single supply LM1875 oscillation

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've soldered up a prototype lm1875 amp on padboard, following the single supply circuit on the datasheet.

With no input connected, when I touch the body of the input cap (large size poly 4uF) and the output cap (3300uF electro) simultaneously, the current draw shoots up - presumably this is the amp going into oscillation? When an input jack cable is connected to + and gnd (but not connected to any source), this doesn't happen. Is this 'normal'? The layout is tight, v. similar to LM1875 Stereo Power Amplifier, but with the output cap on the board.


Also there is a large Thump at turn on. DC at output momentarily reaches nearly full supply voltage, then sits at 1/2 supply voltage until a load (speaker) is connected. Is this normal for the output to have high DC until load is connected? I thought the output cap was there for blocking DC.

First try at a chipamp, so still learning.
 
From your description, it sounds like you're running the LM1875 from a single supply rather than a split (±V) supply, hence the output cap. When you turn on the amp, the output cap needs to charge. When a speaker is connected, the charging happens through the speaker, which results in a turn-on thump. When the speaker is not connected, the cap does not get to charge, so you'll see VCC/2 when you measure on the output. Connect a large resistor (say 10 kΩ) from the output to ground and you'll see the output voltage stabilize faster (probably within a minute or so).

If there is nothing connected to the input of the amp, I suggest connecting the input to ground through a 47 kΩ resistor. A floating input is generally a recipe for disaster.

It is pretty common that if you touch various components with the input floating that mains hum will be picked up by the amp and amplified. That'll cause the current to shoot up (and lots of hum to be emitted by the speaker). I suppose it's also possible that you create a positive feedback loop, which will cause oscillation. So don't touch the input and output cap. :)

Tom
 
From another thread, lm1875:

lm1875 single supply.png

lm1875 schematic.png

and lm3886

lm3886 single supply.png

lm3886 equiv schem.png

1JPG=1000ASCII
 
Thanks all. That's really helpful

Tomchr, it is indeed single supply. Is there a way to avoid the thump caused by the o/p cap charging through the speaker or is this just an unavoidable drawback of single supply designs? The reason I was looking at the single supply version was the DC fault protection given by the cap but if its going to throw VCC/2 through the speaker every time I turn it on it hardly seems worth it anymore.

The 47k resistor to ground you suggested - is this what R4 (1Meg to ground at input) is for on the datasheet schematic?

nigelwright7557, I tried adding an RF filter with a 1k series resistor and 1nF to ground (because that's what I had to hand) - works a treat. No longer oscillates when the caps are touched. My gain is set at 23x so shouldn't be any problem there.

Is there a specific order the components on the input should go in -
47k to gnd --> DC blocking cap --> rf filter

Presumably if the 47k goes after the input cap it forms a HP filter. I have a 4uF cap so the cut off freq would be pretty low anyway.
 
Any issues with putting the input cap last? 47k to gnd --> rf filter --> Input cap? No space on the current layout to put the rf filter after the series cap.


My circuit is shown on attached, with changes from datasheet circuit in red.


Does the 10k pot negate the need for the 47k resistor as the input is no longer floating?

The rf filter as shown has a frequency of about 159kHz - seems lower than i've seen elsewhere. Is this too low?


Any other issues with the attached circuit?
 

Attachments

  • lm1875 single supply.png
    lm1875 single supply.png
    33.6 KB · Views: 493
Any issues with putting the input cap last? 47k to gnd --> rf filter --> Input cap? No space on the current layout to put the rf filter after the series cap.
.......no (if the input filter is not too far away from the opamp)

Does the 10k pot negate the need for the 47k resistor as the input is no longer floating?
............yes

The rf filter as shown has a frequency of about 159kHz - seems lower than i've seen elsewhere. Is this too low?

...........not too low with 1k/1nF, some designs use lower frequencies.
but your 10k pot has a (maximum) output impedance of 2,5k in the mid position,
so you get only ~50kHz (1k+2,5k)...........that's low.

it is good to have a res at the output (paralell to the speaker load) to reduce the turn-on-thump,
something like 330 ohm...........x kOhms.
 
Last edited:
Member
Joined 2009
Paid Member
Tomchr, it is indeed single supply. Is there a way to avoid the thump caused by the o/p cap charging through the speaker or is this just an unavoidable drawback of single supply designs?

The only way to avoid it is to use an output mute circuit, which mutes the output until the cap is charged.

The thump is a drawback of the single supply design.

The reason I was looking at the single supply version was the DC fault protection given by the cap but if its going to throw VCC/2 through the speaker every time I turn it on it hardly seems worth it anymore.

I agree. In addition, large electrolytic caps tend to add audible distortion.

The 47k resistor to ground you suggested - is this what R4 (1Meg to ground at input) is for on the datasheet schematic?

Yep. R4 in Figure 2 of the data sheet.

Does the 10k pot negate the need for the 47k resistor as the input is no longer floating?
............yes

Well... No. When pots fail, the wiper usually opens. Even before they fail, they tend to skip, which results in a scratchy sound when the pot is turned. Having the 47 kΩ there will prevent the amp from going haywire when the pot fails.

...........not too low with 1k/1nF, some designs use lower frequencies.
but your 10k pot has a (maximum) output impedance of 2,5k in the mid position,
so you get only ~50kHz (1k+2,5k)...........that's low.

True. Reducing the 1 nF to 220 pF would bring the cutoff frequency back up in the 150-200 kHz range.

it is good to have a res at the output (paralell to the speaker load) to reduce the turn-on-thump,
something like 330 ohm...........x kOhms.

That'll reduce the thump, but not eliminate it. Lower value resistor in parallel -> more power is dissipated in the "anti-thump" resistor.

Tom
 
A capacitor on the output only helps with failure if the output goes to positive rail.
If it goes to negative rail capacitor is revere biased and becomes a short circuit putting B- on your speaker.

I use dual rail amps but add a speaker protect circuit using a little PIC and a relay.
It doesn't put on relay for 4 seconds on power up to stop thumps.
If it detects DC on output for more than 500mS it shuts off the relay.
 
A capacitor on the output only helps with failure if the output goes to positive rail.
If it goes to negative rail capacitor is revere biased and becomes a short circuit putting B- on your speaker.

That applies to an amp running on a dual/split supply. In OP's case, if the output goes to the negative rail (= ground), the cap will discharge through the speaker. That'll create a thump similar to the turn-on thump, but shouldn't damage the speaker.

I use dual rail amps but add a speaker protect circuit using a little PIC and a relay.

Relays are actually not all that great for speaker protection. They tend to arc over when they break high currents, i.e. when they're needed to protect against a short to the power supply rails. You can see the rather spectacular fireworks here: Speaker DC protection with relays. That's why I use MOSFET switches in my Guardian-86 DC protection circuit.

Some relays also add distortion. This has been characterized both by Douglas Self and Bob Cordell and described in their respective power amplifier design books.

Tom
 
Its looking like the drawbacks of the single supply design are outweighing the benefits. The main reason for using single supply was for the protection from the o/p cap (without having to use complicated protection circuits). Seems like its going to be noisy and thumpy! I tried the resistor in parallel with the o/p but it made no real difference.

The amp will primarily be used for testing speaker drivers with ARTA etc so it doesn't need to be fancy, but I want it to be robust (work with 4ohm load, reasonably low distortion, not fry speakers).
Looks like I risk blowing tweeters with the single supply version. Perhaps a split supply design is in order.

So... how risky is it to use the split supply without a protection circuit (relays or otherwise)? Are chipamps prone to smoking speakers?
 
So... how risky is it to use the split supply without a protection circuit (relays or otherwise)? Are chipamps prone to smoking speakers?

Chipamps tend to be pretty rugged devices. They're protected against a variety of abusive conditions. The LM3886 (and similar Overture Series devices) is protected against SOA, over-temperature, and over-current conditions. You can check the data sheet, but I'm 99% sure these protection circuits are included in the LM1875 as well. The odds of amp failure are pretty darn low. Thus, I generally don't use protection circuits on chipamps.
That said, if you do want a protection circuit that doesn't impair the THD of the amp, I'll be happy to get you hooked up.

As far as smoking speakers goes, the main threat is amplifier oscillations. As long as you follow the data sheet recommendations for decoupling, output snubber (aka Zobel filter), etc. and keep your layout reasonably tight you should be fine.

Tom
 
Last edited:
Right then, redesigned as a single supply. I'll live dangerously and leave off any protection for now.
I'd appreciate any comments on the attached design.

Does the o/p impedance of the pot (max 2.5k) still affect the RF filter if it is upstream of the input cap, as shown? - I assume so, in which case I make the Fc of the filter about 185khz, including the 400R o/p impedance of the soundcard it will be used with (is it right to include that?).

Signal gnd and power ground separated by a 10R resistor. Is this good practice? Is a 1/4w resistor sufficient?

Zobel on the lm1875 datasheet is 1R with 220nF, this gives an Fc of 723kHz doesn't it? - seems very high. The 10R, 100nF (as per lm3886 datasheet) shown gives 159kHz.

Thiele network not located on the board but near the speaker binding posts. Values used as per lm3886 datasheet.

Thanks
 

Attachments

  • lm1875 circuit.jpg
    lm1875 circuit.jpg
    687 KB · Views: 255
Its looking like the drawbacks of the single supply design are outweighing the benefits. The main reason for using single supply was for the protection from the o/p cap (without having to use complicated protection circuits). Seems like its going to be noisy and thumpy!
Hey, you can´t sell your cake and eat it, pick one!!! :D

Balance "noisy and thumpy" (and that if you can´t do it right) to "burnt speaker":eek: and make a decision :rolleyes:

Ok, I thought so :)

In any case, there is a third way to protect speakers: Peavey and many PA amplifiers use an "SCR/Triac Crowbar Protection", search or google it.

Very simple and reliable, it places a Triac in parallel with speaker, and between amp output and Triac you have a regular Fuse.

When circuit detects DC above a dangeros level (so far like other protection circuits do) , Triac triggers and shorts amp Out to Ground so Speaker is protected.
No chattering or slow or fusing welded relay contacts invoved, protection is safe and certain.

In a few milliseconds speaker fuse blows , but speaker is already shorted across by then.

Problems?
Of course, *everything* carries some problem associated to its advantages, this is the Real World.

1) shorting Triac most probably will blow your Chipamp .
No big deal, it´s already dead, so, why worry?

2) sometimes (not always) Triac dies too and needs to be replaced.
Since a $2 Triac protects a $40 to $7000 speaker, again no big deal.
 
Last edited:

PRR

Member
Joined 2003
Paid Member
...Does the o/p impedance of the pot (max 2.5k) still affect the RF filter if it is upstream of the input cap, as shown?

As you assume. Of course it does. The location of the subsonic block cap makes no difference. Since the source is probably more predictable(??), move the supersonic cut there.

...Zobel on the lm1875 datasheet is 1R with 220nF, this gives an Fc of 723kHz doesn't it? - seems very high. The 10R, 100nF (as per lm3886 datasheet) shown gives 159kHz...

The Zobel complements the speed of the internal power devices. Use the values suggested for the chip you are using, not some other chip.
 

Attachments

  • acoustic1-a.gif
    acoustic1-a.gif
    18.3 KB · Views: 242
Hi PRR

My pot will be off the board and connected with wire so I located the RF filter as close to the chip as possible to take care of anything that gets picked up by the wires on the way.

Thanks for the tip about the Zobel - I'd seen others use 10R & 100n with the 1875 in their designs and the lower frequency seemed to make sense. I'll go back to 1R & 220n.
 
My pot will be off the board and connected with wire so I located the RF filter as close to the chip as possible to take care of anything that gets picked up by the wires on the way.

That's a good choice.

Thanks for the tip about the Zobel - I'd seen others use 10R & 100n with the 1875 in their designs and the lower frequency seemed to make sense. I'll go back to 1R & 220n.

I'd use whatever TI uses in their data sheet.

For the decoupling, I'd read the data sheet as well. Keep in mind that lots has happened in the area of ceramic caps since the data sheet was written. Thus, you may find that you can improve the performance of the decoupling network if you up the size of the smallest cap from the 100 nF used by TI. More info is available in my Taming the LM3886 - Supply Decoupling article.

Tom
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.