Thanks,
I intend to replace the PCB power connector at some time in the future but I wanted to make sure the amplifier was all OK before doing any mods. It's a simple job to take it off the board and wire directly to the switch. The case has a removable bottom as well as top so everything is accessible.
The banana jacks are 1" apart to give a bit more space for fat fingers to put wires into the holes. I have never used dual plugs for speakers, I find the springy clip things seem to work OK but there was no room for them on the panel.
I have now managed to do a few tests on the eBay 2.1 board and I think there may be a wrong component specified on the board.
The switching frequency on the left/right chip was 400kHz as it should be but the sub channel was only 119kHz. This gave rise to some noticeable distortion on the sub output. I thought the problem was due to a wrong value resistor being fitted that I mentioned in my previous post (#246). This resistor sets the master/slave mode for the synchronisation as well as the gain so I changed it for the correct value of 75k but this did not synchronise the two chips. The sync. output from the left/right chip is coupled to the sync. in of the sub chip by a 10k resistor with a 1n capacitor to ground. This agrees with the TI application note on the chip for synchronising the two chips, but is at odds with the eval. board which uses 47k and 47pf. The amplitude of the sync. in was only about 1V peak and insufficient to synchronise the sub. chip onto the correct frequency. When the 1n was changed to 100pf it synced up to 400kHz and the distortion disappeared.
I will do some more testing later in the week to see if there are any more issues but so far it looks good.
The switching frequency on the left/right chip was 400kHz as it should be but the sub channel was only 119kHz. This gave rise to some noticeable distortion on the sub output. I thought the problem was due to a wrong value resistor being fitted that I mentioned in my previous post (#246). This resistor sets the master/slave mode for the synchronisation as well as the gain so I changed it for the correct value of 75k but this did not synchronise the two chips. The sync. output from the left/right chip is coupled to the sync. in of the sub chip by a 10k resistor with a 1n capacitor to ground. This agrees with the TI application note on the chip for synchronising the two chips, but is at odds with the eval. board which uses 47k and 47pf. The amplitude of the sync. in was only about 1V peak and insufficient to synchronise the sub. chip onto the correct frequency. When the 1n was changed to 100pf it synced up to 400kHz and the distortion disappeared.
I will do some more testing later in the week to see if there are any more issues but so far it looks good.
Hi John8,
Looks to be an error in the data sheet ckt on page14, figure 27.
BTW, The cap maybe unnecessary, since it acts as a slew rate limiter, another option is a series source terminator, replace the 47K series R with a 22/22/47 ohm, and tune for a reflected wave that suits your fancy.
BTW, sync pin Vih min is 2V!!
Sounds to me that you are doing the designer's job (eBay 2.1 board) , good stuff, nice to see someone else actually testing these ckts out properly.
Suggest to report the sch error to TI, so others do not fall into the trap.
Looks to be an error in the data sheet ckt on page14, figure 27.
BTW, The cap maybe unnecessary, since it acts as a slew rate limiter, another option is a series source terminator, replace the 47K series R with a 22/22/47 ohm, and tune for a reflected wave that suits your fancy.
BTW, sync pin Vih min is 2V!!
Sounds to me that you are doing the designer's job (eBay 2.1 board) , good stuff, nice to see someone else actually testing these ckts out properly.
Suggest to report the sch error to TI, so others do not fall into the trap.
Nice work there John8.Can you try changing the freq select to 1.2MHz and see if it sounds better?
Probably requires the output filter to be disconnected and use ferrite beads and no more than 20-30cm speakers cables to run it at 1.2MHz.
Also some parts are through-hole and placed rather far from the chip. I doubt it'll work (very well) at 1.2MHz.
It's not easy to change the frequency with the board laid out as it is. The frequency select pins are all closely connected to a common ground so it means physically lifting a leg of the IC and tying it to a logic high. Maybe I would have tackled this when I was 30 years younger but all the components are getting much smaller these days. (Bring back DIL packages on 0.1" pin spacing with through hole components.)
As Saturnus said the output filter will need modifying for the higher frequency. All the tests on the data sheet are conducted at 400kHz so I would imagine this gives the best performance.
Switching losses would be greater leading to higher heat dissipation in the chip and the need for a bigger heatsink so I think I will leave that side of the experiment to someone with a more suitable layout.
As Saturnus said the output filter will need modifying for the higher frequency. All the tests on the data sheet are conducted at 400kHz so I would imagine this gives the best performance.
Switching losses would be greater leading to higher heat dissipation in the chip and the need for a bigger heatsink so I think I will leave that side of the experiment to someone with a more suitable layout.
According to TI the 3116 frequency is only tunable fromm 400-600KHZNice work there John8.
The 3118 and 3130 are tunable from 400KHZ to 1.2MHZ
I reported this issue to TI who were aware of the problem; it will be changed in the next revision of the data sheet. TI suggest 4k7 for R73 and 47pF for C41.
I'll leave mine at 10k and 100pF as it seems to work and the fact that I found a 100pF 0805 capacitor in my box of bits
.John8, you are a tweakers tweaker!I reported this issue to TI who were aware of the problem; it will be changed in the next revision of the data sheet. TI suggest 4k7 for R73 and 47pF for C41.
I'll leave mine at 10k and 100pF as it seems to work and the fact that I found a 100pF 0805 capacitor in my box of bits
They are small and they are SMD's. You need magnifying goggles, tweezers, small soldering iron tips, thin solder, and steady hands. But other than that, a lot of fun and a great IC. The reference design works well, as does the design for the eval board (more filters and gain adjustment option). I have tried both and both are good.
Thanks X. I thought they were SMD but not being as familiar as many here, i wasn't sure of the package tssop. To be honest too much real work here for me to look over the pdf thoroughly (something in regularly lambasted for, hey im a forever noob and proud)
I got these free and thought it may be worth a try. Its that or the tda7924 i have waiting some of my free time....
One thing i noticed that I'm not sure i like is the LC output filter. I gather it is necessary, but well...after trying to rid myself of cap and coil in regular amps, its a little bit of a compromise that it would be nice to be able to avoid. Thems the breaks i guess.
I got these free and thought it may be worth a try. Its that or the tda7924 i have waiting some of my free time....
One thing i noticed that I'm not sure i like is the LC output filter. I gather it is necessary, but well...after trying to rid myself of cap and coil in regular amps, its a little bit of a compromise that it would be nice to be able to avoid. Thems the breaks i guess.
Hi tabarddn,
I do understand the Chinese mfg situation, as I just finished worked for a lighting ballast mfg for 5 years. Design was done in Taiwan, mfg was primarily done on the mainland. Anyone heard of Antron Electronics. Quality was all over the place. It is just not wages alone, it is cheap components, unskilled labour, as well.
I asked the EE's what was the target MTBF on some products? = blank and they offered a 5 year free replacement warranty. Did not help the customers, when the fixtures were mounted 40-50 feet in the air, required a $300-500 a day scissors lift + a $50-70/hr electrician to fix. Basically the T5HO lamps had a longer or equivalent life to the ballast.
John8, well done sir, funny that even TI would not put out an errata sheet of this issue.
I do understand the Chinese mfg situation, as I just finished worked for a lighting ballast mfg for 5 years. Design was done in Taiwan, mfg was primarily done on the mainland. Anyone heard of Antron Electronics. Quality was all over the place. It is just not wages alone, it is cheap components, unskilled labour, as well.
I asked the EE's what was the target MTBF on some products? = blank and they offered a 5 year free replacement warranty. Did not help the customers, when the fixtures were mounted 40-50 feet in the air, required a $300-500 a day scissors lift + a $50-70/hr electrician to fix. Basically the T5HO lamps had a longer or equivalent life to the ballast.
I was told by a sales person who visited the factory in China, that they $hit in a whole in the ground too. Good everyone, keep supporting these folks, your job is probably next on the block
John8, well done sir, funny that even TI would not put out an errata sheet of this issue.
The filters are required to meet radiation targets, (RF spray) that is if you are a mfg that is actually required to do this sort of testing , to meet FCC or other regulatory rerquiremnts. For the life of me, I have not figured if there was an FCC equivalent in China. I know the Antron's stuff was certified at UL in the US. I guess because know one accepts any Chinese regulators.
China owns 1.28 trillion of US debt, I do not think you can get to that position by producing crap?
If interest rates ever begin to rise, there will be big trouble for many economies around the world
I agree about the LC for RFI suppression. Where i work we make large electrical machines, and in a similar way, to satisfy certification criteria, the level of generated harmonics is closely tested to meet a very dated spec (aren't they all?) Telephone Interference Factor...
Still, that LC bugs me. I'm not sure id want a complex load on the end of that.
Still, that LC bugs me. I'm not sure id want a complex load on the end of that.