Today I tried to pot one prototype power transformer to test what the result will be... looks very promissing so far.
I wanted to save some potting volume, so needed to incorporate some "filler". First idea was to 3d print something, but then I realised its probably not a good idea to put in something hard as that would not fight resonances. Eventually I have found a rubber mat in hobby market which is allegedly used under dishwashers or washing machines to dampen vibrations and can withstand 100°C. It truly worked well when i tested it with transformer sitting on it. I cut a small squares as transformer distance feets to support transformer in cans. I also cut a stripes which I fitted in bobbin cutouts to hold them in place. I didn't wanted to add too much to not spoil polyurethane holding power to support transformer but still enough to save some potting volume. As a results I was able to save about 200ml per can, which is not a lot but at least something 😀 but was hoping for more .... whatever... here I tried to take a picture of that rubber:
And potting... one can required ~2 litres of polyurethane = 3kg... below a few pictures:
And final result below... after it will harden (I have used Shore 55D poly) I will try to fill also side walls under angle to top (will send pics later) so transformer weight is better supported and as a whole better dampened.
note: you can see some mini bubbles in pictures, as curing time is very long (this is not epoxy...) they disappear after time, and even if some would stay it is only good, the more "foamy" the better because I want it strong to support weight but soft to dampen vibrations.
(The transformer is already vacuum potted in lacquer from factory)
Final results not yet, as it will keep curing for 1-2 days but from what I tested already looks like it helped a bit and is a bit quieter and dampened.
I also took a short clip of how the process looks like 🙂
I wanted to save some potting volume, so needed to incorporate some "filler". First idea was to 3d print something, but then I realised its probably not a good idea to put in something hard as that would not fight resonances. Eventually I have found a rubber mat in hobby market which is allegedly used under dishwashers or washing machines to dampen vibrations and can withstand 100°C. It truly worked well when i tested it with transformer sitting on it. I cut a small squares as transformer distance feets to support transformer in cans. I also cut a stripes which I fitted in bobbin cutouts to hold them in place. I didn't wanted to add too much to not spoil polyurethane holding power to support transformer but still enough to save some potting volume. As a results I was able to save about 200ml per can, which is not a lot but at least something 😀 but was hoping for more .... whatever... here I tried to take a picture of that rubber:
And potting... one can required ~2 litres of polyurethane = 3kg... below a few pictures:
And final result below... after it will harden (I have used Shore 55D poly) I will try to fill also side walls under angle to top (will send pics later) so transformer weight is better supported and as a whole better dampened.
note: you can see some mini bubbles in pictures, as curing time is very long (this is not epoxy...) they disappear after time, and even if some would stay it is only good, the more "foamy" the better because I want it strong to support weight but soft to dampen vibrations.
(The transformer is already vacuum potted in lacquer from factory)
Final results not yet, as it will keep curing for 1-2 days but from what I tested already looks like it helped a bit and is a bit quieter and dampened.
I also took a short clip of how the process looks like 🙂
Last edited:
To match potting angle on both sides I have printed simple 45° stand to support the can during process
Meanwhile I cut remaining rubber mat and made box looking fillers to save volume for output transformers potting as these are way "smaller". Doesn't look like that, but these save about 3 litres = 4,5kg of potting material!
I am not going to pot output transformers yet however. I need to finish power ones first. After that I match each power tranny to output tranny separately for lowest induced noise, I will describe this process in detail once I get to it...
More potting... and I am out of resin so need to order more 😀
I have potted also one output transformer as well to move forward so I can finish at least one monoblock. Potting of output transformer is a bit tricky to get the lowest induced hum for best snr. I mount already potted power transformer can together with output can with internal braces. I always match output to power and mark the spot to know the matched position of cans:
Then I turn on power transformer (does not need to be loaded) and connect output transformer secondary winding (in this case whole transformer winding) to headphones and I listen and try what is the best position. The key is to get the center of windings (power vs output) under 90° angle as well as maintain equal distance against all metal walls. When I get an idea where its best I connect it to soundcard mic input and watch spectrum to get the lowest 50hz possible:
There is always about 1° window where its the best. Then I validate result with headphones again and finally connect some good bass speaker box to it to see what it does. And amazingly it is absolutely as quiet as it would be turned off 🙂 and this is unloaded transformer winding on its own. Loaded with low output impedance of amps it will be even lower... curious what the value is going to be 🤓
today I also had some time to fit some waterslide decals 🙂 its nothing extra, cheap, but works just fine ...
I have potted also one output transformer as well to move forward so I can finish at least one monoblock. Potting of output transformer is a bit tricky to get the lowest induced hum for best snr. I mount already potted power transformer can together with output can with internal braces. I always match output to power and mark the spot to know the matched position of cans:
Then I turn on power transformer (does not need to be loaded) and connect output transformer secondary winding (in this case whole transformer winding) to headphones and I listen and try what is the best position. The key is to get the center of windings (power vs output) under 90° angle as well as maintain equal distance against all metal walls. When I get an idea where its best I connect it to soundcard mic input and watch spectrum to get the lowest 50hz possible:
There is always about 1° window where its the best. Then I validate result with headphones again and finally connect some good bass speaker box to it to see what it does. And amazingly it is absolutely as quiet as it would be turned off 🙂 and this is unloaded transformer winding on its own. Loaded with low output impedance of amps it will be even lower... curious what the value is going to be 🤓
today I also had some time to fit some waterslide decals 🙂 its nothing extra, cheap, but works just fine ...
Kudos to your absolutely excellent craftsmanship, combined with thorough thinking!
WRT to your output transformers: Can you tell me why you chose extra taps to connect the amplifier modules? These taps reflect an impedance of 5.3 Ω, correct? Am I right that, as seen from the center tap GND, the taps are 4 Ω, AMP and 8 Ω for each half? If so, you might have miscalculated the wire cross sections. In this autoformer the biggest current flows from AMP to 4 Ω (sections L1A or L1B). Standardizing this as 1 and calculating with your given voltages, the current between GND and 4 V (sections L2A and L4A or L2B and L4B) is about one third, and the current between AMP and 8 Ω (sections L3A or L3B) is about two thirds. This doesn't exactly fit to your chosen wire cross sections (GND to 4 Ω too thin, AMP to 8 Ω thicker that required).
Best regards!
WRT to your output transformers: Can you tell me why you chose extra taps to connect the amplifier modules? These taps reflect an impedance of 5.3 Ω, correct? Am I right that, as seen from the center tap GND, the taps are 4 Ω, AMP and 8 Ω for each half? If so, you might have miscalculated the wire cross sections. In this autoformer the biggest current flows from AMP to 4 Ω (sections L1A or L1B). Standardizing this as 1 and calculating with your given voltages, the current between GND and 4 V (sections L2A and L4A or L2B and L4B) is about one third, and the current between AMP and 8 Ω (sections L3A or L3B) is about two thirds. This doesn't exactly fit to your chosen wire cross sections (GND to 4 Ω too thin, AMP to 8 Ω thicker that required).
Best regards!
Btw, I was in some error, as the amplifier for sure isn't intended to be loaded by 4 and 8 Ω simultaneously, but by 4 or 8 Ω. If we now assume that the amplifier modules' output current is standardized at 1, the 8 Ω current calculates to (37/45,5)² = .66, but the current through all sections between AMP and GND (and the othe AMP) calculates to 1 - (37/45.5) = 0.34, which is somewhat astounding.
Best regards!
Best regards!
Hey, I don't know what you mean by division of taps (37/45)^2 but I will try to explain from scratch.
answer to your question, for each half it is 2R, amp 2,7R and 4R
I have calculated each situation separately and its not only about power to load, but also about current density, core saturation, lowest frequency etc... there are much more variables involved than just power to load. But I will try as simple as possible just for the sake of load:
Lets omit bridging as push-pull rly results in just the wiring distribution, and lets assume whole transformer for simplification:
for 1000W we need:
91Vrms for 8R
64Vrms for 4R
Amp feeds tap located at 74Vrms (its not exactly in mid-point between 8 and 4 for various reasons, the main one is optimalization of bobbin filling with wire)
1000W / 74Vrms = 13,5Arms current = this is what amp needs to deliver
For 8R tap:
1000W / 91Vrms = approx 11A
(or (1000/8)^-2 )
And as we have autotransformer it has common winding and series winding. So from amp goes 13,5A and its divided:
Thanks to these 2,5 Amperes the series winding is able to transform that remaining voltage that we miss to get to the 8R load...
we miss / needs to transform 17Vrms @ 11A from the 74Vrms @2,5A this is all what is transformed here = 186W. Because its auto-transformer, only the portion of change is transformed.
as you can see to handle those 11A (for current density I used 3,4A/mm^2) I need 3,23mm^2
The wiring is made by bifilar parallel wires, so divided by 2 = 1,6mm^2
and that is wire diameter of 1,4mm (x2)
For 4R tap:
1000W / 64Vrms = approx 15,6
(1000/4)^-2
again 13,5A goes from amp and this is divided:
to your question ... as of current for 4R we need wiring to handle only those 13,5A coming from amp / (3,4A/mm^2) = 4mm^2 of copper needed, again two wires are used so divided by 2 = 2mm^2
to the diameter this is 1,6
(I used 2x1,7mm because I had space in bobbin to be filled)
As for Common winding of transformer, worst case scenario for 8R this is 2,5A, to the diameters with same caluclations this is 0,95mm wire,
I used 1,06
answer to your question, for each half it is 2R, amp 2,7R and 4R
I have calculated each situation separately and its not only about power to load, but also about current density, core saturation, lowest frequency etc... there are much more variables involved than just power to load. But I will try as simple as possible just for the sake of load:
Lets omit bridging as push-pull rly results in just the wiring distribution, and lets assume whole transformer for simplification:
for 1000W we need:
91Vrms for 8R
64Vrms for 4R
Amp feeds tap located at 74Vrms (its not exactly in mid-point between 8 and 4 for various reasons, the main one is optimalization of bobbin filling with wire)
1000W / 74Vrms = 13,5Arms current = this is what amp needs to deliver
For 8R tap:
1000W / 91Vrms = approx 11A
(or (1000/8)^-2 )
And as we have autotransformer it has common winding and series winding. So from amp goes 13,5A and its divided:
- 11A straight to load thru series winding
- remaining 2,5A goes thru common winding to GND
Thanks to these 2,5 Amperes the series winding is able to transform that remaining voltage that we miss to get to the 8R load...
we miss / needs to transform 17Vrms @ 11A from the 74Vrms @2,5A this is all what is transformed here = 186W. Because its auto-transformer, only the portion of change is transformed.
as you can see to handle those 11A (for current density I used 3,4A/mm^2) I need 3,23mm^2
The wiring is made by bifilar parallel wires, so divided by 2 = 1,6mm^2
and that is wire diameter of 1,4mm (x2)
For 4R tap:
1000W / 64Vrms = approx 15,6
(1000/4)^-2
again 13,5A goes from amp and this is divided:
- 13,5A straight to load thru series winding which also transforms remaining current we need
- we lack 2,1A (to get those 15,6A) and this current comes from common winding transformed from that series winding (10V / 13,5A from amp to load, we get current we miss 64V / 2,1A from gnd to load)
to your question ... as of current for 4R we need wiring to handle only those 13,5A coming from amp / (3,4A/mm^2) = 4mm^2 of copper needed, again two wires are used so divided by 2 = 2mm^2
to the diameter this is 1,6
(I used 2x1,7mm because I had space in bobbin to be filled)
As for Common winding of transformer, worst case scenario for 8R this is 2,5A, to the diameters with same caluclations this is 0,95mm wire,
I used 1,06
Last edited:
Slowly putting things together... Inside the chassis I added self adhesive foil below the soft-start board. There is 10mm clearance and device is going to be 1. safety class with earth but just to be extra safe 🙂
built a main capacitor housing
and heatsinks...
finally starting to look like amplifier 😀
however there is a lot of cabling and crimping (especially two thick wires into one faston is a joy to crip 🙄 ) also doing shielded signal cables into molex takes way to much time 😱...
looks like I could finish it already however still going slow... still potting transformers (haven't finished all yet) and pairing for minimum hum takes extra time...
also need to rebuild internal braces, I have found out being them metal works as magnetic "bridge" between transformer can covers and brings more hum. Avoiding internal braces lowers hum by another extra >10dB to sooo low level that I cannot hear hum not even in headphones connected directly to unloaded output transformer 😎 I will order new (same) braces, but made of non-magnetic stainless steel
built a main capacitor housing
and heatsinks...
finally starting to look like amplifier 😀
however there is a lot of cabling and crimping (especially two thick wires into one faston is a joy to crip 🙄 ) also doing shielded signal cables into molex takes way to much time 😱...
looks like I could finish it already however still going slow... still potting transformers (haven't finished all yet) and pairing for minimum hum takes extra time...
also need to rebuild internal braces, I have found out being them metal works as magnetic "bridge" between transformer can covers and brings more hum. Avoiding internal braces lowers hum by another extra >10dB to sooo low level that I cannot hear hum not even in headphones connected directly to unloaded output transformer 😎 I will order new (same) braces, but made of non-magnetic stainless steel
Last edited:
a small weekend step but huge leap for an amplifier look-a-like 😂
finally installed input board, main filter caps and mid cover, only transformers and panel missing...
From other, transformer side:
Caps screws / connections:
finally installed input board, main filter caps and mid cover, only transformers and panel missing...
From other, transformer side:
Caps screws / connections:
Willi, the design is terrible, it's a monster. And the sound through the transformers won't be great either. You would definitely do better to build a new Wolverine amplifier in a 5U modushop chassis. But it's your choice, we all have a different idea of what an amplifier should look like in a listening room...
HRDSTL if you don't like it, just skip it, don't watch it, as simple as that...
Finally mounted front panel today 🙂 a few pics below:
All the profiling meets at front corners where there are many M4 screws, gladly bottom edges of main chassis and front panel are nicely aligned 😎
The whole assembly got incredibly sturdy. Thanks to many profiles at front its hard as rock, and panel could support all amp's weight if the handles would be used. I am not using handles for the sake of look, but just in case... detail of connection below:
Finally mounted front panel today 🙂 a few pics below:
All the profiling meets at front corners where there are many M4 screws, gladly bottom edges of main chassis and front panel are nicely aligned 😎
The whole assembly got incredibly sturdy. Thanks to many profiles at front its hard as rock, and panel could support all amp's weight if the handles would be used. I am not using handles for the sake of look, but just in case... detail of connection below:
New braces, made out of stainless non-magnetic steel 😎 thanks to being them 2nd version I have also optimised a few cable-tie holes etc so its spot-on
To protect wires insulation I have 3d printed plastic grommets
As I am still waiting for remaining transformers at least I did remaining cabling, bridge rectifier etc.... below is how its going to be mounted on front internal brace. Note the cut-out in brace for heatsink fins, so the hot air has a way to escape. By doing this I managed to keep rectifier in between braces in "shielded" compartment, but still able to be cooled properly. The bottom cover has a precisely measured holes to fit this heatsink pattern:
mounted in chassis below, this second brace also covers softstart board AC wiring in nice steel guide hidden behind:
I have also finally tested input board with invertor (opa1611) everything works very nice, the snr is excellent despite quite simple psu for this part with just simple voltage multiplier and a few regulators. THD below 0,0000xyz, very low mains noise, awesome...
However, I've noticed one major fail I did in softstart part 🙄 I was trying so hard to get all "digi" / smps powered / part and analog parts separated, but haven't realised that mute relay's transistor base is powered directly from cmos 🤦♂️and relay for the sake of noise reduction powered from analog source.... ofc this could be easily fixed by simple 100R resistor between digi and analog grounds so no major noise would come thru (confirmed by measurement) but still enough low so transistor base could be triggered. However I want this to be perfect and so much hassle was put into this I want it as planned, so I ordered a new PCB where this fault is resolved - cny17f changed for K827 dual opto. So I still get clock from mains AC as well can switch mute relay, and all galvanicaly isolated. No need to do that, but this is some serious hifi voodoo stuff going on here so this needs to be done! 🤣
See schematic below for better representation:
I have finished new version of soft-starts with galvanically separated circuits between control / analog parts:
However it appeared to be a dead end... will explain why.
by remaking this I have definitely fixed current flow of circuits, however when I did noise measurement of amp I have found out that these circuits separation brings more problems than benefits:
so i did test: wired gnd from both control part and analog part of board to 3-pole switch, including 1k resistor as gnd separation option...
So definitely need to connect these grounds, however I was affraid of that Myrra smps "modern" switching noise in pure analogue amp. (btw definitely better than mini EI transformer producing plenty of 50hz hum everywhere 😀 ) so i measured frequency spectrum up till 400khz with my cosmos and in shorted gnds option, there was visible very small (3uV) 150kHz switching noise. Nothing I should be concerned about, but using that 1k resistor between grounds solved that HF hum and all smps noise was gone, no mains hum, absolutely win-win situation 😀
Amplifier update:
I have finally wired power transformer to the amp and properly tested it!
well, for some specific measurements and numbers you will still have to wait as I want final results to be with fully enclosed enclosure etc... but just preliminary tests looks very promissing, just a short report:
However it appeared to be a dead end... will explain why.
by remaking this I have definitely fixed current flow of circuits, however when I did noise measurement of amp I have found out that these circuits separation brings more problems than benefits:
- temp protection comparator was very sensitive to voltage spikes (f.e. amp with square wave triggered protection) this was fixable with adding some LP filters to all comparator inputs however
- SNR was worse... like 16dB worse! nothing serious, still very low, but if there is possibility to see none mains harmonics why do that right? 😀
so i did test: wired gnd from both control part and analog part of board to 3-pole switch, including 1k resistor as gnd separation option...
- as mentioned above, separated gnd did very ugly 50hz / 100hz mains spike, and a few harmonics
- shorted gnd no mains visible, so deep it dissapeared in noise
- 1k resistor between grounds the same, no mains harmonics
So definitely need to connect these grounds, however I was affraid of that Myrra smps "modern" switching noise in pure analogue amp. (btw definitely better than mini EI transformer producing plenty of 50hz hum everywhere 😀 ) so i measured frequency spectrum up till 400khz with my cosmos and in shorted gnds option, there was visible very small (3uV) 150kHz switching noise. Nothing I should be concerned about, but using that 1k resistor between grounds solved that HF hum and all smps noise was gone, no mains hum, absolutely win-win situation 😀
Amplifier update:
I have finally wired power transformer to the amp and properly tested it!
well, for some specific measurements and numbers you will still have to wait as I want final results to be with fully enclosed enclosure etc... but just preliminary tests looks very promissing, just a short report:
- distortion is very low, lower than original mac
- higher power than expected (>1,1kw)
- output transformers are excellent, go waaay lower than I designed them, my jaw dropped when I saw that! stay tuned 😉 looks like very good winding filling or core properties are better than expected...
- when I wanted to test overcurrent protection 8R tap to 4R load, my dummy load burned with >2kw faster than I saw some current limitation 😀 need to fix and upgrade my dummy 💪
- amp cooling is very good, even I was testing it upside down with heatsinks literally covered on ground without airflow, it was just hot to the touch after some torture tests 🙂 my dummy definitely could not handle more than amp did
Hi Wiliks,
would a X2 capacitor of, say, 0.1 uF, instead of that 1 k resistor also solve your noise issue, while maintaining isolation?
Best regards!
would a X2 capacitor of, say, 0.1 uF, instead of that 1 k resistor also solve your noise issue, while maintaining isolation?
Best regards!
No need for x2 here really, as both circuits are mains isolated anyway… but no, cap would pass HF (150khz) smps noise.
finale is here I was dealing with some hum problems last few days. When I oriented power and output transformers to each other I could get really very good SNR performance, however after I mounted them on chassis everything changed and it very negatively impacted noise performance 🙁 nothing tragic as I cannot hear that (thru my speakers at least) but it can definitely be seen on graphs as obvious issue.
I was suspecting stainless chassis and all the braces however non of that cause the issue. You will not believe, but the problem is in fact thick aluminium front panel! 🙄 even tho its not magnetic material, because its so thick it probably acts as some kind of mirror for magnetic waves and turns it towards output transformer. Maybe thats the reason behind glass front panels on original macs? who knows... if I will be very bored in future maybe I will rebuild front panels to glass ones, but not now.
Strange is that steel front panel cover prevents this partially but not completely. Also on my other amp I have very similar setup with very similar panel with no issues at all, but there are way smaller transformers so maybe thats why. But anyways, cannot comprehend how original macs can have such a good snr ratings with even bigger trannies than I have... 🧙♀️well who knows, I never ever found one measurement of frequency spectrum which includes low noise mains spike to see how good/bad they are in this regard... whatever.
Another issue is that very strong magnetic forces wanted to resonate (non magnetic) steel below mains transformer, I have dampened this with car audio self-adhesive rubber mats which magically resolved the issue. Also between main chassis and transformer cans I used thick textile seal to prevent any resonances.
So finished amp it is, and it weights 45kg (100 lbs)!
A few pictures of finished amp below:
detail of bridge rectifier "cooling duct"
And the majesty...
Video of meter in action:
Measurements below
sine 1k / 8R / 1100Wrms
square 1k / 4R
square 10k / 8R - slewrate 60V/us
square 10k 8R+470nF
Some frequency extremes...
16Hz @ full power = 1kW to 8R!!!! very nice transformer core ❤️
high fr part, 100kHz -3dB
overcurrent protection test...
4R tap to 2R load at a trigger threshold, 1500Wrms! (the power outlet meter showed consumption of 2800W
)
and protection triggered... 4R tap to 2R
Frequency response, take this with a grain of salt as I used focusrite solo for this and I strongly doubt its high frequency capabilities... I measured its loopback and used as compensation but anyway.... measured at about 4W to 4R:
and thd spectrum 1khz, 1kw@8R:
(e1da cosmos + victor gen)
Last edited:
Today I've finished second piece, hum (audibly with headphones testing) was not as problematic as with the first one, maybe because I was testing during day vs night or what, but measurements are very same and front panel had not that negative effect as I first thought, when I did some comparisons... 🤔 afterall it was very same.
To test and compare I did some further measurement and as I am not able to do thd vs power I tested also different lvl testing. I have also lowered bias current for both amps to just 22mA/transistor. This significantly reduced power consumption to below 100W per amp with no distortion change.
4R@16W
4R@100W
8R@1kw
I was also very curious how it compares to big original mac in thermals, as mine is a bit smaller, especially heatsinks etc. I saw this test in stereophile where they loaded it with 400W and state it survived 40minutes 😳
However I am not sure if they use continuous sine wave for this test or normal dynamic music or what 🤷♂️ ?
I tested mine with continuous sine 1kHz / 56Vrms / 8R load (400W) and tried to make a time lapse video of the test but somehow I managed to crop power meter 😀 but it was reading 995W all the time.
My amp turned off after 18 minutes when it reached 82°C 🔥
To test and compare I did some further measurement and as I am not able to do thd vs power I tested also different lvl testing. I have also lowered bias current for both amps to just 22mA/transistor. This significantly reduced power consumption to below 100W per amp with no distortion change.
4R@16W
4R@100W
8R@1kw
I was also very curious how it compares to big original mac in thermals, as mine is a bit smaller, especially heatsinks etc. I saw this test in stereophile where they loaded it with 400W and state it survived 40minutes 😳
However I am not sure if they use continuous sine wave for this test or normal dynamic music or what 🤷♂️ ?
I tested mine with continuous sine 1kHz / 56Vrms / 8R load (400W) and tried to make a time lapse video of the test but somehow I managed to crop power meter 😀 but it was reading 995W all the time.
My amp turned off after 18 minutes when it reached 82°C 🔥