Swapping class AB amp linear supply with a SMPS - incl measurements

[Piere]

Duh??? Of course, if you increase loop gain by an extra op-amp as Tom did, you will improve PSRR as well. Nothing special there.

 

[egellings]

I think I read that the Hypex SMPS isn't regulated per se, the output follows the input. But it may indeed have a lower output impedance.

It could have good load regulation and not very good line regulation. A low output impedance will result in good load regulation even if line regulation is less than stellar.

 

[McFly]

In what sense? If an amp fails, it fails. Most time by some sense of overload or in the long run by wear out of electrolytic capacitors. The Hypex SMPS are designed for their own Class-D amps, not for general use.

So this hypex SMPS is not to be used with a class AB topology?

 

[CleanSound]

Thanks for doing this test and sharing. . .eye opening.

Your experiment leaves a lot of questions to be answered. I wonder if that is a result of just the Hypex SMPS and what would the results be with other SMPS. Saids a lot about modern day SMPS.

 

[Doodski]

So this hypex SMPS is not to be used with a class AB topology?

It supplied amps and voltage so yes, it should be good for class AB stuff. It won't load down/sag like a designed for the task power supply but it will do in a jiffy.

 

[egellings]

So this hypex SMPS is not to be used with a class AB topology?

If the switch mode supply delivers the same DC voltages as the conventional supply does, it should make no difference. Only possible problem could arise with the SMPS is that it is likely to be regulated, and won't sag much under load, whereas the 50/60Hz-xfmr-diode-'lytic one will sag, limiting the power dissipation in the output stage when severely overloaded. So long as that overload/dissipation issue is properly addressed, then both types of supply should work well. Power amplifier output stages don't benefit much from regulation, while small sign stages do.

 

[Piere]

So this hypex SMPS is not to be used with a class AB topology?

Why not? With general use I did mean powering lamps, motors etc. Read the manual. They are designed especially for the characteristic fluctuating momentary load of audio amplifiers. This in contrast with generic SMPS that like to see a constant load. So class-AB wouldn't be a problem provided rail voltages matches that of the amplifier. But you can't power a full class-A with it that draws severe power continously. Then it will smoke likely soon.

 

[Blenderguy]

I am a bit late to this thread. I have the same amplifier and just recently replaced the two filter caps.
Since I have not measured the output, I don't know how high the 60 Hz hum component is on output. But I do know that there is no hum in my speakers. I have ML Electromotion ESL hybrid that uses a woofer.

There is some mechanical hum on the transformer, only audible with cover off and ears within a foot or so. Being in the US the line frequency is 60Hz. The voltage on the secondary is a bit high, 88-90V when line voltage is 122V.
Assuming that the 60 Hz hum component is low enough on the output terminals, I don't see the point of replacing the existing supply with a lower efficiency one.

Even though the Switching supply is very efficient if designed well, it is replacing the isolation transformer with a high frequency transformer and a transistor switching stage. None of those are lossless. In additional you have to power some control circuitry, which may or may not be any lower than the magnetizing current in the transformer for the linear supply. And both use rectifier diodes to convert AC to DC.

I am not an expert on SMPS, but don't they need another rectifier after the switching transformer? And many more capacitors? So I think an unregulated linear supply is about as efficient as you can make. Ripple is a function of the filter capacitor size and in my case I increased caps to 22,000uF each. A quick calculation can show that the energy stored is significantly higher that in a switchmode supply.
So the only conclusion I can draw from the replacement are several things.
1.The cost of the SMPS naybe less that the huge transformer, most likely because the cost of copper has gone up substantially in the last 40 years since it was made.
2. Much less weight. Great for shipping since freight across the ocean is expensive, if it was supplied from factory that way. And great for people that may have trouble moving the amp around.
3. Less magnetic coupling to surrounding circuits. Does not effect twisted signal wires.
4. Increased internal heat from switching transistor. If designed from ground up, the heatsink can be attached to the outside case, a retrofit makes this much harder.

I thought this modification was very interesting. If my transformer turned defective, I would probably do the same as I priced out replacements.

 

[DonH56]

There is no "switching transformer"; SMPS uses a switched inductor followed by filter caps (which can be small due to the high switching speed).

An SMPS is much more efficient than a linear supply, but intrinsically includes regulation, so I don't think you can compare unregulated linear to SMPS. Even so, a SMPS is usually more efficient, but trading diode and transformer losses against how much of the I-V cycle is used gets complicated. And of course "linear" supplies are actually anything but; the diodes switch and generate all kinds of switching noise, transformers have hysteresis, and so forth but marketing does not talk about that.

 

[Blenderguy]

It may not be comparable but nonetheless if you change out the transformer, bridge and filter caps for an SMPS, I am not convinced that the amplifier will use less power. I am actually sure it won't because as you mentioned it is regulated meaning that under extreme loads where the original supply will sag, the SMPS will supply more power by not letting the rails sag as much. There is a switched inductor in the SMPS, but since it only has one set of windings and low number of turns, it will be less lossy than the standard rectifier transformer.
To really benefit this scenario, as I researched it in the last few days, you would have a+/- supply for the output transistors to prevent them for over-stressing, and add an additional boost supply for the VAS section. That would save some power and it does, because I build an amplifier like that. With Mosfet outputs, you need a boost of about 10-12V because of the large Vgs needed to drive them. By reducing the supply to the Mosfets you expose them to less stress and dissipation. There are quite a few Class AB amplifiers like that. The idle current, which I think the OP did not mention is 350mA at 84V x 2. So roughly 120 W idle for two channels. Lower the output transistor voltage by 10V on each rail and you get a saving of 14W. Not really a lot.
For my use, I was looking at this because of the additional Aux voltages, but is not enough power unless I buy one for each channel

SMPS800RE | Connex Electronic

connexelectronic.com

This would be a worthwhile modification for me if I could get a 2000W SMPS

 

[DonH56]

It may not be comparable but nonetheless if you change out the transformer, bridge and filter caps for an SMPS, I am not convinced that the amplifier will use less power. I am actually sure it won't because as you mentioned it is regulated meaning that under extreme loads where the original supply will sag, the SMPS will supply more power by not letting the rails sag as much. There is a switched inductor in the SMPS, but since it only has one set of windings and low number of turns, it will be less lossy than the standard rectifier transformer.

The switching devices in a typical SMPS have much less loss (lower on resistance, less voltage drop, heat, etc.) than rectifier diodes, a factor in achieving higher efficiency.

To really benefit this scenario, as I researched it in the last few days, you would have a+/- supply for the output transistors to prevent them for over-stressing, and add an additional boost supply for the VAS section. That would save some power and it does, because I build an amplifier like that. With Mosfet outputs, you need a boost of about 10-12V because of the large Vgs needed to drive them. By reducing the supply to the Mosfets you expose them to less stress and dissipation. There are quite a few Class AB amplifiers like that. The idle current, which I think the OP did not mention is 350mA at 84V x 2. So roughly 120 W idle for two channels. Lower the output transistor voltage by 10V on each rail and you get a saving of 14W. Not really a lot.

I would not expect much in the way of savings at idle. Like amplifiers themselves, efficiency becomes more important at higher power levels.

If saving power for an AB amplifier was the goal, I'd look at going class G or H, which requires a fancier power supply.

For my use, I was looking at this because of the additional Aux voltages, but is not enough power unless I buy one for each channel

SMPS800RE | Connex Electronic

connexelectronic.com

This would be a worthwhile modification for me if I could get a 2000W SMPS

I do not know, but what does Hypex offer for their high-power amplifiers? Buckeye Amps went with a custom SMPS for his design.

Edit: Hypex has a 3 kW SMPS: https://www.hypex.nl/products/smps-family/smps3ka700

 

[Blenderguy]

Yeah I agree with everything you said, but the Hypex power supply you mentioned does not have 2 additional voltage rails.
I can get a 2kW supply from the other vendor also without the additional rails. Not going to pursue a mod like that.

I think people who rejuvenate vintage equipment, will usually only replace the caps. Frankly I am surprised that my caps lasted 40 years with only a small increase in leakage current and ESR. My concern is that if one wanted to recap their old audio amps, this seems to be the time do do it, as I see that many manufacturers are obsoleting those big caps. I bought special TDK caps with 57Amp ripple current capability. I was mostly impressed with their chart that showed that at 45C instead of 105C and 20% ripple current, the estimated life expectancy was 250,000 hours. These are the components NASA puts in their space probes, lol. With all the disposable equipment being made nowadays, you may not find components rated for more a few years life, before they are destined for the landfill.

 

[DonH56]

Yeah I agree with everything you said, but the Hypex power supply you mentioned does not have 2 additional voltage rails.
I can get a 2kW supply from the other vendor also without the additional rails. Not going to pursue a mod like that.

Thanks, I did not catch that you needed additional rails. If this is to mod a vintage amp I would not bother replacing the power supply with an SMPS, would look at other circuit elements.

I think people who rejuvenate vintage equipment, will usually only replace the caps. Frankly I am surprised that my caps lasted 40 years with only a small increase in leakage current and ESR. My concern is that if one wanted to recap their old audio amps, this seems to be the time do do it, as I see that many manufacturers are obsoleting those big caps. I bought special TDK caps with 57Amp ripple current capability. I was mostly impressed with their chart that showed that at 45C instead of 105C and 20% ripple current, the estimated life expectancy was 250,000 hours. These are the components NASA puts in their space probes, lol. With all the disposable equipment being made nowadays, you may not find components rated for more a few years life, before they are destined for the landfill.

Capacitor lifetime varies wildly IME. Temperature is a big killer of electrolytics. Re-capping is popular but not needed as much as people think IME/IMO. Caps can often be reconditioned, and replacing caps with much larger values can cause other problems like blown fuses, diodes, and transformers if inrush current is not controlled.

A lot of folk also do not realize that common electrolytics are often spec'd at +20/-50% of value over their lifetime.

Satellites and many space applications are actually relatively benign for temperature; it's the radiation that is a killer, and of course surviving the launch. Exterior components can see wide temperature swings, but "inside the box" temperature is often controlled, both for lifetime and performance.

 

[Blenderguy]

There are other sources to consider when seeing that hum is out of spec.. Attached is a schematic of one channel of this amp. It has the rails for the voltage gain stages isolated with additional capacitors. OP did not mention if he looked to see if these were open. I have seen many capacitors not nearly 40 years old that were just an open circuit. In my case when I get more time I will replace them.
You can see the elegance of this circuit, doing its job with just FIVE! transistors in the gain stages. (Taken right out of the Toshiba data book from 1985 - I have a copy. Kind of dates me). It may not have really high PSRR, this is why the exxtra filtering on the supply rail. The description states that to save money, use a smaller transformer with sagging rails and set the rail voltage under load to be able to produce speced output power. Somewhat the antithesis of using a SMPS.

The other interesting test published in their service manual is the ability to handle capacitive loads. This is an ideal amplifier to drive electrostatic loadspeakers with. I have not seen many, if any amplifier tests done in recent timeframe that even test with square wave, let alone with a capacitor in parallel. The ringing shows the ability of the feedback network to keep the output somewhat under control.

Capacitor lifetime varies wildly IME. Temperature is a big killer of electrolytics.

See third picture. Shows the effect of temperature and ripple current. It is very predictable as EPCOS/TDK has made many lifetime tests. They even have confidence on their 250,000 hour (28+ years) life test. Out of all the capacitors on Mouser website this one was amazing.

 

[DonH56]

There are other sources to consider when seeing that hum is out of spec.. Attached is a schematic of one channel of this amp. It has the rails for the voltage gain stages isolated with additional capacitors. OP did not mention if he looked to see if these were open. I have seen many capacitors not nearly 40 years old that were just an open circuit. In my case when I get more time I will replace them.

I must confess I did not go back to the start of the thread; I usually look at "New Posts" which all too often has me jumping into the middle of something I do not belong. Like now.

I have seen caps fail after a couple of years, and last for decades. Much depends upon the operating environment, mainly heat and ripple. Failing tube rectifiers can introduce a lot of ripple, and filter (decoupling) caps mounted too near a heat source (think heat sinks in amps) often fail prematurely.

You can see the elegance of this circuit, doing its job with just FIVE! transistors in the gain stages. (Taken right out of the Toshiba data book from 1985 - I have a copy. Kind of dates me). It may not have really high PSRR, this is why the exxtra filtering on the supply rail. The description states that to save money, use a smaller transformer with sagging rails and set the rail voltage under load to be able to produce speced output power. Somewhat the antithesis of using a SMPS.

Small capacitors are often used to keep supply impedance low over a wider bandwidth. Capacitor self-resonance is pretty low for big electrolytics, and circuit PSRR falls with frequency, so extra smaller caps with higher self-resonant frequency help. Choosing poorly can also cause issues if noise hits at resonant peaks if the filtering is not done properly (just throwing caps at it is not always a good idea). This is something I have actually spent a lot of time on, and have taught fellow engineers who have not as much practical experience. The problem is actually worse with SMPS due to the high speed and need for low ESR but need to avoid resonances.

The other interesting test published in their service manual is the ability to handle capacitive loads. This is an ideal amplifier to drive electrostatic loadspeakers with. I have not seen many, if any amplifier tests done in recent timeframe that even test with square wave, let alone with a capacitor in parallel. The ringing shows the ability of the feedback network to keep the output somewhat under control.

Square-wave testing seems to have gone by the wayside, at least for most published tests. I suspect (hope) it is still done internally. For the past 2-3 decades higher-bandwidth power devices and refined circuit topologies have made all but the very worst designs quite stable with capacitive (etc.) loads.

See third picture. Shows the effect of temperature and ripple current. It is very predictable as EPCOS/TDK has made many lifetime tests. They even have confidence on their 250,000 hour (28+ years) life test. Out of all the capacitors on Mouser website this one was amazing.

Yah, that's impressive! Most of the capacitor manufactures offer lifetime test results as well as enhanced SPICE and/or S-parameter models of their capacitors. C-D, Murata, Panasonic, TDK, etc. are among the ones I researched trying to design broadband (DC to ~10 GHz) supply decoupling networks, but of course the final selection also had to consider size and especially cost. When I first started my last job, they had a decoupling issue, so naturally I suggested a great cap I had used for advanced mil-spec R&D designs. The purchasing guys had heart attacks when they found out the cost, and I had to come up with a cost-effective design that had more parts but was much cheaper.

 

[Blenderguy]

When I get a chance I will measure the 60 Hz hum and replace the decoupl;ing caps and remeasure.

Usually I add another film capacitor across the big electrolytic. I refrained this time, because the wire from there to the amplifier board is quite long. And since I was rushed just to get it working again, I did not want to create an instability. This amp has 3MHz small signal bandwirdth, probably because of so few transistors cascaded. The more transistors in cascade the lower the open loop bandwidth, assuming the same trnasisitors are used.

The purchasing guys had heart attacks when they found out the cost, and I had to come up with a cost-effective design that had more parts but was much cheaper.

The replacement power supply caps cost $69 each. To me that is not a heart attack price, because larger filter capacitors were always expensive, even going back to the 1970s, but others may.