I have not seen this specifically discussed before, so I was looking for some expert opinions and recommendations. It has come to my attention, that perhaps, high feedback regulators may not be the best choice as the final regulation stage for an oscillator module. Right now I am using Twisted Pear Audio Tridents to power the two oscillators (CCHD-957s) on my async USB interface.
Any ideas on powering clocks, and any available oscillators or simple oscillator circuits to try (non feedback perhaps...)
Any ideas on powering clocks, and any available oscillators or simple oscillator circuits to try (non feedback perhaps...)
Well, the oscillators wants 3.3V +- 0.3V so they appear not to be too fussy about their supply.
The Trident looks like 3.3V +- 1% 100mA max with max input of 5.5V
Depending on accuracy and package, I did a search of Digikey and found 387 instock that might be accurate enough for the job. (I would have narrowed it down further but the accuracy column seems to have disappeared and I don't know what package you would like to use.)
I've used MIC5201-3.3YS TR and it seems to work fine. It is an SOT-223 package and I like the pin-out because it is similar to the LM78xx TO-220 types in that the tab is GND and I ground it to the ground plane as a heat sink.
The Trident looks like 3.3V +- 1% 100mA max with max input of 5.5V
Depending on accuracy and package, I did a search of Digikey and found 387 instock that might be accurate enough for the job. (I would have narrowed it down further but the accuracy column seems to have disappeared and I don't know what package you would like to use.)
I've used MIC5201-3.3YS TR and it seems to work fine. It is an SOT-223 package and I like the pin-out because it is similar to the LM78xx TO-220 types in that the tab is GND and I ground it to the ground plane as a heat sink.
potential problems...
Oscillator modules are very, very sensitive to any supply noise, and jitter levels are highly dependent on having a super clean DC supply. Additionally, regulator output impedance may be very important to keep the clock frequency from traveling back through the supply and contaminating the other supplies in the DAC.
So noise, and output impedance are two critical factors, whereas voltage stability may not be such a critical factor.
Oscillator modules are very, very sensitive to any supply noise, and jitter levels are highly dependent on having a super clean DC supply. Additionally, regulator output impedance may be very important to keep the clock frequency from traveling back through the supply and contaminating the other supplies in the DAC.
So noise, and output impedance are two critical factors, whereas voltage stability may not be such a critical factor.
the dac clock used to have an LT LDO as standard in this design but they changed to shunts. hes trying to move away from massive feedback and I presume something better than an LDO. non feedback clock circuits are going to be pretty shithouse IMO, not even sure how you would do it in any meaningful way. I guess this is Loriens board yes barrows?
clocks/regs
Thanks Qusp. I have used batteries (LiFePO) before, and did not like the charging hassle, but for just oscillators it might be worth another try. Are you having good results with LiFePOs nominal 3.2 volts direct to an 3.3V oscillator with no further circuitry? I guess the LiFePO output voltage is stable enough for this application.
Has anyone actually measured the LiFePO battery for noise? I know there is noise caused by the internal chemical reactions, but have never seen a measured spec?
I am using an SOtM async USB receiver board so the layout is what it is-I am just adding separate power supply for the clocks with a dedicated regulator for each of them, and am definiterly looking for something better than IC regs. I hoped to draw out the experts who might be able to discuss the feedback question for clock regulators.
I will also search for Demian's designs, as I would like to read the context of the threads they came up in as well.
Thanks Qusp. I have used batteries (LiFePO) before, and did not like the charging hassle, but for just oscillators it might be worth another try. Are you having good results with LiFePOs nominal 3.2 volts direct to an 3.3V oscillator with no further circuitry? I guess the LiFePO output voltage is stable enough for this application.
Has anyone actually measured the LiFePO battery for noise? I know there is noise caused by the internal chemical reactions, but have never seen a measured spec?
I am using an SOtM async USB receiver board so the layout is what it is-I am just adding separate power supply for the clocks with a dedicated regulator for each of them, and am definiterly looking for something better than IC regs. I hoped to draw out the experts who might be able to discuss the feedback question for clock regulators.
I will also search for Demian's designs, as I would like to read the context of the threads they came up in as well.
http://www.diyaudio.com/forums/power-supplies/214471-anyone-listened-li-ion-batteries.html
some figures on this thread regarding battery noise.
some figures on this thread regarding battery noise.
just quickly, i'll reply to the rest later, but Demian presents his regs earlyish in the fifo thread and there is also some talk of it in the widget thread, as they are included on the PCB as an option. I still havent seen proper testing of lifepo4 noise either, but under such low current draw I really doubt its even worth worrying about, plus the benefits of being totally floating.
i'm just not sure how you would make a non feedback clock at a specific audio related frequency, the output stage of the clocks buffers but also sets the speed by multiplication/division. without feedback you would be stuck with the fundamental frequency of the crystal and they simply dont oscillate at audio frequencies.
i'm just not sure how you would make a non feedback clock at a specific audio related frequency, the output stage of the clocks buffers but also sets the speed by multiplication/division. without feedback you would be stuck with the fundamental frequency of the crystal and they simply dont oscillate at audio frequencies.
yeah but not lifepo4 or anything recent, the stuff referenced there is all pretty old articles and doesnt cover this comparatively new chemistry.
A battery has Z inductive component just as a regulated power supply. The impedance is, in fact, much higher than the Jung regulator. (With a good layout, the Jung regulator achieves Zout of a micro-Ohm or so.)
An externally hosted image should be here but it was not working when we last tested it.
Z out is important because comparators can "stutter" (for lack of a better term) if the supply voltage droops during a transition.
Use a separate regulator for the clock circuitry and local ceramic and electrolytic bypass.
No LiFePo's though on those graphs. 
Their output impedance is reportedly much lower than that of other types of batteries (even at higher frequencies).
I recall jkenny posting a conference paper on the issue but the link was dead last time I checked. :S
Fingers crossed that qusp has kept a copy or something.
Their output impedance is reportedly much lower than that of other types of batteries (even at higher frequencies).
I recall jkenny posting a conference paper on the issue but the link was dead last time I checked. :S
Fingers crossed that qusp has kept a copy or something.
Unlikely that you'll see me shell out $160 for a 12V 7Ah LiFePO4 battery.No LiFePo's though on those graphs.
Their output impedance is reportedly much lower than that of other types of batteries (even at higher frequencies).
Well....
Since the thread is about powering an oscillator, I do not see how a 12V 7Ah pack is relevant. A single LiFePO4 cell will power a clock for good long while, for just a handful of dollars. In fact, a LiFePO4 supply for clocks, including a management circuit and charger can likely be built for less than the cost pf a good transformer based linear supply with a good discrete regulator.
Since the thread is about powering an oscillator, I do not see how a 12V 7Ah pack is relevant. A single LiFePO4 cell will power a clock for good long while, for just a handful of dollars. In fact, a LiFePO4 supply for clocks, including a management circuit and charger can likely be built for less than the cost pf a good transformer based linear supply with a good discrete regulator.
i'm aware of all the decades old comparisons jackinnj, what sort of a comparison is that? 7Ah vs Jung reg? most Jung regs are severely current limited vs LiFePO4, a single $12 cell is 3v3 @2500mAh … we are talking about clocks here right? thats over 30hrs minimum right there and you can then charge it in less than 1hr show me what sort of super reg you can build for $12, show me its current rating …
sorry TheShaman, I dont have that, I know the paper you are talking about though. by comparison batteries present more advantage than just noise and output impedance. for clocks, floating outputs not being the least handy. capable of FAAAAAR higher current bursts.
the difference between a few mOhms and a few µOhms in practical terms since you have to connect them somehow is pretty meaningless IMO. Its a similar argument as OTT high damping factor when you still need to connect with speaker cables/connectors and often through a crossover. AFAIK super regs need to see some impedance or they ring/resonate, batteries will not oscillate, they are unconditionally stable.
i'm not going to try and say Jung regs are bad, I think its an excellent design and have used them quite a lot over the years, it was the first regulator I ever built. but I wouldnt say they are the better choice here
sorry TheShaman, I dont have that, I know the paper you are talking about though. by comparison batteries present more advantage than just noise and output impedance. for clocks, floating outputs not being the least handy. capable of FAAAAAR higher current bursts.
the difference between a few mOhms and a few µOhms in practical terms since you have to connect them somehow is pretty meaningless IMO. Its a similar argument as OTT high damping factor when you still need to connect with speaker cables/connectors and often through a crossover. AFAIK super regs need to see some impedance or they ring/resonate, batteries will not oscillate, they are unconditionally stable.
i'm not going to try and say Jung regs are bad, I think its an excellent design and have used them quite a lot over the years, it was the first regulator I ever built. but I wouldnt say they are the better choice here
qusp...
You just about have me convinced to try a couple of cells for the oscillators. Do you recommend absolutely no decoupling, right now I just use .1uF MLCC on the output of each Trident to each oscillator module.
I would assume that the length of wiring from the cell to the oscillator may be fairly critical, I might have to get creative with layout to keep it short...
Any tips you could provide would be appreciated.
You just about have me convinced to try a couple of cells for the oscillators. Do you recommend absolutely no decoupling, right now I just use .1uF MLCC on the output of each Trident to each oscillator module.
I would assume that the length of wiring from the cell to the oscillator may be fairly critical, I might have to get creative with layout to keep it short...
Any tips you could provide would be appreciated.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.