopedia
http://www.embedded.com/print/4015876
This is a circuit common in power amps. It is also one which I first read about placing across speaker terminals to have same effect on resistance of speaker that the LC /amp sees. I am sure its been mentioned for having in place when using ferrite beads instead of the output circuit of 3116.
Til tomorrow.
Zobel network - Wikipedia, the free encyclopedia
Boucherot cell - Wikipedia, the free encyclopedia
http://www.embedded.com/print/4015876
This is a circuit common in power amps. It is also one which I first read about placing across speaker terminals to have same effect on resistance of speaker that the LC /amp sees. I am sure its been mentioned for having in place when using ferrite beads instead of the output circuit of 3116.
Til tomorrow.
Maybe I'm mistaken but wouldn't the benefit of a Zobel network be mostly related to the speakers more than the amp?
Some may think so. but they are used before LC in power amps. the zobel, boucheret cell, does two things. it stops the coil storing energy through magnetic flux, like a battery through its magnetic flux, it does this in eithe direction.
it also makes the current follow the voltage slope, rather than the other way around. qhich in a voltage output should be an advantage.
So, below here is a PIC of my pre-lc zobel / Boucherot cell fitted to the 3116 as already mentioned.
Please, please read about this circuit before trying it. The bottom of the 3 links I posted clearly states that the values of resistor/cap HAVE TO BE CALCULATED according to the lc after it.
I am no expert in either circuits, or with the 3116 IC. If I destroy my board that's my own stupid novice fault. If you follow me blindly I am not to blame. Don't be a lemming.
(Grrrrrr...my tablet isn't attaching pics.... I'll leave the text above and go to office to use PC and attach the PIC by editing this.....won't be 5mins).
Please, please read about this circuit before trying it. The bottom of the 3 links I posted clearly states that the values of resistor/cap HAVE TO BE CALCULATED according to the lc after it.
I am no expert in either circuits, or with the 3116 IC. If I destroy my board that's my own stupid novice fault. If you follow me blindly I am not to blame. Don't be a lemming.
(Grrrrrr...my tablet isn't attaching pics.... I'll leave the text above and go to office to use PC and attach the PIC by editing this.....won't be 5mins).
So, below here is a PIC of my pre-lc zobel / Boucherot cell fitted to the 3116 as already mentioned.
Please, please read about this circuit before trying it. I am no expert in either circuits, or with the 3116 IC. If I destroy my board that's my own stupid novice fault. If you follow me blindly I am not to blame. Don't be a lemming.
Don't worry about the values I used. refer to the following info...
http://www.embedded.com/print/4015876
As I was only testing for 1 hour max at very low volume levels I just found a resitor and cap which fitted into what author says most manufacturers do as they are lazy or incompetent. seemed to work ok with no ill effects, but needs optimizing according to the equations within the above link.
Please, please read about this circuit before trying it. I am no expert in either circuits, or with the 3116 IC. If I destroy my board that's my own stupid novice fault. If you follow me blindly I am not to blame. Don't be a lemming.
Don't worry about the values I used. refer to the following info...
http://www.embedded.com/print/4015876
As I was only testing for 1 hour max at very low volume levels I just found a resitor and cap which fitted into what author says most manufacturers do as they are lazy or incompetent. seemed to work ok with no ill effects, but needs optimizing according to the equations within the above link.
is it better to instead of 2 caps and resistors use 4 of each and from each output put zobel to gnd?
Your describing the bootstrap mod, but which output? The PIC of my experiment shows the input of inductors, and not going to ground.
This I think is a little confusing. The mod goes from positive signal line to negative signal line. Not, as with bootstrap snubber, going from bootstrap to GND.
I still plan to also add the bootstrap snubber, but as others have already tested and ensure success of that I don't need to test that out. I am only modding in a sequence which allows me to test out different values, or mods which others may have not tried yet. Notice I stripped out the smd 1nf power cap bypass for a leaded replacement. The middle "104" 100nf bypass is next with PP wimas.
Can you please elaborate on how "it stops the coil storing energy through magnetic flux". Whenever current flow thru a coil, there will be induction and therefore magnetisation.
Regards,
Your right of course, and I don't think it removes the magnetic flux. More prevents the "storage of potential energy" that the magnetism creates. I read something along those lines in an electric circuits tutorial when browsing, and that's what it does to speakers as well when placed across terminals. It is showing the IC output legs a stable resistance from the LC circuit, which is what the zobel in the output circuit does between speaker and LC.
For reasons I don't understand, the circuit doesn't even show a potential acrossed it, and i'm pretty sure it said order of resistor/cap in circuit not crucial.....but I felt safer copying order on the 7492 schematic.
The result would be a much improved response. Thus, the improvements I said that I witnessed in sound.
I am not claiming to know how this type of circuit does what it does. I just know it worked.
I suggest people refer to anything, and everything they can find on zobel circuits, Boucherot cells etc.
I am hoping a more respected member, with more knowledge, and testing equipment can try this mod out.
Xrk has those lovely sounding reference speakers and has already loads of test results and sound recordings to compare against. He did suggest he might try it when posting back and forth in the 7492 thread.
Last edited:
I hope to have time to try out tonight. It would be interesting to do this with an alligator clip to a pin so that the Zobel can be quickly out in/out for a listening test in real time.
After hearing the (to me) more dynamic bass and clearer highs on my cheap 7492 (which has OSCON and bootstrap snubber mod) to see if it can make the 3116 sound even better than it already is.
Usually a clearly audible improvement as what Stanleyburbrick describes is real - especially when backed by what another brand does and theory of Zobels or Beaucherot cells.
I am optimistic that it will wrk but we will see. Certainly I will use a mic to measure any difference at the speaker.
After hearing the (to me) more dynamic bass and clearer highs on my cheap 7492 (which has OSCON and bootstrap snubber mod) to see if it can make the 3116 sound even better than it already is.
Usually a clearly audible improvement as what Stanleyburbrick describes is real - especially when backed by what another brand does and theory of Zobels or Beaucherot cells.
I am optimistic that it will wrk but we will see. Certainly I will use a mic to measure any difference at the speaker.
X, Burbick, please give a short summary / description of the mods you're talking about.
I've lost the thread.
I've lost the thread.
The Zobel Mod for the TPA3116 as copy of TDA7492
The mod we are talking about is basically strapping a resistor and cap in series between the two output pins of the amp, BEFORE the inductor stage. This type of filter is also called a Zobel or a Beaucharot cell, and is commonly used on speaker drivers to flatten the rising impedance of a reactive load that the driver presents to the amp. A flatter impedance means less resistance and hence higher current can be applied for same voltage across the speaker terminals as provided by the amp. Higher current will mean more SPL or more dynamic response.
Look at the 7492 circuit and you will notice a 330pF and 22R connected in series and straddling the two output pins from the amp for each channel. I wondered what this was too, and I guess Stanleyburbrick also wondered so he put one on the 3116, which does not have one on ANY circuit schematic I have seen.
What I have observed is that the 7492 sounds more powerful than the 3116 in bass and in HF's. I wonder if this is the cause, although, it is surprising that such a cheap circuit feature was overlooked by very smart TI engineers. Maybe the way the 3116 transistor topology doesn't need a Zobel?
I don't have time to run a Spice (TINA) sim of the Zobel, so will also just use something similar to what was done on the 7492. I will put 330pF and 20R across the legs of the inductors and listen to what happens on my DUG v0 TPA3116D2.
Take a look at the TDA7492 diagram where I circled in red, the 330pF and 33R resistors. This is the circuit that I will apply to the same location on the TPA3116D2 chip.
Hope that helps.
The mod we are talking about is basically strapping a resistor and cap in series between the two output pins of the amp, BEFORE the inductor stage. This type of filter is also called a Zobel or a Beaucharot cell, and is commonly used on speaker drivers to flatten the rising impedance of a reactive load that the driver presents to the amp. A flatter impedance means less resistance and hence higher current can be applied for same voltage across the speaker terminals as provided by the amp. Higher current will mean more SPL or more dynamic response.
Look at the 7492 circuit and you will notice a 330pF and 22R connected in series and straddling the two output pins from the amp for each channel. I wondered what this was too, and I guess Stanleyburbrick also wondered so he put one on the 3116, which does not have one on ANY circuit schematic I have seen.
What I have observed is that the 7492 sounds more powerful than the 3116 in bass and in HF's. I wonder if this is the cause, although, it is surprising that such a cheap circuit feature was overlooked by very smart TI engineers. Maybe the way the 3116 transistor topology doesn't need a Zobel?
I don't have time to run a Spice (TINA) sim of the Zobel, so will also just use something similar to what was done on the 7492. I will put 330pF and 20R across the legs of the inductors and listen to what happens on my DUG v0 TPA3116D2.
Take a look at the TDA7492 diagram where I circled in red, the 330pF and 33R resistors. This is the circuit that I will apply to the same location on the TPA3116D2 chip.
Hope that helps.
Attachments
Last edited:
Xrk971..... Hold off on those values. The info below in the link warns of using values not calculated for the LC circuit.
If your 3116 has same values as above schematic then go ahead. The links advice is to either calculate values using his equations or use a safe value as he notes most manufacturers using the circuit do. A 100nf cap is OK for testing, the resistor was Max 10R.
http://www.embedded.com/print/4015876
In this instance putting it in this place in the circuit makes the current follow the voltage curves, rather than voltage following current curves.
If your 3116 has same values as above schematic then go ahead. The links advice is to either calculate values using his equations or use a safe value as he notes most manufacturers using the circuit do. A 100nf cap is OK for testing, the resistor was Max 10R.
http://www.embedded.com/print/4015876
In this instance putting it in this place in the circuit makes the current follow the voltage curves, rather than voltage following current curves.
Last edited:
Here's what I just posted in the other 7492 thread as there is some doubt because of differences in 3116 and 7492 modulation.
" So, the data sheets and descriptions of AD and BD modulation and the different output filter topologies......
Thats all I should really need to say. The answer is in there. One word........ OUTPUT.
To elaborate, that whole discussion regarding the explanations of AD & BD filter types is based on output filters. If you wish to disagree, I would suggest that in that case the bootstrap snubber mod does not work, and so many of you have carried it out, agree with each other that it improves the sound. When in fact the mod would be "fools gold". But it clearly isn't as a few have posted evidence. But the equivelant circuit of bootstrap mod is in the output filter already, identical it is because it is going to GND. It's sole purpose is to filter the high switching frequencies out of the circuit to GND, so why do it in another place?
Well clearly, its not that simple as in between the the output zobel, and bootstrap mod is the LC filter, and the bootstrap is the other side of LC, which then returns to IC. The bootstrap mod zobel to GND clearly removes high frequency before LC and also prevents from returning into IC gate, so why the need for the output filter zobel?
Looking at it from this direction, relying on what the data sheets etc says, the need for bootstrap mod and output filter together make no sense either.
But, and here's the big difference. The pre-LC zobel mod I am suggesting for 3116, as per 7492 schematic, is not in the output filter as it does not go to GND. The role of the pre-LC is not to filter switching frequencies to ground. Its clearly not there to minimise emi in the speaker wires. That's why those references don't apply.
This layout and placement of zobel/Boucherot cell is exactly as placing it across terminals of speaker. It works in both directions. First, it controls the resistance of the coil that the component/output before it sees. So, a steady state resistance to work against rather than a varying resistance with frequency, as is the side effect of a voice coil, or inductor coil. We have to remember at this stage freq range is only regulated by IC from very low fc right up to switching frequency (apart from top which the bootstrap mod helps removing). And, it is switching frequency at this point, so much happening. It isn't until it goes through LC that you have tamed the frequencies required for your output. That's a lot of varying frequency with varying resistance. So, this is one purpose of the Boucherot cell here. To tame the resistance from variation seen by the IC. Secondly, there is no potential across this circuit, and it acts on the coil with the current passing in either direction in exactly the same way. Which leads to third reason Thirdly, the circuit here also works on the component after, as with speaker. As coils create magnetic flux through inductance they are capable of energy storage, and can act like a battery, for short periods like a huge capacitor I suppose. But we don't want energy stored in the inductor as we need the best rise and fall we can get, i.e. response, nice and fast in low ultra seconds, and accurate. If the coil contains a store of energy from inductance, the energy is in between IC and output filters, and the varying voltage created by IC outputs has to negotiate this, which it will not do accurately. If I pour salt into the top of a half full salt shaker, is it that very same salt which I poured in that comes out first? Even if some of the salt gets through first it will still be muddled together with some salt already in the shaker! This apparently is what the zobel across the coil prevents. It prevents the coil from accumulating a store of energy within its magnetic flux. Which in return should allow the flow of energy, or current straight through, improving response.
I believe this is what's happening with this circuit. It is a cumulation of all the info I have been able to gather on zobels and Boucherot cells, as well as through researching pllxo."
" So, the data sheets and descriptions of AD and BD modulation and the different output filter topologies......
Thats all I should really need to say. The answer is in there. One word........ OUTPUT.
To elaborate, that whole discussion regarding the explanations of AD & BD filter types is based on output filters. If you wish to disagree, I would suggest that in that case the bootstrap snubber mod does not work, and so many of you have carried it out, agree with each other that it improves the sound. When in fact the mod would be "fools gold". But it clearly isn't as a few have posted evidence. But the equivelant circuit of bootstrap mod is in the output filter already, identical it is because it is going to GND. It's sole purpose is to filter the high switching frequencies out of the circuit to GND, so why do it in another place?
Well clearly, its not that simple as in between the the output zobel, and bootstrap mod is the LC filter, and the bootstrap is the other side of LC, which then returns to IC. The bootstrap mod zobel to GND clearly removes high frequency before LC and also prevents from returning into IC gate, so why the need for the output filter zobel?
Looking at it from this direction, relying on what the data sheets etc says, the need for bootstrap mod and output filter together make no sense either.
But, and here's the big difference. The pre-LC zobel mod I am suggesting for 3116, as per 7492 schematic, is not in the output filter as it does not go to GND. The role of the pre-LC is not to filter switching frequencies to ground. Its clearly not there to minimise emi in the speaker wires. That's why those references don't apply.
This layout and placement of zobel/Boucherot cell is exactly as placing it across terminals of speaker. It works in both directions. First, it controls the resistance of the coil that the component/output before it sees. So, a steady state resistance to work against rather than a varying resistance with frequency, as is the side effect of a voice coil, or inductor coil. We have to remember at this stage freq range is only regulated by IC from very low fc right up to switching frequency (apart from top which the bootstrap mod helps removing). And, it is switching frequency at this point, so much happening. It isn't until it goes through LC that you have tamed the frequencies required for your output. That's a lot of varying frequency with varying resistance. So, this is one purpose of the Boucherot cell here. To tame the resistance from variation seen by the IC. Secondly, there is no potential across this circuit, and it acts on the coil with the current passing in either direction in exactly the same way. Which leads to third reason Thirdly, the circuit here also works on the component after, as with speaker. As coils create magnetic flux through inductance they are capable of energy storage, and can act like a battery, for short periods like a huge capacitor I suppose. But we don't want energy stored in the inductor as we need the best rise and fall we can get, i.e. response, nice and fast in low ultra seconds, and accurate. If the coil contains a store of energy from inductance, the energy is in between IC and output filters, and the varying voltage created by IC outputs has to negotiate this, which it will not do accurately. If I pour salt into the top of a half full salt shaker, is it that very same salt which I poured in that comes out first? Even if some of the salt gets through first it will still be muddled together with some salt already in the shaker! This apparently is what the zobel across the coil prevents. It prevents the coil from accumulating a store of energy within its magnetic flux. Which in return should allow the flow of energy, or current straight through, improving response.
I believe this is what's happening with this circuit. It is a cumulation of all the info I have been able to gather on zobels and Boucherot cells, as well as through researching pllxo."
My TPA3116D2 uses a 10uH coild and 680nF caps on the LC stage for the 4ohm amp and 22uH and 680nF caps for the 8ohm amp. I think they are close enough to my current 7492 which uses 22uH and 220nF cap. Although, I am not sure how to model this in Spice - do I pick opposing phase voltage drive sources for the amp pins? Stanley, if you want to take a crack at suggesting what values to use that would be helpful. The TPA3116 is a bulletproof amp as far as shorting the output. That is the worst thing and it is protected and shuts down - so I don't think off-ideal values for the Zobel are going to damage it. We do need to figure out what the optimal value is, if there is an optimal value.
The idea is that it needs to damp AC oscillations between the total dual LC *system" impedance as presented to the speaker driver as both sides are used. 22R or 10R, all the same...
The idea is that it needs to damp AC oscillations between the total dual LC *system" impedance as presented to the speaker driver as both sides are used. 22R or 10R, all the same...