I want the very low distortion of a linear amp combined with efficiency.
And, like a lot of people, I come up with a linear amp driven by a switcher.
Technics seem to have one of the earliest practical patents, then Carver, Yamaha had a variant, lately LabGruppen have been a preferred choice in professional use, there are many others.
This thread is to unify a few ideas I have posted in Power-Supplies State of the Art for Tracker/Down Converter?, Class-D, and Damir's thread 200W Class A amp with high efficency in this forum.
As mentioned in the Power-Supply thread, the proposal is one efficient SMPS for an entire set of home theater bi- and tri-amped speakers, then each individual amp to have its own down converter.
The down converters can incorporate short circuit, thermal and speaker DC protection.
It would be neat to make the converters as an add-on to the power amp boards.
Then the combined module could be a direct replacement for a conventional linear amp, but with much lower power consumption.
Is anybody interested in such a system?
David
I eventually arrived at a simple circuit that looks to have promise, latest version is here Very Efficient Very Low Distortion Linear+ Amp
Assumption is that the amp front end will switch off the SMPS in the event of short circuit, over temp or DC fault.
This simplifies the the rail trackers.
The hi side driver power supply can be an isolated DC-DC converter powered from the lo volt supplies, one per amp.
This works with common SMPS supplies that have auxiliary lo volt +- 12 V (or close) in addition to the main supply.
Or complementary silicon FETs can be used and power drawn from the hi supplies.
And, like a lot of people, I come up with a linear amp driven by a switcher.
Technics seem to have one of the earliest practical patents, then Carver, Yamaha had a variant, lately LabGruppen have been a preferred choice in professional use, there are many others.
This thread is to unify a few ideas I have posted in Power-Supplies State of the Art for Tracker/Down Converter?, Class-D, and Damir's thread 200W Class A amp with high efficency in this forum.
As mentioned in the Power-Supply thread, the proposal is one efficient SMPS for an entire set of home theater bi- and tri-amped speakers, then each individual amp to have its own down converter.
The down converters can incorporate short circuit, thermal and speaker DC protection.
It would be neat to make the converters as an add-on to the power amp boards.
Then the combined module could be a direct replacement for a conventional linear amp, but with much lower power consumption.
Is anybody interested in such a system?
David
I eventually arrived at a simple circuit that looks to have promise, latest version is here Very Efficient Very Low Distortion Linear+ Amp
Assumption is that the amp front end will switch off the SMPS in the event of short circuit, over temp or DC fault.
This simplifies the the rail trackers.
The hi side driver power supply can be an isolated DC-DC converter powered from the lo volt supplies, one per amp.
This works with common SMPS supplies that have auxiliary lo volt +- 12 V (or close) in addition to the main supply.
Or complementary silicon FETs can be used and power drawn from the hi supplies.
Last edited:
Dear David,
IMHO it would be more efficient to go directly to Class-D and try to optimize the available designs. Here at diyaudio there are very good Class D implementations and for example I am using a low distortion design from diyaudio member Chocoholic (SystemD LiteAmp; using low distortion postfilter feedback; schematic at post #223).
Combining SMPS, active crossover, Class-D for Bass, small Blameless for mid/high would be also a high efficient solution with best distortion results.
E.g.: I would use Hypex Audio grade SMPS (400W or 1200W series) at acceptable prices which are able to drive full power at 20Hz.
BR, Toni
IMHO it would be more efficient to go directly to Class-D and try to optimize the available designs. Here at diyaudio there are very good Class D implementations and for example I am using a low distortion design from diyaudio member Chocoholic (SystemD LiteAmp; using low distortion postfilter feedback; schematic at post #223).
Combining SMPS, active crossover, Class-D for Bass, small Blameless for mid/high would be also a high efficient solution with best distortion results.
E.g.: I would use Hypex Audio grade SMPS (400W or 1200W series) at acceptable prices which are able to drive full power at 20Hz.
BR, Toni
Hi Toni
The class D is probably more than acceptable.
But somehow I still want a really low distortion amp, just like your own 200 W linear unit
Only more efficient.Any class-D amp is constrained to have the Unity Gain Crossover Frequency less than half the switch frequency and this seriously limits the feedback and distortion reduction.
The Ncore stuff uses extraordinary efforts to keep the feedback up but the resultant 5th order loop is inevitably only conditionally stable.
This then needs even more efforts to prevent problems when it overloads or clips and for start-up and shut down.
That much R&D investment makes sense for a mass produced product, not for me.
Another sensible solution, but I would prefer to have even the mid amplifiers efficient.
My OPS circuit comes in increments of 100 W and I planned on 2 modules.
(I have JBL compression driver mids and a 200 W amp would be a suitable match, even if not really needed.)
But yes, definitely active crossover, probably SMPS,
The down converter scheme also make possible a neat integration, I want a switch FET for short circuit and speaker DC protection so it's hardly any extra to make it a down converter.
One possibility is to have the down converters as add-ons, after I have the amps built.
Yes, the Hypex are pretty impressive, not such an acceptable price once they are shipped to Australia however.
It's partially the cost of power supplies that has driven me to study this path.
A pile of toroid transformers and 10,000 uF capacitors is not cheap, I can probably save a lot from smaller transformers and capacitors and have better efficiency as a bonus.
Are you familiar with SMPS?
There are circuits to Zero Volt and/or Zero Current Switch the power supply.
This improves both efficiency and reduces EMI.
Seems clever but I don't know the details yet, or whether Hypex already do this.
Best wishes
David
Dear David,
hypex NCore agree!
You could use something similar to the SA2016 with 3 pairs of EXC20N20/P20. Capable to deliver 200W@8R or 400W@4R.
In post #1765 there is a table showing low THD even at low bias. The table was derived from a 3 pair EXC10N20/P20 amp running 70W@8R. IMHO: running 3 pair double die EXC20N20/P20 with low bias total below 120mA (30 - 40 mA per double die device) would result in a amplifier running < 0.01%THD+N at 20kHz@200W@8R.
The Hypex 1200W SMPS is running in my 4 channel SA2015 IRFP240/9240 amp (100W@8R).
No "Hum" (some artefacts are above 100kHz around SMPS switching frequency so you get some lower SNR if we measure full bandwidth, but best results with 80kHz bw during SNR measurement) and the weight and size of the amplifier is extremely reduced compared to a conventional amplifier. Of course actively cooled with 5 ultra silent 60mm diam. fan's.
About efficency of the Hypex power supply there is no data available or I haven't found it. EMI should be low.
For the 3 channel active crossover variant the mid and/or treble amp doesn't need more as 50W output power. EXICON or IXYS powered small amp modules running at lower supply voltage would do the job...
BR, Toni
hypex NCore agree!
You could use something similar to the SA2016 with 3 pairs of EXC20N20/P20. Capable to deliver 200W@8R or 400W@4R.
In post #1765 there is a table showing low THD even at low bias. The table was derived from a 3 pair EXC10N20/P20 amp running 70W@8R. IMHO: running 3 pair double die EXC20N20/P20 with low bias total below 120mA (30 - 40 mA per double die device) would result in a amplifier running < 0.01%THD+N at 20kHz@200W@8R.
The Hypex 1200W SMPS is running in my 4 channel SA2015 IRFP240/9240 amp (100W@8R).
No "Hum" (some artefacts are above 100kHz around SMPS switching frequency so you get some lower SNR if we measure full bandwidth, but best results with 80kHz bw during SNR measurement) and the weight and size of the amplifier is extremely reduced compared to a conventional amplifier. Of course actively cooled with 5 ultra silent 60mm diam. fan's.
About efficency of the Hypex power supply there is no data available or I haven't found it. EMI should be low.
For the 3 channel active crossover variant the mid and/or treble amp doesn't need more as 50W output power. EXICON or IXYS powered small amp modules running at lower supply voltage would do the job...
BR, Toni
Last edited:
I have some of my own that I want to try, that's part of the motivation to use the linear plus switcher rather than just a class-D.
I did my own amps because I have very efficient horn loaded mids @ 115 dB/W so very low noise is an issue.
My solutions are not revolutionary, low noise input transistors, paralleled almost like an MC phono circuit, low value emitter resistors in the inputs, low value feedback resistors.
It simulates very nicely, now I want to build it.
Agree that they are extraordinary or agree that they are too complicated to want to DIY?
I expect you have examined the internals, any comments?
I am keen for information because quite a bit of the SMPS literature seems rather poorly written.
I would really appreciate a Bob-Cordell-quality primer.
Until then I have to work it out myself, I start to understand feedback analysis of switch mode circuits, have you looked at this?
Best wishes
David
Dear David,
both: agree that they are extraordinary and agree that they are maybe too complicated for DIY. Chocoholic is one of the experts to ask about class-D and NCore questions. He has created his own LTSpice simulation to design the class-D feedback loop using a IRS2092. You should have a look at his LiteAmp thread if not already done.
SMPS: I have only limited knowledge about designing switch mode power supplies and would therefore not build myself a 230V powered SMPS. Too much problems to solve e.g. EMI, safety ...
Have fun, Toni
both: agree that they are extraordinary and agree that they are maybe too complicated for DIY. Chocoholic is one of the experts to ask about class-D and NCore questions. He has created his own LTSpice simulation to design the class-D feedback loop using a IRS2092. You should have a look at his LiteAmp thread if not already done.
SMPS: I have only limited knowledge about designing switch mode power supplies and would therefore not build myself a 230V powered SMPS. Too much problems to solve e.g. EMI, safety ...
Have fun, Toni
This is a matter of system design as much as anything else. In order to pull this off, balanced interconnects will be a necessity. Not a bad idea anyway, but still unbalanced won't be an option. (The times when you could design a power amp as IEC Class I and assume that'll be the only such device in the system are long gone. These days I'd hunt down a mains xfmr with shield winding if I really needed unbalanced inputs - alternatively, implement them with balanced receivers, Bruno Putzeys style. I'd rather outfit the preamp with impedance-balanced outputs though, the circuitry is trivial. You just need to replicate output impedance between output ground and "cold". Shield is probably best left disconnected at this end unless you want to implement a ground lift switch.)
I get the motivation for going SMPS, but it's not trivial business. You'll need a PFC circuit (preferably active for efficiency reasons), for one thing. There also is the question of how much there is to be gained in efficiency in the first place. Transformer supplies by themselves can be pretty good (as in 80+%) when quality parts are being used. (There are a handful of high-efficiency transformer design enthusiasts out there, actually. Not sure in terms of commercial availability though.) They're also simple and robust, easy to clean up in EMI, easy to repair, and can be designed to outlast you. In the world of DIY, KISS is generally preferred to a SPOF. There's more than a few commercial Class D amps with linear supplies, too.
It's not like you can get away entirely without any big(-ish) caps anyway - the mains tends not to be a particularly good place to dump high-frequency currents.
If you really want to do it, sure, go ahead, but the really important part will be the DC/DC converters for the modulated supplies. These make the biggest difference in efficiency, that's where I'd focus my efforts.
There also is the question of whether you want to exploit the degrees of freedom that having DC/DC converters affords - suppose you can get, say, a very nice off-the-shelf +48V SMPS with high efficiency and all that jazz. (48 V isn't an uncommon voltage e.g. in telecoms if memory serves.) You could even run the amps on a stack of lead acid / gel cell batteries if the power ever goes out. Even +12V may be an option - while it's a bit of a pain to distribute due to current levels, you can get very nice PC power supplies (modern topologies are primarily +12V with step-downs anyway, and you'd get voltages for other goodies to boot). An off-the-shelf supply means the heavy lifting has already been done for you (electrical safety, EMI, ...), and economics of scale bring down the price. (That's precisely why a lot of equipment comes with wall warts.) Just stuff to think about.
I get the motivation for going SMPS, but it's not trivial business. You'll need a PFC circuit (preferably active for efficiency reasons), for one thing. There also is the question of how much there is to be gained in efficiency in the first place. Transformer supplies by themselves can be pretty good (as in 80+%) when quality parts are being used. (There are a handful of high-efficiency transformer design enthusiasts out there, actually. Not sure in terms of commercial availability though.) They're also simple and robust, easy to clean up in EMI, easy to repair, and can be designed to outlast you. In the world of DIY, KISS is generally preferred to a SPOF. There's more than a few commercial Class D amps with linear supplies, too.
It's not like you can get away entirely without any big(-ish) caps anyway - the mains tends not to be a particularly good place to dump high-frequency currents.
If you really want to do it, sure, go ahead, but the really important part will be the DC/DC converters for the modulated supplies. These make the biggest difference in efficiency, that's where I'd focus my efforts.
There also is the question of whether you want to exploit the degrees of freedom that having DC/DC converters affords - suppose you can get, say, a very nice off-the-shelf +48V SMPS with high efficiency and all that jazz. (48 V isn't an uncommon voltage e.g. in telecoms if memory serves.) You could even run the amps on a stack of lead acid / gel cell batteries if the power ever goes out. Even +12V may be an option - while it's a bit of a pain to distribute due to current levels, you can get very nice PC power supplies (modern topologies are primarily +12V with step-downs anyway, and you'd get voltages for other goodies to boot). An off-the-shelf supply means the heavy lifting has already been done for you (electrical safety, EMI, ...), and economics of scale bring down the price. (That's precisely why a lot of equipment comes with wall warts.) Just stuff to think about.
Last edited:
Hi Stephan
Nice to have a new contributor and thank you for the detailed comments.
Yes, I planned balanced anyway, and to specify a shield for the transformer if I use a 50 Hz supply.
Oh no! I am not yet committed either way, for just that reason.
There are several issues here and you have considered many of them.
A SMPS is not in mind for efficiency reasons, as you say a linear supply can be fine. It's to avoid expensive transformers and capacitors.
This is the key!
I have 3 phase power, I could 3 phase rectify and dispense with "bulk" capacitors, or substantially reduce them.
Then a small hi-frequency transformer and secondary capacitors.
Probably need active PFC to do this properly, not sure the details for 3 phase.
Yes.
There are definite incentives to have a supply from 80 V to ~100 V and use down (buck) DC-DC converters,
Partly it's to keep the currents low, of course.
But there is a more subtle reason, and that is the poorer response of up-converters - they have an RHP zero that makes feedback much harder to do accurately.
AFAIK this is inherent in buck-boost and boost converters so I don't plan to use them.
If that eliminates a second-hand telecom supply then so be it.
I can always fall back on a simple, old fashioned transformer supply because <100 V is a very ordinary requirement.
Best wishes
David
Nice to have a new contributor and thank you for the detailed comments.
Yes, I planned balanced anyway, and to specify a shield for the transformer if I use a 50 Hz supply.
Oh no! I am not yet committed either way, for just that reason.
There are several issues here and you have considered many of them.
A SMPS is not in mind for efficiency reasons, as you say a linear supply can be fine. It's to avoid expensive transformers and capacitors.
This is the key!
I have 3 phase power, I could 3 phase rectify and dispense with "bulk" capacitors, or substantially reduce them.
Then a small hi-frequency transformer and secondary capacitors.
Probably need active PFC to do this properly, not sure the details for 3 phase.
Yes.
There are definite incentives to have a supply from 80 V to ~100 V and use down (buck) DC-DC converters,
Partly it's to keep the currents low, of course.
But there is a more subtle reason, and that is the poorer response of up-converters - they have an RHP zero that makes feedback much harder to do accurately.
AFAIK this is inherent in buck-boost and boost converters so I don't plan to use them.
If that eliminates a second-hand telecom supply then so be it.
I can always fall back on a simple, old fashioned transformer supply because <100 V is a very ordinary requirement.
Best wishes
David
Last edited:
Hi, Dave!
Please, clarify your system components and design goals...
What sensitivity your speakers already achieve, what desired level will be preferred, what available space to jam, music style, how much freq bands are planned?
And so on...
Let’s start from system architect, not from amplifier efficiency...
Please, clarify your system components and design goals...
What sensitivity your speakers already achieve, what desired level will be preferred, what available space to jam, music style, how much freq bands are planned?
And so on...
Let’s start from system architect, not from amplifier efficiency...
Hi Pav
I appreciate that sometimes it's best to clarify overall context.
But if I want to build a 100 W amplifier, for example, I don't really need to know if it's for classical or pop music.
And very often such discussions then become sidetracked as people post their opinions of the choice of speakers, choice of music, size of the room, and so on.
So I don't plan to discuss this too much but will provide a summary.
It's for a system that does both music and home theatre so there's Front, Centre and Side, Rears are planned but not started yet.
All JBL drivers, mostly cinema components.
Subwoofers are two 18" 2245
L,C,R are 3 way
Woofers 15" 2226
Mids are 4" VC 1.5" throat 2453 Horn loaded
Tweeter are slot 2405
SL,SR are 2 way
Woofers are 12" 2214
Mid/Tweeters are 3" VC 1.5" throat 2432 Horn loaded
Plan is to do an active system, so >dozen amps to build.
Amps are 600 W, 300 W, 200 W and 100 W because I have a modular OPS in ~100 W increments.
I have a stock of Thermal Trak OP transistors so a smaller OP module would not really save money.
It's overkill for my current room, about 8 m deep, 3.5 m tall, but I won't ever have to even think about headroom.
Best wishes
David
Last edited:
I would call my proposal essentially class-H, the term is used a bit inconsistently in the literature.
It's just a particular implementation that I think should work well.
Best wishes
David
I know that feeling.
I am an amplifier designer so my interest and contribution is more likely useful in the way you handle the interfacing of your linear amp with the buck converters. So I look forward to when you get around to this part.
Of course we’ll go another way.
Ok, so that’s a relatively Hi-Sensitive system, based on ‘pro drivers, not ‘resin consumer brothers.
Ok.
Let me suppose, that all necessary XO/Delays/Parametric will be properly tuned with measurement mic and we could contrate on amplifier section?
So, amp.
First question - what will be signal source and where volume regulation will be placed?
1. VR in digital domain, so some DAC dynamic range will be thrashed.
2. VR inside XO module, so some influence from low-level interconnection
3. VR inside amp module, so you’ll need very linear and high-CMRR Bal-Unbal converter.
Second question - voltage and power headroom.
1. Your mains module anyway will pick best active PFC practices.
Since your amps will be linear and produce soma idle loading worst case of unloaded SMPS will be excluded. I’ll prefer active PFC and flyback or half-bridge.
2. Your SMPS module will produce known DC level rails. I’ll recommend to have a deep think about. Since your linear amps will be “isolated” from relatively high rails by somekind of DC-DC conversion there are no headashe in rails headroom. Another words - i’ll prefer to choose relatively high rails, say ~+-100 and even 150 Volts.
3. Keep in mind music peak factor around 3, so your amp will long-run around 1/10 of peak levels. All supply section are better to be developed based on sustained power, while peaks will be maintained by high-side capacitance.
4. Let’s design 1, 2 and 4 ThermalTrak pair modules (accordingly high/mid, bass and sub), output power doesn’t highly matter. The main thing is to keep safe phase margin at high-load currents. Bipolars have huge fT/Ic dependance. OPS will be TripleEF? There are two main ways to stabilize it.
Last but not least:
Construction details?
Flat heatsink-paraller, Cubed or linear-straight?
Why not simple slightly tune LabGruppen PCBs?
Main parts of performance:
First half of - PSRR of the linear part itself
Second half of - feedback depth at supply DC-DC working frequency.
First could be designed mostly good, but second is a bit tricky. The higher freq - the more gate charge losses and radiation with lower NFB compensation, the lower freq - the lower conversion efficiency and higher straight signal injection.
This is ultimately going to be a 4 way system, correct? If so there is nothing that says all 4 bands need to use the same topology. Class D or 3- or 4-step class H are certainly viable for woofers or subs, as is the originally intended TD. For mids and tweets I would probably go simpler - converting my favorite ultra-low-distortion class AB to a simple 2-step class H or G. At only 100-200 watts it would be efficient enough, avoiding all high frequency switching artifacts, and pretty darn foolproof. I’ve built a 100 watt per channel class G using two pair of TIP35/6 per channel using two little 20-0-20 toroids in series to power it. Needed very little heat sink, and sound industinguishable from the original host design. And it worked as designed from the very first power up. Try getting that kind of first pass success with any sort of a switchmode circuit.
Yes
Amps will have gain appropriate to their role, for example the mid amp will be low gain to reduce noise.
Second question... Your mains module anyway will pick best active PFC practices.
...worst case of unloaded SMPS will be excluded. I’ll prefer active PFC and flyback...
Active PFC is not yet decided, not as clear cut as I expected, issues of EMI and whether I can use 3 phase mains to avoid the need.
Yes, this does indeed need a deep think.
As you say, the quiescent bias is actually useful to keep some load and stabilize the down converters.
I have not yet worked out the details, I have some articles on LLC SMPS to study, should be similar.
Actually 1, 2, 3 and 6.
Actually EF2, easier to stabilize.
This is one area where I feel pretty happy, lots of simulation done with Tian probes to study Return Ratio as current and Vce varies.
Layout I also have under control, it's the buck converter I want to study
Where's the fun in that?
This stuff advances fast and I have some ideas that are new, AFAIK, and should work even better.
Thank you Jan.
That would be still not quite noiseless with efficient horn loaded mids but more or less what I intend to do.
True, but then I need to research and simulate multiple different circuits.
I prefer a modular approach where they are all essentially similar.
Best wishes
David
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.