One thing is quite obvious in the meanwhile,
It is the TVC that is causing the problems, either input or output.
A first cautious conclusion is that the input side is not impressed by the serial resistors, so it may be the output that wants to see a low ohmic termination load.
the 3 KOhm to ground that you placed improved things, but the two 100 Ohm resisistors seem to play no role.
So try a 470 Ohm resistor to start with, one side in the line from TVC, the other side to ground.
You can also use a pot to find a point where distortion stops.
Hans
Still trying to combine all results, a number of things can be used to exclude possible causes.
1) a generator instead of DAC / TVC results in a normal signal from the filter.
2) placing a resistor in line with the filter, with DAC and TVC in place attenuates the effect somewhat bit distortion is still there
3) placing a buffer before the filter does not solve the problem, so the filter cannot be the cause, the signal is already there before the buffer.
4) The filter cannot produce distortion from itself, unless it is oscillating for which there is no prove at all.
5) So the distortion is caused by the input signal, and although this might look clean, this because the oscilloscope cannot catch HF glitches.
6) steering the filter straight away from the DAC results in a perfect signal.
7) This meand the DAC is also O.K., just as the filters.
8) what remains is the TVC being the cause of the trouble.
9) since the TVC cannot invent new signals at its output, the distortion is already there at its input, pointing clearly in the direction of interaction between TVC and DAC.
10) So either one of following solutions should work:
I) A termination load at the output, working it's way back to influence the DAC to TVC interaction.
II) Buffers at the input to isolate the DAC from the TVC, probably the best solution.
III) Or a combination of both.
Other changes are nothing but cosmetic symptom fighting.
The Filters are O.K and so is the DAC.
The TVC is the cause, so try to tame this beast.
Hans
1) a generator instead of DAC / TVC results in a normal signal from the filter.
2) placing a resistor in line with the filter, with DAC and TVC in place attenuates the effect somewhat bit distortion is still there
3) placing a buffer before the filter does not solve the problem, so the filter cannot be the cause, the signal is already there before the buffer.
4) The filter cannot produce distortion from itself, unless it is oscillating for which there is no prove at all.
5) So the distortion is caused by the input signal, and although this might look clean, this because the oscilloscope cannot catch HF glitches.
6) steering the filter straight away from the DAC results in a perfect signal.
7) This meand the DAC is also O.K., just as the filters.
8) what remains is the TVC being the cause of the trouble.
9) since the TVC cannot invent new signals at its output, the distortion is already there at its input, pointing clearly in the direction of interaction between TVC and DAC.
10) So either one of following solutions should work:
I) A termination load at the output, working it's way back to influence the DAC to TVC interaction.
II) Buffers at the input to isolate the DAC from the TVC, probably the best solution.
III) Or a combination of both.
Other changes are nothing but cosmetic symptom fighting.
The Filters are O.K and so is the DAC.
The TVC is the cause, so try to tame this beast.
Hans
Has he tried driving the TVC and filter from another source which cannot have spikes?
It is possible that the problem is a combination of two issues:
1. the DAC is sending out spikes or other HF
2. the TVC presents an impedance to the filter which causes a near-instability - it isn't quite oscillating but nearly.
Then the spikes trigger ringing in the instability, which manifest themselves as distortion. Changing either of the two issues will change the result.
There is no evidence that the filter is oscillating (apart from the circumstantial evidence that adding an attenuator pad improves things) but there is no evidence that the filter is not oscillating. The buffer could suffer from the same problem, if it is constructed in a similar way.
It is possible that the problem is a combination of two issues:
1. the DAC is sending out spikes or other HF
2. the TVC presents an impedance to the filter which causes a near-instability - it isn't quite oscillating but nearly.
Then the spikes trigger ringing in the instability, which manifest themselves as distortion. Changing either of the two issues will change the result.
There is no evidence that the filter is oscillating (apart from the circumstantial evidence that adding an attenuator pad improves things) but there is no evidence that the filter is not oscillating. The buffer could suffer from the same problem, if it is constructed in a similar way.
When driving the filter from a signal generator or directly from the DAC without TVC in between, there is no distortion making the probability of oscillation very small.
The 3K load termination load on the TVC attenuates things but a buffer isolating the filters doesn't, so it is not the filter that causes the attenuation.
And when using a completely different filter with buffer there is still distortion.
So the filters are not in the line of fire, it is definitely the TVC.
Hans
No. All that shows is that the filter does not misbehave unless the TVC is present. What happens when the TVC+filter is driven by a sig gen?Hans Polak said:
The buffer and alternative filter were built by the same person using the same techniques? I suspect an interaction between the TVC and the following opamp, possibly made more visible by the DAC output.
Aehm, no. Please take another look at my post 82. The distortion is gone with a different filter (KMTech / ebay). Moving the OPA 2134 opamps from the KMTech to the ABACUS board, the distortion decreases considerably with the ABACUS, too (but does not remove it).
Also, remember that the distortion goes away if I change the TVC volume setting to 12 o'clock or so. I'd expect this to change the complex impedances in the TVC, which in turn might remove the interaction with the opamp filter.
What is a more likely explanation for both these observations (swapping bipolar/FET opamps and changing TVC impedances reduce/remove the distortion)? Spikes from the DAC output or complex impedances stressing the opamp filter to misbehave? Yes, I'd need a suitable scope to look for spikes, but I don't have one.
With the 5532 the Bips where back again, so there is still something in the input signal causing this. That one opamp is less sensitive, does not mean the problem is solved, that was what I called cosmetic symptom fighting.
Remember that your scope is the big limiting factor here.
That the output of the OPA2134 looks right, tells very little, since the output of the TVC also looks right, but as we know this is not the case.
O.k. I overlooked this point, this information is new to me.
Probably the spike in the signal has the same amplitude on all the outputs of the TVC because of capacitive coupling, and that's why you don't see it on the 1:1 output because the signal is so much larger there.
Yeti advised you in that same posting just as I did to place a termination at Point A, did you do that ?
The interaction with the filter sounds as if the filter is doing something wrong, although massive evidence is there that the filter is allright.
The real interaction is the difference in loading of the TVC.
When spikes are generated in some form, your TVC might be acting like an ignition coil when not properly terminated.
Yes you are right, a scope might be a big help. But without, several causes can be eliminated from the list. Forget the filters and concentrate on the DAC/TVC.
So my question is, have you terminated your TVC already with a much lower resistance as the 3Kohm that you recently tried and that positively influenced the problems.
It costs no time at all to do this.
To start with, probably the best place for this this termination on the 1:1 output of the TVC.
Hans
I will ask my question again: what happens when the TVC+filter is driven from a sig gen? If the problem is still present then we can probably eliminate the DAC as part of the cause. If the problem disappears then we can suspect the DAC, but exactly what it is provoking would remain a mystery.
Another thing to try is adding a passive low pass filter between the DAC and TVC. The aim is to reduce any spikes without affecting audio. Note that any competently engineered DAC will already have a filter, but maybe you are using a NOS DAC?
Another thing to try is adding a passive low pass filter between the DAC and TVC. The aim is to reduce any spikes without affecting audio. Note that any competently engineered DAC will already have a filter, but maybe you are using a NOS DAC?
And if so, I'd be able to see the spikes with a suitable scope, and remove the effect by inserting a high-pass filter (RC filter) in between the DAC and the TVC.
Questions:
- Would an RC filter with R=47 Ohm and C = 100 nF be ok?
- Any recommendations for a good scope that does not cost an arm and a leg? This problem might serve as a good excuse to get a proper scope...
You placed a 10nf cap on the output and reported it had no effect.
Are you sure it was placed correctly, I find it a bit hard to believe that it had no effect at all.
But to convince you, I simulated a 30nF cap on the output of your DAC and also a 47Ohm with 100nF, assuming that your output resistance is 133Ohm.
The big difference is that the single cap keeps on attenuating wit 6dB/oct and that the RC combination stops at 100kHz. So having an extra resistor in line is not a good idea.
The 30nF that I used attenuates only 0.5 dB at 20kHz, while 47Ohm / 100nF attenuates 5,5dB at 20Khz, way too much.
So when using a filter, the maximum you should use are two 30nF from each output to Gnd or a single 60nF between the two outputs.
I had it soldered directly to the output of the DAC board; so yes, I think it was placed correctly. It did not have any effect on the distortion effect. I have tested this only at 400 Hz, and I haven't checked for changes in the frequency response roll off towards the treble or similar things, as I didn't think of this as a "final solution", but rather as a test for finding the source of the problem.
The output resistance is 2 x 34 Ohm from pos/neg outputs to GND (see my first post).
Why would the simulated RC filter stop at some point?
When you have a 34 Ohm output, that probably means that you have 4 Dac's in parallel, I can't find this information in your first post.
But it actually means that you should have used at least 4 times the amount of 10nF (polystyrene) caps , and not a the output going to the TVC, but as close as possible to the various PCM1794 DAC's, because that's where the spikes are generated.
I do not know the layout of your PCB, but try to go as close as practically possible to Iout A+ (A pos) and Iout A- (A neg) as possible with a 10nF caps, the other side connected to ground.
Going to ground is better because, although the audio signals are of opposite polarity, we don't know whether the spikes coming from the outputs are in phase or out of phase.
Placing a cap too far away, leads to a complete failure to suppress spikes.
Hans
Ok, first I removed the series resistors from the T-pad and left the 3kΩ in place (from signal to GND at the TVC output). The pips were back.
I also tried different resistor values; result didn't change a lot.
Then I removed the resistor from signal to GND and put the series resistors in the connection from the TVC out to the filter input back. Pips went (almost) away.
So it seems the termination of the TVC secondary (signal to GND) is not the issue. Looks more like the coupling of the TVC to the filter to me.
It's in the schematic.
Ok, tried that. I used 33nF from pos to GND and 33nF from neg to GND. Didn't change the pips a bit.
The RC time constant is 34Ω x 33E-9F = 1.1E-6 seconds, corresponding to a cut-off frequency of about 140 kHz. Should be enough to take care of HF spikes, or at least to reduce the effect -- but since the caps didn't change anything I don't think there were any problematic HF spikes to start with. While I don't have a scope to "prove" this in my own setup, I found a few scope shots from the DDDAC output at Doede's website. I don't see any spikes or similar issues.
My conclusion from all this is that the issue is related to the coupling of the TVC to the opamp filter. So I started modifying the KMTech filter to give the exact same filter transfer as the ABACUS (high pass), checking for the appearance of the pips in between the various modification steps. No pips at all.
So unfortunately I didn't learn what's going wrong with the ABACUS. I guess I'll just replace the ABACUS high pass with the KMTech and leave it there. I need to get the system playing music again...
I'm glad you found your way forward.
The information presented to us was rather confusing and did not lead to a solution.
In the first post it was mentioned that a 1KOhm resistor between TVC and filter hardly did anything, but in posting #116 you inserted 200Ohm and the pips were almost gone ??
Complete isolating this connection with a buffer "reduced the distortion, but not to zero".
The KTM board, which aso has a buffer input, had no problems with the OPA2134, but with the 5532 there where again pips visible.
So again, not having a scope seems to be the limiting factor here, because opamps do not "invent" pips.
In posting #110 you mention that when steering TVC + Filter from a Signal generator, distortion is absent, but when placing caps on the Dac output no difference is visible and the pips are there.
All this together, gives the impression of a system where ground connections are not properly made.
Using a transformer makes it possible to use completely separated supplies and prevent ground loops.
I suspect that you are using the same supply for the DAC and for the filters and that the ground of the DAC is connected to the ground of the filter section, instead of having two completely separated supplies.
My first reaction in the beginning of this thread was, "don't mix signal ground and earth, they are not the same"
So when in future you have the time and the possibilities to try two independant supplies, this will be a very easy test.
Hans
The information presented to us was rather confusing and did not lead to a solution.
In the first post it was mentioned that a 1KOhm resistor between TVC and filter hardly did anything, but in posting #116 you inserted 200Ohm and the pips were almost gone ??
Complete isolating this connection with a buffer "reduced the distortion, but not to zero".
The KTM board, which aso has a buffer input, had no problems with the OPA2134, but with the 5532 there where again pips visible.
So again, not having a scope seems to be the limiting factor here, because opamps do not "invent" pips.
In posting #110 you mention that when steering TVC + Filter from a Signal generator, distortion is absent, but when placing caps on the Dac output no difference is visible and the pips are there.
All this together, gives the impression of a system where ground connections are not properly made.
Using a transformer makes it possible to use completely separated supplies and prevent ground loops.
I suspect that you are using the same supply for the DAC and for the filters and that the ground of the DAC is connected to the ground of the filter section, instead of having two completely separated supplies.
My first reaction in the beginning of this thread was, "don't mix signal ground and earth, they are not the same"
So when in future you have the time and the possibilities to try two independant supplies, this will be a very easy test.
Hans
Parenthetical -- L2 and L4 on the LB80ABDT (low dropout 8V regulators) essentially render the regulatory scheme less than useful. The designer put a complex impedance on the output. Remember that the output capacitor on a low-dropout regulator is in the control loop!
C_out should be a minimum of 2uF as ST recommends, and placed as close as possible to the regulator.
btw, what are the rail voltage on the filter opamp?
C_out should be a minimum of 2uF as ST recommends, and placed as close as possible to the regulator.
btw, what are the rail voltage on the filter opamp?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.