Hello Everyone,
This is a review and measurements of the SMSL G1 Clock Generator.
It was kindly sent to me by Aoshidaaudio.com
With this presentation, I'm stepping outside of my comfort zone. So, thank you in advance for your patience and understanding, and please feed me with your comments and requests for more info / data. I guess this initial post will evolve with your feedback.
SMSL G1 - Presentation
The SMSL G1 is a low phase noise high precision 10MHz clock source. It is to be used to synchronize external devices to the same master clock.
Master clocks like the G1 are seen in professional environments much more than at home. The obvious reason is that studios are using several digital devices that need to talk between them or to a controlling console. Every time there is the need to synchronize the communication. It is possible to extract the clock from the signal and resync to the internal clock of every devices, but that multiplies potential deviations from a "perfect" clock, creating phase noise (in frequency domain) aka jitter (in time domain).
Instead of having so many clock generators in every device, with different specs, it is obviously less expensive and more efficient to have only one "Master" that can be of higher precision for a similar total cost.
Many of our devices, such as a DAC or CD transport use an internal clocks, based on a crystal for the modern ones. There are different flavors of them:
The SMSL G1 uses an OCXO and offers 4 outputs at 10MHz. There's nothing to configure, not one single button on the front face. Plug it to AC or DC (9-15V), and after a couple of minutes the OCXO will have reached its optimum stability.
For the measurements, I connected the SMSL G1 to the SMSL D200 DAC and mainly used it that way. I also tested with the SMSL PL200T CD transport, syncing the two of them.
I used the AC and DC to test potential differences, and the only one is that the little front led turns red instead of blue with AC. Oh yes, and there is a longer time for the clock to stabilize with AC (roughly 10min per my measurements).
That's the full stack:
When the D200 or PL200T use the G1, a little square wave shows on the top left of the screen (above only the D200 was using the G1). The back is a bit messy:
And note that the box included one cable to connect to a device, which is a nice attention.
SMSL G1 - Specifications
SMSL shares only two not well documented specs for the G1:
Clock accuracy:
Let's start with the Clock Accuracy. This is often expressed in part per million (ppm). I always provide you with the clock deviations seen from the output of a CD player. For that I use a sine tone at 19'997.00Hz and measure the output of the CD Player, after conversion. For instance, the 19'996.70Hz that I measured from the recently reviewed Pioneer PD-D9, translates to a pitch error (or clock deviation) of -15ppm. With a precision of 0.2ppm or less, I would get the exact 19'997.00Hz tone at the output of the CD Player.
In reality, for audio replay at home, 50ppm is more than good enough, as that would mean getting 400.02Hz instead of 400.00Hz from a tuning fork. And from what I could gather, the very best of them are accurate by +/-0.05Hz, which translates to +/-125ppm...
SMSL claims a crazy 3ppb (part per billion) clock stability, which is 300 times more precise than 1ppm! I can't mesure that, even provided my input interface would be that precise, because the Software I use (REW) has only two decimals when capturing the incoming frequency, allowing at best 0.03ppm measurement (when using a 300'000kHz test tone).
I briefly looked at the competition, and I'll mention only one example of external clocks specifications. dCS has been selling them for years, to use with their separated devices (Transport, upsampler, converter) and for instance the Lina Master Clock is given for a clock accuracy better than +/-1ppm.
Phase Noise:
Let's continue with the Phase Noise. Internally, every single device (eg DAC) will have needs for a different clock speed and so will have to multiply or divide the "base" clock speed that is provided by the internal (or external) crystal to the desired one. This is done by a Phase-Locked Loop.
If we take the example of a DAC, it has to deal with different sampling rates (eg 44.1kHz to 192kHz for PCM, 2.8224MHz for DSD) but also the internal DAC chip have various needs, such as from the oversampling filter (running at x times the sampling rate), but also the delta-sigma modulator which interpolates by a factor of the sampling rate coming from the oversampling filter. All of these varies from DACs to DACs and they all use PLLs to create the required clock speed for the specific activity. We also understand from the above that the multiplier or divider is not always an integer, so it's not that trivial, although PLLs have very much matured in the past decades. In a nutshell, PLL is a simple negative feedback architecture that allows economic multiplication of crystal frequencies by variable numbers (TI Technical Brief SWRA029).
For a PLL, the Phase Noise is an important parameter, an indicator of signal quality. Phase noise and jitter are the same, the latter manifesting itself in time domain as opposed to frequency domain, like I said before.
Specifications of the G1 are given in dBc/Hz, which refers to attenuating phase noise at a specific frequency offset. Phase noise is the unwanted, random phase fluctuations which can be measured as short-term frequency instability, and that I already mentioned when reviewing the SMSLS PL200T.
SMSL mentions that at 1Hz, phase noise is -105dBc/Hz. This means the phase noise power in a 1Hz bandwidth, located 1Hz away from the carrier, is 105dB below the carrier power. That is very good and improves with frequency. In a presentation, SMSL showed a sweep of phase attenuation vs frequency which improves as frequency goes up. At 1kHz, it is already a very high attenuation of -151dBc/Hz, meaning the phase noise power in a 1Hz bandwidth, located 1kHz away from the carrier, is 151dB below the carrier power.
All of this is relevant because the basic feedback loop principles and theory tells us that a multiplication of a given frequency by N raises the signal’s phase noise by 20log(N)dB. And that means a division will reduce phase noise, and therefore jitter as they are linked.
SMSL went for a clock of 10Mhz which means most of the time, the internal PLLs consuming it will need to reduce that speed and so reduce Phase Noise and Jitter.
SMSL G1 - Measurements
Of course, this type of product can't be measured as a DAC. What we expect from this device is to reduce phase noise, therefore reduce jitter, reduce pitch-error...
Before I go into measurements, I had to identify what to measure and how, with the equipment available to me. That was a challenge.
Jitter Test:
Of course the first test that comes to mind is the standardized Jitter Test (24bits 48kHz sampling rate).
For the below test, I used the SMSL D200, with its internal clock, and next the G1. Measurements were done with a punishing lengthy FFT of 512k with 32 averages as to lower the random low level noise so that we can better see potential side bands:
Like I said when reviewing the D200, this is a nailed test, side bands are seen because of the FFT length and 32 averages.
Next is the same measurement, and this time the D200 used the G1 Clock:
Yeah, well, nearly identical, no surprises.
Let's try the same from the CD player at 16 bits. The below is an overlay of 3 tests:
The three traces overlay nearly perfectly and the very small differences are not worth commenting.
Pitch Error:
I've been talking about pitch error for every CD players that I reviewed. I measure it from a 19'997.00Hz test tone and report the output of the CD player. That way, I can measure down to 0.5ppm precision (if the resulting frequency is 19'9997.01Hz), and with all the test I performed for this review, this is below the limit of the Cosmo that I regularly use. So I relied on my Motu Ultralite mk5 which seems to do better, but in the end, no big difference between the two.
First, when the SMSL PL200T feeds the SMLS D200 DAC, I get the below:
This is 1ppm precision, when the two items are used together. That is a very good result.
Now, let's sync the two items with the G1:
We get less precision, but still a very good 4ppm (of no concern). It's the same result if I synced only the D200 or the PL200T to the G1.
EDIT: as suggested by @Rja4000, I used the Leo Bodnar to calibrate the pitch error of the Motu, which happens to be +3.5ppm. So the improvement is better than shown above, as the real result is therefore below 1ppm.
Phase noise reduction:
Much more interesting is what follows. Since the G1 PLL has such a low phase noise, maybe we can benefit from an improved clock reconstruction precision when the DAC is fed by a less precise device than the PL200T?
That would be an indirect positive influence on the DAC processing SPDIF input from a CD Player that suffers clock deviation. Most of the time, when I test a CD Player from its digital output feeding an external DAC, I find the exact same clock imprecision transitioning through the DAC than what I find when testing the analog output of the Player. That means the lack of clock's accuracy from the CD player travels through the SPDIF and the PLL of the DAC does not reduce the phase noise, or very little.
For that test, I used four different CD Players, with variable pitch errors from 4ppm (Onkyo DX-7355) to 100ppm (Revox B-226 S). The test is based on my standard pitch error test tone of 19.997.00Hz. Every time, I first ran the test from the CD player capturing its analog outputs, then the same but connecting to the CD Player to the SMSL D200 via optical output. Last, same but the D200 uses the G1 clock instead of its internal one.
I summarized the results in the below table, rather than flooding this post with FFTs (Edit: table updated to account for the internal pitch error of the Motu):
The results talk by themselves. Where the D200 DAC alone failed to reduce the pitch error coming from the CD Players (as I'm used to see with other DACs too), the G1 stabilized the resulting output to the same below 1ppm deviations whatever is the accuracy of the incoming digital signal.
I was not expecting that benefit when I started this review. I discovered that with measurements, and had to try understanding a little more (asking the AI to teach me), hence all the blahblah above and further testing that required quite some time.
I was especially surprised with the Revox B 226-S that is a very old CD Player, I it like very much. It was restored a year ago at a significant cost, and is "like new" but was still suffering from higher than usual clock deviation. I was sad and thought about potential internal upgrade of its clock, but I got lazy and moved on to something else. Now, seeing that this clock deviation can be reduced to negligible with the help of the G1, well, that changes my perspective on this classic CD player.
Additional tests
And for fun, let me share the before/after from that old Revox (measurement interface is the Motu Ultralite Mk5).
This is my standard measurement of 999.91Hz undithered test tone @0BFS from the RCA analog outputs of the B 226-S:
This is an old Philips TDA1541A-S1 (Crown) conversion, from decades ago. It is still more than decent to enjoy the silver disc. We only loose 1bit of resolution in noise and distortion at full scale (results degrade at lower scale due to non-linearities of this DAC). The output frequency at 1'000.07Hz shows the clock imprecision of 160ppm, which is higher than the previous as obviously the Reveox deviates more when getting into temperature.
Now the same test, but this time the Revox is feeding the Topping D50III DAC via coax:
I used on purpose another DAC than the SMSL D200 to show that the clock deviation transitions to the DAC (1’000.07Hz instead 999.91Hz). It's the same result with the D200.
Now, from the Coax output of the Revox again, feeding the SMSL D200 and G1:
Besides best in class results for the CD Audio, we also get the exact 999.91Hz output frequency, finally!
Conclusions
I am sure I missed something and/or made mistakes, since this is a first for me. So please, be so kind to provide me with feedback so I can improve this review and do better next time a master clock comes to me
I was initially disappointed not to see super low ppm. But I think I was at the limits of what I can safely measure on that perspective. That said, I was more than positively surprised to see improvements in pitch errors with the ancient CD players that I love so much.
In the end, I must say this is a very niche product, but at least it delivered really positive outcomes. I can't say you need the G1 when you use the PL200T, that would be false. But if you use any other player as a transport, you get the assurance to have best in class measurements.
I hope you enjoyed this review and I wish you a nice weekend!
This is a review and measurements of the SMSL G1 Clock Generator.
It was kindly sent to me by Aoshidaaudio.com
With this presentation, I'm stepping outside of my comfort zone. So, thank you in advance for your patience and understanding, and please feed me with your comments and requests for more info / data. I guess this initial post will evolve with your feedback.
SMSL G1 - Presentation
The SMSL G1 is a low phase noise high precision 10MHz clock source. It is to be used to synchronize external devices to the same master clock.
Master clocks like the G1 are seen in professional environments much more than at home. The obvious reason is that studios are using several digital devices that need to talk between them or to a controlling console. Every time there is the need to synchronize the communication. It is possible to extract the clock from the signal and resync to the internal clock of every devices, but that multiplies potential deviations from a "perfect" clock, creating phase noise (in frequency domain) aka jitter (in time domain).
Instead of having so many clock generators in every device, with different specs, it is obviously less expensive and more efficient to have only one "Master" that can be of higher precision for a similar total cost.
Many of our devices, such as a DAC or CD transport use an internal clocks, based on a crystal for the modern ones. There are different flavors of them:
- LC and RC Oscillators (from the past, relatively unstable and not aging well)
- XO : Crystal Oscillator (basic type)
- VCXO : Voltage-Controlled Crystal Oscillators (Crystal oscillator that can have its frequency adjusted by external voltage)
- TCXO : Temperature-Compensated Crystal Oscillators (auto-adjust of frequency to compensate ambient temperature changes)
- VCTCXO : Voltage-Controlled Temperature-Compensated Crystal Oscillator (combines the above two)
- OCXO : Oven-Controlled Crystal Oscillators (use an internal oven to maintain a constant temperature, providing the highest stability)
The SMSL G1 uses an OCXO and offers 4 outputs at 10MHz. There's nothing to configure, not one single button on the front face. Plug it to AC or DC (9-15V), and after a couple of minutes the OCXO will have reached its optimum stability.
For the measurements, I connected the SMSL G1 to the SMSL D200 DAC and mainly used it that way. I also tested with the SMSL PL200T CD transport, syncing the two of them.
I used the AC and DC to test potential differences, and the only one is that the little front led turns red instead of blue with AC. Oh yes, and there is a longer time for the clock to stabilize with AC (roughly 10min per my measurements).
That's the full stack:
When the D200 or PL200T use the G1, a little square wave shows on the top left of the screen (above only the D200 was using the G1). The back is a bit messy:
And note that the box included one cable to connect to a device, which is a nice attention.
SMSL G1 - Specifications
SMSL shares only two not well documented specs for the G1:
- Clock Frequency accuracy of 3ppb
- Phase noise of -105dBc/Hz at 1Hz
Clock accuracy:
Let's start with the Clock Accuracy. This is often expressed in part per million (ppm). I always provide you with the clock deviations seen from the output of a CD player. For that I use a sine tone at 19'997.00Hz and measure the output of the CD Player, after conversion. For instance, the 19'996.70Hz that I measured from the recently reviewed Pioneer PD-D9, translates to a pitch error (or clock deviation) of -15ppm. With a precision of 0.2ppm or less, I would get the exact 19'997.00Hz tone at the output of the CD Player.
In reality, for audio replay at home, 50ppm is more than good enough, as that would mean getting 400.02Hz instead of 400.00Hz from a tuning fork. And from what I could gather, the very best of them are accurate by +/-0.05Hz, which translates to +/-125ppm...
SMSL claims a crazy 3ppb (part per billion) clock stability, which is 300 times more precise than 1ppm! I can't mesure that, even provided my input interface would be that precise, because the Software I use (REW) has only two decimals when capturing the incoming frequency, allowing at best 0.03ppm measurement (when using a 300'000kHz test tone).
I briefly looked at the competition, and I'll mention only one example of external clocks specifications. dCS has been selling them for years, to use with their separated devices (Transport, upsampler, converter) and for instance the Lina Master Clock is given for a clock accuracy better than +/-1ppm.
Phase Noise:
Let's continue with the Phase Noise. Internally, every single device (eg DAC) will have needs for a different clock speed and so will have to multiply or divide the "base" clock speed that is provided by the internal (or external) crystal to the desired one. This is done by a Phase-Locked Loop.
If we take the example of a DAC, it has to deal with different sampling rates (eg 44.1kHz to 192kHz for PCM, 2.8224MHz for DSD) but also the internal DAC chip have various needs, such as from the oversampling filter (running at x times the sampling rate), but also the delta-sigma modulator which interpolates by a factor of the sampling rate coming from the oversampling filter. All of these varies from DACs to DACs and they all use PLLs to create the required clock speed for the specific activity. We also understand from the above that the multiplier or divider is not always an integer, so it's not that trivial, although PLLs have very much matured in the past decades. In a nutshell, PLL is a simple negative feedback architecture that allows economic multiplication of crystal frequencies by variable numbers (TI Technical Brief SWRA029).
For a PLL, the Phase Noise is an important parameter, an indicator of signal quality. Phase noise and jitter are the same, the latter manifesting itself in time domain as opposed to frequency domain, like I said before.
Specifications of the G1 are given in dBc/Hz, which refers to attenuating phase noise at a specific frequency offset. Phase noise is the unwanted, random phase fluctuations which can be measured as short-term frequency instability, and that I already mentioned when reviewing the SMSLS PL200T.
SMSL mentions that at 1Hz, phase noise is -105dBc/Hz. This means the phase noise power in a 1Hz bandwidth, located 1Hz away from the carrier, is 105dB below the carrier power. That is very good and improves with frequency. In a presentation, SMSL showed a sweep of phase attenuation vs frequency which improves as frequency goes up. At 1kHz, it is already a very high attenuation of -151dBc/Hz, meaning the phase noise power in a 1Hz bandwidth, located 1kHz away from the carrier, is 151dB below the carrier power.
All of this is relevant because the basic feedback loop principles and theory tells us that a multiplication of a given frequency by N raises the signal’s phase noise by 20log(N)dB. And that means a division will reduce phase noise, and therefore jitter as they are linked.
SMSL went for a clock of 10Mhz which means most of the time, the internal PLLs consuming it will need to reduce that speed and so reduce Phase Noise and Jitter.
SMSL G1 - Measurements
Of course, this type of product can't be measured as a DAC. What we expect from this device is to reduce phase noise, therefore reduce jitter, reduce pitch-error...
Before I go into measurements, I had to identify what to measure and how, with the equipment available to me. That was a challenge.
Jitter Test:
Of course the first test that comes to mind is the standardized Jitter Test (24bits 48kHz sampling rate).
For the below test, I used the SMSL D200, with its internal clock, and next the G1. Measurements were done with a punishing lengthy FFT of 512k with 32 averages as to lower the random low level noise so that we can better see potential side bands:
Like I said when reviewing the D200, this is a nailed test, side bands are seen because of the FFT length and 32 averages.
Next is the same measurement, and this time the D200 used the G1 Clock:
Yeah, well, nearly identical, no surprises.
Let's try the same from the CD player at 16 bits. The below is an overlay of 3 tests:
- SMSL PL200T and D200 (Red)
- SMSL PL200T synced with the G1 and D200 (Green)
- Both SMSL PL200T and D200 synced with the G1 (Blue)
The three traces overlay nearly perfectly and the very small differences are not worth commenting.
Pitch Error:
I've been talking about pitch error for every CD players that I reviewed. I measure it from a 19'997.00Hz test tone and report the output of the CD player. That way, I can measure down to 0.5ppm precision (if the resulting frequency is 19'9997.01Hz), and with all the test I performed for this review, this is below the limit of the Cosmo that I regularly use. So I relied on my Motu Ultralite mk5 which seems to do better, but in the end, no big difference between the two.
First, when the SMSL PL200T feeds the SMLS D200 DAC, I get the below:
This is 1ppm precision, when the two items are used together. That is a very good result.
Now, let's sync the two items with the G1:
We get less precision, but still a very good 4ppm (of no concern). It's the same result if I synced only the D200 or the PL200T to the G1.
EDIT: as suggested by @Rja4000, I used the Leo Bodnar to calibrate the pitch error of the Motu, which happens to be +3.5ppm. So the improvement is better than shown above, as the real result is therefore below 1ppm.
Phase noise reduction:
Much more interesting is what follows. Since the G1 PLL has such a low phase noise, maybe we can benefit from an improved clock reconstruction precision when the DAC is fed by a less precise device than the PL200T?
That would be an indirect positive influence on the DAC processing SPDIF input from a CD Player that suffers clock deviation. Most of the time, when I test a CD Player from its digital output feeding an external DAC, I find the exact same clock imprecision transitioning through the DAC than what I find when testing the analog output of the Player. That means the lack of clock's accuracy from the CD player travels through the SPDIF and the PLL of the DAC does not reduce the phase noise, or very little.
For that test, I used four different CD Players, with variable pitch errors from 4ppm (Onkyo DX-7355) to 100ppm (Revox B-226 S). The test is based on my standard pitch error test tone of 19.997.00Hz. Every time, I first ran the test from the CD player capturing its analog outputs, then the same but connecting to the CD Player to the SMSL D200 via optical output. Last, same but the D200 uses the G1 clock instead of its internal one.
I summarized the results in the below table, rather than flooding this post with FFTs (Edit: table updated to account for the internal pitch error of the Motu):
| CD Player | Native Pitch Error out of the CD Player | Pitch Error out of SMSL D200 | Pitch Error out of SMSL P200 + G1 |
| SMSL PL220T | -2.5ppm | -2.5ppm | <1ppm |
| Onkyo DX-7355 | <1ppm | <1ppm | <1ppm |
| Pioneer PD-D9 | -19.5ppm | -19.5ppm | <1ppm |
| Sony CDP-337ESD | -32.5ppm | -32.5ppm | <1ppm |
| Revox B 226-S | +160ppm | +122.5ppm | <1ppm |
The results talk by themselves. Where the D200 DAC alone failed to reduce the pitch error coming from the CD Players (as I'm used to see with other DACs too), the G1 stabilized the resulting output to the same below 1ppm deviations whatever is the accuracy of the incoming digital signal.
I was not expecting that benefit when I started this review. I discovered that with measurements, and had to try understanding a little more (asking the AI to teach me), hence all the blahblah above and further testing that required quite some time.
I was especially surprised with the Revox B 226-S that is a very old CD Player, I it like very much. It was restored a year ago at a significant cost, and is "like new" but was still suffering from higher than usual clock deviation. I was sad and thought about potential internal upgrade of its clock, but I got lazy and moved on to something else. Now, seeing that this clock deviation can be reduced to negligible with the help of the G1, well, that changes my perspective on this classic CD player.
Additional tests
And for fun, let me share the before/after from that old Revox (measurement interface is the Motu Ultralite Mk5).
This is my standard measurement of 999.91Hz undithered test tone @0BFS from the RCA analog outputs of the B 226-S:
This is an old Philips TDA1541A-S1 (Crown) conversion, from decades ago. It is still more than decent to enjoy the silver disc. We only loose 1bit of resolution in noise and distortion at full scale (results degrade at lower scale due to non-linearities of this DAC). The output frequency at 1'000.07Hz shows the clock imprecision of 160ppm, which is higher than the previous as obviously the Reveox deviates more when getting into temperature.
Now the same test, but this time the Revox is feeding the Topping D50III DAC via coax:
I used on purpose another DAC than the SMSL D200 to show that the clock deviation transitions to the DAC (1’000.07Hz instead 999.91Hz). It's the same result with the D200.
Now, from the Coax output of the Revox again, feeding the SMSL D200 and G1:
Besides best in class results for the CD Audio, we also get the exact 999.91Hz output frequency, finally!
Conclusions
I am sure I missed something and/or made mistakes, since this is a first for me. So please, be so kind to provide me with feedback so I can improve this review and do better next time a master clock comes to me
I was initially disappointed not to see super low ppm. But I think I was at the limits of what I can safely measure on that perspective. That said, I was more than positively surprised to see improvements in pitch errors with the ancient CD players that I love so much.
In the end, I must say this is a very niche product, but at least it delivered really positive outcomes. I can't say you need the G1 when you use the PL200T, that would be false. But if you use any other player as a transport, you get the assurance to have best in class measurements.
I hope you enjoyed this review and I wish you a nice weekend!
Last edited:
