Hotrodding the UCD modules

[classd4sure]

Hi Chris

for the unregulated psu , i use a 30VA Rcore transformer with primary/secondary screen , a RURP860-based bridge and a CRCRC filter , the last C at the opamp pins is a 47uF ELNA Silmic 2 alone and the 826 is stable
the 826 is biased ( cascode 2SK170 FETs) by the negative rail of course
with a 10K input R at the input stage , i obtain a 50mV DC offset at the module output with no coupling cap

Alain

PS : Is the improvement important with the mkp Panasonic cap in the output filter ? Why have you choose a lower value than 680nF ?

Hi Alain,

Very interesting.

I have a 50VA toroid with static shield as well, 12V secondaries. Even a spare set of BHC T-neworks 10000uF..... hmmm. I'm guessing I'd get about ~18V out of that, too much for the AD8620! Maybe it's another reason to change it.

I'm rather surprised the last cap is so small.. what are the other values if you don't mind saying? From reading up on it seems the general rule of thumb is just to use a smaller value as the first cap to extend/soften the conduction angle of the diode and a bigger storage cap at the end... usually about 1000uF.

I've also seen another very interesting way of creating the current source. That of a typical bjt cascaded mirror.

I had a few reasons behind going with a smaller value filter cap. One was that my speakers are 4ohms and the filter response was designed for six, so I hoped to perhaps flatten it out more given my 4ohm load. Seeing the response of the amp into a 1ohm load though.. it's hardly a concern.

The other reasons were that it was the biggest value available at a 1% tolerance which I wanted to try, and that I thought a slightly higher switching frequency would be beneficial towards lowering THD. It always seem to have that effect in simulation at least.

Those were only the reasons for going to a lower value, certainly the extra precision of such a tight tolerance would not hurt.

I experienced a very big improvement by changing the filter cap.

It would have been much more subtle had I not swapped the decoupling caps for FC's first, but having done so it lost pretty well all the highs, they were entirely suffocated.

For the most part the panasonic cap restored them nicely, and it became very clear in microdetail. A very sharp clearity, yet smooth with no hint of harshness. Sadly I've been unable to try no others to compare it too, but I am very pleased with the result.

 

[Portlandmike]

Hi Alain,

Very interesting.

I have a 50VA toroid with static shield as well, 12V secondaries. Even a spare set of BHC T-neworks 10000uF..... hmmm. I'm guessing I'd get about ~18V out of that, too much for the AD8620! Maybe it's another reason to change it.

I'm rather surprised the last cap is so small.. what are the other values if you don't mind saying? From reading up on it seems the general rule of thumb is just to use a smaller value as the first cap to extend/soften the conduction angle of the diode and a bigger storage cap at the end... usually about 1000uF.

I've also seen another very interesting way of creating the current source. That of a typical bjt cascaded mirror.

I had a few reasons behind going with a smaller value filter cap. One was that my speakers are 4ohms and the filter response was designed for six, so I hoped to perhaps flatten it out more given my 4ohm load. Seeing the response of the amp into a 1ohm load though.. it's hardly a concern.

The other reasons were that it was the biggest value available at a 1% tolerance which I wanted to try, and that I thought a slightly higher switching frequency would be beneficial towards lowering THD. It always seem to have that effect in simulation at least.

Those were only the reasons for going to a lower value, certainly the extra precision of such a tight tolerance would not hurt.

I experienced a very big improvement by changing the filter cap.

It would have been much more subtle had I not swapped the decoupling caps for FC's first, but having done so it lost pretty well all the highs, they were entirely suffocated.

For the most part the panasonic cap restored them nicely, and it became very clear in microdetail. A very sharp clearity, yet smooth with no hint of harshness. Sadly I've been unable to try no others to compare it too, but I am very pleased with the result.

Chris,

I've been playing around with a sim for a op amp supply, or general low power supply.

I happen to have a 35V split secondary transformer laying around, so that was my starting premise.

Some tweaking of the R values will need to happen for lower output voltage transformers.
This is a wordy discription, sorry:

dual bridge supply, but I'll only talk about one side since both are identicle. No need for complementary in this instance.

1N4148 bridge diodes.
220 ohms each end of the bridge to the first bypass cap.
47uF, 100uF, 1000uF whatever you like on the first caps. The resistors are choosen to limit peak currents in the 1N4148's to ~100mA. It simulates very well with 220ohm, 47u, 50Rohms, 47uF, SST511 biased zener driving MPSA18A follower into 2u2 gives great results.
Output impedance if the NPN is biased by 500ohms is about 1.3 ohms to beyond the audio band, a good thing. Absolutely no ringing on the output due to load steps. This is better than tighter load reg with ringing IMO, and any supply disturbances due to load are basically a attenuated version of the supply volts, so very tame and musical.
sub 1mV supply ripple on the output. Doesn't even show on the sim.

AFter you do this twice, tie supplies in series, and you've got a regulator that will be hard to beat. Super regs, eat your heart out.

The transformer coupling of noise should be minimized becuase of the resistors in series into the caps.
The series resistors could, and likely should, be split so that each has one comming into it, one for the +leg, one for the -leg, up to the last regulator.
The base of the NPN is tied to the zener via a 10k resistor bypassed by 100uF.
It takes about 6 seconds to ramp to full rail voltage.

A picture is worth a thousand words.
See attatchment.

Mike

 

[Portlandmike]

LM6172

LM6172 looks stable, at least w/o the amp being powered, but with an aux supply.

Very clean actually, I'm suprised.

Offset is big with inputs open, no suprise, about 7mV between outputs, ~200mV offset from ground. STill not bad.

Offset inputs shorted is sub 1mV.

Totally within the trim range of the latest rev ucd400's.

I'm going to give it a try, but care needs to be taken with single ended AC coupled inputs.

More later.

Mike

 

[Portlandmike]

It would be quite helpful if someone with balanced inputs would try the LM6172. They are not expensive, a couple bucks, and availible from digikey.

I'd even be happy to send a couple LM6172's to someone in N. America wanting to do this with balanced inputs. I suspect it will be quite good.

I might try them in my single ended applications, but it really is sub-optimal and a pain in the ***. I don't know what good it will do to change to a DC coupled preamp that isn't as good from an evaluation standpoint. That, or other complications that might confuse the issue. I could add the AC coupling caps back in, but I didn't like them at all, although I didn't let them break in which I find to be huge in coupling caps.

the offset from a typical balanced input should be quite good as I'm seeing only <1mV offset with a balanced low impedance input.

the power amp stage will add more offset than this.

Regards,

Mike

 

[Stevenacnj]

UCD 400 - on board blue LED

Is there really any need for this LED, other than to indicate if the module is on or not?

Can - should it be removed?

I remember reading somewhere that all LED's introduce some noise into the line. I plan on having an LED on the front of my amp which will be tied into the power supply.

If there is no harm in removing this LED (and possibly some good) do I just remove it or must a wire be inserted across the pads?

Thanks in advance!

 

[Portlandmike]

UCD 400 - on board blue LED

Is there really any need for this LED, other than to indicate if the module is on or not?

Can - should it be removed?

I remember reading somewhere that all LED's introduce some noise into the line. I plan on having an LED on the front of my amp which will be tied into the power supply.

If there is no harm in removing this LED (and possibly some good) do I just remove it or must a wire be inserted across the pads?

Thanks in advance!

I doubt it will make any difference IMO. I'd just leave it, but remove it if you like. LED's are actually pretty low noise in my experience. They make a half decent voltage reference. LED's that are flashing on and off are a different story.

Mike

 

[Stevenacnj]

Mike thanks for the reply.

If the LED is removed, does one just remove it or does a wire need to be inserted across the LED pads?

Thanks

 

[deandob]

Rubycon ZL

There has been some interest in Rubycon ZL caps. I'm discussing with a couple of distributors how to purchase the 1500uF 63v part for a power supply.

I figure 9,000uF per side should be sufficient for the UCD400 which is 12 caps per UCD module. This gives you an equivalent parallel impedence of a 6x1500uF bank of 0.002 ohms. Compare this to a 10,000uF T-Network 4 pole BHC cap of 0.015 ohms.

Looks like I can only purchase 100 minimum quantities, and as I have 5 modules I need to build, I only need 60. Are there other people on this thread willing to buy some of my remaining stock (40). Price should be good, about USD $3-4 per cap.

Regards,
Dean

 

[classd4sure]

What would the point of a lower resistance supply than the Ron of the mosfets themselves?
Maybe the parasitic inductance would be a more interesting parameter?

 

[maxlorenz]

Dear Dean:
I would have to find a use for them 😀 as my amps don't need a cap upgrade for the moment.

I think there could be interest in 470uF/100V caps to replace the module's own PS caps 🙂
I would easilly ask for 20 😎

Unless someone think that the low ESR will be detrimental in some way. I don't need more troubles

What do you think?
My own attempt to get them was frustrating.

Regards
M

 

[deandob]

ClassD4Sure,

Isn't the parasitic inductance measured in the impedence (at a particular frequency)? Including the inductance the impendence of these caps is still very good. Are you saying that having parallel inductance with a bank of low ESR caps is a problem (eg. ringing)?

Regarding the mosfets, yes not much you can do about Ron however having a low impedence power supply still lowers the total series resistance. I dont have the details of the mosfet on the UCD, but assuming its 0.08 ohms then reducing the power supply impedence from say 0.02 ohms to almost 0 will decrease the power supply impedence by a reasonable percentage. However at these low impedences the wiring & even PCB starts having an effect.

I'm open to any further feedback on this topic - I could be going down the wrong path....

I'd prefer to purchase the Jensen or BHC 4 pole caps but the expense is just too much to justify. I'm thinking this low impedence cap array is the next best & cheaper choice.

Regards,
Dean

 

[deandob]

Rubycon ZL 1500uF found!

I just got a quote from an online parts sourcer called 'partsminer', who have a pretty efficient system to locate hard to find parts.

I got a quote back for a minimum quantity of 120 of 63v 1500uF ZL Rubycon for $3.79 ea (would need to pay postage on top of that) which is a decent price. By comparison Newark have ZL 1000uF 50v for $3.16ea but dont stock the 1500uF.

Total price is $450 which is tolerable if there are others interested in making a low impedence power supply out of a capacitor bank and I'll need to get rid of 60 of these (however in a usable power amp capacitor bank you will need quite a few of these).

To keep this thread on topic I'll create a new post in the trading post thread to judge interest.

Regards,
Dean

 

[classd4sure]

Mike thanks for the reply.

If the LED is removed, does one just remove it or does a wire need to be inserted across the LED pads?

Thanks

Steven,

I think your goal would be to relocate it to your case, correct?

I wouldn't worry about the noise emitted from the LED, actually I'll be selling some audiophile window tint exactly for this kind of application

Seriously, properly used they do make good low noise voltage references as Mike was saying but in this case it's simply an indicator and doesn't bias anything to the best of my knowledge.

If you jumper it out with a wire you'll be shorting a circuit for no good reason. Simply use some wires to extend the leads over to your chassis. A small drop of silicon on the wires at the PCB might be a good idea to help support them and keep the solder pads from being stressed and possibly lifting.

Worrying about the noise from it in any way is just being overly paranoid.. so just enjoy the nice cool glow.

Chris

 

[classd4sure]

ClassD4Sure,

Isn't the parasitic inductance measured in the impedence (at a particular frequency)? Including the inductance the impendence of these caps is still very good. Are you saying that having parallel inductance with a bank of low ESR caps is a problem (eg. ringing)?

Regarding the mosfets, yes not much you can do about Ron however having a low impedence power supply still lowers the total series resistance. I dont have the details of the mosfet on the UCD, but assuming its 0.08 ohms then reducing the power supply impedence from say 0.02 ohms to almost 0 will decrease the power supply impedence by a reasonable percentage. However at these low impedences the wiring & even PCB starts having an effect.

I'm open to any further feedback on this topic - I could be going down the wrong path....

I'd prefer to purchase the Jensen or BHC 4 pole caps but the expense is just too much to justify. I'm thinking this low impedence cap array is the next best & cheaper choice.

Regards,
Dean

Hi Dean,

Yeah I believe it is a part of the "ESR" graphs towards higher frequencies... where it rises. // caps will decrease this as well and I think that's possibly where you stand to gain the most.

I'm not at all saying it would be a problem, though if ringing should occure it should definatly be snubbed, and at these low levels of ESR I'd think it very possible something like that could happen.

Good point about the total series resistance.. That's reason enough for me, let the juice flow.

Another excellent point was the wire parasitics, more so than the PCB because you can't do much there. I had the point of diminishing returns in mind, and it would seem at the levels of ESR you projected the wire parasitics are likely start taking over (watch for the ringing). I think it would be very worthwhile giving it at try, but proceed with caution, how you wire it will be critical for best performance, and the need for snubbing could be very real.

Looking forward to hearing about your results.
Chris

 

[Eva]

I find that discussion about capacitors quite funny, because class D amplifiers have a particular current consumption frequency distribution.

First, at audio frequencies, loudspeaker current is drawn from one rail and a small fraction (sometimes not so small) is returned to the supply through the opposite rail. Any capacitor will handle this, as both the frequencies and the RMS currents involved are low, so the a 10.000uF 63V standard capacitor will be absolutely fine (no low ESR is required or even desirable as I'll explain).

Second, at the carrier frequency and their harmonics up to 100Mhz (diode recovery transients) there are quite high RMS AC currents flowing, that *must* be confined into one or two (depending on topology) single, either low-ESR low-ESL or non-electrolytic, capacitors placed very close to the output devices and the output filter, AND NEVER EVER LEFT TO FLOW THROUGH WIRING AND EXTERNAL CAPACITOR BANKS, or even PCB tracks longer than a few centimeters.

To better achieve that HF current confinement into a very small loop around the output stage, filter inductors in series with the supply rails are almost mandatory. These inductors force the incoming power supply rails to show a very high HF impedance as seen from the output stage (should start rising already in the audio band), preventing any HF current to flow outside, otherwise you'll have more of an AM/FM transmitter than a class-D amplifier. With these inductors (that some if not all designs include in a smaller of larger flavour), a third order CLC system is formed in each rail, and high ESR is required to make it damped. High ESR from the inductors is nor quite practical, neither from the local output stage bypass capacitors, so it has to come from the main power supply.

So don't waste time, space and money with low ESR capacitor banks. Even high ESL is useful there to reduce EMI.

 

[classd4sure]

Hi Eva,

Taking your post into consideration, I'd be interested in hearing your theory on why I experienced a notable improvement in bass (lower frequencies, and hitting much harder in general, lower series impedance perhaps?) by going to // Jensen 4 pole caps, and also the same improvement again by going to a low impedance bulk storage cap (panasonic FC) which serves to keep the AC RMS currents on board (or 20% of them anyway).

I really don't think you should have a high ESR supply for doubling as a snubber?

In this case the bulk storage caps + snubbers are already on the board. Anyhow, changing to // caps (which were also more capacitance) and a higher capacitance low impedance bulk storage cap brought on a dramatic improvement in bass response. The bulk storage cap however did it at the cost of high frequency response... thereby it lost a good level of the very extensive holographic effect it had.

I'd appreciate your comments on this.
Chris

 

[Eva]

That's like hearing some water flowing without having a clue of where the river is. Since changes at bass frequencies are much easier to see on the oscilloscope by comparing input to output waveforms than to hear, I would try to trace that phenomena if I was you, it may have an unexpected origin (increased EMI or ground loops causing bass boost??) or it might be just not measurable (plain placebo effect).

Also, which was the control strategy employed in the amplifier(s) in which you observed these changes? Were they pre-filter feedback or post-filter feedback?

 

[deandob]

Eva,

Thank you for the lucid explaination about the HF effects with a low ESR power supply. It looks like there is 2 effects going on here:
1) The high frequency switching where high currents are involved but this is handled by the on board power supply decoupling on the UCD, so the external power supply has little effect.
2) The much lower audio frequency which can also involve large currents over a much longer duration (eg. bass lines) where the setup of the external power supply does have an effect.

I must say I agree with Chris about the improved bass effect with a low impedence capacitor bank. I have explored this with a tripath amplifier (not the UCD) and there was definitely an improvement in the firmness of the bass with the low ESR setup. I also noted that with this setup a snubber on the output of the capacitor bank had more effect (further improvement in bass firmness and mid range clarity) than a set of cheaper 10,000uf capacitors.

I have not measured this on a CRO.

Regards,
Dean

 

[classd4sure]

That's like hearing some water flowing without having a clue of where the river is. Since changes at bass frequencies are much easier to see on the oscilloscope by comparing input to output waveforms than to hear, I would try to trace that phenomena if I was you, it may have an unexpected origin (increased EMI or ground loops causing bass boost??) or it might be just not measurable (plain placebo effect).

Also, which was the control strategy employed in the amplifier(s) in which you observed these changes? Were they pre-filter feedback or post-filter feedback?

Hi Eva,

I'm far too poor for a scope, that's like running a marathon on crutches, but I do what I can with what I have to the best of my ability, using a combination of research, educated guesses, experimentation and listening.

In this case I can assure you, I always try my hardest to never allow myself to be fooled by what I may be expecting to happen or the like, and often get second opinions, which tend to closely mimick my own, save for the exception of two instances where my second opinion told me "there is no difference" and upon a second third and fourth listen I was forced to agree.

I seriously wish I did have a scope, because then I'd have the ability to go alot further with my testing and more importantly verification. So try to not hate me for doing it as I do.

All the same, some "tweaks" offer such a night and day difference even a deaf person would be able to comment intelligently, and I have to say, changing the power supply caps is one of them.

In fairness, I moved up in capacitance as well, from 10 000uF per rail for a stereo ucd180 /4ohm loads (not enough at all) with BHC T-networks up to paralleled Jensen 4 pole caps, 2X15 000uF a rail.

So I believe that's certainly one reason the bass improved, but I'd have to think the lower ESR helped as well. How do I know it did actually improved? It was really night and day, it hit alot lower than ever before, and you could start to feel it in your chest, whereas before you could not.

Then then simple change of the local storage caps, from the stock 470uF's, said to be raited to handle at least 20% of total RMS output current, but not really the best caps, changed them to panasonic FC ~820uF. That cut the high's right down and the caps had a more colored sound from mids to highs, but it gained tremendously in the bass, now _really_ feeling it in the spine and chairs and rattling the walls at even much lower volumes required to feel it at all before.

So the difference is truly very real.

I'd made no changes at all to stock snubbers, which are as you describe I believe, and also said to have something like a ~.05R damping resistor which doubles as overcurrent sense.

I asked the question to Dean wondering myself.... why worry about lower ESR in a supply than the Ron of the mosfets themselves for instance... where's the point of diminished returns. He pointed out total series impedance.... aaaah, that makes sense to me, widden up that river and you get alot more flow.

Maybe it's a case of where just doing your best to ensure the bottleneck isn't at the supply. No doubt the sound could also be changed by tuning the on board snubbers somewhat.

The main complaint you'll find with the UCD is that it is "thin" sounding, lacking bass. The old argument was that with such clean highs it can give the impression that the bass is lacking .... or something to that effect. The truth is in stock configuation it really is lacking bass. The notes for the most part are there, but there's no sense of power behind them at all..... I stated to Jan-Peter a few times when I first got the modules going that it hardly moves my woofers.... it sure moves them now.

I'll also say in fairness to the rest of your questions that I fully agree, wiring and layout is critical, but there's more than one right way to get this amp working, depending on the rest of the system.

I can assure you, there is no such beast as a ground loop in my system. The supply if fully floating, with the only reference to true earth at the signal input of the amplifier. The source was made to float, and the amp's chassis is connected to true earth.

It's a dual IXYS 68A FRED rectifier which is now RC snubbed across the AC inputs (best guess values used), plitron 500VA with static screen connected to true earth at the case, and mu-metal shielding. // Jensen 15 000uF four pole caps as already mentioned, stereo setup where the modules are wired in bi-phase.

The UCD modules are your typical phase shift self oscillator with post filter feedback employed. Wired fully differential from input to output.

I should also state that the homebrew single ended UCD version I built demonstrated a similar effect, in this case however going from // low grade caps to just a set of ELNA's out of an old receiver improved the sound somewhat, but I found what made the biggest difference with bass after that was the size or selection of local bulk storage caps. By messing with them I was able to tune the response of the amp, whereby adding more capacitance got improved low frequency response while diminishing the high frequency response, and also removing capacitance having the opposing effect.

Also, I had a third set of ears who'd only heard the amp on a few occasions, after the upgrade to the // Jensen caps the response to the heavier hitting bass from him was "how'd you get it to do that?".

It's hard to argue a placebo effect when it starts to rattle your spine where it never did before.

If I had a scope I'd sure be looking into it further. 🙁

Cheers,
Chris

 

[BWRX]

Hi Eva,

Taking your post into consideration, I'd be interested in hearing your theory on why I experienced a notable improvement in bass (lower frequencies, and hitting much harder in general, lower series impedance perhaps?) by going to // Jensen 4 pole caps, and also the same improvement again by going to a low impedance bulk storage cap (panasonic FC) which serves to keep the AC RMS currents on board (or 20% of them anyway).

Isn't the reason for the bass improvement easily seen by looking at the voltage and current relationships for capacitors and inductors? i=C(dV/dt) and V=L(di/dt).

Larger caps can supply more current with less voltage drop over time. Class D just switches the load between the supply rails, so larger capacitances with low esr have increased current capability with less voltage sag than smaller caps with higher esr. The more stable that voltage is while driving an inductive load (output filter and the driver's voice coil) the more stable di/dt will be through the inductive load. And as well all know, the speaker makes sound by the interaction between a fixed magnetic field and the induced magnetic field of the voice coil - which is solely determined by the current flowing through it.

As to why many people say bass is improved while the highs seem lacking due to a larger bank of supply caps - could it be a result of the higher esl of larger caps?