My (very) mini review of the Singxer SU-1 DDC

[elan120]

Hi @Energy,

 

12 minutes ago, Energy said:

Why didn't you use a single module to drop 7V to 5V? Is the voltage drop too great where it heats up the chip?

 

I had just one module that go from 7V to 5V since early this year, the experiments in recent months proved with two modules setup gains a little more in overall sound quality, so I incorporated here as well.

 

18 minutes ago, Energy said:

So you recommend connecting USB input 5V to one module which outputs 3.3V for the XMOS chip then remove the switch mode dc-dc voltage regulator and use the other module which outputs 1V output to where the previous inductor was?

 

Is there any need to cut off the USB input power anywhere by unsoldering anything else or do you just simply solder the LT3045's input to the USB voltage?

This seem to be the simplest way to get the mod completed for the purpose of supplying better quality DC to both XMOS and the USB input oscillator.

 

20 minutes ago, Energy said:

Kevin since i'm like you and only upsample my music to DSD512, only replacing the bottom ADP150 is needed correct?

Does this ADP150 regulator take 5V on it's voltage input or does it draw 3.3V from the nearby regulator (Belleson/Sparkos/LT3045)?

What voltage does it output for the crystal oscillator?

 

So you simply solder the VIN/GND to the LT3045 and connect it's output to where Vout would be?

Yes, that will be the only one needed, although I might consider changing out both ADP150's, but that is only because I test/compare in 48KHz families once in a while.

 

The two ADP150 draw voltage from the 5VDC input.  In my case, I supply 7V from LPS-1 externally to the two series module of LT3045 regulator boards to step it down to 5V, and this 5V input not only go to the regulator that many people have replaced (Belleson/Sparkos/LT3045), the same 5V also goes to these two ADP150 regulators that I have not yet seen mention of changing here.  These two ADP150 regulators take 5V input and step it down to 3.3V that goes to pin-4 of both Crystek CCHD-575 oscillators (in my case, the two sorted NDK oscillators now).

 

By changing from ADP150 to LP5907 earlier this evening proved to be a good upgrade.  Although the difference in spec voltage output noise some might consider to be low (9uV vs. 6.5uV), but the result is clearly noticeable.  I look forward to install the two LT3045 regulators once I receive them, and will just remove this LP5907 and solder the 3.3V output to the pin-5 (Vout) soldering pad.

 

38 minutes ago, Energy said:

Just making sure I have everything down. So it looks like I'll need three LT3045's. Any other possible upgrades? I want to change out my Crystek HD575's as well but I'll wait until you finish experimenting with the Pulsar's. Are they technically better than the NDK's?

To make these upgrades I outlined above, three LT3045's is all you need, unless you want to do the 7>6>5V input voltage "upgrade" as well.  I originally was planning to try out Pulsar oscillators, but likely will be trying out SOtM clock first instead.

 

From spec standpoint, Pulsar oscillator does have better spec than NDK, but the cost is much higher as well.  In terms of how much difference Pulsar oscillator really is compare to other oscillators in the similar price range such as Neutron Star 2 from NewClassD will be something I would love to know first hand as well.  If you are interested, I will let know more once I get the SOtM clock project going.

 

Kevin

 

[elan120]

1 hour ago, scan80269 said:

Hi @elan120,

 

When you replaced the 1.0V switching regulator and 3.3V linear regulator on the upstream (XMOS) side of the SU-1, did you measure the current consumption over USB +5V?

 

I checked the XMOS XU208-256-TQ64 datasheet, and it shows a max IDD current on the VDD (1.0V) rail being 375mA.  If you run the USB +5V bus power through a pair of LT3045 regulators to get to +1.0V, the total current consumption is likely to push closer to what a USB 2.0 bus-powered device is allowed to consume: 500mA max.

 

I suspect the SU-1 designer may be aware of this, and chose a switching regulator with much higher efficiency to generate the 1.0V VDD rail for the XMOS.  Generating 1.0V from 5V via linear regulators will consume much more total power.  There's definitely a SQ vs. power consumption tradeoff.

 

My SU-1 draws about 98mA over USB +5V when streaming DSD512.  I'm curious as to how much USB +5V current your dual-LT3045 mod ends up consuming.

 

Hi @scan80269,

 

Happy holidays.  I took a quick measurement using a portable USB meter, and it show a little over 40mA (0.19~0.2W) as the consumption, and both LT3045's are running cool to touch running DSD512.

 

[elan120]

1 hour ago, scan80269 said:

Happy holidays!

 

Hmmm, I don't understand how your current reading can be lower than my SU-1 with the stock switching regulator in the upstream section.

 

I inserted a YZXstudio USB 3.0 Power Monitor on the USB cable to read the current.

 

I used a similar tester (USB Satety Tester), but not sure how accurate it is.  By measuring stock SU-1 at idle, it is hovering around 0.02A (0.149W), while charging my phone is showing 0.88A.  I think this little gizmo will only give an idea about the consumption, but not anywhere close to being accurate.  In any case, the true consumption should be well within the maximum limit of 500mA, and I will see about getting a better tester after the holidays to see if I can get a more accurate measurement.

[elan120]

11 minutes ago, Energy said:

So you suggest the following for the XMOS.

1. one LT3045 module with 3.3V output (remove inductor and switch mode dc-dc voltage regulator)

2. one LT3045 module with 1.0V output (remove previous voltage regulator)

Very close, except the components to remove are just the opposite, where you will do the following instead.

1. one LT3045 module with 3.3V output (remove stock voltage regulator)

2. one LT3045 module with 1.0V output (remove inductor and switch mode dc-dc voltage regulator)

Please note, I connect these two LT3045 modules in series 5V>3.3V>1.0V in my setup.

 

14 minutes ago, Energy said:

 So you suggest the following for the Crystek HD-575 oscillator.

1. one LT3045 module with 3.3V output (remove ADP150)

 

Since both of the crystal oscillators are powered by separate ADP150's but are connected to the 5V, is it possible to power both oscillators with a single LT3045 or should the two be isolated? Between the 22.5792MHz and 24.576MHz, which one is used for DSD playback?  The bottom one is for DSD but which frequency is that? They are only used either or right? Not both at the same time ever? Are the oscillators used for various bitrate families? If you could elaborate that for me that would be appreciative. 

 

I don't think I will do the 7>6>5V input voltage upgrade unless I'm given more information on it. Having two extra modules in line after the LPS-1 only adds more to total noise would it not?

 

It's good that the LP5907 offers mild improvements over the ADP150 so I am excited to go the LT3045 for even further benefits. At first I was concerned about the overall power consumption but since measurements doesn't seem to hit that of 500mA for USB 2.0, I will go ahead with the upgrades. I wanted add the oscillator upgrade into the mix and replace the Crystek CCHD-575 oscillators however the Pulsar's are insanely expensive. I didn't realize until now that they were over $400 EU.

From the power consumption standpoint, it is well within the spec of either single LT3045 board (500mA) or dual LT3045 board (1A) to use just one to supply power to both oscillators in parallel, but since I plan to stack them when I received mine and doing the mod, I will use two separately to avoid any possible noise cross talk.

 

DSD playback only use 22.5792MHz oscillator (ie. DSD512 =>22579200Hz/44100Hz=512), and only one oscillator is being used at a time.  The bottom oscillator from the pictures attached in my previous post is the 22.5792MHz oscillator, and in addition to all DSD playback, this same oscillator also being used for all 44.1KHz music file playback (ie. 8X oversampling of 44.1K => 44.1Kx8=352.8K), and the other oscillator (24.576MHz) is only used for 48KHz family music files (ie. 8X oversampling of 48K => 44Kx8=384K).

 

The improvement for 7>6>5 setup is, although subtle, but is clearly there in my system, and can be a later project to consider.

41 minutes ago, Energy said:

How much are the improvement specification of the NDK's over the stock CCHD-575's? Are there really none in the same package size and a carrier board is needed? How much are they for each including carrier board? I listen to DSD up-sampled from both 44.1KHz and 48KHz so that means i'll need to change out both oscillators 22.5792 MHz for 44.1 and 24.576MHz for 48 right? Unless i'm getting it wrong.

 

Can you point me into the right direction on where to get the NDK's and carrier board if the same package is not available?

 

Is there any other oscillators that are better than the NDK and only marginally more expensive? I do not want to spend over $50.

The standard NDK oscillators having similar spec like CCHD-575, but the sorted NDK can offer around -10dBC/Hz at 10Hz improvement over Crystek.  Unfortunately, there is only one physical size available, that is the reason why I had to make those carrier boards to retrofit them to where the Crystek's located.

 

The only other oscillator I am aware of that can be used as replacement that also fits the budget range you specified is NDK NZ2520SDA, but the spec isn't necessary better than the sorted NDK NZ2520SD I used, and it can be challenging to acquire them since NDK sells in bulk quantity only and no other online seller I know carry the two frequency oscillators we need here.

 

If you plan to upgrade to these NDK NZ2520SD oscillators, the seller I used to purchase mine is here:  http://www.ebay.com/usr/analog.research

 

There are different options to get the carrier boards put together, it does involve some work, but doable especially I know you are very good with soldering work.

 

Hope this helps.  Let me know if you have additional questions.

[elan120]

46 minutes ago, Energy said:

Wouldn't having both modules connected to 5V be better by lowering the noise of an additional regulator in series or does the series offer better input rejection from the incoming 5V?

When connecting in series, I am not certain whether it is the better PSRR or more filtering that leads to the quieter background, I like the series connection setup for another reason, it will keep the drop out voltage lower for the 1V module, keeping it cooler.

 

1 hour ago, Energy said:

Thanks for explaining the oscillator layout. I will go with NDK NZ2520SD's. What did you use to retrofit it? I was looking using a simple DIP8 PCB and soldering a wire of the oscillator corners to each of the corners on the DIP8 socket and connecting those into the soldering pad.

 

 

 This simple DIP8 PCB could work if the physical size is correct, as Crystek CCHD-575 has physical size of 7.5mmx5.08mm, which the physical carrier board should be a little smaller to expose more of the soldering pads.

 

When I retrofit mine, I make two carrier boards to make sure it has correct physical size, as well as the PCB trace location to match the soldering pads from the removed Crystek oscillators.  Below are some pictures of the retrofit work I did, hopefully be some help visualizing what needs to be done.

 

Carrier board with NDK NV2520SD oscillator mounted.

 

Stock Crystek oscillators removed.

 

NDK NV2520SD 22.5792MHz oscillator installed.

 

Both NDK NV2520SD oscillators installed.

 

 

1 hour ago, Energy said:

Your method seems better though as it's more form fitting. Teach me senpai. Also I wanted to add, did you hear more improvements from the oscillator change or the linear voltage regulator one?

 

I got the oscillator mod done few month ago, and like the result quite well.  Other regulator mods like 3.3V and 1.0V on USB input side, as well as the regulator for 22.5792MHz oscillator was just completed few days ago; in between, I have done other improvements to my system, so it is difficult to say which mod done more to the overall improvement, and I can only comfortably say that all of them more of less contribute to the sound quality I got now...not what you were asking, but it is somewhat difficult to identify over this long period.

[elan120]

6 hours ago, Energy said:

I've realized the DIP8's are a little bigger and according to your photos it will be better to have something slightly smaller enough to reveal the soldering pads. How did you make your boards? Did you just cut it out from an existing PCB? Somehow the pads or copper plate beneath the oscillator fits well for it's four corners and leads well to the existing soldering pads. Did you cut out four metal pieces in the same shape? What metal piece did you use to form those eyelet looking corners?

 

I might just cut out a small PCB and solder some pure silver solid-core wires to it but the copper plate idea you used seems like it'll be better for impedance. I'm just confused how you got it stuck to the PCB fiberglass so well... 

In my case, it was a little bit more straight forward, I drew the PCB trace pattern and cut it with a CNC machine using a blank PCB, inserted those eyelets to four corners, and bent the legs out to create soldering points.  Once oscillators are mounted, I solder the board to the exposed soldering pads from removed stock Crystek oscillators.  This retrofit can also work without using carrier board, but will be more challenging controlling the heat while soldering.  By using carrier boards will minimize the overheating possibility, and also making the retrofit easier.  If you are having difficulty getting the carrier board put together, send me a PM, I will see if I can help cutting couple more.  Below are some pictures I took while making these carrier boards, again, hopefully be helpful visualizing the board making steps.

 

Cutting carrier board.

 

Eyelets inserted, but not soldered yet.

 

Eyelets soldered.

 

 

[elan120]

2 hours ago, Energy said:

Thank you for going beyond in terms of explanations! You're a lifesaver! 

 

That would be too much work for you. I can't ask that much from someone, makes me feel bad. The explanations has already been loads of help. Besides, I believe you live outside of the United States so shipping would take awhile. Someone should make a project later on to make these oscillator adapter PCB's to sell hehe. Wouldn't be a bad idea.

 

Kevin. I've noticed that you used the LT3045-A's for the USB input and the LT3045-S's at the 5V DC input, why is that? Shouldn't that be opposite? I know the unit doesn't take anywhere over 500mA either way but that seems to be restrictive on the 1A max output of the LPS-1 by using a 500mA max module. I went with the 500mA's for the USB input side as well as for the ADP150's. Should be enough right?

You are very welcome, again, if you are set to do the NDK oscillator mod, and needs some help with the carrier board, feel free to send me a PM, and I can see about getting couple more cut as we might only be an hour apart if I am not mistaken.  Making carrier board as a business, based on my experience, likely be a proposition most will decline, as economy of scale will be challenging to justify, but cutting few for fun is another story.

 

I had the 7>6>5V for 5V input setup since several month ago, and that was before they start making the 1A version, once the 1A board is available, since it has slightly lower output noise as well as output impedance spec, I went with them for the USB side voltage regulator mod, and now waiting for the same boards to replace ADP150's.  I don't think there will be noticeable difference between these two different version boards, but as a standard practice of mine, I went with the better spec parts.

 

Using LT3045-S board for USB side regulator and ADP150 regulator mod will certainly be fine, and it won't be restrictive for LPS-1 since the current consumption for the down stream devices are well within the maximum 500mA current output from these boards.  As a reference, typical current consumption @ 25 degree C for NDK NV2520SD is 1.8 to 8.5mA, where typical input current for Crystek CCHD-575 is 15mA, and maximum at 25mA.

[elan120]

44 minutes ago, Energy said:

 

Haha I meant to have one made on a circuit board and sold from somewhere that isn't hefty.

But funny story. One of my buddies (@camel) found the NDK to Crytek adapter designed by Analog Research (whom you buy the NDK's from). It is open sourced from OSHPARK and I ended up buying 9 pieces for only $3. 

 

Size: 0.37 x 0.57 inch (9.4 x 14.5 mm) 2 layer board.

 

Front:

Back:

 

Source: http://analogresearch.co/ndk.html

Purchase Link: https://oshpark.com/shared_projects/5bdTrMZl

 

No wonder you didn't go for the 1A version then. I did notice what you said regarding the improved specifications and the "<0.8µVrms" within the listing also gave it away. Since it's just a couple bucks more I'll  go for the LT3045-A as well. Thanks!

 

It's good to hear that the smaller packaged NDK's offer more efficiency. Now with the easy to access adapters available I will use them to change any existing Crystek CCHD-575 & CCHD957's within my setup. I will probably use the extra adapter to see if I can change the crystal oscillator within my UltraRendu to NDK. 

Great to know Pat from Analog research finally made carrier (adapter) boards available.  I remember the conversation with him about making it, but was not aware this has already become available for purchase.  One thing you might want to discuss with either Pat or OSH is physical size for the two Crystek CCHD-575 on SU-1 is 7.5mm X 5.08mm, where the link you provided for purchase is for Crystek CCHD-975 that has a different physical size (9.14mm X 14.2mm).  Pat can supply different drawing file to OSH to have the correct size adapter made (unless one is already available).

 

To modify UltraRendu, you will need likely a 24.000MHz oscillator (for USB input).  I have not yet be able to source any in reasonably small quantity, so if you can locate one, please do let me know, as I will be very interested in getting them as well.

[elan120]

13 hours ago, Energy said:

NZ2520SD's are much different than NX2520SD's huh. Found a bunch of those on Digi-key. I will check Dove Electronics or NDK directly to see if they can send me two samples, one for each of us.  If you can find it before me, please also let me know.

If I recall, NZ2520SD is a oscillator (combined crystal unit and oscillation circuit), NX2520SD is a crystal unit, and they are not interchangeable.

[elan120]

14 hours ago, Energy said:

 

Thanks Kevin. 

 

Question. If i use LT3045-A to where Sparkos/Belleson would be, should I still remove the three MLCC ceramic's or leave them there? Also are there anything better than NZ2520SD's (other than Pulsar) that doesn't cost an arm or a leg?

I have removed all four on my SU-1 since they are not needed.

 

The only other oscillator I know that is comparable in both cost and performance to sorted NZ2520SD would be NZ2520SDA, but it is somewhat challenging to source in small quantity purchase.

[elan120]

4 minutes ago, bratsson said:

Can I use this one to connect lps-1 or sbooster to my SU-1? 

https://www.ebay.com/itm/Silver-5-5mm-x-2-1mm-DC-Power-Socket-Female-Panel-Mount-Connector-Plug-Copper/282379016199

 

Or should I buy a a more expensive one?

This is the correct size to use with 5.5mmx2.1mm plug.  Hopefully they will have a good snog fit with the plug you intend to use.

[elan120]

I have some time today to work on changing out the voltage regulator for the 22.5792MHz oscillator, and is happy to report that the result from this mod is very much worth the effort, and certainly would recommend anyone comfortable with soldering to give it a try.

 

The first mod for this regulator was done late December last year.  It was a direct pin out regulator change from ADP150 to LP5907, and the result was quite good, which gave me enough confidence to order couple of the 1A LT3045 to replace the LP5907, thinking if the spec noise dropped from 9uV to 6.5uV can be easily detected in my system, dropping the spec noise to 0.4uV should improve even further.  After completing the mod earlier, letting system warm up for couple of hours, proved this idea is very much true.  A better, cleaner power feeding the oscillator will indeed result in better SQ.

 

Below is a picture of the mod, hopefully it is clear enough to be self-explanatory.

[elan120]

1 hour ago, jjh1585 said:

 

@elan120

I am trying to figure out how to connect a Pulsar VR as you have done with the LT3045.

Looking at the ADP150 data sheet 1=VIN 2=GND 5=OUT.

I was planing to connect my wires to the board where the ADP150 was removed.

I see you have done it a different way. Can you explain your connections for the LT3045?  Thank You.

@jjh1585, 

 

Are you using the Pulsar VR from:  http://www.pulsarclock.com/Power.html  ?

 

If yes, this may not be a good choice to replace ADP150 unless you plan on modifying the VR circuit board.  You need pin-3 connected as shown in the mod picture I attached from my earlier post, but Pulsar VR have ADM7150 EN pin (pin-7) tight together with Vin (pin-8).  If you still want to use Pulsar VR, you will need to cut open the trace and connect ADM7150 pin-7 to ADP150 pin-3 location directly.  This is part of the reason why I used LT3045 VR on my mod.

 

Meanwhile, I by-passed all components by connecting Vin and GND directly to the input connector instead of connecting to ADP150 pin-1.

 

Hope this helps.

[elan120]

1 hour ago, jjh1585 said:

@elan120

 

Thank you for the clear explanation.

 

I am gathering some parts presently. I placed an order for Pulsar Clocks and bought the Pulsar VR's. I have also ordered LT3045-A boards as well. I thought I would test both. 

 

Looking at your photo I see a wire at 3 EN ( ADP150 ). Where is the other end connected?

 

What do you think about adding a 2nd LPS-1 @5v to power the XMOS Mod (3.3v and 1v LT3045-A boards).

 

Thanks.

@jjh1585

 

Good that you also got some LT3045 coming as well, they will make the ADP150 replacement mod much simpler.

 

The other end of the EN wire is connected to the LT3045 EN pin.  You should be able to see it from the bottom center portion of the picture I attached from my earlier post, where both Vin and GND pin were left open, and the wire you see soldered to ADP150 pin-3 location is soldered to the LT3045 EN pin from the bottom.

 

I do think running another LPS-1 to power USB input side will contribute further benefits to the overall SQ, and this is the same power supply I am using to feed my USB card presently, which in term feeding the 3.3V and 1V LT3045 boards I used for USB input side voltage regulator modification I did last December.  I am sure you saw the new LPS-1.2 announcement by @Superdad recently, if you don't have a LPS-1 already,  this new one will be a great supply to have for this application.

[elan120]

3 hours ago, drmike said:

to elan 120,

is the LT 3045 A  a special order or will the 1.2A work? do you need 2? if so where does the second one go?

thanks,

drmike

@drmike

 

I am not clear on some of your questions, so I can only answer part of the questions.

 

The LT3045 board is not a special order, and you can check the website where I purchased mine here:  http://www.ldovr.com/category-s/118.htm

 

As you can see from the website link above, they have either 0.5A or 1.0A version available; unfortunately, I am not familiar with the 1.2A version you mentioned.

 

As for the "do you need 2" question, can you clarify what this means?  Once this is cleared up, I will likely be able to help with the "if so where does the second one go? question.

[elan120]

@jjh1585

 

Glad that I can be a help.

 

It would be great to hear your impressions once you have oscillator VR mod and the oscillator upgrade completed.  I am working on plans to use SOtM clock for USB input 24.000MHz and 22.5792/24.576MHz sampling oscillators.  Hopefully this will be a successful venture and will help gain further improvements.

[elan120]

6 hours ago, Energy said:

 

You should also connect a bypass/decoupling capacitor with low ESR to pin 2 and 4 of your NDK NZ2520SD. The value should be 0.01uF to work as a power filter for the Ground and Power Line so that the noise from the oscillator does not leak into the clock circuit and the power noise from the circuit does no leak into the oscillator. Better isolation is what I'm saying. The leads should be cut as short as possible prior to soldering in order to minimize parasitic inductance from being formed which would equivalate to an increase in impedance for higher frequencies. It is written in the NDK datasheet as well as recommended by Analog Research. No changes are needed for the build out resistors as they are already in place on the SU-1 circuit board.

 

I recommend WIMA FKP2 for better quality since it's film/foil rather than metalized film however I'm sure even the metalized MKP2 should be fine for this purpose.

 

Circuit Design for Crystal Oscillator (and optimizations)

NDK NZ2520SD DataSheet

@Energy,

 

I also agree having the by-pass capacitor is a must, but I a bit not sure the reason why you want to add another by-pass capacitor to the existing by-pass capacitors.  Is it because you find them not quite up to task or WIMA FKP2 will yield better result?

 

I didn't do anything more than replacing both Crystek oscillators with NDK NZ2520SD thinking the stock by-pass capacitors used for both Crystek CCHD-575 oscillators should be good enough, but if you think WIMA will work even better, I will get some and give them a try.

 

Below is a picture I outlined the stock by-pass capacitors for your reference.

[elan120]

3 hours ago, jjh1585 said:

@elan120

 

I received the Pulsar clocks. They are larger than the 575's.

 

What do you think would be the best way to mount them?

 

Thanks.

 

 

@jjh1585,

 

When I was first looking at them last year, I was a bit concerned about the physical size being so much larger than the existing Crystek CCHD-575's, but thought if I was to use them, I will likely use a carrier board to solder the DIL14 socket came with the Pulsar oscillators to the carrier board, then solder the carrier board to the existing soldering pads after removing the Crystek oscillator.  The height of the carrier board will need to be raised about 2mm to clear the components carrier board+Pulsar oscillator board will covered up.  This method will keep the shortest signal path between Pulsar oscillator and the main board, yet having enough mechanical strength to support the Pulsar oscillator.

 

Other method will have longer path by mounting the Pulsar oscillator on the side and solder wires to the existing soldering pads.

 

I drew a draft carrier board (below) as a visual reference of the carrier board idea.  This is a 17mmx10mm size board, with 4 small holes 15.24mmx7.62mm apart that will be used to solder the DIL14 socket, the other 4 holes will be used to solder either a via or wire to the board, and then solder the carrier board to the existing soldering pads that are the 4 square blocks in the center.  The 4 PCB traces will be 7mmx2.2mm, and a 12mmx5.6mm hole in the center is used to solder the carrier board to the 4 soldering pads.  Once the carrier board is soldered to the main board.

 

 

[elan120]

39 minutes ago, jjh1585 said:

@elan120

 

I like your carrier board. It could make many possibilities for clock upgrades in various devices.

 

Is the a company that would make some?

 

Thanks.

I normally will make the carrier board drawing and cut it myself, so I am not too up to date on who the possible supplier might be, but this place might be a possible source you can contact:  https://oshpark.com/

 

[elan120]

18 hours ago, Energy said:

 

@elan120 Do you by any chance have a drawing so I could have a PCB made on OSHPARK to use as a carrier for the NDK NZ2520SD's? Having read more into decoupling of oscillators I can understand now why you chose to angle your traces. 

Hi @Energy,

 

I will take a look at the drawing requirements.

 

Ran into a bit of issue with my computer yesterday that has the files, and is now in recovery mode.  I should find out more later today to know where it stands.  In the meantime, I will take a look at re-do the drawing using the CAD programs OSHPARK requires and will let you know more later.

[elan120]

On 1/29/2018 at 12:27 PM, Energy said:

Thank you for looking into it. I believe according to the majority of optimizations for oscillator fixtures is to have a ground plane directly beneath die and eliminate the use of vias (changes impedance and causes reflections) and 90 degree angles.

 

I've been experimenting with trying to make my own carrier board using CAD software but it's a little over my level of expertise.

Hello @Energy,

 

I spend some time the last few days to play with KiCAD, which is one of the PCB layout softwares needed by OSHPARK, but the learning curve is a bit more than I had anticipated.  I am able to get the soldering pads done on both top and bottom side of the carrier board, but to lay down tracks linking top side soldering pads to each corners became an obstacle I have yet to overcome.  I didn't want to hold you up too long, so I thought I will give you this much update.  I will play with this some more, but not sure if I will get past this issue soon.

 

The carrier board I use was done using CAD program, convert to .DXF file, and then convert to g-code to have them cut on CNC machine, and that was much simpler than learning KiCAD.

 

Below are both top and bottom 3D view of the board I got so far to show what I was referring to.

[elan120]

56 minutes ago, R1200CL said:

So my question is, has anyone done a combination of the LT3045 and LPS-1 on the SU-1 ?

 

I would expect it should be possible to add these two boards inside the SU-1. Anyone did ? Pictures?

 

I was planning to go for the 1A boards, but do I really need them ? Will not the 500 mA  enough to feed the SU-1 ?

 

 

 

 

Short answer is "Yes".  I have done this few month ago.

 

Either 0.5A or 1A version will work, but the 1A version will have lower noise level.

 

[elan120]

1 hour ago, manishex said:

Upgrade the 5 main board caps to the Nichicon FPCAP (R7) 220μF 6.3V

Upgrade the USB VR to a LP5907MFX-3.3/NOPB

 

Any other advice on offer? I will also need to make a cable to connect LPS-1 via DC.

Other caps you can consider is OS-CON 16SEPC220MD.  I am currently using 4 of them on mine.

 

Instead of using LP5907 for the 3.3V voltage regulator, consider going with LT3045 since you are going to do the work, might as well go for better spec'ed VR.  I tried both and liked the LT3045 a little better.  In addition, consider change out the 1.0V switch mode voltage regulator with LT3045, you can refer to the pictures I attached in previous posts to see further details.

 

Finally, consider change out the two 3.3V voltage regulators feeding the two Crystek oscillators with LT3045's.  It will help lower the noise going into the oscillators to the minimum.  The mods you plan on doing with power supply section and the VR for the main board 5V input mostly is to help lower the noise feeding the oscillators, and these are the two final VR that can be improved prior to powering the oscillators.  More detailed mod and pictures is also in my previous posts.

 

Hope this helps.

[elan120]

23 minutes ago, manishex said:

I read that if I power the su-1 and iso regen both with a single lps-1 I will lose galvanic isolation.

Also the iso regen needs 7V, so by making the lps-1 output 7V, can I use a  LT3045 for the su-1 to step it down to 5v? Should 1A be enough? 

If you keep ISO Regen in ISO mode, you should still have galvanic isolation.

 

LT3045 1A board will be more than enough to step 7V down to 5V, since SU-1 main board will only consume a little over 110mA.

[elan120]

3 hours ago, RickyV said:

 It is NOT these caps!!!!

This is in case of an xo and the 575 is an oscillator!!!!

I do not know why elan120 removed those caps. 

Well...somehow my name is mentioned here...for the record, the picture I posted was to point out where the bypass caps are located on SU-1 when the discussion was on replacing the stock Crystek oscillators with NDK, and I never did remove them.  Now that I have moved on replacing NDK oscillators with SOtM sCLK-EX clock board, these caps are still there.