Please go to Linear Audio website and download the article.
It has detailed explanation of the distortion mechanism of the circuit.
http://www.diyaudio.com/forums/digi...minimalistic-iv-converter-14.html#post2969654
Patrick
It has detailed explanation of the distortion mechanism of the circuit.
http://www.diyaudio.com/forums/digi...minimalistic-iv-converter-14.html#post2969654
Patrick
Last edited:
I've read it. In my balanced SEN, I'm taking the output across the top of the I/V resistors which means that not all of the current from the DAC has to go through them.
Spice indicates that distortion is sensitive to the load impedence - in this case increasing with decreasing load impedence.
Also I think this makes the distortion sensitive to imbalance in the two balanced halves.
As I said earlier, I don't really know what I'm doing here so I might be completely wrong.
Is there a better way to get a balanced output?
Any other thoughts about this?
I am not sure what you mean by balanced output.
If you want to use XLR, then you need 3 wires (+ve phase, -ve phase and Gnd).
The two phases you take off the R_iv of the two circuits, Gnd is Gnd on the PCB.
But if you have a balanced I-out DAC and what single ended output,
then you either only use the positive phase, or you need a Balanced to SE converter.
As I said, the distortion of the circuit was not only simulated but measured with reliable equipment.
So we know what we are doing.
Patrick
If you want to use XLR, then you need 3 wires (+ve phase, -ve phase and Gnd).
The two phases you take off the R_iv of the two circuits, Gnd is Gnd on the PCB.
But if you have a balanced I-out DAC and what single ended output,
then you either only use the positive phase, or you need a Balanced to SE converter.
As I said, the distortion of the circuit was not only simulated but measured with reliable equipment.
So we know what we are doing.
Patrick
When using two singled ended circuits in balanced mode, provided the gain is well matched (R_iv in this case), the even harmonics will cancel (or reduce by at least 20dB or so), leaving only the odd harmonics. This has been discussed in the article.
In your measurements, the following uncertainties cannot be distinguished from the distortion of the IV circuit alone :
1) Distortion of your reference signal (i.e. from digital data to balanced DAC analog current output)
2) Distortion of your measurement equipment.
Until those have been individually verified, I consider any such measurements unreliable.
This was precisely why we did not want to use a DAC as a source in our measurement, but instead use the high voltage capability of th AP1 in series with a resistor. Even then, the small distortion of the resistor in combination with the non-linear Zin of the circuit still mask the results somewhat.
As I also mentioned in the article, measuring below 90dB is a difficult task. And until one gets one's equipments right, the measurements are going to be masked by intrinsic errors in the reference source / measurement chain that the real distortion of the DUT cannot be truely revealed.
As far as I am concerned, the distortion of the circuit itself is lower than the measurement capability that I have access to, and I am sure Jan Didden will support that statement.
Cheers,
Patrick
In your measurements, the following uncertainties cannot be distinguished from the distortion of the IV circuit alone :
1) Distortion of your reference signal (i.e. from digital data to balanced DAC analog current output)
2) Distortion of your measurement equipment.
Until those have been individually verified, I consider any such measurements unreliable.
This was precisely why we did not want to use a DAC as a source in our measurement, but instead use the high voltage capability of th AP1 in series with a resistor. Even then, the small distortion of the resistor in combination with the non-linear Zin of the circuit still mask the results somewhat.
As I also mentioned in the article, measuring below 90dB is a difficult task. And until one gets one's equipments right, the measurements are going to be masked by intrinsic errors in the reference source / measurement chain that the real distortion of the DUT cannot be truely revealed.
As far as I am concerned, the distortion of the circuit itself is lower than the measurement capability that I have access to, and I am sure Jan Didden will support that statement.
Cheers,
Patrick
This link has some information concerning the intrinsic distortion of the Roland UA-25EX that you were using for your measurements :
baudline solution - Edirol UA-25
The measured intrinsic THD of the sound card is about -86dB if I read correctly.
The manufacturer did not provide any THD figures or distortion specifications at all.
So I personally would not compare that to an AP1.
Patrick
baudline solution - Edirol UA-25
The measured intrinsic THD of the sound card is about -86dB if I read correctly.
The manufacturer did not provide any THD figures or distortion specifications at all.
So I personally would not compare that to an AP1.
Patrick
What you say is true - I think you have proved with your measurements that the distortion of SEN in isolation is great which is why I decided to build it - but what matters to me is the distortion when using it in the chain I will be listening to including the DAC.
I know the DAC is capable of better results than I posted above because I have measured very low distortion using it to drive a transformer (bass response was poor for that I/V).
So, somewhere in the chain I have a problem - and yes, it might be some compression in the ADC I'm using to measure the signal.
There are a few things I can think of to try: measure the loopback performance of the ADC with a test signal; match the jfets better, investigate to see whether I can fine tune the balanced halves to be better balanced; reduce Riv; try more jfets to reduce the input impedance; use an attenuator between SEN and the ADC so I can test the DAC at 0dB.
I know the DAC is capable of better results than I posted above because I have measured very low distortion using it to drive a transformer (bass response was poor for that I/V).
So, somewhere in the chain I have a problem - and yes, it might be some compression in the ADC I'm using to measure the signal.
There are a few things I can think of to try: measure the loopback performance of the ADC with a test signal; match the jfets better, investigate to see whether I can fine tune the balanced halves to be better balanced; reduce Riv; try more jfets to reduce the input impedance; use an attenuator between SEN and the ADC so I can test the DAC at 0dB.
More measurements of PCM1794 with balanced SEN
I have replaced a toasted op-amp on the sound card I normally use for measurements (which is better than the UA-25 aside from a couple of spikes of noise) and it is working again.
I also worked out how to get it to measure a balanced signal using L+R channels for each half and then calculating the difference in software afterwards.
My normal measurement card is the one with my DAC so there is a chance that the toasted op amp had been affecting the performance of the DAC chip in my previous results. Probably not though because it was on the ADC input and nowhere near the DAC.
Here are the new results including a comparison with my previous IV which is a balanced OP-AMP design I hand-crafted.
There's an argument to be made that the lowest of the three lines at any frequency indicates an upper bound for any distortion introduced by the ADC (the counter argument would be that the signal might be canceling out some of the ADC distortion at that frequency).
If you subscribe to the former argument then that would indicate that most of those early peaks are measured distortion values for the PCM1794/SEN combination where it is worse than the PCM1794/OP-AMP solution.
I haven't done any work on optimizing my SEN circuit yet.
While testing, I did notice a little distortion either introduced by the servo or perhaps by the regulated power supply: the sudden change from no signal to a full amplitude 1kHz sine wave results in a little dip in the signal level followed by a correction - the little dip contains clipping distortion. I did a warm-up stage before getting the results for the graph so they don't include this.
The ADC has a couple of OP-AMP buffers in front of it which are the same type as the ones in my OP-AMP IV so there might very well be some cancellation going on.
AP1 is good is it?
I should mention that the 12V and 18V measurements are for my left and right SEN channels, they were not done on the same channel so it's not clear whether the difference in distortion is due to the supply voltage or the different circuit under test - I was comparing the power supplies for noise rather than distortion - sorry.
I have replaced a toasted op-amp on the sound card I normally use for measurements (which is better than the UA-25 aside from a couple of spikes of noise) and it is working again.
I also worked out how to get it to measure a balanced signal using L+R channels for each half and then calculating the difference in software afterwards.
My normal measurement card is the one with my DAC so there is a chance that the toasted op amp had been affecting the performance of the DAC chip in my previous results. Probably not though because it was on the ADC input and nowhere near the DAC.
Here are the new results including a comparison with my previous IV which is a balanced OP-AMP design I hand-crafted.
There's an argument to be made that the lowest of the three lines at any frequency indicates an upper bound for any distortion introduced by the ADC (the counter argument would be that the signal might be canceling out some of the ADC distortion at that frequency).
If you subscribe to the former argument then that would indicate that most of those early peaks are measured distortion values for the PCM1794/SEN combination where it is worse than the PCM1794/OP-AMP solution.
I haven't done any work on optimizing my SEN circuit yet.
While testing, I did notice a little distortion either introduced by the servo or perhaps by the regulated power supply: the sudden change from no signal to a full amplitude 1kHz sine wave results in a little dip in the signal level followed by a correction - the little dip contains clipping distortion. I did a warm-up stage before getting the results for the graph so they don't include this.
The ADC has a couple of OP-AMP buffers in front of it which are the same type as the ones in my OP-AMP IV so there might very well be some cancellation going on.
AP1 is good is it?
I should mention that the 12V and 18V measurements are for my left and right SEN channels, they were not done on the same channel so it's not clear whether the difference in distortion is due to the supply voltage or the different circuit under test - I was comparing the power supplies for noise rather than distortion - sorry.
Attachments
Last edited:
AP1 was at its time industrial reference. It is still one of the best you can get with reasonable money.
The red curve (SEN IV 18V) in your measurement above indicates 2nd & 3rd harmonics in the order of -95~-100dB.
This is about the same order as what we published.
http://www.diyaudio.com/forums/digi...on-minimalistic-iv-converter.html#post2688676
I am fully aware that you can find opamp circuits that can give better figures.
Just as you can find opamp based amplifier circuits with exceeding low THD figures.
If that is the most important criterion for you then maybe you should stay with your original IV design.
I am sure there are also many nice sounding opamp designs if done right.
Best regards,
Patrick
The red curve (SEN IV 18V) in your measurement above indicates 2nd & 3rd harmonics in the order of -95~-100dB.
This is about the same order as what we published.
http://www.diyaudio.com/forums/digi...on-minimalistic-iv-converter.html#post2688676
I am fully aware that you can find opamp circuits that can give better figures.
Just as you can find opamp based amplifier circuits with exceeding low THD figures.
If that is the most important criterion for you then maybe you should stay with your original IV design.
I am sure there are also many nice sounding opamp designs if done right.
Best regards,
Patrick
Last edited:
Music sounds great with SEN - see #696
My power amp is a pair of KT88 mono-blocks with no global feedback - somehow I don't think I'll be kicking everything off with a pair of op-amps for much longer.
Let's see where I get to with tuning your lovely SEN circuit. The results above are just baseline numbers. I haven't even started yet.
> For this version is 2x9V batteries a still a good starting point ?
Yes.
> I don't have a model of the 2sk369 ?
MODEL 2SK369_V NJF (VTO=-574.395M BETA=95.173M LAMBDA=1M IS=10F RS=3.11187
+ CGD=36.6561P CGS=60.0451P PB=1.99682 KF=0.001F AF=500M )
.MODEL 2SK369BL NJF (VTO=-398.254M BETA=85.9633M LAMBDA=1M IS=10F CGD=36.6561P
+ CGS=60.0451P PB=1.99682 KF=0.001F AF=500M )
.MODEL 2SK369GR NJF (VTO=-337.112M BETA=80.3593M LAMBDA=1M IS=10F CGD=36.6561P
+ CGS=60.0451P PB=1.99682 KF=0.001F AF=500M )
Just build it. It is quicker than any sim, and you have real results.
Patrick
Yes.
> I don't have a model of the 2sk369 ?
MODEL 2SK369_V NJF (VTO=-574.395M BETA=95.173M LAMBDA=1M IS=10F RS=3.11187
+ CGD=36.6561P CGS=60.0451P PB=1.99682 KF=0.001F AF=500M )
.MODEL 2SK369BL NJF (VTO=-398.254M BETA=85.9633M LAMBDA=1M IS=10F CGD=36.6561P
+ CGS=60.0451P PB=1.99682 KF=0.001F AF=500M )
.MODEL 2SK369GR NJF (VTO=-337.112M BETA=80.3593M LAMBDA=1M IS=10F CGD=36.6561P
+ CGS=60.0451P PB=1.99682 KF=0.001F AF=500M )
Just build it. It is quicker than any sim, and you have real results.
Patrick
I was thinking the same thing. That would be another interesting baseline but I really don't want an unattended battery charging circuit. Particularly not with li-pos and absolutely not one designed by any amateur (myself included).
No need to use LiPo's. NiMH is just fine.
I have been leaving my charger on for my preamp since 2007, non-stop.
You can use the one I published here. It is simple, and quite safe.
http://www.diyaudio.com/forums/digi...minimalistic-iv-converter-13.html#post2952183
Patrick
I have been leaving my charger on for my preamp since 2007, non-stop.
You can use the one I published here. It is simple, and quite safe.
http://www.diyaudio.com/forums/digi...minimalistic-iv-converter-13.html#post2952183
Patrick
Last edited:
I am the same way. My theory was use 18650 battery holders in the circuit and just pop them out and put on a proper charger.
However with one project I inserted one of the batteries the wrong way when I put it back in the holder after charging, big stupid mistake
I like the idea of 9V rechargeables 'casue you can't put them in wrong
I think there are lithium rechargeable 9V's now.
Last edited:
PCM1794/SEN on batteries with servo op-amps removed
I was a bit suspicious of all the high order distortion in my previous measurements - normally it's the result of lots of negative feedback somewhere - so I pulled out the servo op-amps during the measurements and took the spectrum from just afterwards before the circuit had drifted.
The difference is particularly clear in this second set of measurements which I have taken whilst running the floating supply on batteries (one channel is batteries, the current drain and other channel are still regulated 12V supplies).
I don't believe the 2nd harmonic distortion result in these measurements - the only way they can all be the same is if it's from the ADC. A lot of the other even harmonics are also less than the odd ones. The rest of the harmonics are more interesting.
So my hacked together servo is introducing some high order distortion - not that surprising given that it is amplifying the signal way past the point of clipping and then applying that to the input (even if through a low-pass filter).
I need to design a better servo.
I also measured the balance of the floating supplies at rest and they are out of balance by about 5% because I didn't match my jfets well enough (I should probably have waited for them to warm up; there were 100 though). Spice indicates this is enough to cause some distortion as well.
I have a design for a second servo to fix the mismatch which works in spice (3rd harmonic down to -140dB with mismatched jfets and Riv reduced to 250R) but I also need to work on that one to make sure it doesn't introduce high order distortion.
So it looks like there is plenty of scope for SEN to beat my OP-AMP IV on measurements as well as the listening test
I was a bit suspicious of all the high order distortion in my previous measurements - normally it's the result of lots of negative feedback somewhere - so I pulled out the servo op-amps during the measurements and took the spectrum from just afterwards before the circuit had drifted.
The difference is particularly clear in this second set of measurements which I have taken whilst running the floating supply on batteries (one channel is batteries, the current drain and other channel are still regulated 12V supplies).
I don't believe the 2nd harmonic distortion result in these measurements - the only way they can all be the same is if it's from the ADC. A lot of the other even harmonics are also less than the odd ones. The rest of the harmonics are more interesting.
So my hacked together servo is introducing some high order distortion - not that surprising given that it is amplifying the signal way past the point of clipping and then applying that to the input (even if through a low-pass filter).
I need to design a better servo.
I also measured the balance of the floating supplies at rest and they are out of balance by about 5% because I didn't match my jfets well enough (I should probably have waited for them to warm up; there were 100 though). Spice indicates this is enough to cause some distortion as well.
I have a design for a second servo to fix the mismatch which works in spice (3rd harmonic down to -140dB with mismatched jfets and Riv reduced to 250R) but I also need to work on that one to make sure it doesn't introduce high order distortion.
So it looks like there is plenty of scope for SEN to beat my OP-AMP IV on measurements as well as the listening test
