The two regulators start chasing each other. The final reg. increases it's power when the audio level increases, then the first reg. reacts by increasing it's output, now the fial reg. has more input so it lowers it's demand.
I first read about this type of problem when Bob Carver built his first high-power amplifiers. Because they were very powerful and very light, they were often used for PA work at remote locations. With AC power coming from a portable generator. The amp's regulator and the generator's regulator would chase each other until the amp blew-up.
Henry W. Ott has a short section on this problem, let me look.
I first read about this type of problem when Bob Carver built his first high-power amplifiers. Because they were very powerful and very light, they were often used for PA work at remote locations. With AC power coming from a portable generator. The amp's regulator and the generator's regulator would chase each other until the amp blew-up.
Henry W. Ott has a short section on this problem, let me look.
Properly implemented series regulators will have sufficient phase margin for stability. But folks do sometimes fail to check all the design corners and hence can end up with implementations which work at some output levels but not others. However, the nCore is a rather performant module, meaning attached DIY regulators may degrade performance. Not something I'd suggest changing unless one has access to the measurement gear to determine whether one's making things better or worse. Some of us do, but I think we're in the minority. This has probably been discussed plenty on the main nCore thread anyway.
Try here. As for your other questions in post 57 what the datasheet says is the nCore module uses the same supply arrangement as any other half bridge driver. You can look at the design collateral for any bridge driver to understand how that works. But audio specific ones such as the IRS20957 will be the most illuminating as they've the same rail arrangement and dual level shifts. See the schematic for IRAUDAMP4, for example. I'd also search the main nCore thread.
Last edited:
From the Henry W. Ott book "Electromagnetic Compatibility Engineering"
EMC Books
Section 13.6 'Power Supply Instability' this paragraph is about power conditioners and power line filters but regulators have the same problem.
Consider what could happen if a power supply having a negative input
impedance were connected to a power-line filter with a positive output impedance.
If the source voltage decreases slightly, then the power supply will have to
draw more current to keep the power constant. The increased current will cause
an increased voltage drop across the filter’s output impedance, which lowers the
input voltage to the power supply. This in turn will cause the power supply to
draw even more current and will lower the input voltage even more and so on
and so on. Hence instability!
EMC Books
Section 13.6 'Power Supply Instability' this paragraph is about power conditioners and power line filters but regulators have the same problem.
Consider what could happen if a power supply having a negative input
impedance were connected to a power-line filter with a positive output impedance.
If the source voltage decreases slightly, then the power supply will have to
draw more current to keep the power constant. The increased current will cause
an increased voltage drop across the filter’s output impedance, which lowers the
input voltage to the power supply. This in turn will cause the power supply to
draw even more current and will lower the input voltage even more and so on
and so on. Hence instability!
Thanks. I had observed people dropping pre-regulators in front of Jan Didden's super regulator and had not thought this an issue. Learning from you guys!
So to summarise, all share the same power supply typology but:
Power supply voltage
UcD400: 45-65V unregulated; 57V typical
UcD700: 75-95V unregulated; 90V typical
NC400: 35-75V unregulated; 64V typical
Signal supply:
UcD400/700: +/- 12V regulated
NC400: 16-25V unregulated
Driver supply voltage (each referenced to the neg power supply voltage*):
UcD400: 13V regulated
UcD700: 15V regulated
NC400: 16-25V unregulated
* Understanding why this is the case requires knowledge on the operation of the amp I don't yet possess.
I'm slowly making my way through the massive main NCore thread...
So to summarise, all share the same power supply typology but:
Power supply voltage
UcD400: 45-65V unregulated; 57V typical
UcD700: 75-95V unregulated; 90V typical
NC400: 35-75V unregulated; 64V typical
Signal supply:
UcD400/700: +/- 12V regulated
NC400: 16-25V unregulated
Driver supply voltage (each referenced to the neg power supply voltage*):
UcD400: 13V regulated
UcD700: 15V regulated
NC400: 16-25V unregulated
* Understanding why this is the case requires knowledge on the operation of the amp I don't yet possess.
I'm slowly making my way through the massive main NCore thread...
Last edited:
Any further comments on Piisami's circuit or is the DC-error protection all correct now?
(I am wading through the main N-Core thread, up to post #3750 (about halfway). At this stage Bruno seems to have long ago left the building and Julf you are one of the main guiding lights for N-Core enthusiasts. ;-) It's a challenge to sift through the audiophile bashing on the one hand and, in some cases, idolisation of Bruno and Hypex on the other, but I am taking notes as to the more useful posts as I go along. It's interesting that after 3750 posts the topic of linear versus switching has not really been discussed much at all. The closest we get (at this point in the thread at least) are posts #1523 and 1524 in which Bruno suggests "this is discussed at length somewhere in this thread" but indeed it wasn't - at least not as far as I could find. By #3705, the fact that Bruno designed an SMPS to work well with the NC400 is used to dismiss prospects that a linear power supply could be better. ;-) You guys certainly know more about this stuff than I do and I will keep reading. I will come back to this question in the NCore thread and be prepared for taking a bashing daring to raise the issue again. No doubt, I will also take a bashing for owning high-end audio gear. In the interim, I would very much like to see Piisami complete his power supply!)
(I am wading through the main N-Core thread, up to post #3750 (about halfway). At this stage Bruno seems to have long ago left the building and Julf you are one of the main guiding lights for N-Core enthusiasts. ;-) It's a challenge to sift through the audiophile bashing on the one hand and, in some cases, idolisation of Bruno and Hypex on the other, but I am taking notes as to the more useful posts as I go along. It's interesting that after 3750 posts the topic of linear versus switching has not really been discussed much at all. The closest we get (at this point in the thread at least) are posts #1523 and 1524 in which Bruno suggests "this is discussed at length somewhere in this thread" but indeed it wasn't - at least not as far as I could find. By #3705, the fact that Bruno designed an SMPS to work well with the NC400 is used to dismiss prospects that a linear power supply could be better. ;-) You guys certainly know more about this stuff than I do and I will keep reading. I will come back to this question in the NCore thread and be prepared for taking a bashing daring to raise the issue again. No doubt, I will also take a bashing for owning high-end audio gear. In the interim, I would very much like to see Piisami complete his power supply!)
Switchmode half bridges have 0dB PSRR so all PSRR comes from the amplifier's loop gain. That's typically around 30dB in most class D implementations, which is really quite poor (120dB PSRR is not particularly difficult when implementing class AB with ICs; refer to link post 63 for requirements), so a supply with load regulation is useful if one's looking for sound in = sound out for music reproduction. As opposed to, say, guitar amps where sag can be a feature.
Hypex's SMPSes are all a bit strange in that they lack line regulation. I've never been able to find anything as to why Bruno didn't use a bandgap reference the way normal SMPSes do. But, so far as I can tell, it's a bug and not a feature.
Hypex's SMPSes are all a bit strange in that they lack line regulation. I've never been able to find anything as to why Bruno didn't use a bandgap reference the way normal SMPSes do. But, so far as I can tell, it's a bug and not a feature.
I will have to read that thread and you post again. I'm new to all this. I'm still learning how designs for each of the 3 power supplies (Vcc/Vee, Vsig, Vdr) are best designed.
LM317 based regulation can be improved greatly by having an SPX431 set the reference voltage and an J310 anchor its current. The same could likely be applied to the 15V rail. (I've not looked at this in terms of neg voltage.) The thread I previously linked to involved a discussion of enhancing an LM317 circuit with a TL431 (as per the latter's data sheet).The SPX431 is a much lower noise replacement for the TL431. The addition of the J310 was a logical next step in the discussion. I built this little setup as a learning exercise following the discussion in that thread.
R1 can be omitted. Some suggest a cap between J310 source and SPX431 ref but my Spice modelling suggested this wasn't a good idea so I left it out.
Lasse did some great work testing all the different combinations. Results here.
Of course none of this helps a supply for the NC400 unless one can get comfortable with the dual regulation issue warned of above. In this respect it is interesting that when the question was raised, Bruno did not say don't regulate the Vdr supply, rather he said that regulation wasn't necessary because of the on-board HxR12 regulation.
LM317 based regulation can be improved greatly by having an SPX431 set the reference voltage and an J310 anchor its current. The same could likely be applied to the 15V rail. (I've not looked at this in terms of neg voltage.) The thread I previously linked to involved a discussion of enhancing an LM317 circuit with a TL431 (as per the latter's data sheet).The SPX431 is a much lower noise replacement for the TL431. The addition of the J310 was a logical next step in the discussion. I built this little setup as a learning exercise following the discussion in that thread.
An externally hosted image should be here but it was not working when we last tested it.
R1 can be omitted. Some suggest a cap between J310 source and SPX431 ref but my Spice modelling suggested this wasn't a good idea so I left it out.
Lasse did some great work testing all the different combinations. Results here.
Of course none of this helps a supply for the NC400 unless one can get comfortable with the dual regulation issue warned of above. In this respect it is interesting that when the question was raised, Bruno did not say don't regulate the Vdr supply, rather he said that regulation wasn't necessary because of the on-board HxR12 regulation.
Last edited:
Yah, the 317+431 thread is an unusually good one (I've been aware of it since a bit before post 11). The underlying concept of composite amplifiers is a rich one and happens to be where most effort in performance pushing audio is getting invested these days---using a 431 as an error amplifier to a regulator, for example, is the same basic notion as applying an error amplifier to overdrive the comparator in a class D amp. A more difficult, but more important, question to answer is identifying what level of performance no longer yields subjective improvement. Design effort beyond that is wasted, after all. The super regulator crowd seems both really good and really bad about this. The typical level of standalone regulator characterization is about the best for any class of DIY project. But before and after measurements of a system showing a performance improvement (or even a difference) from hotting up the regulator are few and far between. And I've never come across an ABX test of a measured regulation improvement to determine if it's subjectively meaningful.
You may have noticed Bruno's justifiably proud of the nCore having 56dB loop gain at 20kHz. That's one reason the module hits 23uV RMS noise, which is quite competitive with class AB amplifiers---for the most part, the way noise in audio circuits works is it's a microvolts or a few tens of microvolts happen kind of thing. However, if one doesn't need the output power it's actually easy to beat this. I've multiple amp channels sitting in my living room with 83dB loop gain at 20kHz. One of them is a low noise implementation that comes in around 300nV RMS over the audio band. It runs off a linear supply with regulation from a Fairchild LM7812ACT+LM7912CT pair in standard configuration---input caps, regulators, output caps, shutdown protection diodes, nothing else. I've ABXed it against a noisier implementation with the same loop gain that's about 6uV RMS with drivers that go to 100dB SPL W/m and could get no discrimination whatsoever between the two channels. I could hear a cold solder joint I inadvertently left on one pin of one part on the board (when assembling boards on painkillers after surgery, hey, it happens) but rewetting it got things back to no discrimination.
What this suggests to me is one shouldn't expect much, if any, subjective improvement from pulling the noise limited portion of the NC400's THD+N curve to the left. Partly because the module's pretty quiet already. Partly because it has 31.8dB noise gain. In particular, the latter means that if, say, you improved the module to 10uV RMS the source would have to be around 100nV RMS to cleanly expose the module limit. Headline spec on the quietest audio DACs currently available (ESS901x in mono) is 15dB noisier than that and mainstream pro audio generally lands around 35dB noisier. So if you want to get more out of the NC400 you're probably better off changing it to a lower signal gain so as to make life easier on the source. I'm unsure of the NC400's gain structure but the manual does mention a gain set resistor similar to that on the UcD modules. If the difference amplifier part of the nCore has the same 4.5 gain as the UcD modules the minimum noise gain easily obtainable from the module would be 19dB.
I'm not saying don't have fun messing with regulators. Just pursue it with the knowledge ABXing is most likely to only show a difference if regulator changes degrade rather than improve performance. It is possible a linear supply will be enough worse than the Hypex SMPS the NC400 was presumably measured on that improved regulators will be of value. But as I do the maths back of the envelope that's one of those maybe, maybe not kinda deals.
You may have noticed Bruno's justifiably proud of the nCore having 56dB loop gain at 20kHz. That's one reason the module hits 23uV RMS noise, which is quite competitive with class AB amplifiers---for the most part, the way noise in audio circuits works is it's a microvolts or a few tens of microvolts happen kind of thing. However, if one doesn't need the output power it's actually easy to beat this. I've multiple amp channels sitting in my living room with 83dB loop gain at 20kHz. One of them is a low noise implementation that comes in around 300nV RMS over the audio band. It runs off a linear supply with regulation from a Fairchild LM7812ACT+LM7912CT pair in standard configuration---input caps, regulators, output caps, shutdown protection diodes, nothing else. I've ABXed it against a noisier implementation with the same loop gain that's about 6uV RMS with drivers that go to 100dB SPL W/m and could get no discrimination whatsoever between the two channels. I could hear a cold solder joint I inadvertently left on one pin of one part on the board (when assembling boards on painkillers after surgery, hey, it happens) but rewetting it got things back to no discrimination.
What this suggests to me is one shouldn't expect much, if any, subjective improvement from pulling the noise limited portion of the NC400's THD+N curve to the left. Partly because the module's pretty quiet already. Partly because it has 31.8dB noise gain. In particular, the latter means that if, say, you improved the module to 10uV RMS the source would have to be around 100nV RMS to cleanly expose the module limit. Headline spec on the quietest audio DACs currently available (ESS901x in mono) is 15dB noisier than that and mainstream pro audio generally lands around 35dB noisier. So if you want to get more out of the NC400 you're probably better off changing it to a lower signal gain so as to make life easier on the source. I'm unsure of the NC400's gain structure but the manual does mention a gain set resistor similar to that on the UcD modules. If the difference amplifier part of the nCore has the same 4.5 gain as the UcD modules the minimum noise gain easily obtainable from the module would be 19dB.
I'm not saying don't have fun messing with regulators. Just pursue it with the knowledge ABXing is most likely to only show a difference if regulator changes degrade rather than improve performance. It is possible a linear supply will be enough worse than the Hypex SMPS the NC400 was presumably measured on that improved regulators will be of value. But as I do the maths back of the envelope that's one of those maybe, maybe not kinda deals.
Last edited:
Piisami, how are you getting on?
Your latest links no longer work. I was looking at your previous schematic. Haven't you got the pin orientation on the VOM1271 around the wrong way?
I find the datasheet for these a little confusing. In your circuit is the voltage into the input side of the driver 5V? (It would appear to be.) What current is required/appropriate for the VOM1271 to turn on and drive the mosfets to conduct?
In addition to this project I have another where the use of VOM1271 and mosfets would seem to be applicable. It's a linear ATX power supply. The 12V, 5V and 3.3V rails should not deliver current until the a PWR_ON signal comes from the motherboard. I was planning to use a TPS3510 to provide the necessary PWR_OK signal to the motherboard. I'm thinking this 5V (5mA or so) signal could turn the main power rails in similar fashion as to what you have done here.
Your latest links no longer work. I was looking at your previous schematic. Haven't you got the pin orientation on the VOM1271 around the wrong way?
I find the datasheet for these a little confusing. In your circuit is the voltage into the input side of the driver 5V? (It would appear to be.) What current is required/appropriate for the VOM1271 to turn on and drive the mosfets to conduct?
In addition to this project I have another where the use of VOM1271 and mosfets would seem to be applicable. It's a linear ATX power supply. The 12V, 5V and 3.3V rails should not deliver current until the a PWR_ON signal comes from the motherboard. I was planning to use a TPS3510 to provide the necessary PWR_OK signal to the motherboard. I'm thinking this 5V (5mA or so) signal could turn the main power rails in similar fashion as to what you have done here.
The current through the led's of the VOM1271 is about 18 mA's as the max output current of the picaxe is 20mA's . If that's enough to turn the Fets on and off and fast enough? time will tell .
My hobby room is being painted and redecorated at the moment , so can't test it at the moment for another 2 weeks or so . See post 39 :" I'm not an expert in Fet formulas , maybe one of the smart guys here ( Andrew T ?, Chocoholic ?, Esperado ? ), can do the math on driving four IRFB4115's with a single VOM1271"
Or two after the revision .
Cheers ,
Rens
Slowly.... The next thing I'm going to do is improving those regulators. I'm thinking of one of these circuits: LCAudio typology or RJM Audio X-reg typology. I'll try to find some time during the weekend to finish the schematics...
Then, I'll start redesigning the board layout
I have no idea why the links don't work anymore, haven't touched those. Anyone can suggest a better free sharing solution. Obviously dropbox sux..
Last edited:
Here is the new version, with X-reg regulated buffer PSU and improved driver psu.
SGK, You mentioned that the resitor between Source and Gate of J310 can be omitted. Do you mean that it should be shorted or leave the source unconnected? How about the resistor value for 12V?
Oh, and I think VOM1271 is connected correctly. Please note that I could not find proper component from eagle, I used random smd4 component instead...
An externally hosted image should be here but it was not working when we last tested it.
SGK, You mentioned that the resitor between Source and Gate of J310 can be omitted. Do you mean that it should be shorted or leave the source unconnected? How about the resistor value for 12V?
Oh, and I think VOM1271 is connected correctly. Please note that I could not find proper component from eagle, I used random smd4 component instead...
Last edited:
I placed a resistor between source and gate in my circuit simply because of what I read in The Art of Electronics, Horowitz and Hill (pg 127 and 128)
but the testing Lasse did seemed to suggest it made no difference and so perhaps I am misapplying Horowitz and Hill's text to this application. By "leave it out" I mean wire through the resistor (source to gate).
Resistor values for 12V: Vout = (1+R1/R2)Vref where R1 and R2 are the voltage divider and Vref = 2.5V. So you want R1/R2 = 3.8. Perhaps 931 and 243? (There's probably some impedance calculation which would yield the most optimal absolute values.)
For the protection diode D1 it was suggested to me in the LM317+TL431 thread that this could be improved further by placing two of them in series with a 0.1uF cap from the junction of the two diodes to ground. Easy to implement.
On the VOM1271, here's a clip from the data sheet oriented so it matches your top one.
Your circuit is fine but if you take it to a board and lay down the VOM1271 as you have depicted won't you run into a problem because the pins aren't in the right place? (you have in-bound into what would be pin 2) In any event, I think it less confusing and error prone to lay it out like the pictorial in the data sheet.
but the testing Lasse did seemed to suggest it made no difference and so perhaps I am misapplying Horowitz and Hill's text to this application. By "leave it out" I mean wire through the resistor (source to gate).
Resistor values for 12V: Vout = (1+R1/R2)Vref where R1 and R2 are the voltage divider and Vref = 2.5V. So you want R1/R2 = 3.8. Perhaps 931 and 243? (There's probably some impedance calculation which would yield the most optimal absolute values.)
For the protection diode D1 it was suggested to me in the LM317+TL431 thread that this could be improved further by placing two of them in series with a 0.1uF cap from the junction of the two diodes to ground. Easy to implement.
On the VOM1271, here's a clip from the data sheet oriented so it matches your top one.
An externally hosted image should be here but it was not working when we last tested it.
Your circuit is fine but if you take it to a board and lay down the VOM1271 as you have depicted won't you run into a problem because the pins aren't in the right place? (you have in-bound into what would be pin 2) In any event, I think it less confusing and error prone to lay it out like the pictorial in the data sheet.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.