Hope everyone understand this it not plug and play.
It's Eq and active two-way with measurement.
It's Eq and active two-way with measurement.
PA3S works quite well for a hacky active speaker setup.
My CD+Seos12 combo is very very very very sensitive so realistically I need to lower the noise floor more. Linear PSU is out because the SMPS is already very low noise. Maybe just a 10db attenuator on the hi outs. These 107dB CDs in 12" horns are crazy sensitive.
Played with REW and the DBX software and got this. Green line is where I am at so far. If I do a 20db attenuator I can probably use the +4dbV input setting and get more bang for the buck, though pushing the volume past 70% is likely to result in people outside my home enjoying my music as well.
My CD+Seos12 combo is very very very very sensitive so realistically I need to lower the noise floor more. Linear PSU is out because the SMPS is already very low noise. Maybe just a 10db attenuator on the hi outs. These 107dB CDs in 12" horns are crazy sensitive.
Played with REW and the DBX software and got this. Green line is where I am at so far. If I do a 20db attenuator I can probably use the +4dbV input setting and get more bang for the buck, though pushing the volume past 70% is likely to result in people outside my home enjoying my music as well.
Quite good actually. Horrific package for DIY though.
Still a bridge rectifier and a 10.000 to 15.000 µF filter cap are needed. Or is the plan to use a switcher before the switcher (and have double fun 🙂)?
This one will work Ok I guess:
https://de.farnell.com/en-DE/bourns/srp1245c-3r3m/leistungsinduktivit-t-3-3uh-gesch/dp/3384955
Still a bridge rectifier and a 10.000 to 15.000 µF filter cap are needed. Or is the plan to use a switcher before the switcher (and have double fun 🙂)?
This one will work Ok I guess:
https://de.farnell.com/en-DE/bourns/srp1245c-3r3m/leistungsinduktivit-t-3-3uh-gesch/dp/3384955
Last edited:
It looks pretty good. I think it's your ver of the low EMI 5V application example, right? I am guessing you will leave the 26V to the discretion of the end user, which I think is a good move. The only thing that bothers me about 400khz switching is the possible interference since most of the power modes have sideband resonance in the 1.2mhz range, which is an integer multiple of 400khz. This may upset sensitive drivers.
The application noted show a 5V 2MHz setup. But what I am curious about is the 1MHz frequency. Looking at Kyocera AVX datasheets for X5R and X7R dielectrics, the lowest ESR tends to be in the 1MHz range. This is also outside of any possible resonance stacking on the MA12070. THe lowest ESr point as a switching frequency may also reduce the ripple even further. Not sure if it would make any audible difference, but numbers bragging is numbers bragging 😛
The application noted show a 5V 2MHz setup. But what I am curious about is the 1MHz frequency. Looking at Kyocera AVX datasheets for X5R and X7R dielectrics, the lowest ESR tends to be in the 1MHz range. This is also outside of any possible resonance stacking on the MA12070. THe lowest ESr point as a switching frequency may also reduce the ripple even further. Not sure if it would make any audible difference, but numbers bragging is numbers bragging 😛
No one serious would use a 42V 15A capable switcher for 5V 0.25A. Sorry.
Of course this is the switcher for 26V PVDD.
Of course this is the switcher for 26V PVDD.
Wow that is so embarrassing. For some reason I thought I read 5V. How did I miss the output voltage notes? Hahahaha. I feel dumb.
Hi JPS, for what purpose it he 5V generated by the MCP1793? It looks good but I am pretty sure it will be expensive. I built ideal rectifiers with LT4320 and these cost as much as a cheap complete ready made MA12070 amplifier.
It really looks like a device as I would want it to be honest.
It really looks like a device as I would want it to be honest.
Last edited:
Hi JP,
it‘s for the bias of the switcher (< 25V@100mA).Option to connect the bias to ground on board, needs to be tested regarding performance.
I know, not cheap, but depends on MOQ; and why not for other amplifiers? We‘ve 15A!
JP
You have an MA12070 that you have listened to?
I mean it have something that is good?
It's going on the out of a 400KHz switcher. Per spec sheet it should be .10 uV/sqr(Hz)
sqrt(400000)*.01 is only about 6.32uV rms.
PSRR also peaks at 400kHz at about -45dB for 5vO. But if we calculate PSRR as log20 (4.4mV /Ripple Out) = 45dB we're looking at 24 uV ripple or ~178x attenuation. Somewhere around there. (10^ (45/20)) =~ 177.8X ripple attenuation.
It doesn't seem especially noisy considering it's dropping 26ish V to 5V. Even going TI and running 6V - 5V you only gain 20dB PSRR headroom. The Analog ADP7142 has a 20dB PSRR gain over the MCP and actually looks to have less noise and incorporates the -20dB PSRR gain of the TI part. Just for biasing the switcher the noise really doesn't seem that bad at all.
Considering that the ASR-measured PA3S and DA-6 were 85dB and 81db SINAD respectively while the Aiyima A07 was 75dB (all three using so-so layouts) I think this might be a real winner as long as people only need a few watts.
sqrt(400000)*.01 is only about 6.32uV rms.
PSRR also peaks at 400kHz at about -45dB for 5vO. But if we calculate PSRR as log20 (4.4mV /Ripple Out) = 45dB we're looking at 24 uV ripple or ~178x attenuation. Somewhere around there. (10^ (45/20)) =~ 177.8X ripple attenuation.
It doesn't seem especially noisy considering it's dropping 26ish V to 5V. Even going TI and running 6V - 5V you only gain 20dB PSRR headroom. The Analog ADP7142 has a 20dB PSRR gain over the MCP and actually looks to have less noise and incorporates the -20dB PSRR gain of the TI part. Just for biasing the switcher the noise really doesn't seem that bad at all.
Considering that the ASR-measured PA3S and DA-6 were 85dB and 81db SINAD respectively while the Aiyima A07 was 75dB (all three using so-so layouts) I think this might be a real winner as long as people only need a few watts.
MCP1793 has a typical output noise of its own of 608 uV @5V output. That is 15 times more than an Onsemi 7805.
The amplifier chip goes well till around 15W in BTL continuously so it puts out more than a few Watts. I really don’t know what people do at home but it goes for sure way louder than I’ll probably ever need. I use amplifiers to drive 2 normal efficiency loudspeakers in a 5 x 7 meter living room to moderate volume levels.
The amplifier chip goes well till around 15W in BTL continuously so it puts out more than a few Watts. I really don’t know what people do at home but it goes for sure way louder than I’ll probably ever need. I use amplifiers to drive 2 normal efficiency loudspeakers in a 5 x 7 meter living room to moderate volume levels.
Last edited:
Not a real issue with the MCP. Could replace them, but need to know what do you want, I hate continuous changes.
What do you think about implementing something like the TI Post Filter FeedBack PFFB TPA3255 solution to improve THD+N?
@Solve
French guys saying: c´est pour l´exercice! I like making layouts.
Well, no I don´t have a MA12070 Class D amplifier but does it matter?
I would like a few watts Class D but no Chinese product. I´m also curious about the results expected?!
JP
What do you think about implementing something like the TI Post Filter FeedBack PFFB TPA3255 solution to improve THD+N?
@Solve
French guys saying: c´est pour l´exercice! I like making layouts.
Well, no I don´t have a MA12070 Class D amplifier but does it matter?
I would like a few watts Class D but no Chinese product. I´m also curious about the results expected?!
JP
Well trust the french guy I think he knows what he is talking about.
🙂
EMI, even if its not that we thinking about but hearing if I get close to the speaker.
And oufcousre we should get rid of it.
Allo Volt+D was dead quiet.
But this is an very interesting chip and sound way better then anything cheap I heard.
So please keep going!
🙂
EMI, even if its not that we thinking about but hearing if I get close to the speaker.
And oufcousre we should get rid of it.
Allo Volt+D was dead quiet.
But this is an very interesting chip and sound way better then anything cheap I heard.
So please keep going!
Well I am not the one that decides but what does a "way more than normal" noisy regulator bring when used for bias voltage? When looking at the rest of the schematic details are fine so I thought it was odd to use a regulator so noisy. Sorry to say but "continuous changes" are a part of life when you are the one choosing such a part and then ask for opinions 😀
Datasheet of LT8648 is IMO not very clear about the bias pin 1:
"BIAS (Pin 1): The internal regulator will draw current from BIAS instead of VIN when BIAS is tied to a voltage higher than 3.1V. For output voltages of 3.3V to 25V this pin should be tied to VOUT. If this pin is tied to a supply other than VOUT use a 1μF local bypass capacitor on this pin. If no supply is available, tie to GND. However, especially for high input or high frequency applications, BIAS should be tied to output or an external supply of 3.3V or above."
Well what is it, Linear Technology? 🙂
BTW no opinion on TI Post Filter FeedBack PFFB TPA3255 solution as I never used that chip or a product with it.
Last edited:
My reading of the datasheet just says "for HF or high input applications, do not tie to ground" since they list ground as an option earlier. I still think the MCP would be fine.
L7805 are OOS at Mouser and Digikey with an ETA of 2023 at the earliest. NCV and MC7800 series are available but they measure worse on noise except for PSRR which is already very low.
L7805 is also noisier than the MCP at 400KHz.
OnSemi provides no actual measurement graphs that show anything beyond an idealized mathematical function to make a pretty curve. For the 400KHz application the MCP is superior. The STMicro, TI, and OnSemi datasheets for 7805__ products could literally be the same document. All with nonsense "measurements" which are just identical to the limited information they provide above with no real tools to extrapolate noise levels beyond the shown 100kHz.
The tables at the top of the datasheet are trumped by actual measurements provided further down in the Microchip documentation. Most of the datasheets only show up to 10kHz but that is not where @JPS64 has set the operating point.
I think the answer highly depends on goals. A lot of discussion goes into MA12070 vs TPA3255/TPA3116D2. The Sylph measures well and delivers 35W clean into 8-Ohms which is more than a PBTL MA12070 can claim. If the goal is boutique MA12070, the design would be a bit different than trying to leverage the I2C selectable power modes and power mode profiles which allow the MA12070 to be adapted to the physical layout. Every MA12070 board which has been measured contains a microcontroller. I think your design as you have it will still outperform anything existing.
But for real, DSP is the future. DSP usually means 2 channels of amplification per speaker in a 2-way, 3 channels in a 3-way, etc.
Here is a DIYSG Fusion 10 Pure as measured by the designer (tuxedocivc) versus the exact same speaker built by an amateur (me 10 years ago, still pursuing an ME degree) but with active DSP XOver. There is just no comparison. But to take advantage of DSP, you need to build to the most demanding channel in the physical speaker coupled with gain staging losses (no adjustable gain outside of 2 profiles to select on the MA12070) so a clean 20W isn't really a clean 20W compared to higher potential (read: voltage potential) amps because the MA12070 will drop the Vpp way low (1.8Vpp max, aka .6XX Vrms). In theory, 13W at 8 Ohms should be plenty to be uncomfortable, but I am not the only person to say "it depends." My 12Wpc EL84 amp feels like it has a bit more grunt than my PBTL MA12070 amps. By every measure the MA12070 should get louder; faster... but it just doesn't.
Analog vs DSP XOver as applied to DIYSG Fusion 10 Pure. The DSP XOver is also able to go 200+Hz lower than the passive XOver (1.2kHz compared to passive @ 1450Hz). Bi-Amping allows more usable range from the CD+Waveguide with its own dedicated channel.
Anyway, just make whatever makes you happy. Based on the multi-tap toroid (size of implementation) it seems you are targeting passive speakers which means that the 13W/channel into 8 Ohms and 30W into 4 Ohms under -80dB SNR is probably fine for that application as long as the speakers are over 90-93dB/W@1m.
It is your baby, make of it what you want! After all, that is the joy of design.
Also RE; constant changes to the design... As somebody in the engineering dept at my place of employment, continuous changes are what keep things moving forward (and me employed!). They are what also keep my pet electronics projects from advancing because I keep making changes as I learn (again, I am not an electrical engineer). Typical time from application preliminary design (even for pre-built applications / systems we integrate into our environment) is over one year, usually closer to pushing 2.5. But we pride ourselves on exploring all options, and usually the only limitation to service life of the design is vendor hardware support. So if it takes a while, do not be discouraged! That just means you are evaluating all your options and carefully examining every use-case scenario.
Sorry, I like to be wordy. I am a tech nerd.
L7805 are OOS at Mouser and Digikey with an ETA of 2023 at the earliest. NCV and MC7800 series are available but they measure worse on noise except for PSRR which is already very low.
L7805 is also noisier than the MCP at 400KHz.
OnSemi provides no actual measurement graphs that show anything beyond an idealized mathematical function to make a pretty curve. For the 400KHz application the MCP is superior. The STMicro, TI, and OnSemi datasheets for 7805__ products could literally be the same document. All with nonsense "measurements" which are just identical to the limited information they provide above with no real tools to extrapolate noise levels beyond the shown 100kHz.
The tables at the top of the datasheet are trumped by actual measurements provided further down in the Microchip documentation. Most of the datasheets only show up to 10kHz but that is not where @JPS64 has set the operating point.
I think the answer highly depends on goals. A lot of discussion goes into MA12070 vs TPA3255/TPA3116D2. The Sylph measures well and delivers 35W clean into 8-Ohms which is more than a PBTL MA12070 can claim. If the goal is boutique MA12070, the design would be a bit different than trying to leverage the I2C selectable power modes and power mode profiles which allow the MA12070 to be adapted to the physical layout. Every MA12070 board which has been measured contains a microcontroller. I think your design as you have it will still outperform anything existing.
But for real, DSP is the future. DSP usually means 2 channels of amplification per speaker in a 2-way, 3 channels in a 3-way, etc.
Here is a DIYSG Fusion 10 Pure as measured by the designer (tuxedocivc) versus the exact same speaker built by an amateur (me 10 years ago, still pursuing an ME degree) but with active DSP XOver. There is just no comparison. But to take advantage of DSP, you need to build to the most demanding channel in the physical speaker coupled with gain staging losses (no adjustable gain outside of 2 profiles to select on the MA12070) so a clean 20W isn't really a clean 20W compared to higher potential (read: voltage potential) amps because the MA12070 will drop the Vpp way low (1.8Vpp max, aka .6XX Vrms). In theory, 13W at 8 Ohms should be plenty to be uncomfortable, but I am not the only person to say "it depends." My 12Wpc EL84 amp feels like it has a bit more grunt than my PBTL MA12070 amps. By every measure the MA12070 should get louder; faster... but it just doesn't.
Analog vs DSP XOver as applied to DIYSG Fusion 10 Pure. The DSP XOver is also able to go 200+Hz lower than the passive XOver (1.2kHz compared to passive @ 1450Hz). Bi-Amping allows more usable range from the CD+Waveguide with its own dedicated channel.
Anyway, just make whatever makes you happy. Based on the multi-tap toroid (size of implementation) it seems you are targeting passive speakers which means that the 13W/channel into 8 Ohms and 30W into 4 Ohms under -80dB SNR is probably fine for that application as long as the speakers are over 90-93dB/W@1m.
It is your baby, make of it what you want! After all, that is the joy of design.
Also RE; constant changes to the design... As somebody in the engineering dept at my place of employment, continuous changes are what keep things moving forward (and me employed!). They are what also keep my pet electronics projects from advancing because I keep making changes as I learn (again, I am not an electrical engineer). Typical time from application preliminary design (even for pre-built applications / systems we integrate into our environment) is over one year, usually closer to pushing 2.5. But we pride ourselves on exploring all options, and usually the only limitation to service life of the design is vendor hardware support. So if it takes a while, do not be discouraged! That just means you are evaluating all your options and carefully examining every use-case scenario.
Sorry, I like to be wordy. I am a tech nerd.
Attachments
I completely forgot to put the link to Sylph measurements. I believe you said it was your layout? I thought for some reason the JLE is just a prebuilt XKR board https://www.audiosciencereview.com/...ics-sylph-d200-amplifier-module-review.25295/
Maybe I am mistaken in that belief, but datasheets tend to be good at guiding folks towards good implementation. Your posted schematics so far lead me to believe it would be competitive with $100-$200/channel amps which are currently class-leading (Topping PA5, Topping L90) assuming the end user assembly was also equally meticulous.
(can you tell I enjoy design theory and data analysis?)
Maybe I am mistaken in that belief, but datasheets tend to be good at guiding folks towards good implementation. Your posted schematics so far lead me to believe it would be competitive with $100-$200/channel amps which are currently class-leading (Topping PA5, Topping L90) assuming the end user assembly was also equally meticulous.
(can you tell I enjoy design theory and data analysis?)