Amplifier hiss with no input signal

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I have a beginner question. I am using 3 ClassDAudio SDS-250 amp modules to drive a pair of 3-way active speakers. When there's no signal going into the amps, I can hear a faint hiss coming out of each of the speaker drivers. It's not loud, normally I don't notice it. At night, when everything is quiet and the music stops playing, that's when it becomes obvious. Is this kind of hiss normal, generally speaking, for class D amps? Is there an amp specification that would indicate the level of hiss the amp will produce? Is it possible that the way I have things connected is causing the hiss, or making it worse? Thanks!
 
I have a beginner question. I am using 3 ClassDAudio SDS-250 amp modules to drive a pair of 3-way active speakers. When there's no signal going into the amps, I can hear a faint hiss coming out of each of the speaker drivers. It's not loud, normally I don't notice it. At night, when everything is quiet and the music stops playing, that's when it becomes obvious. Is this kind of hiss normal, generally speaking, for class D amps Is there an amp specification that would indicate the level of hiss the amp will produce?

Random hiss (noise) is poduced by all active amplifiers when powered. It usually is psychoacoustically masked when music is playing. The louder the music signal, the less you notice the hiss noise, the lower the music signal, the more you notice the hiss noise. The level of hiss remains constant while the signal level is free to change. The engineering specification related to this is termed, 'signal to noise ratio' (SNR). Generally speaking, the higher the ratio (given in dB) the better.

Hiss (noise) is normal for all active amplifiers to produce, and is usually audible with ears pressed against the speakers and no signal. The question here is whether the level of hiss you are experiencing is normal for your model amplifier. Two ways to know for certain is to either perform an instrumented measurement of the SNR parameter and verify that it is within published specs. for your amp, or to compare the level of hiss by ear to the level produced by an identical model amp.

My advice is not to be concerned, so long as you don't notice any hiss from your normal listening position with no signal. Any hiss audible from from your listening chair is too high, IMO. Even so, hiss high enough to be audible from your listening chair still might be normal for your model amp, especially if you have high sensitivity speakers.

Is it possible that the way I have things connected is causing the hiss, or making it worse? Thanks!

It is possible that the way your system is connected is provoking ground-loop type noise, but that typically sounds more like a lower frequency hum than a wideband hiss.
 
Last edited:
Its quite possible that if you're driving your amp directly into your tweeters (as you said its an active set-up) that you don't have a fully optimized gain structure. Tweeters are often a few dB more sensitive than bass or mid units and the usual correction for that in a passive speaker is to fit a 'pad' resistor in series with the tweeter. With active speakers you could use a ferrite cored transformer for feeding the tweeter which would likely give better results subjectively than a series resistor.

<edit> Also work checking that your tweeter is presenting the correct termination impedance to the amp. ClassD amps normally have an LC output filter which relies on the speaker to provide damping. Insufficient damping (due to too high termination impedance) gives a rising HF response which will accentuate any hiss.
 
Last edited:
Hi Ken, thanks for the answer. Of the three drivers in my speakers, the lowest sensitivity is 93 db. If I disconnect the other two I can hear hiss from the listening position. But I live out in the country where the ambient noise is very low, especially at night when the refrigerator isn't running... I'm not able to find an SNR for these amps, not online and not in the material I received from the manufacturer. Curious, I looked at a few different amp modules and found only one with a published SNR, the Hypex NC400. (It's typical SNR is listed at 125dB. I'm going to guess that that's pretty good.) Is there another specification which is sometimes used instead? For example, the spec sheet for Hypex UcD400HG does not list a SNR value but it does list "Output noise" at 30uV (typical.) Are these be related?
 
Hypex does present a very good SNR - in part that's due to accepting a very high input level, and they also have typically lower output noise than the norm.

With the output noise at 30uV its possible to estimate the maximum output voltage given the SNR figure - it should be 10^(125/20)*30uV which I get to be 53VRMS. This matches fairly well with my vague memory of Hypex being specified around 400W/8ohms. So the answer to your last question is 'yes' they are indeed related by a straightforward formula.
 
Its quite possible that if you're driving your amp directly into your tweeters (as you said its an active set-up) that you don't have a fully optimized gain structure. Tweeters are often a few dB more sensitive than bass or mid units and the usual correction for that in a passive speaker is to fit a 'pad' resistor in series with the tweeter.

I'm using miniDSP's miniShark for the crossover, and correct the gain structure there.

<edit> Also work checking that your tweeter is presenting the correct termination impedance to the amp. ClassD amps normally have an LC output filter which relies on the speaker to provide damping. Insufficient damping (due to too high termination impedance) gives a rising HF response which will accentuate any hiss.

Hmm. Would it be possible to give me a quick lesson in how to do that?
 
The gain structure is a system-wide thing - you'd want to check that the DAC's peak output voltage (at the onset of digital clip) matches your amp's input sensitivity (at output clip).

I'd be able to help out on your question but I can't access the classDaudio webpage you referenced. It could be its blocked by virtue of me being in China or it could just be down.

You'll want to get the spec of your tweeter and check the output inductor and capacitor values in your amp to verify that you've got the optimum amount of damping. You could check this for yourself if you have the appropriate test equipment - a signal generator and an AC voltmeter that goes up to at least 20kHz (not all DMMs do). In the absence of test kit you'd need to rely on investigation of the component values in the output filter in your amp.
 
The gain structure is a system-wide thing - you'd want to check that the DAC's peak output voltage (at the onset of digital clip) matches your amp's input sensitivity (at output clip).

I'd be able to help out on your question but I can't access the classDaudio webpage you referenced. It could be its blocked by virtue of me being in China or it could just be down.

You'll want to get the spec of your tweeter and check the output inductor and capacitor values in your amp to verify that you've got the optimum amount of damping. You could check this for yourself if you have the appropriate test equipment - a signal generator and an AC voltmeter that goes up to at least 20kHz (not all DMMs do). In the absence of test kit you'd need to rely on investigation of the component values in the output filter in your amp.

The ClassDAudio web page loads for me...

I'd have to do some digging to find the specs for the tweeter, if they're available at all. They were made by Hawthorne Audio who has since gone out of business. There were only a few of them ever made. (The AMT 700, designed for open baffle and a crossover at 700hZ. I love them.) I don't know the sensitivity, but it isn't that much higher than the sensitivity of the woofers. They don't hiss much louder than the woofers either. The mids are the sensitive drivers in this system. 100dB.

In the miniShark software, where the crossover is defined, I made an adjustment of -5 dB to the output for the mids. Somewhat surprisingly, the woofers and the tweeters are equivalent with zero adjustment. This after measuring the responses of each driver using REW. (I do have two woofers in parallel, making 4 ohm impedance and doubling the power delivered by the amp.) Does this say anything about either the gain structure, or whether the damping is optimum? I should say as well that when music is playing I never hear clipping or anything sounding "off". It always sounds very good to my ears.
 
The 5dB adjustment in the software tells me you likely have 5dB too much noise in the mid, by virtue of its high sensitivity. The software can only reduce the digital gain on the mid channel, the noise will stay relatively constant.

Given the variety of loudspeaker load impedances, designers of classD amps which don't have PFFB (post-filter feedback) have to make a compromise between 4 and 8ohms loading so they might design for 6ohm. In such a case a 4ohm load is over-damped and an 8ohm, under-damped.
 
The 5dB adjustment in the software tells me you likely have 5dB too much noise in the mid, by virtue of its high sensitivity. The software can only reduce the digital gain on the mid channel, the noise will stay relatively constant.

So then I think you would feed this driver with a ferrite cored transformer to balance the gain, instead of doing it in software? The transformer would drop the voltage between the amp and the speaker. Is there a straight-forward way to figure out the needed transformer specs?

Would this excess noise be present at all volume or is it only an issue at high volumes?


Given the variety of loudspeaker load impedances, designers of classD amps which don't have PFFB (post-filter feedback) have to make a compromise between 4 and 8ohms loading so they might design for 6ohm. In such a case a 4ohm load is over-damped and an 8ohm, under-damped.

The specs state RMS Power per channel is 125W @ 8 ohms and 250W @ 4 ohms. Does this imply that he designed for both, so that damping would be correct in either case?
 
Doing the 5dB attenuation in hardware rather than software will certainly improve the noise. However while I'm sure a ferrite transformer is well suited to the tweeter, I'm not so certain about the mid, owing to the lower frequencies. Ferrite is best suited to higher frequencies though it may indeed work OK at mid-range freqs, depending on where the lower cut-off is.

No, the two output power specs imply that he probably compromised between 4 and 8ohms. The only way I know of to get the damping correct in both cases is from using post-filter feedback. As far as I know only Hypex is doing that, though there may be others turning up now as some of TI's later designed chips can be operated in this fashion.

As to whether the noise is present at all levels, it depends whether you're using digital volume control for your system volume. If so, then yes.
 
Doing the 5dB attenuation in hardware rather than software will certainly improve the noise. However while I'm sure a ferrite transformer is well suited to the tweeter, I'm not so certain about the mid, owing to the lower frequencies. Ferrite is best suited to higher frequencies though it may indeed work OK at mid-range freqs, depending on where the lower cut-off is.

What about putting a resistor in series between the input signal and the amp? Would that be a good solution? If yes for the mids, would this approach also be good for the higher frequencies of a typical tweeter?

No, the two output power specs imply that he probably compromised between 4 and 8ohms. The only way I know of to get the damping correct in both cases is from using post-filter feedback. As far as I know only Hypex is doing that, though there may be others turning up now as some of TI's later designed chips can be operated in this fashion.

Trying to get my head around damping. What is considered "correct" damping?

ClassD amps normally have an LC output filter which relies on the speaker to provide damping. Insufficient damping (due to too high termination impedance) gives a rising HF response which will accentuate any hiss.

This confuses me. Perhaps I just need you to clarify what "termination impedance" is in reference to. I want to think it is the speaker impedance, but that wouldn't make sense given the definition of damping factor. I.e. higher speaker impedance equals higher damping factor.
 
What about putting a resistor in series between the input signal and the amp?

You'd need two resistors to form a 'potential divider'. This would certainly work yes.

Would that be a good solution? If yes for the mids, would this approach also be good for the higher frequencies of a typical tweeter?

In my experience use of a resistive attenuator is likely to reduce dynamics as you're driving your amp from opamps inside your MiniSharc and you're feeding a classD amp. Normally a classD has lowish input impedance and performs best when driven from a low source impedance. But its worth trying to see how you get on. Yes you can use this approach for the tweeter too, but you have no need to adjust its sensitivity right?

Trying to get my head around damping. What is considered "correct" damping?

'Correct' damping is when you have a Butterworth alignment, i.e. a flat passband leading to a smooth (monotonic) roll-off with no peaking at the corner frequency. In such an arrangement the Q = 1/SQRT(2).

This confuses me. Perhaps I just need you to clarify what "termination impedance" is in reference to. I want to think it is the speaker impedance, but that wouldn't make sense given the definition of damping factor. I.e. higher speaker impedance equals higher damping factor.

The damping I'm talking about has nothing at all to do with 'damping factor' which is an entirely misleading concept. 'Damping factor' is ostensibly about damping of the bass resonance of the speaker, here we're about damping the HF resonance of the amp's LC output filter. You can find a paper or two about LC filter design for classD amps on TI's website, this will fill in the details for you. App note sloa119b is as good a place to start as any : ignore section 1 and dive straight into 2.
 
You'd need two resistors to form a 'potential divider'. This would certainly work yes.

I hope you don't mind but I need to back up. I'm not seeing how, in terms of what the amp sees, doing volume control digitally is different than putting resistors between the signal and amp. When volume is lowered digitally, doesn't the DAC respond by reducing the voltage of its output? So it seems to me that the amp should see essentially the same signal in either case. What am I missing?
 
I am using 3 ClassDAudio SDS-250 amp modules to drive a pair of 3-way active speakers. When there's no signal going into the amps, I can hear a faint hiss coming out of each of the speaker drivers.

You may want to read An Amplifier for all Mankind: Class D Audio SDS-440C Amplifier Review | Featured Reviews, Hi-Fi Systems Reviews | HighFidelityReview - Hi-Fi systems, DVD-Audio and SACD reviews .

In the manufacturers reply at the bottom of the page note " The amplifiers can easily be adjusted for super-high efficiency speakers, but in my opinion, if we make this adjustment to all amps, it might make the sound more like other class D amps on the market, meaning the sound would be a little flat compared to the exciting, lively sound and dynamics our amplifiers are capable of. "

In other words you can likely tame the hiss, but at a price.

Grant.
 
I hope you don't mind but I need to back up.

Quite the opposite of minding, I'm encouraged that you're concerned to get clarity.

I'm not seeing how, in terms of what the amp sees, doing volume control digitally is different than putting resistors between the signal and amp. When volume is lowered digitally, doesn't the DAC respond by reducing the voltage of its output? So it seems to me that the amp should see essentially the same signal in either case. What am I missing?

I'll take the opportunity to back up too, as I've made an assumption about your system which might not be valid. I've assumed that you want your system's noise to be primarily determined by the noise on the recording and that you may wish to play higher resolution music than RBCD. If you only want to play RBCD (16bits depth recordings) and you don't at all care about noise when no music is playing then you probably have no need to be concerned about gain structure.

Since your opening post was about noise with nothing playing it seemed to be a reasonable assumption to make. You're correct about what happens when the volume is lowered digitally - the DAC is then given smaller (amplitude) numbers to convert back to analog. However the noise of the DAC when no music is playing comes from the DAC's output circuits and not from the digits fed to it, since with no music playing the DAC's converting a value of zero constantly. It cannot generate a true zero in response to this input, there will always be some noise and its this noise which is most likely responsible for any hiss you hear under quiescent conditions. To put it another way, your DAC's ultimate signal-to-noise ratio is the bottleneck in the system.

When you attenuate digitally your maximum signal is reduced (by 5dB in your case) and so the signal-to-noise ratio (SNR) decreases by the same amount, 5dB because the DAC's output noise is constant. With a resistive divider at the DAC's output, provided the resistor values are low enough not to contribute significant noise of their own, the DAC's residual noise will also go down by 5dB, along with the signal. So the SNR remains the same.
 
Here is another thread in case you didn't find it in your search. It's been asked and discussed before. It's not the way you have things connected I suspect.

Class-D and The Elephant in the Room = Hiss

Grant.

My takeaway from that thread is that Class D amps are not inherently hissy simply because they are Class D, and that hiss can be caused by a lot of things.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.