What if you measure at the MM’s amp output with your 7.33volt + 110k at the MC input, then you have the whole chain.
This should give 63x32.5mV = 2.05Vrms.
Hans
This should give 63x32.5mV = 2.05Vrms.
Hans
Just checked it now - 2.06 Volts on the scope and the QA401 is reading 1.944 - given the two completely different set-up's this is quite good correlation IMV
😕
Strange issue here
😕
Strange issue here
Last edited:
This rules out any gain accounting error in the QA401.
Is noise being calculated as we think by the 401? My understanding is that it takes the reading in each bin ('resolution'), squares it, sums it and then takes the square root.
Let me check this by changing the resolution and make sure we still get the same noise reading
Is noise being calculated as we think by the 401? My understanding is that it takes the reading in each bin ('resolution'), squares it, sums it and then takes the square root.
Let me check this by changing the resolution and make sure we still get the same noise reading
Here are the noise measurements at different resolution settings. The Left channell is open circuit, the right channel has a 47 Ohm SMD TF soldered across the input to ground. Resolution is bin width in Hz L & R readings in uV RMS on QA401
Resolution/L/R
.732/10.52/15.84
1.46/11.5/17.36
2.92/12/18
5.8511.5/19.5
23.4/20/33
You will note there are quite large discrepancies in the noise voltages read from day to day, since I was getting 27-30 uV RMS over the last two days.
It seems at the higher resolutions, the readings are consistent between resolutions.
Resolution/L/R
.732/10.52/15.84
1.46/11.5/17.36
2.92/12/18
5.8511.5/19.5
23.4/20/33
You will note there are quite large discrepancies in the noise voltages read from day to day, since I was getting 27-30 uV RMS over the last two days.
It seems at the higher resolutions, the readings are consistent between resolutions.
Do you have a scope with an FFT to look at the spectra of MM with shortened input and the spectra taken at at the same point at the output of the MM but now with MC + MM having resp 47R and open input.
Hans
Hans
Last edited:
Hans, I’ll have to have a look - the MDO 3024 has a FFT function but I have not used it.
If I use the inverse RIAA function with the QA401 it’s flat to within +- 0.15 dB on both MC and MM.
Are you thinking there’s a response anomaly?
If I use the inverse RIAA function with the QA401 it’s flat to within +- 0.15 dB on both MC and MM.
Are you thinking there’s a response anomaly?
The Hermitian symmetry of the FFT is often overlooked and negative frequencies ignored. FFT implementations vary in how they deal with back and forth scaling and when you use a canned library you have to note it.
I guess what you are saying is that for stochastic signals like thermal noise, this is not the ideal tool.
I have been a bit surprised at the fact that the reading wrt noise can vary over a c. 2:1 range - so from 17 uVRMS to 30 uVRMS and then it is about half of what the sim says.
The signal readings look pretty accurate, so clearly no issue there.
Looks like I'll have to go with the sim results and put those in the spec wrt noise (I'm still pretty happy with them BTW!), and use the measurements for distortion, FR etc.
I will have to look up 'Hermitian Symmetry' ! 🙂
I have been a bit surprised at the fact that the reading wrt noise can vary over a c. 2:1 range - so from 17 uVRMS to 30 uVRMS and then it is about half of what the sim says.
The signal readings look pretty accurate, so clearly no issue there.
Looks like I'll have to go with the sim results and put those in the spec wrt noise (I'm still pretty happy with them BTW!), and use the measurements for distortion, FR etc.
I will have to look up 'Hermitian Symmetry' ! 🙂
What I'm saying is that any FFT library has to return the original signal when you do a forward transform and then the backward one. The scaling of the transforms does not have an absolute standard so that when you do only the forward transform there can be a factor that you must account for.
ok - I assume here that this applies to noise, since the signal readings are accurate when I check them with the scope.
It looks like the spot noise - i.e. the noise floor - on the QA401 is accurate, but the RMS reading for some reason gives results that we cannot correlate back to the sim - they are c. 6dB out. Later today I will try to limit the bandwidth on both the sim and the QA401 and see what I get.
It looks like the spot noise - i.e. the noise floor - on the QA401 is accurate, but the RMS reading for some reason gives results that we cannot correlate back to the sim - they are c. 6dB out. Later today I will try to limit the bandwidth on both the sim and the QA401 and see what I get.
To clarify my earlier post, I assume the forward and backward transforms are working correctly with a signal, but not with noise because if I measure a signal (1kHz c. 0dB) with two different instruments, the results tally.
Andrew,
I have caused you some work because of inconsistencies in the figures, sorry for that, but in the end all seems well. Your QA400 seemed to have caused this all.
When taking the simulation output giving the most reliable figures, I get 77dB S/N for your 550mH+1k35 MM Cart ref 5mV@1Khz and 74 dB S/N for a 47R MC cart ref 0.5mV@1Khz, right ?
Above figures translate resp. into 80dB(A) for MM and 79dB(A) for the MC Cart.
Both absolute excellent results, top of the bill.
Hans
I have caused you some work because of inconsistencies in the figures, sorry for that, but in the end all seems well. Your QA400 seemed to have caused this all.
When taking the simulation output giving the most reliable figures, I get 77dB S/N for your 550mH+1k35 MM Cart ref 5mV@1Khz and 74 dB S/N for a 47R MC cart ref 0.5mV@1Khz, right ?
Above figures translate resp. into 80dB(A) for MM and 79dB(A) for the MC Cart.
Both absolute excellent results, top of the bill.
Hans
Hans
Your feedback much appreciated. Yes, I agree with your numbers and will go with this. I will update the spec accordingly.
My thanks to Scott as well.
Ive certainly been on a learning journey here!
🙂
Your feedback much appreciated. Yes, I agree with your numbers and will go with this. I will update the spec accordingly.
My thanks to Scott as well.
Ive certainly been on a learning journey here!
🙂
Isn't it a peak instead of rms reference value they use to calculate noise. I once experienced such issue in one SW.
Noise is just noise and is exactly defined. Many SW applications like Arta and Spectralab measure rms noise over defined delta F precisely. If QA401 gives wrong rms noise voltage value, there must be an algorithm error. As it it is a semi-diy product, I would not be surprised.
I don’t think that’s entirely fair.
I am sure Instruments like this measure correlated signals very accurately and certainly the QA401 does because I checked it.
If I understand Scott’s comment correctly, uncorrelated, stochastic signals are a bit more difficult to deal with, and I would think especially so below a -130 dBV. The noise floor on the QA401 with the inputs shorted is at -155 dBV. We are talking about a 6 dB delta at -134dBV uncorrelated. In the big scheme of things, not a train smash. If you are trying to measure the last ounce of noise as is the case here, then of course we have to dig deeper. Luckily the calculations and the sims agree, so this is what we can proceed with.
Just my view - I respect yours may differ.
🙂
(See S.O. Rice ‘Mathematical Analysis of Random a Noise’ - it’s a complex subject IMV)
I am sure Instruments like this measure correlated signals very accurately and certainly the QA401 does because I checked it.
If I understand Scott’s comment correctly, uncorrelated, stochastic signals are a bit more difficult to deal with, and I would think especially so below a -130 dBV. The noise floor on the QA401 with the inputs shorted is at -155 dBV. We are talking about a 6 dB delta at -134dBV uncorrelated. In the big scheme of things, not a train smash. If you are trying to measure the last ounce of noise as is the case here, then of course we have to dig deeper. Luckily the calculations and the sims agree, so this is what we can proceed with.
Just my view - I respect yours may differ.
🙂
(See S.O. Rice ‘Mathematical Analysis of Random a Noise’ - it’s a complex subject IMV)
Last edited:
Bonsai, reading posts over the past few days it seems that if one did not need to have a MM phono input, the second gain stage could be redesigned for lower noise - the nominal 47k MM input is no longer needed - and a better overall Ein might be obtained. if so, might not a lower noise MC only option be possible?
I realise this is premature as the design has not been officially been released, but thought it might be worth discussing.
Just thinking out loud...
I realise this is premature as the design has not been officially been released, but thought it might be worth discussing.
Just thinking out loud...
@AL2002
Since it's me who probably caused this confusion, based on previous data of -124dBV noise output with shortened MM input, Let me be the one to correct this because the specified noise turned out to be -135dBV.
This -135dBV means that by going back to the input of the MM amp, that its input noise is 237nV within the 20Khz BW.
In all cases, output of the MC stage will be around 5mV meaning that the gain will be 10x for a 0.5mV cart or even more for a Cart with a lower output.
The input noise of the MC amp alone is ca. 31nV to which the noise of the Cart's resistance will be added, giving >>10x31nV output noise or way above the 237nV input noise from the MM stage.
So your concerns are totally ungrounded, this is an excellent design.
Hans
Since it's me who probably caused this confusion, based on previous data of -124dBV noise output with shortened MM input, Let me be the one to correct this because the specified noise turned out to be -135dBV.
This -135dBV means that by going back to the input of the MM amp, that its input noise is 237nV within the 20Khz BW.
In all cases, output of the MC stage will be around 5mV meaning that the gain will be 10x for a 0.5mV cart or even more for a Cart with a lower output.
The input noise of the MC amp alone is ca. 31nV to which the noise of the Cart's resistance will be added, giving >>10x31nV output noise or way above the 237nV input noise from the MM stage.
So your concerns are totally ungrounded, this is an excellent design.
Hans