Dear Ian,
The huge picture above makes me think the two DIL16 layout arrengement will not allow to plug Andrea Mori DIL adaptators with their osscillator boards but for the right area near the power supply.
Are you aware of a supplier of OCXO with DIL16 arrengement or with a compatibility with your SMD XOs adaptors or the CHDD-957 adaptors?
Thank you very much for your advices.
The huge picture above makes me think the two DIL16 layout arrengement will not allow to plug Andrea Mori DIL adaptators with their osscillator boards but for the right area near the power supply.
Are you aware of a supplier of OCXO with DIL16 arrengement or with a compatibility with your SMD XOs adaptors or the CHDD-957 adaptors?
Thank you very much for your advices.
Hi Ian,
Perhaps the answer to this is in one of the earlier posts, but I could not quite get it out and want to be sure: I have an ES9038 DAC that does not have a MCK input (only LRCK, BCK and DATA). In the FiFoPi manual I read "external DACs at synchronized master clock mode". Can I use FiFoPi to feed my DAC or does that not make sense?
Maarten
Perhaps the answer to this is in one of the earlier posts, but I could not quite get it out and want to be sure: I have an ES9038 DAC that does not have a MCK input (only LRCK, BCK and DATA). In the FiFoPi manual I read "external DACs at synchronized master clock mode". Can I use FiFoPi to feed my DAC or does that not make sense?
Maarten
@maartn,
Not Ian, but I can answer here. Yes feeding your ES9038 DAC that does not have a MCK input with a FiFoPi does make sense. The FiFoPi will reduce jitter on the 3 I2S lines your DAC uses to good effect.
I have used a similar setup (Ian IsolatorPi with Allo Kali) and gotten good results on a number of DACs that do not use the MCK feed, such as the Mamboberry (ES9023) and Ian's ES9028K2M & ES9028Q2M prototypes.
I DID prefer the sound with them operated in synchronous mode using the MCK feed, but they addition of a FIFO reclocker improved them even without an MCK feed.
I also use one of Ian's original FIFOs before each of my 2 Twisted Pear BuffaloIIIPro DACs which currently are not using the MCK from the FIFO, again with good results.
Greg in Mississippi
Not Ian, but I can answer here. Yes feeding your ES9038 DAC that does not have a MCK input with a FiFoPi does make sense. The FiFoPi will reduce jitter on the 3 I2S lines your DAC uses to good effect.
I have used a similar setup (Ian IsolatorPi with Allo Kali) and gotten good results on a number of DACs that do not use the MCK feed, such as the Mamboberry (ES9023) and Ian's ES9028K2M & ES9028Q2M prototypes.
I DID prefer the sound with them operated in synchronous mode using the MCK feed, but they addition of a FIFO reclocker improved them even without an MCK feed.
I also use one of Ian's original FIFOs before each of my 2 Twisted Pear BuffaloIIIPro DACs which currently are not using the MCK from the FIFO, again with good results.
Greg in Mississippi
Last edited:
Hi Greg, are u using The Buffalo-IIIse Pro (Stereo Edition) 2-Channel DAC this version of buffalo DAC? It seems that this DAC has a master clock input so why are u using it wthout? Which is your general impression regarding the audio quality ? Which I/V stage are u using ?
I'm looking it to use after che fifopi but i've some doubts considering the price.
Paolo
@paoloilpizzo, I have 2 TP BIIIPro setups... the ES9028PRO feeding a Legato, the ES9038PRO feeding a Mercury. I have also used a custom single-stage OPA1632 I/V (similar to the TP IVY, but using only 1 OPA1632 stage) on the ES9028PRO.
I found the OPA1632 stages to be competent technically, but they did not speak to me emotionally. The Mercury is VERY good technically, the work TP did is quite evident. BUT the setup with the Legato provided a greater sense of the music to me.
Note that I have not spent much time refining these setups as yet. I am sure I can wring more out of each and that may change my impressions. BUT that is where they stand as now.
I posted more details of these earlier in this thread, posts 4663, 4664, 4665, 4668, and 4669.
The main reason I am not YET running them in full-sync mode using the MCK output from the FIFOs is that I've been waiting for updated firmware from TP to enable this (& that reminds me I need to go check if they've made that available yet!). Also that I've been focusing on other projects. BUT once I can do this, I intend to run them this way. PLUS I have a number of other upgrades to do and/or try.
One thing to clarify for those who may not understand this... ALL ESS-based DACs that I am aware of (ES9008, 9018, 9022, 9023, 9028, 9028K2M, 9028Q2M, 9038, and 9038Q2M... hopefully I didn't leave any out!) DO have a MCK input AND it must be used. The typical way to use them is to feed an appropriate non-audio sampling frequency-related clock (typically 50mHz, 80mHz or 100mHz) into that input and the DAC uses the built-in Asynchronous Resampling engine in the reconstruction process. BUT one can also feed an appropriate audio sampling frequency clock (often 11.2896 MHz for 44.1-based music and 12.2880 MHz for 48-based, but they may be integer multiples or divisions of these). In the earlier chips, this supposedly 'bypassed' much of the ASRC processing. In the later chips, you feed the appropriate clock frequency for the music signal AND can actually turn off the ASRC in the DAC settings (via I2C). In my experience with these DAC chips, I find I prefer the sound qualities when run this way. BUT in the same latest ESS chips people have reported that the ASRC sonics are much closer to that of 'full sync mode' and there is less of a reason to enable it, which of course requires something like Ian's FIFO to detect the sampling frequency and chose which clock to use for a given sampling family.
Sorry, this got long!
Greg in Mississippi
I found the OPA1632 stages to be competent technically, but they did not speak to me emotionally. The Mercury is VERY good technically, the work TP did is quite evident. BUT the setup with the Legato provided a greater sense of the music to me.
Note that I have not spent much time refining these setups as yet. I am sure I can wring more out of each and that may change my impressions. BUT that is where they stand as now.
I posted more details of these earlier in this thread, posts 4663, 4664, 4665, 4668, and 4669.
The main reason I am not YET running them in full-sync mode using the MCK output from the FIFOs is that I've been waiting for updated firmware from TP to enable this (& that reminds me I need to go check if they've made that available yet!). Also that I've been focusing on other projects. BUT once I can do this, I intend to run them this way. PLUS I have a number of other upgrades to do and/or try.
One thing to clarify for those who may not understand this... ALL ESS-based DACs that I am aware of (ES9008, 9018, 9022, 9023, 9028, 9028K2M, 9028Q2M, 9038, and 9038Q2M... hopefully I didn't leave any out!) DO have a MCK input AND it must be used. The typical way to use them is to feed an appropriate non-audio sampling frequency-related clock (typically 50mHz, 80mHz or 100mHz) into that input and the DAC uses the built-in Asynchronous Resampling engine in the reconstruction process. BUT one can also feed an appropriate audio sampling frequency clock (often 11.2896 MHz for 44.1-based music and 12.2880 MHz for 48-based, but they may be integer multiples or divisions of these). In the earlier chips, this supposedly 'bypassed' much of the ASRC processing. In the later chips, you feed the appropriate clock frequency for the music signal AND can actually turn off the ASRC in the DAC settings (via I2C). In my experience with these DAC chips, I find I prefer the sound qualities when run this way. BUT in the same latest ESS chips people have reported that the ASRC sonics are much closer to that of 'full sync mode' and there is less of a reason to enable it, which of course requires something like Ian's FIFO to detect the sampling frequency and chose which clock to use for a given sampling family.
Sorry, this got long!
Greg in Mississippi
Hi Greg, did you used also an I/V Transformer stage with (for eg.) an LL1544A Transformer ?
@Supersurfer, I bet there'd be a lot of opinions on what clocks are the for the FifoPi. Ian (and I) both recommend the Crystek CCHD-957 units as very good, not too expensive selections. Others have had good results with the NDK SDA series. AND there are other, MUCH more expensive choices you can select.
@paoloilpizzo, I believe Ian tried the LL1544A transformers, but I have not. Hopefully he'll see this and provide his feedback. I have a pair of LL1684 transformers to try, but haven't had a chance to assemble them onto a board and try them yet. The Onetics I like clearly bested a pair of LL1590XL, but I suspect the Onetics will be difficult to procure for most.
I hope this all helps!
Greg in Mississippi
@paoloilpizzo, I believe Ian tried the LL1544A transformers, but I have not. Hopefully he'll see this and provide his feedback. I have a pair of LL1684 transformers to try, but haven't had a chance to assemble them onto a board and try them yet. The Onetics I like clearly bested a pair of LL1590XL, but I suspect the Onetics will be difficult to procure for most.
I hope this all helps!
Greg in Mississippi
@Greg Stewart,
Hi,
I didn't find on the Net any other boards with embeded oscillator & good XO. Pulsar Clock is off and was expensive. NDK-SA is not better than CCHD-957, certainly in the same league :differences heared due to the decoupling caps & powersupply around more than the XO itself imo.
It's off topic and I appologize but has anyone removed the CCHD-957 casing to have acess to the Pierce oscillator board around its XO ? IanCanada answered me he has a much better Dil-14 adaptator board for this Crysteq board than before : 6 x 100 nF X7R + 2x 10 pF C0G decoupling areas. But what about the embeded caps of the Pierce board itself below the casing?
It's pretty sure to be X7R as well and I bet a swap with C0G might improve something ? But I assume IanCanada designed his CCHD-957 adaptator to avoid any oscillations because the distance with the FiFoPi local caps.
I don't know, maybe just a waste of time. Crystek CCHD-957 | H i F i D U I N O
Hi,
I didn't find on the Net any other boards with embeded oscillator & good XO. Pulsar Clock is off and was expensive. NDK-SA is not better than CCHD-957, certainly in the same league :differences heared due to the decoupling caps & powersupply around more than the XO itself imo.
It's off topic and I appologize but has anyone removed the CCHD-957 casing to have acess to the Pierce oscillator board around its XO ? IanCanada answered me he has a much better Dil-14 adaptator board for this Crysteq board than before : 6 x 100 nF X7R + 2x 10 pF C0G decoupling areas. But what about the embeded caps of the Pierce board itself below the casing?
It's pretty sure to be X7R as well and I bet a swap with C0G might improve something ? But I assume IanCanada designed his CCHD-957 adaptator to avoid any oscillations because the distance with the FiFoPi local caps.
I don't know, maybe just a waste of time. Crystek CCHD-957 | H i F i D U I N O
Hi Greg,
Thank you for chiming in. I will certainly buy the CCHD-957 clocks and look out for a fitting OCXO to upgrade.
Regards,
up to you... could be the photograph of the Year.
Everyone would like to know about the caps and if they can be swapped by better NPO for instance Well if the X7R we can see from Hifiduino are as low capacitance than the NPO range/casing size.
Think they did their job at Crystek, but sometimes financial winns against engineers (always in fact...)
Did you have performed scope tests as well ? I'm a little worried by the capacitance value of the NPO cap of the new CCHD-957 black adaptator (in the pico range) from IanCanada : what happen if the Crystek in board X7R (i saw ther eis an oap as well) are from a superior capacitance value ?
Does it mean whatever the FiFoPi or MC board, the capacitance of the adaptator board was measured for both boards ? For instance the Supraclock board has also its embeded oscillator and there is no need of a further board when stacked on the FiFo II/MC/Pi boards ????
Sorry, questions have no sense perhaps on a technical side... On a hifi-porn side, sure everybody would like to lurk without the top!
Everyone would like to know about the caps and if they can be swapped by better NPO for instance Well if the X7R we can see from Hifiduino are as low capacitance than the NPO range/casing size.
Think they did their job at Crystek, but sometimes financial winns against engineers (always in fact...)
Did you have performed scope tests as well ? I'm a little worried by the capacitance value of the NPO cap of the new CCHD-957 black adaptator (in the pico range) from IanCanada : what happen if the Crystek in board X7R (i saw ther eis an oap as well) are from a superior capacitance value ?
Does it mean whatever the FiFoPi or MC board, the capacitance of the adaptator board was measured for both boards ? For instance the Supraclock board has also its embeded oscillator and there is no need of a further board when stacked on the FiFo II/MC/Pi boards ????
Sorry, questions have no sense perhaps on a technical side... On a hifi-porn side, sure everybody would like to lurk without the top!
Last edited:
@diyiggy,
Cap values and types need to be more carefully engineered for RF than you might think. NPO/C0G can cause more problems than help in some cases. So can using larger value caps than needed. The problems have to do with self-resonance due to ESL and insufficient ESR to provide adequate damping. Even the smallest SMD .01uf X7R or X5R caps will self resonate at 50MHz or somewhere up there. Very little can be done to eliminate it in some cases, but choosing values and types carefully can go a long way to making sure they can do an adequate job for their intended purpose.
Cap values and types need to be more carefully engineered for RF than you might think. NPO/C0G can cause more problems than help in some cases. So can using larger value caps than needed. The problems have to do with self-resonance due to ESL and insufficient ESR to provide adequate damping. Even the smallest SMD .01uf X7R or X5R caps will self resonate at 50MHz or somewhere up there. Very little can be done to eliminate it in some cases, but choosing values and types carefully can go a long way to making sure they can do an adequate job for their intended purpose.
Thank you Markw4,
well I surmise IanCanada calculated and measured it, I was just surprised than the caps arrengement was different from the previous SMD adaptator which was made first for the CCHD-957 as well (FiFo II generation). Stacking clock board should be a bad thing and further decoupling was usefull. Anyhow with 6 x X7r + 2x NP0 will improve ESR & ESL a lot.
I mainly use near a PPS cap (1 uF as close current source for the clock with good sounding result (better to me than a same value but with tantalum or ceramic class II as X7R). But sure those 0607 size case on the IanCanada black adaptaor board should has something (very low inducrance).
Well I will leave the top where it is for the honor of the CCHD-957...
well I surmise IanCanada calculated and measured it, I was just surprised than the caps arrengement was different from the previous SMD adaptator which was made first for the CCHD-957 as well (FiFo II generation). Stacking clock board should be a bad thing and further decoupling was usefull. Anyhow with 6 x X7r + 2x NP0 will improve ESR & ESL a lot.
I mainly use near a PPS cap (1 uF as close current source for the clock with good sounding result (better to me than a same value but with tantalum or ceramic class II as X7R). But sure those 0607 size case on the IanCanada black adaptaor board should has something (very low inducrance).
Well I will leave the top where it is for the honor of the CCHD-957...
Last edited:
Here's the pic old mate.
I'm not sure if the questions were for me, but just in case; I didn't test anything just plugged and played. I think measurements would require some special equipment for phase noise below 1Hz, and I don't know how consistent the measurements would be across Crystek samples.
Pulsar was a good improvement but I don't think that means the CCHD957 is not good, IMO it is perhaps the better balance for price and performance, at least with the sample that I had, and I think that might be important - sample variation; what is their test criteria for pass/fail at Crystek QC dept?.
FWIW the clock was non-functional so no CCHD's were harmed in this experiment, the crystal might have been okay, but I wanted to see inside that too - was anti-climax of the week
Go with A.Mori clock?..
Kind regards,
Shane
Attachments
Last edited:
For me, the investment in a Pulsar clock is worth it.
There is a nice improvement compared to the CCHD-957 and NDK SDA.
There were offers at Pulsar at the end of 2018, I had taken advantage of it!
To limit costs, I only bought one Pulsar and set the sampling frequency at 96 Khz at the source (I don't listen an improvement in 192 KHz).
nounouchet
There is a nice improvement compared to the CCHD-957 and NDK SDA.
There were offers at Pulsar at the end of 2018, I had taken advantage of it!
To limit costs, I only bought one Pulsar and set the sampling frequency at 96 Khz at the source (I don't listen an improvement in 192 KHz).
nounouchet
Too bad that Pulsar is out of business. They had a nice compact and good performing clock, I was too late to order
Andrea Mori is now finshing a group buy, I ordered a bunch pcb’s and 4 oscillators to experiment with.
Regards,
thanks for the hints
@ HK,
Thanks to have removed the roof of this Crysteq fun house.
Sure it's not bad I have two and it sounds even better with a close energy decoupling cap : 1 uF Panasonic PPS or even better (subjective listening) with a Cornel Dublier acrylic cap. All the guys saying it has a bad sound have imho a bad design or do not know how to measure and power it or design things around. At least I'm not a tech and I had just luck with this decoupling cause the improvement was big... maybe just because of that I will try the SMD/CCHD-957 adaptator kit of IanCanada just cause it has one 0807 area for bigger platic smd cap I'm talking about (seems it was for a 1 uF 0807 X7R cap in the IanCanada kit). I certainly going to test the black adaptator side to side, would have liked it has 0807 decoupling area as well, but inductance seems important as said by Markw4 and small foot print might be important. Well at 0402 size I can't solder it : too small for my hands., it's good IanCanada stayed with 0605 size case. I understand people like choice but it should be better to solder the Crystek on the pcb directly with these high frequencies/speed...
@ Nounouchet,
I'm sure it worth it but it's a too big money for my pocket. I'm on the way to test what Andrea Mori have made here with Laptech sc-cut & discroll oscillator.
Sure most of us have 44.1 KHz materials, better to choose a XO multiple of that.
What do you use to over/upsample please? A Pi like renderer I assume (or a pc), but which playback software ?
@ HK,
Thanks to have removed the roof of this Crysteq fun house.
Sure it's not bad I have two and it sounds even better with a close energy decoupling cap : 1 uF Panasonic PPS or even better (subjective listening) with a Cornel Dublier acrylic cap. All the guys saying it has a bad sound have imho a bad design or do not know how to measure and power it or design things around. At least I'm not a tech and I had just luck with this decoupling cause the improvement was big... maybe just because of that I will try the SMD/CCHD-957 adaptator kit of IanCanada just cause it has one 0807 area for bigger platic smd cap I'm talking about (seems it was for a 1 uF 0807 X7R cap in the IanCanada kit). I certainly going to test the black adaptator side to side, would have liked it has 0807 decoupling area as well, but inductance seems important as said by Markw4 and small foot print might be important. Well at 0402 size I can't solder it : too small for my hands., it's good IanCanada stayed with 0605 size case. I understand people like choice but it should be better to solder the Crystek on the pcb directly with these high frequencies/speed...
@ Nounouchet,
I'm sure it worth it but it's a too big money for my pocket. I'm on the way to test what Andrea Mori have made here with Laptech sc-cut & discroll oscillator.
Sure most of us have 44.1 KHz materials, better to choose a XO multiple of that.
What do you use to over/upsample please? A Pi like renderer I assume (or a pc), but which playback software ?
Last edited:
Well the Crysteq has both Class I NPO/C0G and Class II ceramics on the pcb... better to leave as it is .
Or could be fun to remove the crystal destroyed case and put an Andrea Mori's Laptech at-cut or low frequency sc-cut (no need perhaps of an oven for this last at low frequency : 5.xx K Hz)
Superb photograph, thanks again, think it's the first time one can see it on the www.
Or could be fun to remove the crystal destroyed case and put an Andrea Mori's Laptech at-cut or low frequency sc-cut (no need perhaps of an oven for this last at low frequency : 5.xx K Hz)
Superb photograph, thanks again, think it's the first time one can see it on the www.
Last edited: