I have completed the assembly and some testing of a pair of Salas Shunt Regs for the LME end of the amplifier.
Add 15Vac to the main 40Vac to get ~55Vac fed into the Salas Reg. Input voltage range ~71Vdc to 81Vdc. Output variable from 58Vdc to 65Vdc at upto 70mA.
Add 15Vac to the main 40Vac to get ~55Vac fed into the Salas Reg. Input voltage range ~71Vdc to 81Vdc. Output variable from 58Vdc to 65Vdc at upto 70mA.
but why? with current draw so constant and with PSRR being so large, it just seems like a waste to me, adding heat to the case for no real benefit
Last edited:
do we know psrr at 100kHz?
do we know the lme current is nearly constant?
I know that all ICs have a pssr that specfies fantastically @ around -80 to -110dB, but when one looks at the 100kHz psrr for the -ve supply it is often down around -40dB.
I know that the current draw of single ended ClassA amplifiers is not nearly constant. It varies just like the output current varies.
I know that there are capacitive loads on the LME output that will draw big currents if fast slew rates are demanded.
do we know the lme current is nearly constant?
I know that all ICs have a pssr that specfies fantastically @ around -80 to -110dB, but when one looks at the 100kHz psrr for the -ve supply it is often down around -40dB.
I know that the current draw of single ended ClassA amplifiers is not nearly constant. It varies just like the output current varies.
I know that there are capacitive loads on the LME output that will draw big currents if fast slew rates are demanded.
I cant see any of that requiring a shunt with the amount of decoupling onboard. is there anything in the measurements of the actual amplifier that makes you think the LME needs a better PSU? what sort of sustained power drain are you expecting at 100khz that the caps cant handle? what sort of drain at 100khz would you expect a remote shunt reg to supply more effectively than the local decoupling caps would?
Apparently electrolytic caps have a higher ESR at lower audio frequencies than a regulator. Low inpedance caps are designed to work best at, eg, 100kHz. Most other caps at 100 to 120Hz. Not at all audio frequencies. Their ESR apparently rises meteorically at lower frequencies.
Either way, on a Hafler amp years ago I used 7815 and 7915 to regulate the early stage including the pre-driver but not including the driver as that's the OE configuration in that respect. It seems normal that designers don't include the driver with the early stage making this LME chip different as it does have those built in. Had the original 100uF after each regulator on the amplifier board and fitted four 330uf 100V caps, or 470uF, etc, before the regs, after the bridge. Also very little current draw into that early stage. All caps were run for three months 24/7 before testing. I could tell the difference each make of pre-reg caps had, and even combinations of caps. I don't know what the PSRR is of that discrete circuit, but if such effects were so easily audible the regs couldn't have been doing all that much apart from stopping 100Hz. Maybe this LME chip is less susceptible but I think that whatever the measurements show, it'll take some listening tests on a revealing system to know for sure. In that Hafler test, it was the caps claimed as designed for audio that were best, IMO.
Graph of a wall wart SMPS vs 317 vs teddy reg: TeddyPardo
Teddy Pardo and others have said that it appears evident that very high frequency noise has an effect on the audio band.
We're striving for the highest quality sound we can get so it's worth having a go with some of these things as past experience says sometimes they're efficacious.
Either way, on a Hafler amp years ago I used 7815 and 7915 to regulate the early stage including the pre-driver but not including the driver as that's the OE configuration in that respect. It seems normal that designers don't include the driver with the early stage making this LME chip different as it does have those built in. Had the original 100uF after each regulator on the amplifier board and fitted four 330uf 100V caps, or 470uF, etc, before the regs, after the bridge. Also very little current draw into that early stage. All caps were run for three months 24/7 before testing. I could tell the difference each make of pre-reg caps had, and even combinations of caps. I don't know what the PSRR is of that discrete circuit, but if such effects were so easily audible the regs couldn't have been doing all that much apart from stopping 100Hz. Maybe this LME chip is less susceptible but I think that whatever the measurements show, it'll take some listening tests on a revealing system to know for sure. In that Hafler test, it was the caps claimed as designed for audio that were best, IMO.
Graph of a wall wart SMPS vs 317 vs teddy reg: TeddyPardo
Teddy Pardo and others have said that it appears evident that very high frequency noise has an effect on the audio band.
We're striving for the highest quality sound we can get so it's worth having a go with some of these things as past experience says sometimes they're efficacious.
Last edited:
the teddy reg is unlikely to work at this voltage, few solid state solutions do. the LME has already been stated in the DS and by opc to be a pretty benign load, so something like a shunt seems odd to me is all.
I'm just trying to get to the bottom of the reasoning thats all, nothing against people trying things out and Andrew would most definitely do the same for me, so I would hope there is no ill feelings there. a remote shunt reg transient response for high frequency versus the large caps for any likely extended transient, plus the ceramic caps for any HF transient already seems overkill for what is stated as a pretty much constant current draw.
the teddy is rated for 3-35V, I havent looked at the schematic, so perhaps it can be used floating if the SOA as far as voltage drop is respected, but on the surface it wouldnt seem a prudent choice, with the shunt being better, given it can be made to work at quite high voltages with the right parts onboard.
I'm just trying to get to the bottom of the reasoning thats all, nothing against people trying things out and Andrew would most definitely do the same for me, so I would hope there is no ill feelings there. a remote shunt reg transient response for high frequency versus the large caps for any likely extended transient, plus the ceramic caps for any HF transient already seems overkill for what is stated as a pretty much constant current draw.
the teddy is rated for 3-35V, I havent looked at the schematic, so perhaps it can be used floating if the SOA as far as voltage drop is respected, but on the surface it wouldnt seem a prudent choice, with the shunt being better, given it can be made to work at quite high voltages with the right parts onboard.
Sorry i was a bit unclear again. The picture it not of my sink, it was to show the type of fins only (might have been a bad idea that caused more confusion).
I couldn't measure the sinks directly since i need to keep them in the package for a while longer to avoid scratches and other stuff since i don't have the space right now. The sinks to the amp looked just like the one measured except for the wavy vs straight fins which is why i thought it was reasonable to measure it instead. It was sealed in very thin plastic so i could see the fins directly but i can't open it to measure and seal it again...
The sinks are 40 mm in total hight which leaves about 32 mm for the fin hight, so they are a bit smaller than the conrad sinks but not much.
I haven't seen their chassis before so i can't say about that quality change but the picture on the site shows straight fins for the chassis so there is definitely some change in the process.
The heatsink measured (300x120x40) is specified at 0.4 C/W. The sinks i use for the chassis measure 400x120x40 and is specified at 0.25 C/W according to the website (if that's correct is another problem). So the sinks for the chassis should be better but exactly how they perform in C/W i will not comment about.
Edit: The heatsink measured also came from modushop in the same shipment and should be used for the 300 mm depth dissipante chassis. So the ones for my 400 mm depth chassis should be equivalent and that seems to be the case (by looking at the sizes).
Last edited:
I'll ask the reverse question.
Why adopt a regulated smps with low noise output for this amplifier project?
I'll try and answer:
Probably because the results posted so far show that the amplifier performance changes when fed with clean current supplies.
I'll ask another question.
Did the improved performance come from the clean current supplied to the LME, or did it come from the clean current supplied to the output stage, or did the changed performance come from a bit of both, i.e. clean current supplies for the LME and the output stage? That then prompts my final question: What proportion of the changed performance came from each part of the cleaned up current supplies, the +ve LME, or the -ve LME, or the +ve output stage, or the -ve output stage.
I don't have any of those answers.
I do have a few answers to some of my earlier questions:
the LME has a stated 105dB for typical psrr. The graphs for +ve and -ve rail psrr show approximately -135dB to -50dB for typical. These same graphs also show that the -ve rail psrr is slightly worse than the +ve rail.
If the LME can output ~40mApk then that peak current whether continuous or sinewave or transient will come from ONE supply rail. The two supply rails will alternate supply the half-wave output currents.
The current draw of the LME is not constant. It behaves (for current draw) just like any other ClassA input and output stages.
Why adopt a regulated smps with low noise output for this amplifier project?
I'll try and answer:
Probably because the results posted so far show that the amplifier performance changes when fed with clean current supplies.
I'll ask another question.
Did the improved performance come from the clean current supplied to the LME, or did it come from the clean current supplied to the output stage, or did the changed performance come from a bit of both, i.e. clean current supplies for the LME and the output stage? That then prompts my final question: What proportion of the changed performance came from each part of the cleaned up current supplies, the +ve LME, or the -ve LME, or the +ve output stage, or the -ve output stage.
I don't have any of those answers.
I do have a few answers to some of my earlier questions:
the LME has a stated 105dB for typical psrr. The graphs for +ve and -ve rail psrr show approximately -135dB to -50dB for typical. These same graphs also show that the -ve rail psrr is slightly worse than the +ve rail.
If the LME can output ~40mApk then that peak current whether continuous or sinewave or transient will come from ONE supply rail. The two supply rails will alternate supply the half-wave output currents.
The current draw of the LME is not constant. It behaves (for current draw) just like any other ClassA input and output stages.
Last edited:
Mainly because of the clean regulated supply to the output stage, the LME is already fed with a regulated supply in the stock linear format (at least for me, supplying unregulated to both was never on the cards) so there is little/no benefit for the LME using the dps600.
the output stage has much lower PSRR and presents a far more dynamic and transient power demand; thus benefits much more from having a stiff, low noise supply.
OK, perhaps you should email Ti and tell them the datasheet is wrong =) the main point i'm making is that I do not see how this transient loading will be handled by the shunt reg and not the decoupling caps with any expected loading. Secondly i'm wondering where, in the measurements of the actual amplifier being tortured under the microscope with full power transients, at 100hz, 1khz and 10khz, you can see room for improvement that a shunt reg will provide? the DPS600 VAS regulator is no different from the stock lm317, so you can expect similar performance there. even your worst case PSRR coupled with the PSRR of the linear PSU is pretty safely inaudible
I support you to experiments though, dont get me wrong, i'll be watching to see how it goes, as I have spare shunt regs I can bring into play. do you have the equipment to tell whether you have made an improvement?
Andrew, which version of the salas regs have you built? "salas shunt reg" is a touch ambiguous ...
My guess is sshv1.2
My guess is sshv1.2
You're right, well spotted, the one I linked to and most of them don't stretch to more than 35 Volts due to the capacitor choice and I think mainly the low voltage low noise jfet. But he did do a workaround employing a bypass transistor for higher voltages. I don't have a link to that. There are at least 5 threads on Pink Fish Media and many pages on his site perhaps other places and I haven't yet collated all the gems from it all. I seem to recall reading from the threads that they found this 'fetlington' sounded better than whatever shut regs they'd tried, even though it may not regulate as well. The Teddy reg seems to have been, as much as his own work, partly as he owned the expensive wave testers, an open source community development circuit where various people had fundamental input to the development and resultant end circuit.
Personally I'll be happier if the 317 /337 prove to be not audibly beatable as they're a cheap fairly easy solution.
Just in case anyone was wondering how I did it with a 7815, I simply added two Zeners between the earth pin and ground to effectively lift the earth at the 7815 to about 60 volts. It's a bit messy as the Zeners shift a lot with temperature and I measured the voltage drop across each one once all fully running, then took them all out and swapped them about to get the most even voltages. There is also a 39k Ohm resistor from pre reg power rail to 7815 earth pin so as to supply some current to get the Zeners working.
Last edited:
Hi,
I have a problem on some components for the circuit of the LM317. this circuit was expected to 25V on original version, so I mounted capacitors smd-805. I tried to replace those 4 capacitors with discrete 2,2uF100V 5x6mm (considering the voltage lme) aspect is not good. What we decide now? SMPS are ready (22). Perhaps you have all the PCB regulators opc? I continue with discrete components? please, reply fast
alternative is assemble smd-1206 1uF100v, buy today and have in two day.
I have a problem on some components for the circuit of the LM317. this circuit was expected to 25V on original version, so I mounted capacitors smd-805. I tried to replace those 4 capacitors with discrete 2,2uF100V 5x6mm (considering the voltage lme) aspect is not good. What we decide now? SMPS are ready (22). Perhaps you have all the PCB regulators opc? I continue with discrete components? please, reply fast
alternative is assemble smd-1206 1uF100v, buy today and have in two day.
Last edited:
Quanghao PCBs using SSR V1.1 with three work arounds to get >>35Vdc. But I needed a fourth work around to solve the capacitor blowing the lsk170 on a shorted output. That was not a problem with normal LV operation, but at 65V there is plenty of energy available.
See Salas Regs for details soon.
if you can test mine with 2.2uf 5 x 6mm, I will replace with 1206 or 1210 (I have many times successfully and neatly replaced 0805 with 1206. which position in the standard regulator circuit is this? reference bypass? output bypass? alternatively if it really only means a couple of days, I guess we can wait, but this sounds more like another week delay
Classic diagram lm317,then use one in bypass(ref) and two (one input,one output). input of regulator have LC filter from.
the heat sink, preventing the installation of capacitors, and I'm thinking for others is not easy. proceed with the 1206 ,is better that smps is ready for use.
I've already bought on Distrelec now.
the heat sink, preventing the installation of capacitors, and I'm thinking for others is not easy. proceed with the 1206 ,is better that smps is ready for use.
I've already bought on Distrelec now.
78xx series have very poor internal scheme, it is good only for fix current output. (without modifications)