Hello,
I'm returning to one of my old project. A CK2III amp.
I had interesting experience with it. On my bench as it was built, I was quite happy with it, but when I boxed it, I did not like it any more. I did not understand, but after a while I realized, the difference was the power supply. In both case I've used the on board 7815/7915 pait, but, in the case I've run it from a small transformer, on th bench, I did run it from a huge old HP rack psu (HP 6627A).
Than, for weeks, months I was testing different PSUs, but, I just could not get close to the HP psu. I did tried a large toroid transformer - no difference. LM317/337 pair - improvement, but not close to the HP. Than, it just landed in shelf for years. Recently, based on suggestions, I built some "superreg" voltage regulators for DAC, tried and it is not bad.
I also saw recently that there are small surface mount PCBs for using instead of 78xx regulators. That would fit the CK2III PCBA nice. Do you have experience with those? Do you know where to get from?
Do you have any other suggestion?
Thanks,
JG
I'm returning to one of my old project. A CK2III amp.
I had interesting experience with it. On my bench as it was built, I was quite happy with it, but when I boxed it, I did not like it any more. I did not understand, but after a while I realized, the difference was the power supply. In both case I've used the on board 7815/7915 pait, but, in the case I've run it from a small transformer, on th bench, I did run it from a huge old HP rack psu (HP 6627A).
Than, for weeks, months I was testing different PSUs, but, I just could not get close to the HP psu. I did tried a large toroid transformer - no difference. LM317/337 pair - improvement, but not close to the HP. Than, it just landed in shelf for years. Recently, based on suggestions, I built some "superreg" voltage regulators for DAC, tried and it is not bad.
I also saw recently that there are small surface mount PCBs for using instead of 78xx regulators. That would fit the CK2III PCBA nice. Do you have experience with those? Do you know where to get from?
Do you have any other suggestion?
Thanks,
JG
Attachments
LDOVR is one of the first to provide a 78xx/79xx compatible package using the LT3045/3094.
All credit to them for that.
But in the bill of material listed in :
LT3045-78XX Ultra Low Noise LDO Voltage Regulator
LT3045 x1
T495C226K025ATE300 x 1
GRM31CR71E106KA12L x2
GRM31MR71E475KA01L x1
CRCW0603xxxxFKEA x1
1)
I personally have had bad experience with tantalum electrolytics.
They fail by short circuit. And I had such disaster before.
There are recommendations that one need to specify 3x more voltage rating.
But I rather avoid them.
2)
The GRM series of MLCC from Murata has the same problem as most other ceramic caps.
They lose 75% of their capacitance when e.g. a 25V cap sees 15V DC.
See the Murata simulator :
ds.murata.co.jp/simsurfing/mlcc.html?partnumbers=%5B"GRM31CR71E106KA12"%5D&oripartnumbers=%5B"GRM31CR71E106KA12L"%5D&rgear=suaykx&rgearinfo=com
This is why the datasheet recommends using the GJ8 series.
See Fig. 6 of :
https://www.analog.com/media/en/technical-documentation/data-sheets/LT3045.pdf
Unfortunately Murata is stopping production of the GJ8 series.
So everything is on NRND.
If you want the best results from the LT's, especially for analogue, then you have to go PPS and PMLCAP :
https://www.diyaudio.com/forums/dig...1862-tht-i2s-input-nos-2r-32.html#post6557584
Comparing "Good" SMT Capacitors - C0G vs. PMMA (Acrylic) vs. PPS
The difference is audible.
3)
Thick film 0603 has both higher noise and distortion than a thin film 0805 from e.g. Susumu.
The low noise LDO does not save the day if the voltage setting resistor is noisy.
Thin and thick film | Resistor materials | Resistor Guide
Just using a LT3045/3094 is no guarantee that you get the performance as listed in the datasheet.
Everythig else around it also counts.
And of course we make our own.
Cheers,
Patrick
All credit to them for that.
But in the bill of material listed in :
LT3045-78XX Ultra Low Noise LDO Voltage Regulator
LT3045 x1
T495C226K025ATE300 x 1
GRM31CR71E106KA12L x2
GRM31MR71E475KA01L x1
CRCW0603xxxxFKEA x1
1)
I personally have had bad experience with tantalum electrolytics.
They fail by short circuit. And I had such disaster before.
There are recommendations that one need to specify 3x more voltage rating.
But I rather avoid them.
2)
The GRM series of MLCC from Murata has the same problem as most other ceramic caps.
They lose 75% of their capacitance when e.g. a 25V cap sees 15V DC.
See the Murata simulator :
ds.murata.co.jp/simsurfing/mlcc.html?partnumbers=%5B"GRM31CR71E106KA12"%5D&oripartnumbers=%5B"GRM31CR71E106KA12L"%5D&rgear=suaykx&rgearinfo=com
This is why the datasheet recommends using the GJ8 series.
See Fig. 6 of :
https://www.analog.com/media/en/technical-documentation/data-sheets/LT3045.pdf
Unfortunately Murata is stopping production of the GJ8 series.
So everything is on NRND.
If you want the best results from the LT's, especially for analogue, then you have to go PPS and PMLCAP :
https://www.diyaudio.com/forums/dig...1862-tht-i2s-input-nos-2r-32.html#post6557584
Comparing "Good" SMT Capacitors - C0G vs. PMMA (Acrylic) vs. PPS
The difference is audible.
3)
Thick film 0603 has both higher noise and distortion than a thin film 0805 from e.g. Susumu.
The low noise LDO does not save the day if the voltage setting resistor is noisy.
Thin and thick film | Resistor materials | Resistor Guide
Just using a LT3045/3094 is no guarantee that you get the performance as listed in the datasheet.
Everythig else around it also counts.
And of course we make our own.
Cheers,
Patrick
PS
The basic circuit of the CK2III is a good one.
But there are better active devices than those specified.
If you can find 2SA970/2SC2240BL, I certainly prefer them to BC550/560.
All 4 are not in active production anyhow.
The Toshibas are not pin compatible, but for TO92 you can easily get round that.
And both the BD139/40 and MJE243/253 are not particularly linear in hfe vs Ic.
I prefer TTA004/TTC004 instead.
Just my own preference,
Patrick
The basic circuit of the CK2III is a good one.
But there are better active devices than those specified.
If you can find 2SA970/2SC2240BL, I certainly prefer them to BC550/560.
All 4 are not in active production anyhow.
The Toshibas are not pin compatible, but for TO92 you can easily get round that.
And both the BD139/40 and MJE243/253 are not particularly linear in hfe vs Ic.
I prefer TTA004/TTC004 instead.
Just my own preference,
Patrick
And regarding your power supply experience :
The HP6627A is a very good supply.
Better than any transformer + rectifier on its own.
So it is not a fair comparison.
To make optimum use of any ultra-low-noise LDO's, you need to use a pre-reg to get rid of most of the ripples first.
For +/-15V after the LT3045/3094, you need a prereg of +/-17~18V.
Either 7818/7918 or LM317/337 set to 17Vout is OK.
The LT3045/3095 comes after those.
And you can play with an additional inductor between prereg out and 3045/3094 in.
But we also have interesting experience with dry batteries.
This is an easy test, just use 20x AA cells.
Will last you for a few hours, enough to tell if you like it or not.
A Battery-Powered Class A Headphone Amp | audioXpress
The Borbely EB602-200 Revisited
Patrick
The HP6627A is a very good supply.
Better than any transformer + rectifier on its own.
So it is not a fair comparison.
To make optimum use of any ultra-low-noise LDO's, you need to use a pre-reg to get rid of most of the ripples first.
For +/-15V after the LT3045/3094, you need a prereg of +/-17~18V.
Either 7818/7918 or LM317/337 set to 17Vout is OK.
The LT3045/3095 comes after those.
And you can play with an additional inductor between prereg out and 3045/3094 in.
But we also have interesting experience with dry batteries.
This is an easy test, just use 20x AA cells.
Will last you for a few hours, enough to tell if you like it or not.
A Battery-Powered Class A Headphone Amp | audioXpress
The Borbely EB602-200 Revisited
Patrick
Just take a look at LT3045/94 evaluation boards and notice how they route input ground and power to not have them couple any rubbish into the output. It is pretty much impossible to do this on a daughterboard this small with 3 pins. I tried it a while ago and made little PCBs, also for TPS7A47/33. But considering how expensive these ICs are, I don't really use them on these daughterboards. They deserve a proper implementation. If you want to make such PCBs yourself and you only use 2 layers, I can recommend Oshpark for that. You can these PCBs in low quantities for less than a dollar a piece (in multiples of 3), because Oshpark only charges by board size. And you can get 0,8mm thickness and 2oz copper at no extra cost, which is a no-brainer for compact power applications.
Edit: Now that you bring this to my mind, I should probably dig up my design files and have a look at making a revision 2, because revision 1 was probably suboptimal even considering the restrictions.
Edit: Now that you bring this to my mind, I should probably dig up my design files and have a look at making a revision 2, because revision 1 was probably suboptimal even considering the restrictions.
Last edited:
If I remember correctly most of the details about layout are not in the datasheets but in the evaluation board documentation. Not all of it is equally relevant depending on the actual implementation, but since I use these ICs also for post regulation of switching regulators I decided to not just discard the recommendations made there.
What width is your PCB? I restricted myself to the width of the TO-220 package for maximum compatibility with heatsinks, but even a few mm extra width gives more freedom in layout. Looks like you chose to do that? I would probably do the same for my revision 2, if there is going to be one that is
What width is your PCB? I restricted myself to the width of the TO-220 package for maximum compatibility with heatsinks, but even a few mm extra width gives more freedom in layout. Looks like you chose to do that? I would probably do the same for my revision 2, if there is going to be one that is
> If I remember correctly most of the details about layout are not in the datasheets
LT3045 Datasheet Fig. 8.
> What width is your PCB?
The pins are 2.5mm pitch. So you can scale.
These ones are for an application with low dissipation.
I have TO220 compatible layouts which are 10.5mm wide.
Will fit any TO220 heatsink.
In any case I never use LT3045/3094 without pre-reg.
IMHO the only way to get the most out of them.
And then the dissipation is mostly not an issue.
Cheers,
Patrick
LT3045 Datasheet Fig. 8.
> What width is your PCB?
The pins are 2.5mm pitch. So you can scale.
These ones are for an application with low dissipation.
I have TO220 compatible layouts which are 10.5mm wide.
Will fit any TO220 heatsink.
In any case I never use LT3045/3094 without pre-reg.
IMHO the only way to get the most out of them.
And then the dissipation is mostly not an issue.
Cheers,
Patrick
One last suggestion :
Replace all the 1N4004s with SBYV26C.
See :
I heard about the Linear Audio article on Soft Recovery diodes, and ...
And I would replace all ceramic caps with either WIMA MKS, MKP, FKP or Silver Mica.
Film caps preferred.
Patrick
Last edited:
About the component selection and layout for my PCBs (ldovr)
Max capacity for GJ8 X7R/25V is 4.7UF only, GRM X7R 10uf/25V (same 1206 case)
Max input for LT3045 is 20VDC only, max output 15VDC you can compare final capacitance for GRM 10uF vs CJ8 4.7uF @15V or @20V
I have no issues with the tantalum caps I use (but again it is rated for 125C)
Temp is the most important factor IMO, a small board can not dissipate heat efficiently
Max capacity for GJ8 X7R/25V is 4.7UF only, GRM X7R 10uf/25V (same 1206 case)
Max input for LT3045 is 20VDC only, max output 15VDC you can compare final capacitance for GRM 10uF vs CJ8 4.7uF @15V or @20V
I have no issues with the tantalum caps I use (but again it is rated for 125C)
Temp is the most important factor IMO, a small board can not dissipate heat efficiently
I prefer GJ8 10µ 25V X5R 1206 to GRM X7R.
As can be seen from the Murata simulator :
SimSurfing
The GJ8 X5R retains 70% of its capacitance at 15V, compared to 25% of the GRM.
And of course one can use GJ832ER7YA106KA12L if X7R is a must.
I use PMLCAP for analogue anyhow.
But each one can make his own choice of components.
I made mine based on technical information from the manufacturers and own experience of component failure.
I have no problems dissipating heat from a 10.5x30mm board with full thermal contact at the back.
Cheers,
Patrick
As can be seen from the Murata simulator :
SimSurfing
The GJ8 X5R retains 70% of its capacitance at 15V, compared to 25% of the GRM.
And of course one can use GJ832ER7YA106KA12L if X7R is a must.
I use PMLCAP for analogue anyhow.
But each one can make his own choice of components.
I made mine based on technical information from the manufacturers and own experience of component failure.
I have no problems dissipating heat from a 10.5x30mm board with full thermal contact at the back.
Cheers,
Patrick
If it appears twice, I'm sorry. Somehow it did not show up when I sent it first.
So, I did ordered a pair of the TPS7xxx modules, will try, but I will replace the tantalum caps, because I have bad experience with them in audio circuits. A bit of ringing in highs. Similar with large ceramic caps, what are designed for switch mode power supplies.
I did not hear about the LT3045/3094, but I will see.
Anyway, do you guys experience that these ultra low noise regulators are better? I mean, it is not the noise what is the problem. When I run it from the HP, there is life in it, with the 7815/7915 it is perfect, but dead.
Thanks a lot!
JG
So, I did ordered a pair of the TPS7xxx modules, will try, but I will replace the tantalum caps, because I have bad experience with them in audio circuits. A bit of ringing in highs. Similar with large ceramic caps, what are designed for switch mode power supplies.
I did not hear about the LT3045/3094, but I will see.
Anyway, do you guys experience that these ultra low noise regulators are better? I mean, it is not the noise what is the problem. When I run it from the HP, there is life in it, with the 7815/7915 it is perfect, but dead.
Thanks a lot!
JG
The CK2III is not particularly bad with PSRR.
So an ultra-low-noise regulator may not be most effective.
But no one can tell you about subjective impressions.
It also differs with the amplifier circuit.
The Borbely, for example, has quite a bit of NFB, yet sounds best with batteries.
The Pioneer SL, on the other hand, is zero global feedback, but can be happy with a Didden Super Reg.
You will just have to try different ones and decide for yourself what suits you best.
If I were you, I would start with batteries, and then maybe a simple cap multiplier.
And very wise, IMHO, to avoid Tantalum and ceramics for analogue.
Patrick
So an ultra-low-noise regulator may not be most effective.
But no one can tell you about subjective impressions.
It also differs with the amplifier circuit.
The Borbely, for example, has quite a bit of NFB, yet sounds best with batteries.
The Pioneer SL, on the other hand, is zero global feedback, but can be happy with a Didden Super Reg.
You will just have to try different ones and decide for yourself what suits you best.
If I were you, I would start with batteries, and then maybe a simple cap multiplier.
And very wise, IMHO, to avoid Tantalum and ceramics for analogue.
Patrick
I still prefer X7R and it is based on my experience, my board designed as a direct replacement of LM78xx/TO220, in most cases, you can not adjust the input voltage (10V+ drop is not very rare) and very often boards operate in a very hot environment (closed chassis, high dissipated power), and 85C limit is a weak point. About capacitance, 90% of my sales for 1.2v/3.3/5V regulators, and in most case existing LM78xx circuits comes with the onboard output capacitor (10-100uf lytic)
We agree that for 3.3~5V one can use GRM 25V without issues.
But we are talking about the CK2III headphone amp here with +/-15V rails.
One can still purchase your regulators and swap out 3 caps.
Size compatible. Not really an issue.
Or you might reconsider using GJ8 just for your regulators >5V.
Cheers,
Patrick
But we are talking about the CK2III headphone amp here with +/-15V rails.
One can still purchase your regulators and swap out 3 caps.
Size compatible. Not really an issue.
Or you might reconsider using GJ8 just for your regulators >5V.
Cheers,
Patrick
Quoting LT3045 datasheet, P.17 Fig.6 :
https://www.analog.com/media/en/technical-documentation/data-sheets/LT3045.pdf
"While capacitance change due to DC bias for X5R and X7R is better than Y5V and Z5U dielectrics,
it can still be significant enough to drop capacitance below sufficient levels.
As shown in Figure 6, capacitor DC bias characteristics tend to improve as component case size increases,
but verification of expected capacitance at the operating voltage is highly recommended.
Due to its good voltage coefficient in small case sizes, LTC recommends using Murata’s GJ8 series ceramic capacitors."
Patrick
https://www.analog.com/media/en/technical-documentation/data-sheets/LT3045.pdf
"While capacitance change due to DC bias for X5R and X7R is better than Y5V and Z5U dielectrics,
it can still be significant enough to drop capacitance below sufficient levels.
As shown in Figure 6, capacitor DC bias characteristics tend to improve as component case size increases,
but verification of expected capacitance at the operating voltage is highly recommended.
Due to its good voltage coefficient in small case sizes, LTC recommends using Murata’s GJ8 series ceramic capacitors."
Patrick