I see.
Well the over temperature warning is going to be tripped at different points due to the thermal difference in our PCBs anyway and layout.
I wouldn't have thought it would be that different re running them without a heatsink though.
The efficiency of the amplifiers is presumably effected by the voltage it is being powered by. 12V is pushing it to the limit and I cannot run mine this low as the drop out of my linear + switching regs would prevent that. I have the voltage after my switcher set to 14.5V to give some head room for improved performance of the linear regs.
I need about 15-15.5V for mine to operate correctly.
Powered from a 20V rail though my implementation is drawing 130mA.
This is running with two channels switching, the switching reg, the two linear regs and 4 fully differential opamps.
Have you tried seeing how the over temp warning comes in if you lower the main rail down to 15V or so? I only say this because 12V might be quite sub optimum. They say it will function on 12V but that might be with limited performance in some areas.
Yeah, might be that the Outputstage is not switching perfectly at 12V. acally using 100n rather than 33nF at the charge pumps/bootstrap.
What makes me wonder is, the OT does like a clipping-indication, so only flickers up at bass when using music. When using fixed sine waves, the amp will run in thermal runaway when operating at the thermal trip point. This let me think that the Rds_on is increasing a lot at higher temps. Pushing it then way further, like up to 3A at 12V, the amp shuts down into latched fault mode.
Quiescent current here is actually: 4*6mA opamp + 6mA vref + 2*2mA LDO ground current + amp. So 34mA + amp. Leads to less than 100mA for the amp at minimum voltage 12V.
Edit:
I actually don't use a step-down for the LDOs, so 12V is possible due to the (relatively) low dropout voltage. Using some AZ1117C at the moment.
What makes me wonder is, the OT does like a clipping-indication, so only flickers up at bass when using music. When using fixed sine waves, the amp will run in thermal runaway when operating at the thermal trip point. This let me think that the Rds_on is increasing a lot at higher temps. Pushing it then way further, like up to 3A at 12V, the amp shuts down into latched fault mode.
Quiescent current here is actually: 4*6mA opamp + 6mA vref + 2*2mA LDO ground current + amp. So 34mA + amp. Leads to less than 100mA for the amp at minimum voltage 12V.
Edit:
I actually don't use a step-down for the LDOs, so 12V is possible due to the (relatively) low dropout voltage. Using some AZ1117C at the moment.
Last edited:
I have quickly run through this thread, -indeed interesting!
I noticed in the schematic posted that there were RC snubbers used on the device outputs (C27, 28, 29, 30 and R33, 34, 35, 36). These are not recommended, as the power stage is designed to run without (they will not help you in any way, just consume power). They are not shown in neither the datasheet nor the EVM user guide.
There has been quite some discussion of PSRR, which, I believe, is often a source for misunderstanding.
Considering a BTL output, PSRR only express how symmetric the two output halfbridges/channels are. Often PSRR is used to express how immune the amp is to PSU ripple, but as often PSRR is measured in idle, where the PSU usually does not have ripple in the audio band.
A class-D power stage works as a multiplier (inputs: duty cycle and PSU voltage), so when the music plays (duty cycle modulated) and ripple is present in the PSU this result in modulation sidebands which will be suppressed by the feedback loop. How well these modulation components are suppressed depends on how the amplifier is designed. TAS5613A/30B uses a forward correction of the PSU ripple wich results in being very robust to PSU ripple (audio pretty unaffected by moderate levels of ripple). TPA3251D2 uses same mechanisms and will be at least as robust to PSU ripple.
CSTART voltage cannot be used directly to set the CM of an OPA1632 (which is preferred for input opamp over SE opamps in general), as it is not same voltage as the INPUT pins' DC level. However, the INPUT DC level is AVDD/2 and is, different from TAS5613A/30B, constant over PVDD range. The CM of the OPA1632 can be set to basically any voltage, as the AC coupling caps are still needed, if not a DC servo circuit is used.
I have listened a few times to TPA3251D2 with different speakers, and so far I find the sound more smooth than TAS5630B. In fact it provides lots of detail and often I have seen the /CLIP indicator flashing, without noticing I was playing that loud.
I noticed in the schematic posted that there were RC snubbers used on the device outputs (C27, 28, 29, 30 and R33, 34, 35, 36). These are not recommended, as the power stage is designed to run without (they will not help you in any way, just consume power). They are not shown in neither the datasheet nor the EVM user guide.
There has been quite some discussion of PSRR, which, I believe, is often a source for misunderstanding.
Considering a BTL output, PSRR only express how symmetric the two output halfbridges/channels are. Often PSRR is used to express how immune the amp is to PSU ripple, but as often PSRR is measured in idle, where the PSU usually does not have ripple in the audio band.
A class-D power stage works as a multiplier (inputs: duty cycle and PSU voltage), so when the music plays (duty cycle modulated) and ripple is present in the PSU this result in modulation sidebands which will be suppressed by the feedback loop. How well these modulation components are suppressed depends on how the amplifier is designed. TAS5613A/30B uses a forward correction of the PSU ripple wich results in being very robust to PSU ripple (audio pretty unaffected by moderate levels of ripple). TPA3251D2 uses same mechanisms and will be at least as robust to PSU ripple.
CSTART voltage cannot be used directly to set the CM of an OPA1632 (which is preferred for input opamp over SE opamps in general), as it is not same voltage as the INPUT pins' DC level. However, the INPUT DC level is AVDD/2 and is, different from TAS5613A/30B, constant over PVDD range. The CM of the OPA1632 can be set to basically any voltage, as the AC coupling caps are still needed, if not a DC servo circuit is used.
I have listened a few times to TPA3251D2 with different speakers, and so far I find the sound more smooth than TAS5630B. In fact it provides lots of detail and often I have seen the /CLIP indicator flashing, without noticing I was playing that loud.
Can you show me where you've managed to get the EVM datasheet for the TPA3251D2? As it isn't out yet on the TI website
I assume they are waiting for your design being completed
I assume they wait for your design being completed
Haha I can assure you that TI will know how to get the most out of their chip. Maybe not under all circumstances, which user experience can be crucial for as TI cannot test every implementation out there, but I am eager to see the EVM myself too
Doc I think the clip warning is supposed to be sensitive. In TI's previous designs it was implemented as a way of informing a system controller that the PSU voltage needed to be raised. This all being within the interest of low power consumption where you would run the chip from 15V most of the time then have the PSU up the voltage only when needed. To be effective this would need to be sensitive and go off on a hair trigger.
Can you show me where you've managed to get the EVM datasheet for the TPA3251D2? As it isn't out yet on the TI website
The preliminary EVM user guide ships with the preliminary EVM. I think the final version will go public within a month or so.
If I remember correctly, they had dedicated /CLIP and /OVERTEMP in the past. Indicating clipping with the /OVERTEMP is a crude way of implementation.
If I remember correctly, they had dedicated /CLIP and /OVERTEMP in the past. Indicating clipping with the /OVERTEMP is a crude way of implementation.
The data sheet actually says that the clip warning goes off when the signal is approaching clipping, not that clipping has occurred. I agree that it is strange for the clip and soft over temperature warning to be a shared flag, the last thing you would want to do is increase the rail voltage if you were approaching over temp, but there you are!
I don't suppose it was the close to clipping warning that was being triggered when you. were running on 12V as that doesn't give a lot of voltage head room, rather than over temperature.
Well, the triggering is only fraction of seconds and it dims with less power. The DS shows a hysteresis of 25K, strange.
But yes, if you raise voltage, tripping goes away.
Edit: Now I found the clipping warning..
Would guess a uC must interpret/compare this signal against a timer/threshold to decide for clipping/overtemp.
But yes, if you raise voltage, tripping goes away.
Edit: Now I found the clipping warning..
Would guess a uC must interpret/compare this signal against a timer/threshold to decide for clipping/overtemp.
Last edited:
It sounds like it could be clipping then. The peaks of a 30Hz wave last longer than higher frequencies so would be more prone to trip the indicator.
To be fair if your heatsink is adequate then you are very unlikely to hit an over temperature warning when running at say 15V. I'd imagine that if the flag triggered in that case then it would be due to clipping under normal operating conditions.
Once you've reached the full rail voltage however your still faced with a problem. But in either case clipping generally sounds bad so should be avoided as should an over temperature event. I'd imagine that if the flags were far apart and for very short duration I'd let them pass, otherwise I'd turn down the volume. Over temp could damage the board and sustained clipping your loudspeakers.
Last edited:
That is true. /CLIP is asserted when pulses starts to miss in the PWM stream out of the analog comparator, but since the feedback loop is still able to correct this to some extend (when one pulse skipped, the feedback loop can correct for this) until the input signal gets high enough to cause true clipping. You see the assertion of /CLIP to be just where the THD starts to rise rapidly with output power. This is very visible on the older TAS5613A/30B, but hardly noticeable on TPA3251D2. All closed loop (fixed switching frequency) amplifiers have this behavior.
Thanks for clarification. The somewhat, to me, strange thing is, I can virtually put out more current for higher frequencies. At fixed voltage and gain setting, 30hz "clips" earlier than i.e. 250Hz.
Anyway, testing goes on when the rest of the missing inductors arrive (StepDown and second channel)
So far a great little chip.
Anyway, testing goes on when the rest of the missing inductors arrive (StepDown and second channel)
So far a great little chip.
Agreed I did more listening to mine running the tweeters and enjoyed it immensely.
Higher frequencies get through because their peaks don't last as long and may not trip whatever system is in place.
Probably be at least a couple weeks before I start on it. So no chinese boards yet? If I beat the chinese to it, i'ma do a happy dance
Well there's no rush as there's a shortage of chips 😁