Ok so nothing obvious came up in my searches specific to class-D, and its an ideas as old as the hills..
I always *knew* that you should never run an amp without a load attached.
ive seen this vaguely repeated many times over the years.. however its never been explained to me why.
now i have a rather silly situation where im testing my 6 channel amp, and the speakers are not ready, i only currently have 2 bookshelf speakers, and a pile of old drivers from various old projects.
so i have 4 of the channels muted, and the other two running my bookshelf speakers. but to be safe, i have 4 old 4 inch drivers connected to the muted channels during this initial test.
is this necessary? is it true its harmful to run an amp without a load? specifically class D? even with no signal on input?
very specifically we are talking about 3255 EVM..
thanks, for helping me understand an old wives tale better!
cheers.
I always *knew* that you should never run an amp without a load attached.
ive seen this vaguely repeated many times over the years.. however its never been explained to me why.
now i have a rather silly situation where im testing my 6 channel amp, and the speakers are not ready, i only currently have 2 bookshelf speakers, and a pile of old drivers from various old projects.
so i have 4 of the channels muted, and the other two running my bookshelf speakers. but to be safe, i have 4 old 4 inch drivers connected to the muted channels during this initial test.
is this necessary? is it true its harmful to run an amp without a load? specifically class D? even with no signal on input?
very specifically we are talking about 3255 EVM..
thanks, for helping me understand an old wives tale better!
cheers.
Class AB amplifiers can run without load.
Class D amplifiers should not run without load because the load is used to damp the LC output filter and prevent the filter-chokes from generating excessive voltages. Some class D amplifiers may survive running without a load, in particular if they are protected with catch-diodes.
Class D amplifiers should not run without load because the load is used to damp the LC output filter and prevent the filter-chokes from generating excessive voltages. Some class D amplifiers may survive running without a load, in particular if they are protected with catch-diodes.
When testing an amplifier, you always should use a noninductive load resistor of the proper value
and power rating on each channel when powered. Only when the amplifier is found to be operating
normally should it then be connected to a speaker.
i assume you mean rated for the actual max output of the amp? blimey if so.. id have to find /hack some nice 300+watt resistors..
in any case, the amp is basically a pre-constructed evaluation board, and even the user guide suggests connecting speakers (or a resistive load) for first test.
i have momentarily run the amp without loads connected ( max 5 seconds) during construction, with no noticeable ill effects, and no triggering of the inbuilt safety shutdown.
in any case, at the stage im at now ( amp is finished and screwed shut in its case) im testing the thermal characteristics and long term behavior of the amp at gradually increasing loads.. Once i get the proper speakers built (likely a month or so) i can start really winding it up..
so far ive confirrmed it remains cool with 2 of the 6 channels running driving circa 50w loads, and its been left on overnight with no ill effects...
baby steps!
It isn't safe to use under-rated load resistor(s), since a fault could cause full power output,
or a high DC output. Then the under-rated load resistor(s) would burn up.
well in that case, for this project, the most catastrophic failure would be a full power dc output to the speaker terminals from the 800w/1.2kw peak switchmode power supply..
thats a heck of a big resistor.. possibly need to be in a bucket of mineral oil!
this is very interesting.. im glad im getting this advice now ive basically finished, or id have been in a right flap about getting really enormous power resistors for six channels.
thats a heck of a big resistor.. possibly need to be in a bucket of mineral oil!
this is very interesting.. im glad im getting this advice now ive basically finished, or id have been in a right flap about getting really enormous power resistors for six channels.
The zobel should dissipate some energy from the LC. Either way doubt the Q of the LC is high enough to destroy things.
Is the early tube amps that really hate having no load. Those can be damaged.
Anyway the dummy load is also needed for you to measure the actual performance of the amp while designing.
Is the early tube amps that really hate having no load. Those can be damaged.
Anyway the dummy load is also needed for you to measure the actual performance of the amp while designing.
1) What part would be destroyed by doing so?
I should have killed all my class D boards if this is true as I tested them all with 1 speaker connected.
We should have seen many more instances of class D boards dying if this would be true, so I guestimate the opposite is true; you can kill some Class D boards by running them without a load.
2) Do you have a hands on experience bbq-ing a board? Which one?
3) Do you realize that all anti plop (relay) boards have no load when you turn them on?
I found the irs2092 with no load would often latch at 17VDC.
Also the output LC would ring at high voltages.
I put a higher voltage cap on the output to cope with this.
I sold one online and the buyer complained that it was outputting 17 volts with no speaker and he was afraid to connect his speaker. I had to explain the 2092 sometimes needed a load to oscillate properly.
Also the output LC would ring at high voltages.
I put a higher voltage cap on the output to cope with this.
I sold one online and the buyer complained that it was outputting 17 volts with no speaker and he was afraid to connect his speaker. I had to explain the 2092 sometimes needed a load to oscillate properly.
I have BBQed a TDA8954. There are two means to protect the chip: catch-diodes, preventing the voltage on the choke terminals going outside of the supply voltage range, and snubber circuits. The catch diodes may be implemented on the chip.
ddapkus is Audio Apps Manager at TI and he wrote a posting on 16.06.13 about a TPA3116 problem:
TPA3116D2 Amp
Just after he made a second posting (following the first) where he shows a schematics with catch-diodes implemented (8x1N4004).
The problem seems to be that a rather undamped series LC resonant circuit can generate important voltages on the choke. The voltage generated may cause an over-voltage at the output switches or the filter capacitor. In my TDA8954 case, the snubber circuit was wrongly implemented (values without effect). My upper output switch got permanently stuck to the supply rail, probably because the driving circuit for that switch was exposed to an over-voltage through the boot-strap capacitor.
It is rather an inductive over-voltage problem than an excessive power problem.
I have seen catch-diodes implemented on TDA7498E boards and TA2022 boards, so it is not a new discovery.
You question 3) is very relevant both for protective relays and for A<>B testing. No load on a class D amplifier is a risk, even for a short moment. The relay may have both make and brake terminals and a dummy load can be arranged on the terminal opposite of the speaker.
Catch-diodes may be alternative solution as proposed by ddapkus.
I always try to connect the load before turning ON a class D amplifier, and I am happy to hear that you got away with turning them ON without a load (perhaps there was a snubber circuit or catch-diodes included?). I do not say it will go wrong, I say it may go wrong. ddapkus from TI should know.
ddapkus is Audio Apps Manager at TI and he wrote a posting on 16.06.13 about a TPA3116 problem:
TPA3116D2 Amp
Just after he made a second posting (following the first) where he shows a schematics with catch-diodes implemented (8x1N4004).
The problem seems to be that a rather undamped series LC resonant circuit can generate important voltages on the choke. The voltage generated may cause an over-voltage at the output switches or the filter capacitor. In my TDA8954 case, the snubber circuit was wrongly implemented (values without effect). My upper output switch got permanently stuck to the supply rail, probably because the driving circuit for that switch was exposed to an over-voltage through the boot-strap capacitor.
It is rather an inductive over-voltage problem than an excessive power problem.
I have seen catch-diodes implemented on TDA7498E boards and TA2022 boards, so it is not a new discovery.
You question 3) is very relevant both for protective relays and for A<>B testing. No load on a class D amplifier is a risk, even for a short moment. The relay may have both make and brake terminals and a dummy load can be arranged on the terminal opposite of the speaker.
Catch-diodes may be alternative solution as proposed by ddapkus.
I always try to connect the load before turning ON a class D amplifier, and I am happy to hear that you got away with turning them ON without a load (perhaps there was a snubber circuit or catch-diodes included?). I do not say it will go wrong, I say it may go wrong. ddapkus from TI should know.
Last edited:
Thk for sharing the catch-diodes (8x1N4004) as chip-protector.
I have similar concern: am using a Fx502spro - it is a pre-build a TPA3250 Class D configure to BTL, without schematic cant re-config to PBTL. Sometimes need 1 of channel without load.
Currently I have inputs muted (shorted), and having a 8ohm 2watt resistor connected to output without speaker, be also sufficiently protect the chip from damage?
I believe any dummy load with a relevant value (4/8 Ohm), that is not getting so hot it risks being damaged, will ensure the necessary damping.
thks for affirming, it gives me some peace.
clarify a little further regarding catch-diode: Would need 4 x 1N4004 for each channel? ie. 2 x 1N4004 at L-Spkr-out positive terminal & 2 x 1N4004 at L-Spkr-out negative terminal?
clarify a little further regarding catch-diode: Would need 4 x 1N4004 for each channel? ie. 2 x 1N4004 at L-Spkr-out positive terminal & 2 x 1N4004 at L-Spkr-out negative terminal?
I'm a bit surprised of 1N400x's being recommended. If anything, i'd rather expect some fast-recovery diodes in those positions (since you have the several-hundred-kHz switching waveform going on, before the output filter). Something like UF400x's or MUR1x0's or something.
For my TDA8954 I used Schottky diodes with a 1A continuous rating. They are fast and can handle surges much higher than 1A. 100V types I believe.
I ran the blue TPA3118 without load and C18 exploded. C34 is its counterpart on the other channel. It is 680nF in the spec sheet, but the installed value is 1nF.
To clarify this problem: The output LC-tank has a resonant frequency in the range of 50 kHz. If you connect a 50kHz-signal with the input, the output of the amp sees a very low impedance of the series resonant tank. In that case
1) The voltage at the resonant cap (680nF) may rise far above supply voltage thus destructing the cap.
2) Output current exceeds limits and over-current protection trips.
But...
If you do not connect 50kHz to the input, the resonant tank is not excited and these strange things do not happen.
With TIs TPA-chips I could observe these things in reality and I can say that no single chip died this way until now - over current protection does its job.
Clamp diodes limit output voltage at resonant overdrive thus reduce current flow as well. And their power dissipation is far below an appropriate dummy resistor - all in all a preferable bet imho.
Besides the proposals above it is possible to dampen output resonance with a simple RC- post-feedback filter with neglible power loss. As a result square response is nearly perfect without ringing - at any load.
Attached are measurements of my TPA-3255-pcbs, measured with 5Ohm and with no load
Last edited:
Attachments....
Top trace: input signal
Bottom traces: output signals of bridge legs
Top trace: input signal
Bottom traces: output signals of bridge legs
Attachments
Last edited:
- Status
- Not open for further replies.