Mine still seems to be ok running the JLH Chiarra classA headphone amp🙂
Longest I have run it is just over an hour because those heatsinks get very hot although the voltage stays spot on +/-15v
All the 1uf caps are Wima MKS
The transformer has 18v secondarys 120vA, theres no oscillation on mine as far as I can see on the scope
The only thing is those 470uf caps get very warm too because they are right next to the heatsink
Longest I have run it is just over an hour because those heatsinks get very hot although the voltage stays spot on +/-15v
All the 1uf caps are Wima MKS
The transformer has 18v secondarys 120vA, theres no oscillation on mine as far as I can see on the scope
The only thing is those 470uf caps get very warm too because they are right next to the heatsink
I haven't had time to do any further testing, (i'm laying floorboards at the moment) but perhaps just removing the film cap parallel to the elco will do the trick.Another scenario is adding a small resistor in series with the film cap.
Kees.
Kees.
Kees,
Get that floor in place so you can get back to fun. 😉 I hope that you will be able to eliminate the oscillation by removing or changing parts, not having to make a snubber.
T. -
How much current does the Chiara draw? Are you using a single supply for both channels? Are you using 50mm high sinks? I have extra boards if you aren't already dual mono.
If you can get a 15VAC transformer you'd have less heat. You could also use a 5W resistor in series with each transformer lead to drop the input voltage. Size it to leave at least 15VAC at the inputs.
Jacco, you sure are colorful. You're supposed to use a hammer. 🙂
Get that floor in place so you can get back to fun. 😉 I hope that you will be able to eliminate the oscillation by removing or changing parts, not having to make a snubber.
T. -
How much current does the Chiara draw? Are you using a single supply for both channels? Are you using 50mm high sinks? I have extra boards if you aren't already dual mono.
If you can get a 15VAC transformer you'd have less heat. You could also use a 5W resistor in series with each transformer lead to drop the input voltage. Size it to leave at least 15VAC at the inputs.
Jacco, you sure are colorful. You're supposed to use a hammer. 🙂
BobEllis said:
But i did.
As for brains, the g/f told me three weeks ago i broke my fingers.
She arranged a plastic surgeon collegue a week ago to break them again, i'm getting modern colorfull cast tomorrow.
I did put the boards on the bench to check if they kept the voltage with increasing load. On my old discrete class A headphone amp boards they sounded pretty ok, compared to the TO3 LT117/LT137AK regulators used previously.
(picture chop cleaned, Elso said it was obscene.
)Attachments
Hi,
tried removing c13 today.
With no cap // to r4 the result is the best yet.
All noise and hum and if any oscillation below 100uV.
With r4//1uF and c13 removed worse than with just r4//1uF but oscillation about 200uV with noise and hum total still no worse than 300uV.
Reminder, these voltages are approximate due to being right at the limit of my uncalibrated scope.
Early days, since only one board tested so far at Vout=15V and 18V.
tried removing c13 today.
With no cap // to r4 the result is the best yet.
All noise and hum and if any oscillation below 100uV.
With r4//1uF and c13 removed worse than with just r4//1uF but oscillation about 200uV with noise and hum total still no worse than 300uV.
Reminder, these voltages are approximate due to being right at the limit of my uncalibrated scope.
Early days, since only one board tested so far at Vout=15V and 18V.
Well, that's good news. If Kees or someone else confirms that works, I'll consider it case closed.
At this point:
With plenty of input voltage (AC>15V/ DC input greater than 21V) use the standard BOM. The group buy transformers will provide sufficient input voltage.
If you are cutting the minimum input close, omit one of the output film caps (C13, C23).
At this point:
With plenty of input voltage (AC>15V/ DC input greater than 21V) use the standard BOM. The group buy transformers will provide sufficient input voltage.
If you are cutting the minimum input close, omit one of the output film caps (C13, C23).
Hi Bob,
I know I said early doors, but I think it does not matter whether Vin>>Vout or Vin=Vout+2V3.
The cap solution is there somewhere I/we have just to try and make sense of it.
Part of the difficulty is that the PCB and schematic are not exactly the same.
C12 decouples/bypasses T2 collector to ground between C10 & P1
C13 decouples/bypasses T4 collector to ground between R11 and R10. Neither of these ground connections are near the output and both very effectively contaminate the ground reference of R10, R11 and D7.
The schematic would lead one to believe that C12 & C13 are augmenting the output caps C10 & C11.
I know I said early doors, but I think it does not matter whether Vin>>Vout or Vin=Vout+2V3.
The cap solution is there somewhere I/we have just to try and make sense of it.
Part of the difficulty is that the PCB and schematic are not exactly the same.
C12 decouples/bypasses T2 collector to ground between C10 & P1
C13 decouples/bypasses T4 collector to ground between R11 and R10. Neither of these ground connections are near the output and both very effectively contaminate the ground reference of R10, R11 and D7.
The schematic would lead one to believe that C12 & C13 are augmenting the output caps C10 & C11.
I am by no means qualified to speak to the finer points of grounding, board design, ground plane use and high frequency parasitics. All I know is that they can be a problem and that using a ground plane (like Jens did) can help reduce the issues.
If you feel that the caps mentioned are contaminating the ground references mentioned, by all means remove them. The regulator should operate well without them (and by your analysis better than as designed).
As I understand what Jens found in his testing, for the majority of users no changes are needed if you stick to the BOM parts (Panasonic FC, MKT film caps except the 100 pF which are FKP) unless you want to run the regulator with less than 22 V DC (sorry I misspoke last post).
The AVEL Y236203 http://avellindberg.com/transformers/y23_range_specs.htm supplied in the group buy has typical regulation of 11%, meaning at no load you are likely to see 16.65 VAC or more => 22.7 VDC after the rectifiers. Since this transformer is rated at 50 VA, at few hundred mA the DC voltage will likely remain over 22 VDC for most users.
Using caps with lower ESR could be the source of or contributing to your problems.
If you feel that the caps mentioned are contaminating the ground references mentioned, by all means remove them. The regulator should operate well without them (and by your analysis better than as designed).
As I understand what Jens found in his testing, for the majority of users no changes are needed if you stick to the BOM parts (Panasonic FC, MKT film caps except the 100 pF which are FKP) unless you want to run the regulator with less than 22 V DC (sorry I misspoke last post).
The AVEL Y236203 http://avellindberg.com/transformers/y23_range_specs.htm supplied in the group buy has typical regulation of 11%, meaning at no load you are likely to see 16.65 VAC or more => 22.7 VDC after the rectifiers. Since this transformer is rated at 50 VA, at few hundred mA the DC voltage will likely remain over 22 VDC for most users.
Using caps with lower ESR could be the source of or contributing to your problems.
Hi,
stolen from thread "power supply problems"
stolen from thread "power supply problems"
I'm away to solder in some output diodes to see what happens.
BobEllis said:
Sorry I'm late🙂
The Chiara draws about 200mA, I've just measured the positive rail.
You can only use a single +/- supply for this unit.
I've also got a some ALW super regs so it will be interesting to see how these compare sonically when I get time😉
I don't have a 15v transformer at the minute but I'll pick one up
So far I'd say these are good, well done guys😎
Forgot to say, yes I'm using 50mm heatsinks
The 100pf cap is NPO
Finally got some time for further testing.
I first tried different size bypass caps for R4 with little result.
Changing or removing C7 gave no results.
Removing C13 and C12 lowered the signal below the capabillities of my scope (2mV/div-2uS/div).
There is still some signal there but i can no longer make out its properties,being it to small for my scope to trigger on.Most of the time it drowns in the noisefloor of my scope.
Perhaps when i do a complete overhaul of my scope which requires a HV tester (i have a full service manual but no HV tester) it will trigger on the signal but i doubt it.
Someone with a better scope can perhaps do further testing but i think that any improvement will be in track layout etc.
As i said before i am now satisfied with the PSU and will do no further testing.(except from trying how it sounds when i finish my filterboards.)
Kees.
I first tried different size bypass caps for R4 with little result.
Changing or removing C7 gave no results.
Removing C13 and C12 lowered the signal below the capabillities of my scope (2mV/div-2uS/div).
There is still some signal there but i can no longer make out its properties,being it to small for my scope to trigger on.Most of the time it drowns in the noisefloor of my scope.
Perhaps when i do a complete overhaul of my scope which requires a HV tester (i have a full service manual but no HV tester) it will trigger on the signal but i doubt it.
Someone with a better scope can perhaps do further testing but i think that any improvement will be in track layout etc.
As i said before i am now satisfied with the PSU and will do no further testing.(except from trying how it sounds when i finish my filterboards.)
Kees.
Hi,
Assembled my PSU board with all group buy parts. I measured 15.11v and -15.09v. AC in is 17.87v.
I read recently on Aspen amplifier's site about great benefits to be had by using separate PSU for each channel. Does anyone here agree with that? Thinking about buying another PSU.
Assembled my PSU board with all group buy parts. I measured 15.11v and -15.09v. AC in is 17.87v.
I read recently on Aspen amplifier's site about great benefits to be had by using separate PSU for each channel. Does anyone here agree with that? Thinking about buying another PSU.
Seperate PSU's per channel are more valid for power amps than for pre-amps.Most power amps use a unregulated supply where as pre-amps usually have a regulated supply.On top of that most pre-amp are pure class A which means that the current draw is always the same no matter what signal is passing through it.The only thing speaking for using seperate PSU's for a pre-amp is to prevent crosstalk.A decent designed pre-amp doesn't put much signal on the powerrails and if there is a small signal there,the other channel should see it as a power supply ripple.A pre-amp with a decent PSRR should have no problems coping with that at all.
kro5998
kro5998
Hi,
an update of my latest experiences.
Blew my first output attaching scope ground to output OOPs.
The ratio of r10 to r1 is very relevant to start up conditions.
I have been using a 1.5 to 2.0 ratio in all my earlier, higher voltage, versions. This made my start up worse.
Increasing r1 to match r10 improved start up considerably. Still not starting properly though.
Caution! if r4 is removed then T4 can be reverse biased before/at start up. The BC5xx have a 5v absolute limit on reverse bias (Veb0).
Grounding.
I separately grounded D7, D8, C14, C15, R10, R11 direct to output ground. Almost all the noise, ripple, and if any oscillation became immeasureable on the scope. Just discovered that all my earlier voltage measurements were with scope probe set to 10:1. This latest measurement was done at 1:1 meaning the grounding alteration reduced all the rubbish by about a factor of 20! Now it really is about <=100uV with Vout=45V
I still have a number of other implimenatations to work around and test. I may take a while. The thunderstorm took out my ethernet adaptor (and it's integrated) so I'm on dial up at the moment.
an update of my latest experiences.
Blew my first output attaching scope ground to output OOPs.
The ratio of r10 to r1 is very relevant to start up conditions.
I have been using a 1.5 to 2.0 ratio in all my earlier, higher voltage, versions. This made my start up worse.
Increasing r1 to match r10 improved start up considerably. Still not starting properly though.
Caution! if r4 is removed then T4 can be reverse biased before/at start up. The BC5xx have a 5v absolute limit on reverse bias (Veb0).
Grounding.
I separately grounded D7, D8, C14, C15, R10, R11 direct to output ground. Almost all the noise, ripple, and if any oscillation became immeasureable on the scope. Just discovered that all my earlier voltage measurements were with scope probe set to 10:1. This latest measurement was done at 1:1 meaning the grounding alteration reduced all the rubbish by about a factor of 20! Now it really is about <=100uV with Vout=45V
I still have a number of other implimenatations to work around and test. I may take a while. The thunderstorm took out my ethernet adaptor (and it's integrated) so I'm on dial up at the moment.
Andrew, I am glad you seem to have resolved your issue with noise .
I still don't understand your startup problems - I had NO issues with startup with my 65V version. It uses MPSA42/92 and R1=1K, R10=2K.
Correct me if I am wrong, but D6 should prevent reverse biasing of T4 to any significant extent. The regulators in my Pass A75s use a similar, but not as refined topology (http://www.passdiy.com/pdf/a75p2.pdf) without anything in out R4 or D6 positions.
Mine run at 64 volt rails, all I did was swap in 12V zeners, use 1/2W resistors in the feedback loop and bypass the zeners with 100nF. In 3 or 4 years of nearly daily use they have not failed. I suspect that reverse biasing of T4 does not happen, although I see it as a possibility. (I could not find the MPSA42 reverse bias limit in my first cup of coffee state)
Comparing the two circuits, it may be that T4 must carry more current than T3 for reliable startup. Pass uses 3K3 in our R1 position (and no R5) and 2K2 for our R10 for 50V operation. This is still about the ratio that you are having problems with, though.
I still don't understand your startup problems - I had NO issues with startup with my 65V version. It uses MPSA42/92 and R1=1K, R10=2K.
Correct me if I am wrong, but D6 should prevent reverse biasing of T4 to any significant extent. The regulators in my Pass A75s use a similar, but not as refined topology (http://www.passdiy.com/pdf/a75p2.pdf) without anything in out R4 or D6 positions.
Mine run at 64 volt rails, all I did was swap in 12V zeners, use 1/2W resistors in the feedback loop and bypass the zeners with 100nF. In 3 or 4 years of nearly daily use they have not failed. I suspect that reverse biasing of T4 does not happen, although I see it as a possibility. (I could not find the MPSA42 reverse bias limit in my first cup of coffee state)
Comparing the two circuits, it may be that T4 must carry more current than T3 for reliable startup. Pass uses 3K3 in our R1 position (and no R5) and 2K2 for our R10 for 50V operation. This is still about the ratio that you are having problems with, though.
Pixelpusher -
Whether independent supplies is a benefit if probably best left to listening tests. Kro's argument against for low signal level circuits makes sense, but as I mentioned earlier I saw a bit of signal on my rails when driving my crossover to very high levels. The crossover uses NE5532s and should have decent a PSRR.
I am using a single supply for my active crossover at the moment because I haven't had a chance to hook up the second that has been ready to go for quite a while.
If you decide to try dual supplies, I do still have boards and parts kits.
Whether independent supplies is a benefit if probably best left to listening tests. Kro's argument against for low signal level circuits makes sense, but as I mentioned earlier I saw a bit of signal on my rails when driving my crossover to very high levels. The crossover uses NE5532s and should have decent a PSRR.
I am using a single supply for my active crossover at the moment because I haven't had a chance to hook up the second that has been ready to go for quite a while.
If you decide to try dual supplies, I do still have boards and parts kits.
Bob if you want to try seperate PSU's please go ahead and let us know if there is a audible difference.I am always carefull making statements one way or the other.In theory there should be some gain in using seperate PSU's.If you can hear the difference is an other matter.The mind is a tricky thing and wanting to hear some difference often makes you hear a difference that others don't hear.
Kees.
Kees.
Hi Bob,
I have been poring over circuit diagrams and measuring voltages on scalable for weeks now and beginning to inderstand how the reg starts up.
T4 draws no current before start up. When Vout is low it has a tiny Vbe voltage which holds it open.
But r10 has a voltage and this reverse biases t4 if Vr10>Vr11.
Prior to start up Vd7 is less than Vr10+Vbe of Tt. For start up to occur Vd7>Vr10+Vbe of t3
This is the source of the start up problem.
At start up the resistor ratio r3:r4 & output load controls the voltage on d7.
Working back from R10, the Voltage on r10=1.3*r10/r1. (1.3~=Vbe of t1 &t2)
Voltage on t3 base =Vr10+0.6 (0.6~=Vbe of t3)
Voltage across r9 is negligible.
Vd7=Vt3base=Vr10+0.6=1.3*r10/r1+0.6
putting some numbers on that first part. r1= r10=1k
Vd7~=1.3+0.6=1.9V. This is approximate, due to the effect of various Vbe voltages (t1,t2,t3) and the effect of d6 r7 pulling down r3 voltage since it is in parallel to r4.
For start up to occur the supply at the intersection of r3 & r4 must exceed Vd7.
With no load connected and t4 kept open then the r3:r4 ratio allows easier starting since the only "loads" present are the r2,r11 pair and LED+r6. However with a load connected the r3:r4 ratio causes the supplied voltage from the transformer to be reduced due to the resistive ladder.
The supply voltage =r4/(r3+r4)*Vd7=Vstart
Again some numbers show what is happening.
r3=10k, r4=1k, then Vstart~=11times Vd7.
With Vd7=1.9V therefore Vstart >21V if t1 + t2 Vbe total =1.2V then Vstart ~= 19.8V
A 15V transformer will generate about 21.2v less the diode drops of the bridge, about -1.6V for the 1n4004, leaving just 19.6V.
But the pass transistor is open and so the transformer will output a voltage approaching unloaded Vout or about 110% of rating say, 23.3-1.6=21.7Vdc after smoothing.
Wow, that is close, the approx starting voltage for a 15Vdc regulator is about 1volt or 2volts below the unloaded transformer voltage. Luck or design?
Now increase the r10:r1 ratio to reduce dissipation in t4.
You have used r1=1k and r10=2k. with r3=10times r4, then
Vstart~=(10+1)*(1.3*2/1+0.6)=35.2V. Vr10~=2.6V. This appears as as Veb0 on t4, no problem since most transistors can survive 5V and your MPS allows about 7V. But remove R4 and destructive Veb0 occurs on t4.
So it appears that Vstart is substantial and depends heavily on the resistor ratios chosen to suit dissipation and suitable device currents.
The start voltage can be reduced substantially by selecting the resistors specifically on the basis of Low Vstart but keeping an eye on dissipation as secondary.
I am looking at a two transistor CCS to replace r10 to keep t4 dissipation sensible at elevated Vout and still maintain a good r1 current to switch on the t1 early. CCS minimum voltage is about 0.6V to 0.7V, much less than 1.3*r10/r1.
That circuit is in the process of measuring at the moment and first indications are Vstart<5V for Vout=61V. But still not low enough to allow a pre regulator between c5 and r1 etc.
A lot of content, hopefully without typos.
I have been poring over circuit diagrams and measuring voltages on scalable for weeks now and beginning to inderstand how the reg starts up.
T4 draws no current before start up. When Vout is low it has a tiny Vbe voltage which holds it open.
But r10 has a voltage and this reverse biases t4 if Vr10>Vr11.
Prior to start up Vd7 is less than Vr10+Vbe of Tt. For start up to occur Vd7>Vr10+Vbe of t3
This is the source of the start up problem.
At start up the resistor ratio r3:r4 & output load controls the voltage on d7.
Working back from R10, the Voltage on r10=1.3*r10/r1. (1.3~=Vbe of t1 &t2)
Voltage on t3 base =Vr10+0.6 (0.6~=Vbe of t3)
Voltage across r9 is negligible.
Vd7=Vt3base=Vr10+0.6=1.3*r10/r1+0.6
putting some numbers on that first part. r1= r10=1k
Vd7~=1.3+0.6=1.9V. This is approximate, due to the effect of various Vbe voltages (t1,t2,t3) and the effect of d6 r7 pulling down r3 voltage since it is in parallel to r4.
For start up to occur the supply at the intersection of r3 & r4 must exceed Vd7.
With no load connected and t4 kept open then the r3:r4 ratio allows easier starting since the only "loads" present are the r2,r11 pair and LED+r6. However with a load connected the r3:r4 ratio causes the supplied voltage from the transformer to be reduced due to the resistive ladder.
The supply voltage =r4/(r3+r4)*Vd7=Vstart
Again some numbers show what is happening.
r3=10k, r4=1k, then Vstart~=11times Vd7.
With Vd7=1.9V therefore Vstart >21V if t1 + t2 Vbe total =1.2V then Vstart ~= 19.8V
A 15V transformer will generate about 21.2v less the diode drops of the bridge, about -1.6V for the 1n4004, leaving just 19.6V.
But the pass transistor is open and so the transformer will output a voltage approaching unloaded Vout or about 110% of rating say, 23.3-1.6=21.7Vdc after smoothing.
Wow, that is close, the approx starting voltage for a 15Vdc regulator is about 1volt or 2volts below the unloaded transformer voltage. Luck or design?
Now increase the r10:r1 ratio to reduce dissipation in t4.
You have used r1=1k and r10=2k. with r3=10times r4, then
Vstart~=(10+1)*(1.3*2/1+0.6)=35.2V. Vr10~=2.6V. This appears as as Veb0 on t4, no problem since most transistors can survive 5V and your MPS allows about 7V. But remove R4 and destructive Veb0 occurs on t4.
So it appears that Vstart is substantial and depends heavily on the resistor ratios chosen to suit dissipation and suitable device currents.
The start voltage can be reduced substantially by selecting the resistors specifically on the basis of Low Vstart but keeping an eye on dissipation as secondary.
I am looking at a two transistor CCS to replace r10 to keep t4 dissipation sensible at elevated Vout and still maintain a good r1 current to switch on the t1 early. CCS minimum voltage is about 0.6V to 0.7V, much less than 1.3*r10/r1.
That circuit is in the process of measuring at the moment and first indications are Vstart<5V for Vout=61V. But still not low enough to allow a pre regulator between c5 and r1 etc.
A lot of content, hopefully without typos.
I chose to let the regulator start at about 17.5V DC (From my simulation) when the load is about 30 Ohms and up. I wanted to make the design able to withstand a short circuit by simply not letting any power be dissipated in the pass devise during a short circuited output situation. If you measure and touch the heatsink of the regulator while keeping the output shorted to ground you will see that it does not get hot at all. The reason is that the ref voltage is pulled down and thus the pass devise is turned off.
You can change the allowed power dissipation in the pass devise, but I advise not to do so as the pass devise will fail quickly if you run the regulator at high input voltages and short the output to ground (I'm sure you know what I mean)
EDIT: Note that short circuit protection often calls for a power resistor of maybe 0.5 ohm to generate a voltage drop in the shortcircuit protection detection circuit. This takes up space and gives you more power loss during operation.
Another requirement was to make a BOM that uses as few different parts as possible and thus make the group buy easier to handle. This approach sometimes scarifies other design parameters, but as I simulated and build the prototype I found no issues regarding the power up.
Anyhow keep up the spirit, and please provide a BOM for the high voltage regulator when it is ready.
\Jens
You can change the allowed power dissipation in the pass devise, but I advise not to do so as the pass devise will fail quickly if you run the regulator at high input voltages and short the output to ground (I'm sure you know what I mean)
EDIT: Note that short circuit protection often calls for a power resistor of maybe 0.5 ohm to generate a voltage drop in the shortcircuit protection detection circuit. This takes up space and gives you more power loss during operation.
Another requirement was to make a BOM that uses as few different parts as possible and thus make the group buy easier to handle. This approach sometimes scarifies other design parameters, but as I simulated and build the prototype I found no issues regarding the power up.
Anyhow keep up the spirit, and please provide a BOM for the high voltage regulator when it is ready.
\Jens
Hi all,
the RC filter formed by r3 & c14 attenuates the ripple from the rough supply fed into the base of t3.
The dynamic impedance of the 4040 Voltage reference also attenuates the ripple.
How does one calculate the attenuation effectiveness of the Vref?
When the Vref and the Cap combine, how does one calculate the combined attenuation effect?
If I can't get the knowledge here I will open a new thread. I'll give it a couple of days for readers to see the request.
the RC filter formed by r3 & c14 attenuates the ripple from the rough supply fed into the base of t3.
The dynamic impedance of the 4040 Voltage reference also attenuates the ripple.
How does one calculate the attenuation effectiveness of the Vref?
When the Vref and the Cap combine, how does one calculate the combined attenuation effect?
If I can't get the knowledge here I will open a new thread. I'll give it a couple of days for readers to see the request.
- Status
- Not open for further replies.