Hi All,
I was measuring the DC supply in my DAC and am surprised to find that there appears to be significant broadband noise.
Measuring using a DMM the ripple is below 1mV - nice and stable. However using a 100Mhz Tek scope I can see ~50mVpp noise at 20Mhz bw and ~100mVpp noise at full bw. FFT plot shows that these are wideband noises.
What is even more interesting is that the noise is still there even if I connect *both* the tip and the ground clip of the probe on the same ground bus! By contrast, shorting the probe *without* touching the ground bus did not show the noise, powering off the DAC also eliminated the noise so the noise must be generated by the DAC circuit.
I am puzzled as to if this is something wrong or this is typical of digital circuits where the noises are radiated by the DAC circuit?
The DAC functions normally otherwise and does not show noticeable noise through the speakers.
Any ideas or suggestions are appreciated. Thanks.
I was measuring the DC supply in my DAC and am surprised to find that there appears to be significant broadband noise.
Measuring using a DMM the ripple is below 1mV - nice and stable. However using a 100Mhz Tek scope I can see ~50mVpp noise at 20Mhz bw and ~100mVpp noise at full bw. FFT plot shows that these are wideband noises.
What is even more interesting is that the noise is still there even if I connect *both* the tip and the ground clip of the probe on the same ground bus! By contrast, shorting the probe *without* touching the ground bus did not show the noise, powering off the DAC also eliminated the noise so the noise must be generated by the DAC circuit.
I am puzzled as to if this is something wrong or this is typical of digital circuits where the noises are radiated by the DAC circuit?
The DAC functions normally otherwise and does not show noticeable noise through the speakers.
Any ideas or suggestions are appreciated. Thanks.
This is either common mode noise, or more likely, stray pickup due to the probe's ground lead.
Sometimes the only way to measure cleanly in a situation like this is to remove the probe tip
and ground wire, and use a makeshift ground probe. Sometimes these parts come with new probes.
https://www.youtube.com/watch?v=-4q8geE5ef8
terminology - What is the name of this springy type oscilloscope probe accessory? - Electrical Engineering Stack Exchange
Last edited:
Sometimes the ethernet or whatever is showing up. Switch off as much as you can. I often solder the test cables to the thing I measure. Try a few cables as often one is about to fail. Measure a PP3 battery. It might have hiss. It should not have hum nor RF. One I tested recently did. That's a testing problem. I was very pleased as the PP3 did have more of this noise than the PSU I had just made. It was the high internal resistance of the PP3 allowing in the environment.
Thanks for the ideas guys - my responses below.
> This is either common mode noise, or more likely, stray pickup due to the probe's ground lead.
Thanks. The thing is the noise went away it when the DAC is switched off. So if this is stray pickup it is picking it up from the DAC's RF radiation?
The low inductance ground lead in the video is interesting but it won't work in my situation as the ground point is a few inches away from the various measurement points.
> Measure a PP3 battery. It might have hiss. It should not have hum nor RF.
I did measure some other power supplies in the past and they did not exhibit such noises. However the big difference is that those measurements were done when the PSU is *not* connected to the device in use so perhaps that's the key. But then if one cannot measure the PSU noise while in use the results won't be very useful!?
> Is your Tek scope battery powered?
No. It's AC powered - the TDS2000 series with two channels.
> This is either common mode noise, or more likely, stray pickup due to the probe's ground lead.
Thanks. The thing is the noise went away it when the DAC is switched off. So if this is stray pickup it is picking it up from the DAC's RF radiation?
The low inductance ground lead in the video is interesting but it won't work in my situation as the ground point is a few inches away from the various measurement points.
> Measure a PP3 battery. It might have hiss. It should not have hum nor RF.
I did measure some other power supplies in the past and they did not exhibit such noises. However the big difference is that those measurements were done when the PSU is *not* connected to the device in use so perhaps that's the key. But then if one cannot measure the PSU noise while in use the results won't be very useful!?
> Is your Tek scope battery powered?
No. It's AC powered - the TDS2000 series with two channels.
I think the DAC being a noise source is logical. I wonder if the I to V converter is where it happens? I don't know enough about DAC's to speculate further. A friends powers his DAC with a 12 V motorcycle battery bypassed with 10 uF polyester 100 V. It might be two. He also has this for the phono stage. I built him an ultra high grade charger which he admits sounds no different when forced to charge and listen.
> I think the DAC being a noise source is logical.
Indeed. I did an experiment by removing the power to the digital board while keeping power to the rest of the analog circuits - the broadband noise went away in that case.
So the digital circuit is verified to be the noise source. Perhaps noises in the Mhz range do not pose any ill effect - it doesn't show up at the amplifier output so the intermediate stages must have filtered off such noises.
I also found that the noise spectrum changes depending on whether USB or SPDIF is selected as the digital input.
Interesting results. Wish my scope has a low-pass filter so that I can measure noises in the audio band.
Indeed. I did an experiment by removing the power to the digital board while keeping power to the rest of the analog circuits - the broadband noise went away in that case.
So the digital circuit is verified to be the noise source. Perhaps noises in the Mhz range do not pose any ill effect - it doesn't show up at the amplifier output so the intermediate stages must have filtered off such noises.
I also found that the noise spectrum changes depending on whether USB or SPDIF is selected as the digital input.
Interesting results. Wish my scope has a low-pass filter so that I can measure noises in the audio band.
Intesting you say that. I wanted to measure a Hypex UCD 180. I built a 12 pole inverting active design. Strangely the chip that worked was TL074 using from memory 186 kHz. This gave about - 3dB at about 50 kHz. " Better " chips were not happy. This sort of supports the idea that J Fets cope better in RFi siituations. Here I show an 8 pole Hybrid. Without this no meaniful tests could be done. TL074 will not give a total picture. It still is closer than guessing. When I added the Hypex SMPS it was again difficult. The SMPS sounds brighter and this may be by the same route. These tests were done with what I had in the shed, doubtless you can do better. A passive version was mostly the same. These are my notes to myself so not really for impressing anyone or to be definitive.
If one were to bolt/solder a BNC socket direct to the measuring PCB, then one could plug the oscilloscope direct to the measuring jig using BNC cables. This also brings the very high bandwidth of the BNC connections to the scope input.
But that restricts one to using X1 equivalent probe and it's inherent 1M impedance.
Is there a way to mimic the X10 probe with it's 10M impedance and get the direct BNC cable connection into the measuring jig?
One thing is to test at many points that should be similar. Using a bare wire soldered on is the only way ( BNC to 50 ohm coax ). If you get various results try reinforcing this PCB area. If this resolves the problems this is a victory. The simple explanation is the various points are at subtly different voltages. As amplifiers of any type need absolute points of reference this will show performance improvements as the refrences points become tighter. This is why Dead-bug or Gainclone devices sound good. The path is more direct. I had exactly this problem with a PSU I recently built dead-bug. As already said it measures better than a zinc PP3 even for hum. Alas the PCB is not quite as good. It has a beautiful ground plane that covers 90% of the underside. If I solder a 2.5 mm thick link between the reservoir caps I might get very close. This is for a friend. He hopes to sell a few. I think just to solder in that bar which will take a minute will be OK. Best of both worlds. The effects are at - 120 dB to - 140 dB refered to 12 V. Thinking about this I would guess every piece of hi fi that is mass produced will respond to this improvement. I did this to a Rotel RA 931 I repaired recently. Of all the amplifers I ever saw this was outstandingly good. If it had an op amp upgrade and extra transformer etc for the preamp who knows how good it could be? In fact upgrade the power amp PSU and use the internal one for the preamp. Where you get your best test results is the best place to exit the PCB. On my PSU the output cap is the best place. I have removed the PCB pads to discourage trying to find a convienient place, the capacitor is fine.
Also the RA931. I replaced the wire output fuses with thermal type 3A. The amp has survived master Dan and me after the mod and has tripped many times. I bypassed the fuses with 1 uF polyester. The fuses are in the feedback path. Dan says it sounds better ! Nothing I can think of would sound worse than bimetal, that cap is a having your cake and eat it. Rotel told me these fuses are sonically the weakest part of the amp. I dare say wire with 1 uF would be ideal. It also will say if the amp is OK as the tweeters will still produced sound. I also reset the standing current to 20 mA as all older Rotels were ( 44 mV / 0R22 ), it was a bit lower.
Also the RA931. I replaced the wire output fuses with thermal type 3A. The amp has survived master Dan and me after the mod and has tripped many times. I bypassed the fuses with 1 uF polyester. The fuses are in the feedback path. Dan says it sounds better ! Nothing I can think of would sound worse than bimetal, that cap is a having your cake and eat it. Rotel told me these fuses are sonically the weakest part of the amp. I dare say wire with 1 uF would be ideal. It also will say if the amp is OK as the tweeters will still produced sound. I also reset the standing current to 20 mA as all older Rotels were ( 44 mV / 0R22 ), it was a bit lower.
Please post picture of DAC board (both sides) and other info if you can.
On a digital supply you'll always get a little bit of noise, that's normal. It is a bit difficult to decide between "a bit of noise" and "too much noise"... that depends on the circumstances.
If you measure between 2 GND points close by, and you still get lots of noise, either you're measuring wrong (f.ex using the huge ground lead), or the board is crap, or both.
If you want to measure noise on a power supply, then it is not a problem. Just solder a coax and a 50R resistor.
Coax shield goes to GND plane. If there is no GND plane, close the case and put it on ebay.
Make a resistive divider like 450 ohms + 50 ohms. That will terminate your cable and only load the circuit with 500 ohms, it's a poor man's high speed x10 probe. Logic gates dont care much about the extra 500 ohms loading. Don't use that for high impedance analog nodes of course...
On a digital supply you'll always get a little bit of noise, that's normal. It is a bit difficult to decide between "a bit of noise" and "too much noise"... that depends on the circumstances.
If you measure between 2 GND points close by, and you still get lots of noise, either you're measuring wrong (f.ex using the huge ground lead), or the board is crap, or both.
If you want to measure noise on a power supply, then it is not a problem. Just solder a coax and a 50R resistor.
Coax shield goes to GND plane. If there is no GND plane, close the case and put it on ebay.
Make a resistive divider like 450 ohms + 50 ohms. That will terminate your cable and only load the circuit with 500 ohms, it's a poor man's high speed x10 probe. Logic gates dont care much about the extra 500 ohms loading. Don't use that for high impedance analog nodes of course...
Last edited:
I think my PP3 test is not bad , not least the PSU I built was better even on hum. My question about this would be is this DAC noise getting back into the main amplifer. I recently helped my friend build a simple preamp. He is using the cheap Lorlin 4 pole switch ( make before brake ). This removes the amplifier from unused sources as both ground and hot is broken. To have more inputs is a Makaswitch with extra wafers or relays. The Makaswich is expensive and complex for him to work out. Looking through model making books I have always tried to find low cost gears to link the Lorlin switches , never saw ideal ones. The fact one cog goes backwards is unimportant. One could use an intermediate gear. Seems more complicated for no gain. As cheap as they are the Lorlin switch is not bad.
You cant have a ground point inches away when measuring anything digital, the ground point HAS to be next to the signal or as near as possible.....
The 10x probe doesn't have 10M impedance, it has 10M // about 10pF... ie, 1kOhm @ 15MHz.....
You can always use a FET probe ($$$$$$$) or make one... I've never needed one though...
You can always use a FET probe ($$$$$$$) or make one... I've never needed one though...
I was going to add look up slots in ground planes and how it effects noise, this is what you are creating with a long ground lead, a return path disconuity....
It is not only the supplies that create noise in digital circuits, the digital circuit switching create noise (simultaneous switching noise) which is often what people see.
Always remember noise like any electrical current will flow in a loop, without details of the layout etc it is hard to make any judgements on the possible loops. This is basic signal integrity/EMC engineering, that said the high frequency noise if it is a problem can be solved by inter circuit block filtering, good layout etc.
But on most digital designs there will be some HF noise visible on other signals, the higher the scope frequency the more you can see, but it can be limited, but has to be done at the initial design stage and layout is critical, often have to do designs where this is very critical often up to 18GHz for immunity.
Hi Pierre, going to do my emails tonight, still doing the nucleonic's boards so internet access is limited and often in a hotel as data has to be hand delivered!!!!!!!
It is not only the supplies that create noise in digital circuits, the digital circuit switching create noise (simultaneous switching noise) which is often what people see.
Always remember noise like any electrical current will flow in a loop, without details of the layout etc it is hard to make any judgements on the possible loops. This is basic signal integrity/EMC engineering, that said the high frequency noise if it is a problem can be solved by inter circuit block filtering, good layout etc.
But on most digital designs there will be some HF noise visible on other signals, the higher the scope frequency the more you can see, but it can be limited, but has to be done at the initial design stage and layout is critical, often have to do designs where this is very critical often up to 18GHz for immunity.
Hi Pierre, going to do my emails tonight, still doing the nucleonic's boards so internet access is limited and often in a hotel as data has to be hand delivered!!!!!!!
One bit of fun is to measure between ground points. You will measure all kinds of junk between that which is thought to be silent. Some of this is measuring error. Much is truth.
A friend said years ago measure across the decoupling cap to a chip. Suddenly all is better. If you hand solder a new decoupling cap as close as possible you might be surprised at what you get. Crystal gave this advice on their first 20 bit chips. They also said to have different digital and analogue ground returns. The same friend said to have a choke between them. The decoupling cap will have inductance in the wires so even this is frought with problems.
A friend said years ago measure across the decoupling cap to a chip. Suddenly all is better. If you hand solder a new decoupling cap as close as possible you might be surprised at what you get. Crystal gave this advice on their first 20 bit chips. They also said to have different digital and analogue ground returns. The same friend said to have a choke between them. The decoupling cap will have inductance in the wires so even this is frought with problems.
One solid contiguous ground is often better than split planes...
Through hole components are no good for decoupling digital circuitry SMD devices in the smallest possible case size is what you need....
Henry Ott is one good source of info regarding all aspects of this (also Ralph Morrison); as is doing a Google search on decoupling.....
home page
One thought, does this board have a ground plane?
Through hole components are no good for decoupling digital circuitry SMD devices in the smallest possible case size is what you need....
Henry Ott is one good source of info regarding all aspects of this (also Ralph Morrison); as is doing a Google search on decoupling.....
home page
One thought, does this board have a ground plane?
Hi Marce 😀
Yeah, but it is fake. The chip cares about what happens on its pins, not on the capacitor...
For DAC, 1 GND plane is easiest but correct placement is needed. Split planes are difficult to pull off. Splitting in the wrong place makes it worse.
Example is hifimediy uae23 DAC... placement fail : spdif output LED GND return shares analog output GND. As a result, SPDIF leaks into analog output, among other things.
Yeah, but it is fake. The chip cares about what happens on its pins, not on the capacitor...
For DAC, 1 GND plane is easiest but correct placement is needed. Split planes are difficult to pull off. Splitting in the wrong place makes it worse.
Example is hifimediy uae23 DAC... placement fail : spdif output LED GND return shares analog output GND. As a result, SPDIF leaks into analog output, among other things.
What kind of DAC is it? Chip that is? What's the clock speed?
They're always chuntering on about how modern DACs upsample internally and use Sigma-Delta architectures, and shift (audible) noise into the MHz range. Maybe some radiated noise is a consequence, especially if the layout is not ideal. I guess you might even exclude it from your performance measurements, if you consider it inaudible.
They're always chuntering on about how modern DACs upsample internally and use Sigma-Delta architectures, and shift (audible) noise into the MHz range. Maybe some radiated noise is a consequence, especially if the layout is not ideal. I guess you might even exclude it from your performance measurements, if you consider it inaudible.
- Status
- Not open for further replies.