Do you have a decent regulated 12 to 15V supply available inside your amp? If not you will need another board for that. have you considered something like this
SparkFun THAT 1646 OutSmarts Breakout - BOB-14003 - SparkFun Electronics Wouldn't take up much space and the IC is very well designed. Not quite as good as Tom's but no slouch either.
You will get a doubling of voltage potential with the conversion, I think that is +6dB in voltage terms.
Excerpt from the That 1646 data sheet
"The 1606 and 1646 both provide +6 dB gain
(factor of 2) between their inputs and differential
outputs. This is appropriate, since with a balanced
output, twice the voltage between the power supply
rails is available at the output of the stage".
I can ask Tom for above, and for below will simulate SE to ballanced and the other way around, but think voltage thing is as simple whenever we convert it happens automatic, so you will not get any gain over a ballanced lines higher voltage because as soon it convert to SE voltage is in half again. And about you Jriver seems sounds best minus those 6dB regained with ATOM, that could be because one of your many plugins works optimal at that particular lower level or could be because DAC loose 1bit resolution. Say its a plugin then it could be cheated or tested in one of the PEQ containers is placed most upstream set those -6dB and with the other PEQ container placed most downstream +6dB, that way plugin will not see if set Jriver volume 6dB higher as it was before ATOM, btw that + or - 6dB can also be done using Jriver standard graphic EQ in its left slider is not a filter but a pre amp slider. Will get back in few days after some email talk with Tom and voltage simulation in TI Tina.
I know the doubling of the voltage would be undone for the regular SE amps, but it would be valid with the subwoofers, as there I put in that single ended RCA socket, where there should be a proper differential input. That was the question
.I agree that for the Goldmund and Ambience amp, it would make no difference. As on the other side you lose the same amount converting back to SE.
I think the reason for the gain having a positive influence is because I do use up a lot of headroom inside of the digital chain. The boost that is used does influence how everything works. A 2 volt average RMS out would yield what kind of peak values? I think the observed difference is in the peaks traveling through the digital chain. Why else would it sound way more dynamic?
It's true for both the Xonar and the M1, so it's not the DAC at fault here.
Inside JRiver everything happens in 64 bit, I can't imagine running into a difficulty there, I did try to cut (before processing) and boost (after) as you mention, that didn't change anything sound wise.
With the old Pioneer I used to have the volume barely open and use the JRiver loud. As I sometimes hit the maximum of the input (quite obvious limit) I turned up the volume of the Pioneer and turned down the volume of the chain within JRiver: way better. I may use up to 9 dB internal of headroom, when comparing the DSP IR vs a regular Dirac pulse.
While on average not much watts will be used up, the peaks in dynamic recordings eat up power very quickly once you boost with such numbers.
If one recons the line array is ~ 96 dB sensitivity (over a small limited bandwidth), remember that even Roger Russell said he used up to 20 dB of boost:
So at bass noted it would not be sensitive at all. Which is why we boost that part, but it does eat up headroom fast. Which is why a powerful amp is a good idea, but it seems a good idea to have enough headroom inside of the digital chain. What can I say, it just works better that way...
The +6 dB figure makes sense, the power doesn't double, the voltage does.
Last edited:
Basically what I said about the sub amps, with a SE feed RCA input, could never get them to full power. Is that true? I think it is.
On the other hand, having a Universal buffer at the input of an SE input amp would halve the line in voltage again. So any win there (with differential long lines) is purely a possible reduction in noise rejection only.
On the other hand, having a Universal buffer at the input of an SE input amp would halve the line in voltage again. So any win there (with differential long lines) is purely a possible reduction in noise rejection only.
You should be covered in that buffer is flexiable unit, its input can be single-ended or differential and it always have dual output in one single-ended and one differential, then look yellow marker below whatever gain should be covered, we just need to find out how voltage levels works after conversion when i have ran that TI Tina spice simulation
Attachments
Last edited:
Think about it ballanced should be better in other features but its not the only way to run subwoofer amp up to max potential, its the value of voltage level so if a SE output is gained high enough it can bring that sub amp into cliping levels, so in principle if you want to save on cost and buffer boards then buy one to use get mains pro ballanced and move Atom amp to turbo sub line, ballanced for all three lines mains/sub/ambience is costly and not garuantee be audioable better but electric it will be better when common modes is fully broken and the noise rejecting higher voltage levels do their work.
ballanced should be better in other features but its not the only way to run subwoofer amp up to max potential, its the value of voltage level so if a SE output is gained high enough it can bring that sub amp into cliping levels, so in principle if you want to save on cost and buffer boards then buy one to use get mains pro ballanced and move Atom amp to turbo sub line, ballanced for all three lines mains/sub/ambience is costly and not garuantee be audioable better but electric it will be better when common modes is fully broken and the noise rejecting higher voltage levels do their work.
It could, but it would need double the voltage compared to balanced. Not that smart of a choice if you ask me.
I know you love an all balanced situation, but that's not what my questions are about.
This isn't about saving cost, anywhere. I'm just not totally on the path you'd like to see. Because there is the problem of where to put
a Universal Buffer at the other (amp) side and the open question of which DAC to use, where one of the two possible solutions does have balanced output
and the other does not. I will reduce all channels equally inside JRiver, so I will apply similar on the Universal Buffer side and each pair of outputs does
get such a buffer board. The question remains if there will be 2 more Universal buffers at the input of both main and ambience amps, but getting 3 boards to
start, that can get it's own case at the PC side will allow me to investigate this further. All channels, in- and output, can be run as SE as well as
Differential, so I don't have to make any final decisions. The one thing that is clear for me is the Subs will go balanced right there from the start.
The other two amps could get that long run upgraded a little later, this means one buffer board behind the M1, if I decide that's the route, will be extra.
But it still gives me all the options I want or could wish for. Not a question of money, as time can solve that, but what I want it to do and keeping all options
open as much as possible. Getting one Universal Buffer for the subs would solve everything there at once. Because there the ground problems from running
a PC for audio are potentially higher anyway. A 50 Hz hum is one of the biggest problems PC supply gan give.
Heck, I'm even thinking if I should reserve one more UB as I do have 8 channels available. For future expanding/experimenting. I'm using 6 channels now.
I know you love an all balanced situation, but that's not what my questions are about.
This isn't about saving cost, anywhere. I'm just not totally on the path you'd like to see. Because there is the problem of where to put
a Universal Buffer at the other (amp) side and the open question of which DAC to use, where one of the two possible solutions does have balanced output
and the other does not. I will reduce all channels equally inside JRiver, so I will apply similar on the Universal Buffer side and each pair of outputs does
get such a buffer board. The question remains if there will be 2 more Universal buffers at the input of both main and ambience amps, but getting 3 boards to
start, that can get it's own case at the PC side will allow me to investigate this further. All channels, in- and output, can be run as SE as well as
Differential, so I don't have to make any final decisions. The one thing that is clear for me is the Subs will go balanced right there from the start.
The other two amps could get that long run upgraded a little later, this means one buffer board behind the M1, if I decide that's the route, will be extra.
But it still gives me all the options I want or could wish for. Not a question of money, as time can solve that, but what I want it to do and keeping all options
open as much as possible. Getting one Universal Buffer for the subs would solve everything there at once. Because there the ground problems from running
a PC for audio are potentially higher anyway. A 50 Hz hum is one of the biggest problems PC supply gan give.
Heck, I'm even thinking if I should reserve one more UB as I do have 8 channels available. For future expanding/experimenting. I'm using 6 channels now.
Last edited:
Ah this is similar, but way less pricey. Tom used to have a THAT driver board that did the same, I'd need the other variant though, this one:
SparkFun THAT 1206 InGenius Breakout - BOB-14002 - SparkFun Electronics
As this is a Balanced to SE convertor for the amp input. The reason I like Tom's boards because it leaves all options open. Even moving toward all an balanced setup etc. It has both, but it isn't cheap. It doesn't have to be if it's worth it due to being of sufficient quality though.
I don't have a 15 volt source. So I would need to use one of the options mentioned in the Universal Buffer thread.
According to the datasheet the input sensitivity for a 4 Ohm load is 2.35V for 500W output. So 2V SE would get you very close.
In that case balanced to single ended is really just an opamp combiner as shown above in Tom's drawing, plenty of good quality options there. Tom's board is a step up for the single ended to balanced operation not really so much the other way round.
I was just trying to give you an option in case the Universal buffer and power board didn't fit.
Attachments
I can sense the confusion here, that people think I'm bonkers with what I want. How could I want more than that 2 volt RMS, it should be plenty, right?
I'll present a more compelling case tomorrow to explain. Actually that should read: later today. It's just too much work to type it all out before heading off to bed.
I'll present a more compelling case tomorrow to explain. Actually that should read: later today. It's just too much work to type it all out before heading off to bed.
Interested to hear your case when you wake up
Just to be clear that 2.35V volts balanced is for 500W at 4 ohms, which is the practical limit of your sub amp. That is your input voltage limit. If you feed it balanced you will have to attenuate the input to keep it below 2.35V.
Attachments
I know, I do know those graphs, but I won't ever play the subs louder than they do right now, so yes... it will get attenuated inside JRiver to match the SE levels. That's the whole point! Reducing levels inside JRiver!
Let's talk about the arrays first:
I have often said my average listening levels are around 85 to 87 dB. So let's see how the situation of the Goldmund compares to the old Pioneer.
The way the Pioneer was setup, I had the volume dial set to about 3/4 with JRiver's internal volume attenuated to -14 dB to reach that 85 to 87 dB.
Once the Goldmund was inserted for the first time, we needed to add a pré amp to set the levels the same as before, measured with a song playing and a Radioshack SPL meter.
When that pré amp was removed, I needed to up the level inside JRiver to end up at those same levels. I ended up at -3 to -4 dB internal levels to get to the same output. Now for at least a large part of the frequency spectrum, there was quite a bit of attenuation at those levels. I use boost at the bottom and top part, but also cut the midrange by about the same number.
I know many here consider boost a no go. However even upping the volume level would theoretically count as boost, so if I had used only cut in equalizing, the total SPL level should rise to get to the same output numbers.
This ends up exactly the same as using both cut and boost in the end.
The one warning about using boost EQ should be: do not try to boost a null. Lots of people are scared away from any boost at all for no good reason. Once you boost a null, that's where trouble starts. As the more energy you bring in, the output level does not change by an equal amount, So you only add energy while the end levels don't match up, as a null happens where more than one wave front are subtracting from each other. Say a bounced wave from a wall aligns with the wave front from the speaker, but it's phase is reversed... that's a null at that point in the room.
Anyway, when setting the output, so I get 85 to 87 dB in the room at the listening spot, what happens? Around 200-400 Hz, where the array is working in its most efficient part of the frequency spectrum. To reach those numbers we don't even need a single watt, not even at a listening distance of 2.7 to 3 meter. As said, the top and bottom end are nowhere near as efficient, so there quite a bit of boost is applied to reach the same numbers.
More at the bottom end than at the top though.
However, if we were to assume 95 dB at 1 watt, it would only be valid for a restricted area of the frequency spectrum. In comparison, the bass part probably is only reaching 75 dB at best, maybe even less.
So in the situation as described, while the lower midrange is functioning at a fraction of a watt, the bass is using way more watts to keep up. Same goes for the top end to a lesser extend.
The -3 DB volume level isn't the only thing used. Without any other setting active it would blow my ears off. Within JRiver the Volume level setting is turned on, wich (after analyzing) makes all songs play about equally loud. Never mind the way it was recorded, the R128 algorithm has determined an average loudness of the song and adjusts the volume to reach my desired 85 to 87 dB levels.
So let's play a song. I'll pick Frank Zappa - Find Her Finer as that is the last song I've played and see what happens...
When it's analyzed with JRivers algorithm it comes up with the following numbers:
We have a track that has a dynamic range of 5.5 LU (R128) or to match the Dynamic Range database we see that DR number of 11 dB.
The current JRiver attenuates the songs based on the R128 Volume Level, meaning in playback this song is attenuated 8.5 dB. Let's check that.
Hit play gives us this:
See? The attenuation is indeed -8.5 dB for volume leveling and a further -4 dB that I have chosen to end up at my desired 85 to 87 dB average listening level.
Depending on which number we trust, we could have peaks above this average of either 5.5 dB (R128) or 11 dB (DR as noted by Replaygain)
If I were to play a sine wave with the same average numbers, I could play it all day long without ever surpassing that 2 volt RMS number anywhere in my chain. My maximum boost at low frequencies used to be (prior to subs) set to 14 dB. The maximum cut at the most sensitive part of the array was set to about -10 dB. So my total swing here: 24 dB.
If we look at the attenuation, -8.5 dB + 4 dB, and consider that dynamic range of 11 dB valid, it fits under that total of -12.5 dB of attenuation. Which is why I said it would form no problem at all with a sine wave.
Music is different though. We can have a bass line with a mid frequency wave riding on top of that, which really is what the dynamic range is showing us. So when playing music, we could end up with dynamic peaks that eat away from the 12.5 dB of attenuation. Especially considering that to shape the frequency curve flat we used 14 dB of boost!
Would that be a problem? Let's look at what the specs of the Goldmund say:
GOLDMUND - Products - Telos 400
There's the 350 watt into 8 ohm figure, it also says: 175 watt RMS. But the next line is interesting: Instantaneous power (8 ohms): bigger than 400 watt
Stating a further maximum of 1000 watt! So for short peaks, we have more power available, not all the time, just a burst of energy. Well that's what music contains, sudden bursts of energy, but where exactly in it's frequency content is a lot more difficult to find. Spectrum analyzers may give a clue here.
Let's move on to another song, this time Pink Floyd, Another brick in the Wall (part 1):
Quite a bit of change in this one, more dynamic range, actually less attenuation as determined by the R128 algorithm, -2.5 LU. Dynamic range according to both algorithm's is higher than the last one. 9.9 LU and 13 (DR).
Let's hit play again:
The attenuation is even less than the number found by R128, -2.1 dB. Together with -4 dB internal level as set by the user (me) it has a total attenuation of -6.1 dB. Now depending on where it's dynamic content is, this song potentially hits well above the 2 volt RMS figure of my soundcard!
Remember: a lot of the frequency curve has a further attenuation up to 10 dB. But at bass frequencies, I'm potentially well above that 2 volts number if there are notes there that hit hard!
Even though this scenario seems to be a journey towards disaster it really isn't. I have been playing with similar settings for the most part of 4-5 years with a measly 100 watt amplifier, the Pioneer A757 MKII. Though to keep it sane I had to lower the internal levels and have the pré amp of the Pioneer pick up the difference. Only one or two songs ever got me into real trouble.
That meant it actually had those hard hitting notes right where there's a lot of boost in my EQ settings.
Now I ask: why would using the Atom as a pré amp give me a better feel of dynamics? From loopback tests I know that anything hitting above that 2 volt RMS generates a sudden increase of distortion. By reducing the internal levels and creating headroom to allow even the peaks to stay below it, I shift the problem towards the analog part of the chain.
In a comparison a number like that Pink Floyd song simply sounds more dynamic and real with a reduction in level inside JRiver, which is made up for with the Atom pré amp.
When I don't use that pré amp, and run into peaks higher than they should be, the clip protection from JRiver most probably saves the day:
Early on with the Pioneer I ran JRiver with higher than the mentioned -14 dB SPL levels. The volume knob was at about 10 o clock at that time and lots of songs could upset something. Remember, the Pioneer had way lower input levels than even the 2 volt RMS. Looking at the analyzer I noticed it's behaviour and adjusted towards that latter -14 dB standard.
I probably have clipped in inputs a couple of times with the earlier volume position as that would bring loud farting noises (never heard with the Goldmund at similar settings).
Now I know I will get some comments that might suggest all kind of things. But please read first, let it sink in and realise that if we want a 20 dB headroom (I'm shooting for about ~15 dB) it should mean we actually can play the wild peaks too! Without hard (or soft clipping) limits stopping them!
Let's talk about the arrays first:
I have often said my average listening levels are around 85 to 87 dB. So let's see how the situation of the Goldmund compares to the old Pioneer.
The way the Pioneer was setup, I had the volume dial set to about 3/4 with JRiver's internal volume attenuated to -14 dB to reach that 85 to 87 dB.
Once the Goldmund was inserted for the first time, we needed to add a pré amp to set the levels the same as before, measured with a song playing and a Radioshack SPL meter.
When that pré amp was removed, I needed to up the level inside JRiver to end up at those same levels. I ended up at -3 to -4 dB internal levels to get to the same output. Now for at least a large part of the frequency spectrum, there was quite a bit of attenuation at those levels. I use boost at the bottom and top part, but also cut the midrange by about the same number.
I know many here consider boost a no go. However even upping the volume level would theoretically count as boost, so if I had used only cut in equalizing, the total SPL level should rise to get to the same output numbers.
This ends up exactly the same as using both cut and boost in the end.
The one warning about using boost EQ should be: do not try to boost a null. Lots of people are scared away from any boost at all for no good reason. Once you boost a null, that's where trouble starts. As the more energy you bring in, the output level does not change by an equal amount, So you only add energy while the end levels don't match up, as a null happens where more than one wave front are subtracting from each other. Say a bounced wave from a wall aligns with the wave front from the speaker, but it's phase is reversed... that's a null at that point in the room.
Anyway, when setting the output, so I get 85 to 87 dB in the room at the listening spot, what happens? Around 200-400 Hz, where the array is working in its most efficient part of the frequency spectrum. To reach those numbers we don't even need a single watt, not even at a listening distance of 2.7 to 3 meter. As said, the top and bottom end are nowhere near as efficient, so there quite a bit of boost is applied to reach the same numbers.
More at the bottom end than at the top though.
However, if we were to assume 95 dB at 1 watt, it would only be valid for a restricted area of the frequency spectrum. In comparison, the bass part probably is only reaching 75 dB at best, maybe even less.
So in the situation as described, while the lower midrange is functioning at a fraction of a watt, the bass is using way more watts to keep up. Same goes for the top end to a lesser extend.
The -3 DB volume level isn't the only thing used. Without any other setting active it would blow my ears off. Within JRiver the Volume level setting is turned on, wich (after analyzing) makes all songs play about equally loud. Never mind the way it was recorded, the R128 algorithm has determined an average loudness of the song and adjusts the volume to reach my desired 85 to 87 dB levels.
So let's play a song. I'll pick Frank Zappa - Find Her Finer as that is the last song I've played and see what happens...
When it's analyzed with JRivers algorithm it comes up with the following numbers:
We have a track that has a dynamic range of 5.5 LU (R128) or to match the Dynamic Range database we see that DR number of 11 dB.
The current JRiver attenuates the songs based on the R128 Volume Level, meaning in playback this song is attenuated 8.5 dB. Let's check that.
Hit play gives us this:
See? The attenuation is indeed -8.5 dB for volume leveling and a further -4 dB that I have chosen to end up at my desired 85 to 87 dB average listening level.
Depending on which number we trust, we could have peaks above this average of either 5.5 dB (R128) or 11 dB (DR as noted by Replaygain)
If I were to play a sine wave with the same average numbers, I could play it all day long without ever surpassing that 2 volt RMS number anywhere in my chain. My maximum boost at low frequencies used to be (prior to subs) set to 14 dB. The maximum cut at the most sensitive part of the array was set to about -10 dB. So my total swing here: 24 dB.
If we look at the attenuation, -8.5 dB + 4 dB, and consider that dynamic range of 11 dB valid, it fits under that total of -12.5 dB of attenuation. Which is why I said it would form no problem at all with a sine wave.
Music is different though. We can have a bass line with a mid frequency wave riding on top of that, which really is what the dynamic range is showing us. So when playing music, we could end up with dynamic peaks that eat away from the 12.5 dB of attenuation. Especially considering that to shape the frequency curve flat we used 14 dB of boost!
Would that be a problem? Let's look at what the specs of the Goldmund say:
GOLDMUND - Products - Telos 400
Goldmund Telos 400 spec said:
There's the 350 watt into 8 ohm figure, it also says: 175 watt RMS. But the next line is interesting: Instantaneous power (8 ohms): bigger than 400 watt
Stating a further maximum of 1000 watt! So for short peaks, we have more power available, not all the time, just a burst of energy. Well that's what music contains, sudden bursts of energy, but where exactly in it's frequency content is a lot more difficult to find. Spectrum analyzers may give a clue here.
Let's move on to another song, this time Pink Floyd, Another brick in the Wall (part 1):
Quite a bit of change in this one, more dynamic range, actually less attenuation as determined by the R128 algorithm, -2.5 LU. Dynamic range according to both algorithm's is higher than the last one. 9.9 LU and 13 (DR).
Let's hit play again:
The attenuation is even less than the number found by R128, -2.1 dB. Together with -4 dB internal level as set by the user (me) it has a total attenuation of -6.1 dB. Now depending on where it's dynamic content is, this song potentially hits well above the 2 volt RMS figure of my soundcard!
Remember: a lot of the frequency curve has a further attenuation up to 10 dB. But at bass frequencies, I'm potentially well above that 2 volts number if there are notes there that hit hard!
Even though this scenario seems to be a journey towards disaster it really isn't. I have been playing with similar settings for the most part of 4-5 years with a measly 100 watt amplifier, the Pioneer A757 MKII. Though to keep it sane I had to lower the internal levels and have the pré amp of the Pioneer pick up the difference. Only one or two songs ever got me into real trouble.
That meant it actually had those hard hitting notes right where there's a lot of boost in my EQ settings.
Now I ask: why would using the Atom as a pré amp give me a better feel of dynamics? From loopback tests I know that anything hitting above that 2 volt RMS generates a sudden increase of distortion. By reducing the internal levels and creating headroom to allow even the peaks to stay below it, I shift the problem towards the analog part of the chain.
In a comparison a number like that Pink Floyd song simply sounds more dynamic and real with a reduction in level inside JRiver, which is made up for with the Atom pré amp.
When I don't use that pré amp, and run into peaks higher than they should be, the clip protection from JRiver most probably saves the day:
Early on with the Pioneer I ran JRiver with higher than the mentioned -14 dB SPL levels. The volume knob was at about 10 o clock at that time and lots of songs could upset something. Remember, the Pioneer had way lower input levels than even the 2 volt RMS. Looking at the analyzer I noticed it's behaviour and adjusted towards that latter -14 dB standard.
I probably have clipped in inputs a couple of times with the earlier volume position as that would bring loud farting noises (never heard with the Goldmund at similar settings).
Now I know I will get some comments that might suggest all kind of things. But please read first, let it sink in and realise that if we want a 20 dB headroom (I'm shooting for about ~15 dB) it should mean we actually can play the wild peaks too! Without hard (or soft clipping) limits stopping them!
Attachments
In a more normal setup. with a WAW or a 2 or 3 way, we usually use drivers in their own strong area. Meaning we don't need to use boost above normal levels. So why do we get away with it? Because notes in music mostly hit harder within the (lower) midrange and there we have plenty of output, even on a fraction of a watt.
The subs increase headroom too, as with them added the need for 14 dB boosts at lower frequencies is gone. It is lowered by as much as 10.5 dB so the maximum boost is now 3.5 dB.
Yet, I still feel the same way about the Atom pré-amp. Shifting anything higher than average out of the digital chain, into the analog part. With the Atom set to 6 dB of gain, everything in the digital sense falls under a dirac pulse run through the same system. So in other words, no boost above 'normal' levels within the digital environment. So any swing that might be above that level will trigger the Atom to amplify it, avoiding any soft clip inside JRiver. Where the small signal will be cleaner than if it were generated from within JRiver. Up to the amplifier if it can handle it or not...
Original schematics of the Goldmund included a pré-amp circuit with 6 dB of gain, so I think it will be able to handle that. It will hardly ever be above the nominal levels even after that pré-amp gain. But when it does reach higher, it isn't cut off or left in there with a flat line overflow (the two choices within JRiver to handle peaks).
So the idea here is to keep enough headroom inside the digital side of the system to be able to play anything.
Lowering the internal levels and making up for that with clean amplification outside of the digital realm.
It worked before, tests with the Atom amp indicate it still works better than the alternative. Does this make it more clear?
Think of a Linkwitz boost with dipole bass, he is boosting the woofers for as much as they can take it. After the normal playback device in the analog realm.
This isn't that much different, however digital headroom seems important to have to me, that's the conclusion I've come to. Twice, with two different amplifiers.
Both sound better with analog gain instead of running them to the max inside the digital environment.
End result? A more dynamic sound, more pleasing and natural. Engaging and fun. No strain as if one is up to the limit, making you want to turn it down.
Instead you hardly notice it is playing as loud as it is. That wasn't the case with the Goldmund for a long while, making me believe I had done something wrong
with my processing. It wasn't bad though, just not as good as it could be. I turned the music down more often than ever before and wondered why.
That's when this experiment with the Atom amp came in.
The subs increase headroom too, as with them added the need for 14 dB boosts at lower frequencies is gone. It is lowered by as much as 10.5 dB so the maximum boost is now 3.5 dB.
Yet, I still feel the same way about the Atom pré-amp. Shifting anything higher than average out of the digital chain, into the analog part. With the Atom set to 6 dB of gain, everything in the digital sense falls under a dirac pulse run through the same system. So in other words, no boost above 'normal' levels within the digital environment. So any swing that might be above that level will trigger the Atom to amplify it, avoiding any soft clip inside JRiver. Where the small signal will be cleaner than if it were generated from within JRiver. Up to the amplifier if it can handle it or not...
Original schematics of the Goldmund included a pré-amp circuit with 6 dB of gain, so I think it will be able to handle that. It will hardly ever be above the nominal levels even after that pré-amp gain. But when it does reach higher, it isn't cut off or left in there with a flat line overflow (the two choices within JRiver to handle peaks).
So the idea here is to keep enough headroom inside the digital side of the system to be able to play anything.
Lowering the internal levels and making up for that with clean amplification outside of the digital realm.
It worked before, tests with the Atom amp indicate it still works better than the alternative. Does this make it more clear?
Think of a Linkwitz boost with dipole bass, he is boosting the woofers for as much as they can take it. After the normal playback device in the analog realm.
This isn't that much different, however digital headroom seems important to have to me, that's the conclusion I've come to. Twice, with two different amplifiers.
Both sound better with analog gain instead of running them to the max inside the digital environment.
End result? A more dynamic sound, more pleasing and natural. Engaging and fun. No strain as if one is up to the limit, making you want to turn it down.
Instead you hardly notice it is playing as loud as it is. That wasn't the case with the Goldmund for a long while, making me believe I had done something wrong
with my processing. It wasn't bad though, just not as good as it could be. I turned the music down more often than ever before and wondered why.
That's when this experiment with the Atom amp came in.
Attachments
Last edited:
I understand what you are aiming for and ultimately if you prefer the sound in whatever configuration it makes sense to go that way
In a digital system the only way to exceed 0dB is either to clip or from an intersample over that some dacs handle better than others.
The dynamic peaks are included in that. That is the dynamic range. The peak as analysed by Jriver is the absolute peak of the music in that track, sometimes that exceeds 0dB because of bad mastering or intentional loudness attempts. The 0dB signal will usually give the 2V RMS output, to avoid any chance of intersample overs affecting the output applying -4dB of attenuation digitally at the start is a good idea. If you need to make that back up to make the amp loud enough then some analogue gain is needed.
So it is totally understandable that reducing the output on the DAC by 6 to 10dB and adding that gain back via a preamp results in an improved sound.
Currently I have my DAC set to -11.5dB for my normal listening level. I probably should swap some of that cut at the DAC for a cut in Jrivers output, as with the Sabre it's digital anyway.
In a digital system the only way to exceed 0dB is either to clip or from an intersample over that some dacs handle better than others.
The dynamic peaks are included in that. That is the dynamic range. The peak as analysed by Jriver is the absolute peak of the music in that track, sometimes that exceeds 0dB because of bad mastering or intentional loudness attempts. The 0dB signal will usually give the 2V RMS output, to avoid any chance of intersample overs affecting the output applying -4dB of attenuation digitally at the start is a good idea. If you need to make that back up to make the amp loud enough then some analogue gain is needed.
So it is totally understandable that reducing the output on the DAC by 6 to 10dB and adding that gain back via a preamp results in an improved sound.
Currently I have my DAC set to -11.5dB for my normal listening level. I probably should swap some of that cut at the DAC for a cut in Jrivers output, as with the Sabre it's digital anyway.
I've been following this discussion but didn't have anything to contribute. I knew that balanced wasn't the obvious answer unless the problem was noise/interference picked up on the link. I think the focus on dynamic headroom is appropriate. We see all these level ratings in terms of Vrms that leave the peak undefined while its compression or clipping of the peaks that is problematic.
Its impressive that you can indeed quantify the peaks instead of assuming something like an 8 to 10 db ratio from RMS to peak for music. You have a value for the amp, but what is it for the preamp and line drivers? Simply speaking, its their power supply voltages. Those kinds of circuits are normally able to drive at most rail to rail and no more. So if limited by the typical +/-12V or +/-15V supply rails typical for op amp circuits how does one improve the dynamic headroom? Quite simply going balanced will give you 6 db more!
Its impressive that you can indeed quantify the peaks instead of assuming something like an 8 to 10 db ratio from RMS to peak for music. You have a value for the amp, but what is it for the preamp and line drivers? Simply speaking, its their power supply voltages. Those kinds of circuits are normally able to drive at most rail to rail and no more. So if limited by the typical +/-12V or +/-15V supply rails typical for op amp circuits how does one improve the dynamic headroom? Quite simply going balanced will give you 6 db more!
Well that's the thing, the output of the Xonar is 2 volts RMS. If it goes any higher a distortion raise is obvious as seen in loopback measurements that I've done.
By lowering the internal levels inside of JRiver by the same amount that we substitute as gain in the line driver, we lower it to a value way well below that 2 volts limit. The line driver, for instance a Universal Buffer, then adds gain getting it back to that desired value (Which was somewhere near 2 Volts) but has the potential to do swings up to nearly 10 volts cleanly on 15 volt rails. (that's well over 12 dB more headroom than what the Xonar or M1 can handle)
It would not need to go that high at all in daily use, but when it does, it doesn't clip inside JRiver and has that >12 dB more (clean) headroom.
See: https://www.ti.com/lit/ds/symlink/lm4562.pdf
and:
http://www.ti.com/lit/ds/symlink/lme49720.pdf
Balanced only works if the amp is designed that way, otherwise a second conversion would still be needed on the other side, bringing the level back to single ended so you're right back where you started, at least for amps with single ended input. The advantage is the higher signal during long transports and it's common mode rejection along that long line.
The subs is a different story. Balanced in should indeed gain headroom, right? At least, that's what I've been wondering.
Is it true that if I were to put in 2 volts RMS SE to the Hypex it would be half of what a 2 volts RMS balanced input signal would supply? Thus the 6 dB gain difference?
That's still a bit fuzzy for me.
So first and foremost the plan is to use line drivers with gain(*), probably using it single ended at the start, feeding the RCA's of the Goldmund and Fetzilla.
Optional in this configuration is using the balanced out (4.4 volt) from the M1 to the input of the UB line driver feeding the Goldmund. That's why BYRTT wants
me to put that Universal Buffer inside the Goldmund amplifier. I don't disagree with that, but placement inside above the toroids does not seem wise/make sense.
The linedriver powering the subs will convert SE to Differential and feed the sub that signal.
By lowering the internal levels inside of JRiver by the same amount that we substitute as gain in the line driver, we lower it to a value way well below that 2 volts limit. The line driver, for instance a Universal Buffer, then adds gain getting it back to that desired value (Which was somewhere near 2 Volts) but has the potential to do swings up to nearly 10 volts cleanly on 15 volt rails. (that's well over 12 dB more headroom than what the Xonar or M1 can handle)
It would not need to go that high at all in daily use, but when it does, it doesn't clip inside JRiver and has that >12 dB more (clean) headroom.
See: https://www.ti.com/lit/ds/symlink/lm4562.pdf
and:
http://www.ti.com/lit/ds/symlink/lme49720.pdf
Balanced only works if the amp is designed that way, otherwise a second conversion would still be needed on the other side, bringing the level back to single ended so you're right back where you started, at least for amps with single ended input. The advantage is the higher signal during long transports and it's common mode rejection along that long line.
The subs is a different story. Balanced in should indeed gain headroom, right? At least, that's what I've been wondering.
Is it true that if I were to put in 2 volts RMS SE to the Hypex it would be half of what a 2 volts RMS balanced input signal would supply? Thus the 6 dB gain difference?
That's still a bit fuzzy for me.
So first and foremost the plan is to use line drivers with gain(*), probably using it single ended at the start, feeding the RCA's of the Goldmund and Fetzilla.
Optional in this configuration is using the balanced out (4.4 volt) from the M1 to the input of the UB line driver feeding the Goldmund. That's why BYRTT wants
me to put that Universal Buffer inside the Goldmund amplifier.
I don't disagree with that, but placement inside above the toroids does not seem wise/make sense.The linedriver powering the subs will convert SE to Differential and feed the sub that signal.
Last edited:
Do you know the Xonar output circuit? I did a little research and found its ESS DAC operation with only a 3.6V power supply. That would account for your observation. It needs some gain between the DAC and the amp so that the peaks out of the DAC don't need to be more than 3V or so. I agree with your plan
This is no reflection on the ESS design but a consequence of the advanced semiconductor process its built in that doesn't tolerate higher voltages.
ESS Technology :: ES9023
In balanced, the peak to peak swing is 2x the peak to peak swing of single ended because one side swings opposite to the other so the net swing is indeed double. Double maximum voltage swing is +6db headroom
This is no reflection on the ESS design but a consequence of the advanced semiconductor process its built in that doesn't tolerate higher voltages.
ESS Technology :: ES9023
In balanced, the peak to peak swing is 2x the peak to peak swing of single ended because one side swings opposite to the other so the net swing is indeed double. Double maximum voltage swing is +6db headroom
Correct... I don't have the exact schematics, I just noticed the limitations there due to wondering what I was hearing and why.
A bit of useful information was found in this test: ASUS Xonar Essence ST DeLuxe review - Electronics Overview
The M1 DAC has 2.2 Volts RMS SE or 4.4 Volts Balanced out. But it needs to feed a single ended input on the Goldmund amplifier.
Basically it has similar limits as the Xonar card for my use. Loopback tests looked very similar.
It does sound rather different as I've mentioned, but part of that difference could be the upsampling process in that DAC.
Aside from the differing components used that is...
The test setup, long ago, with the HP-1 headphone amp as a pré-amp proved that the Goldmund could really deliver. The differences in bass compared
to all other amps we tested was huge. There was between 6 to 12 dB of gain set with the HP-1 to level it with the Pioneer I was using at that time.
We used the same song sequence with the Radioshack SPL meter to match them.
That HP-1 was an extremely powerful and clean beast and one very expensive pré-amp.
A bit of useful information was found in this test: ASUS Xonar Essence ST DeLuxe review - Electronics Overview
The M1 DAC has 2.2 Volts RMS SE or 4.4 Volts Balanced out. But it needs to feed a single ended input on the Goldmund amplifier.
Basically it has similar limits as the Xonar card for my use. Loopback tests looked very similar.
It does sound rather different as I've mentioned, but part of that difference could be the upsampling process in that DAC.
Aside from the differing components used that is...
The test setup, long ago, with the HP-1 headphone amp as a pré-amp proved that the Goldmund could really deliver. The differences in bass compared
to all other amps we tested was huge. There was between 6 to 12 dB of gain set with the HP-1 to level it with the Pioneer I was using at that time.
We used the same song sequence with the Radioshack SPL meter to match them.
That HP-1 was an extremely powerful and clean beast and one very expensive pré-amp.
Last edited: