Ambika -Noise on all voices [Solved]


I just finished building an Ambika with 6 SMR4 voicecards. All is working fine but the sound is really noisy and disappointing.

When I play a pure sine wave on oscillator one with filter freq 127 and resonance 0 I hear strange noise in the background which sounds a little bit like a high frequency switching power supply. It appears on all voices for single output and sum.

When I build the motherboard I discovered that the 8V regulator delivers only 7,93 V. Might this regulator probably the source?

I checked the spectrum with baudline and can see a second strong frequency.
Spectrum chart:


> Might this regulator probably the source?

No. That’s less than 1%, perfectly fine for this kind of regulators. The actual value of the supply rail is not relevant to the good performance/operation of the filter circuits.

In the clip you have posted, the background noise is 80dB below the level of the signal in the sustained part. It’s lower than on most analog synths I’ve played…

The third harmonic is due to the distortion brought by the filter - you can attenuate it by reducing the level of the signal going into the filter.

Do you more audio clips showing problems? If you have a Shruthi, what are the differences you hear?

Ok, thanks a lot for the clarification.

I really thought there is an issue because I never discovered this on my Shruti SMR4.
The Ambika sound differs from Shruti SMR4, and I thought it would sound nearly the same except the resonance.

A 1to1 comparison shows that the Ambika is crisper, that might be the reason why I discovered noise as an issue. I will give it a try and see how it does with real patches.

Here is a side by side comparison with open filter, no resonance and a triangle wave. The Ambika has a metallic background noise in comparison to the Shruti.

Shruti Ambika Shruti- Ambika

Here is an example the noise adds up if I play a melodie with simple sine waves on Ambika (open filter, no resonance)

With your triangle wave, they sound pretty close to me with my ancient ears, except the Ambika version has a resonant peak that sits about 10dB higher at ~7.5kHz than the triangle’s harmonics (are you sure you’ve got your filter resonance turned fully down and the frequency turned right up?)

Also, the Ambika version has slightly higher levels of aliasing noise (about 12dB higher @18kHz for instance) - I think this accounts for the metallic background noise you’re hearing, and it becomes fairly obvious if you notch out the fundamental and just compare the contributions of the harmonics and aliasing noise.

However, this aliasing noise is still fairly low in both examples - Shruthi is about 92dB down compared with the fundamental at this frequency, while Ambika is 80dB down.


Hi Martin,

The filter was completely open and reso at 0.

I had to tune the SMR4 filter cards of the Ambika to a really high tone at C4 until I achieved double frequency over 2 octaves. This was much lower at the Shruti and might be a reason for the sound difference.

I also did a lot of side by side comparison of patches today.

The sound of the Ambika is much brighter than the Shruti SMR4. If I have a noisy patch on Ambika it’s also noisy on Shruti but a more attenuated in the high frequencies.

I think we can close this thread as it seems to be a normal behavior.

Ah well, I remember Olivier saying that the Shruthi SMR4 filter is always closed down a little (~16kHz at maximum setting did he say?), whereas the Ambika is apparently cleaner, so this might indeed account for you hearing this small amount of extra background noise.

It doesn’t sound as if it’s worrying you though, which is all that matters


The Shruthi uses PWM instead of DACs. To make it sure that the deadly 39kHz spike in the spectrum (from PWM) is attenuated, the maximum filter cutoff is limited to around 15kHz.

Ambika uses a DAC so there’s no such limitation, and the filter reaches higher frequencies - in which case it is almost fully open and it reveals all the high-frequency trash (8-bit wavetables + 39kHz aliasing) present in the signal.

You can probably get both to match by lowering the cutoff by 6 or 7 units on Ambika. For very round/pure sine/triangle sounds, don’t hesitate to lower the cutoff to clean the digital mess…

Thanks a lot for the explanation.

One last question for this thread.
I missed some parts in the order and purchased the op amps for the voicecards locally.

I used:
TL074IN instead of TL074CN
TL072CN instead of TL072ACP

Does this effect the sound?

No, the effect of op-amp choice is negligible compared to noise/distortion coming from the filter OTAs and from the 8-bit digital synthesis itself.

Wow, thanks a lot for the insights. I love this forum and the great support.
It’s now time to go for polyphonic madness with my new little friend :slight_smile:

Olivier: “the effect of op-amp choice is negligible compared to noise/distortion coming from the filter OTAs and from the 8-bit digital synthesis itself.”

What he means of course is “the effect of op-amp choice is negligible compared to the carefully designed-in extra character coming from the filter OTAs and from the 8-bit digital synthesis itself.”

Yes, this forum is great isn’t it?


Actually - i doesn’t get better, does it?