advertisement


Pre-amp Nirvana…?

Well it's equally an elegant and effective approach, at least from where I sit. I think its success might be due to the fact that it's approaching both sides of the power supply problem from a single point, and/or a single solution, and providing both with an equal level of noise/performance. Perhaps anything that impacts one side of the PS equation affects the other side equally (or nearly equally), which might retain a sense of balance in how the resulting music is relayed. Just my speculation. (I guess that is sort of how a split supply works, isn't it.)

One problem has appeared: heat. The 32 case is very warm to the touch. I don't think it's affecting performance of the music, but I will research more over the weekend.

One other comment about the mod as I continue to listen to a range of sources: the pre is now revealing and almost ruthless with regard to the _sonic_ aspects of different sources, but it seems to unlock music from all of them. Low bit-rate MP3s, Internet Radio, and live recordings all immediately reveal tonality problems and noise, more so than previously, but the music in each seems to comes through, better than before in most cases.

M
 
Sounds like a job really well done, Mike, your comment:
the pre is now revealing and almost ruthless with regard to the _sonic_ aspects of different sources, but it seems to unlock music from all of them.
struck home with me - it's exactly what I prize in mods I consider to have gone 'right'.

2R + Oscon (i presume 47R) looks like a really good mix, starting to roll-off about the point that the three-pin regs own noise starts to rise (-3dB at about 1.7Khz) and will kill a lot of noise feedthrough, giving up to 35db rejection of feedthrough at HF - limited only by the Oscon's ESR. NB you can go as low as 0.33R in series and still have critical damping of a 47uF oscon with the 317's effective output inductance, i.e no unwanted ringing. Drop below that and hardness appears in the midrange, followed by a lisping treble.
 
One problem has appeared: heat. The 32 case is very warm to the touch. I don't think it's affecting performance of the music, but I will research more over the weekend.

Mike, If I've read your earlier posts correctly, then apart from the converters, you've not had local reg's inside the 32 until now.

The reg's plus the converters could be creating about 60% more heat (in addition to that created by the amp components) in the preamp. You can easily work out the total amount of heat being generated, as almost all of the current going in is turned into heat. For instance, if the snaps is delivering say 300mA @ 24V, then approx 7W of heat is being created inside the 32, with the converters and reg's making up about 2.5W of that. Measure the current leaving your snaps and we'll get a better idea of what's going on (and see if it seems excessive).

Another thought - can you measure the case temp? (a cheap thermometer will do).

Mr Tibbs
 
Got to open the case (and took pix, will upload those later).

The general temp inside wasn't as warm as I expected it would be. But the SMPS units get very, very hot. The 317s are warm but not nearly as warm as the SMPS. Touching the SMPS for more than a second is uncomfortable. Can't measure precise temp.

The power rating on these is 250-300ma each, and I have 1 running 2 x phono and 2 x 729, the other running 2x gain cards and 2 x tape buffers. I think in terms of load that should be within the limits of each SMPS module, as the Traco specs are very similar.

Also, a brief scare when I plugged it all back in, the pre exhibited a low-freq steady "thumping" sound. Reseated all the cards and the wiring and it seemed to resolve it, but I don't know the source of that issue. Any thoughts immediately come to mind for you now? It's back to sounding great as ever.

One more listening comment (stop me please if these bore anyone): I played a WAV-format recording of an LP done before the SMPS/0v update, and was intrigued to very clearly hear low-end feedback in the TT; it seemed to be at the very lowest frequencies, the heavyweight stuff, and I hadn't heard or felt this otherwise before. It was clearly recorded with some feedback, that the speakers previously had just sort of grouped together as a "thump". What I found most interesting was how this improved setup resolved the some aspects of the low-end energy very well, that I hadn't heard until now, yet it exposed a very recognizable weakness in the same freq range that was there all along, hidden/masked by noise or some other issue. What I wonder is if the signal inside the preamp before the SMPS/0v change was good/clean enough to have captured the feedback along with the rest of the low-end sound, (and dump it to the WAV file), but the output to the amp (via the PSU) clouded things enough to mask the feedback. (Hope that makes sense, typing in a rush here.)

And it's all warming up, silky smooth top end, perhaps the 2r/47u combo is really helping with that.

Pix later tonight, hopefully.

M
 
DSC_0051.JPG

DSC_0052.JPG

DSC_0047.JPG


Not as elegant as the board itself. I powered the 729 cards from the side opposite the closest set of 317s because the closer pair were a bit noisier, and I wired the tape output with those instead.

The final close-up shows how I wired up the cards to keep them modular. I removed the 24v power pin from the main board, inverted it and attached it to the wire from the 24v LM317 line, and then inserted it into the top of the connector on the daughter card. (I did all of this for the 6 x 24v external supply I built several years ago. I was able to reuse all of that work, just redirecting the source of the 24v from a DIN jack to the new SMPS board.) This makes it possible for me to easily remove and rework the individual cards, but is a challenge to install due to cramped spaces working around the cards inside the 32. The main gain cards (321/5) have the "Xmas" FB cap mod also.

I have more feedback/reactions on sound quality also, but may save that post until morning as I'm dead tired. Short version is the rest of my family noticed the change, but not consciously.

M
 
The general temp inside wasn't as warm as I expected it would be. But the SMPS units get very, very hot. The 317s are warm but not nearly as warm as the SMPS. Touching the SMPS for more than a second is uncomfortable. Can't measure precise temp.

The converters are approx 80% efficient, so a converter rated at 5W will be warming itself to the tune of about 1W at full output. With little or no air movement going on inside the closed 32 case, the converters may indeed get hot. The Wall spec suggests a max ambient operating temp of 71degC (pretty hot) but I would think the lifespan would shorten if they are continuously run at or near that temp.

Two options are open. 1) Drill some holes in the case (sorry, that's a bit drastic!). 2) Get hold of a couple of appropriately sized heatsinks and attach them to the smps cases using thermal bonding epoxy. I would go for option 2.

Also, a brief scare when I plugged it all back in, the pre exhibited a low-freq steady "thumping" sound. Reseated all the cards and the wiring and it seemed to resolve it, but I don't know the source of that issue. Any thoughts immediately come to mind for you now? It's back to sounding great as ever.

Sounds like low freq oscillation - hard to say what caused it (possibly one of the smps didn't start up properly)

One more listening comment (stop me please if these bore anyone):

Do please keep the listening comments coming as they are so important in getting a feel for how things are working. (will have to think about what you said before replying though).

Good looking job with the install. A couple of things that may get you some more performance; If you can, split the output 0V on the PS board into two and connect each separately to the preamp 0V point, using as short a link as possible. If splitting is not easily done, replace the existing single 0V wire with a heavier gauge wire and keep it short as possible.

Also, I find it improves performance if you arrange the power feeds so that each converter and associated reg's feed the boards of a single channel.

HTH

Mr Tibbs
 
Also, a brief scare when I plugged it all back in, the pre exhibited a low-freq steady "thumping" sound. Reseated all the cards and the wiring and it seemed to resolve it, but I don't know the source of that issue.
Sounds like motorboating - are there any decoupling caps on the input to the SMPS units? WIthout, the SMPS might just have a problem interacting with inductance from the long interconnecting wires back the raw PSU, causing LF oscillation (which is why the LM317 recommends an input cap if the preceeding cap is more than just 6" away. ) Might be worth trying, say, 10uF here anyway.
 
My first pass on summarising the thread is now up here.

I've also cached the thread up to yesterday also for future reference. If anyone is unhappy about (their) content being taken off PFM, please let me know and I'll edit as required.
 
Thanks much Mr Tibbs & Martin for your input and help on this!

2) Get hold of a couple of appropriately sized heatsinks and attach them to the smps cases using thermal bonding epoxy.

Already with you on this, and have a couple options in the odd parts box. Will try hopefully this week.

Sounds like low freq oscillation - hard to say what caused it (possibly one of the smps didn't start up properly)

I remember seeing some pretty intense oscillations on SMPS boot up when I was 'scoping them. Took a second or two for them to settle then, so this may be the case. Martin's 10uf suggestion is an easy test/solution.


If you can, split the output 0V on the PS board into two and connect each separately to the preamp 0V point, using as short a link as possible.
Not easy to do. My original board layout had two completely isolated SMPS modules like your setup, but then I thought I'd try to move the preamp star point to the board itself. That seemed to necessitate joining it all into one point. Also, both my SNAPS and DIY HiCap aren't electrically isolated split supplies, just one big supply with multiple outputs hanging off it.

If splitting is not easily done, replace the existing single 0V wire with a heavier gauge wire and keep it short as possible.

Yes, good thought. It has a few turns in it now (mostly for looks I guess) that I could remove/shorten.

Also, I find it improves performance if you arrange the power feeds so that each converter and associated reg's feed the boards of a single channel.
Given that it's single supply feeding both, do you think this is still important? Would involve rerunning 4 lines, not a huge deal, but I'm of the mind to leave well enough alone unless you think it's holding back even more music.

Now for more listening feedback. This the opposite of what I typically expect for hi-fi listening feedback, but it was eye-opening for me.

We have 2 boys, 5 and 3. They are into all sorts of music but have their favorites. Yesterday afternoon I put on They Might Be Giants' "Here Come the ABCs" (great fun for kids and adults) and they started the usual dancing around. About the second song in, both started singing along almost perfectly on pitch with these songs...which I hadn't heard them do before. When the CD was over, the younger said "Dad the radio didn't play any more songs. Make it start again". A little later that evening, the same kid was at the fridge playing with magnetic letters, singing one of the songs from the TMBG collection, again with amazing melodic accuracy and flow. My wife turned and said, "When did he learn to do that?!"

The kicker for me came even later, when I hit "shuffle" on the Squeezebox and let it run through whatever came up. Joe Jackson's "Sunday Papers" came on, which I think was the first time the kids had ever heard it. The 5yr old sang the first line of the chorus the second time it came around, with the right words and melody. Then he paused and asked "Dad why is he singing about a newspaper?" Another first as far as I know. What do these say about the improvement? My kids had never done anything beyond just dancing around, really, when music was playing. Sometimes they'd sing a song later, but I'd never heard them sing along, until today.

In all of this testing and listening, I wonder if my stock preamp wasn't at the same performance level to begin with as others who mod/tweak, and so perhaps what I'm hearing as an improvement may be greater than what others would hear. I don't know if newer designs with better grounding schemes and board layouts than the old 32 (ie 102, 72, 62 or even the 32-5) may be an improvement in and of themselves, without the SMPS/0v work.

Hope this is useful for the rest of you.

M
 
Yes, good thought. It has a few turns in it now (mostly for looks I guess) that I could remove/shorten.

I'd be very interested to hear what you think after you shorten/upgrade this link - it doesn't seem that intuitive, but this is a critical area (IMO!). It won't be night-and-day, as the common 0V line has already been shortened by a large amount... (I'll say no more and let you come back on this one).

Given that it's single supply feeding both, do you think this is still important? Would involve rerunning 4 lines, not a huge deal, but I'm of the mind to leave well enough alone unless you think it's holding back even more music.

I do think it's worth it, because it means each smps has a slightly less complex load and can therefore perform better (you could try a temporary arrangement without pulling everything asunder). Leave it for a while if you want - it's like money in the bank, you can take advantage of it when it suits you.

Now for more listening feedback. This the opposite of what I typically expect for hi-fi listening feedback, but it was eye-opening for me.

Great feedback! (I can't claim the same results in Tibbsville)

I don't know if newer designs with better grounding schemes and board layouts than the old 32 (ie 102, 72, 62 or even the 32-5) may be an improvement in and of themselves, without the SMPS/0v work.

The newer designs do have improved local grounding, but this mod is more about fixing the grounding between the pre and the power supply, so may not be that different in function across the range.

Mr Tibbs
 
One more listening comment (stop me please if these bore anyone): I played a WAV-format recording of an LP done before the SMPS/0v update, and was intrigued to very clearly hear low-end feedback in the TT; it seemed to be at the very lowest frequencies, the heavyweight stuff, and I hadn't heard or felt this otherwise before. It was clearly recorded with some feedback, that the speakers previously had just sort of grouped together as a "thump". What I found most interesting was how this improved setup resolved the some aspects of the low-end energy very well, that I hadn't heard until now, yet it exposed a very recognizable weakness in the same freq range that was there all along, hidden/masked by noise or some other issue. What I wonder is if the signal inside the preamp before the SMPS/0v change was good/clean enough to have captured the feedback along with the rest of the low-end sound, (and dump it to the WAV file), but the output to the amp (via the PSU) clouded things enough to mask the feedback. (Hope that makes sense, typing in a rush here.)

I think it must be because when recording, the signal has a less complex route through the preamp compared to when playing back. If you made that recording over again with the modded 32, you'll find that the recording will also be improved and the end result will be better again.

Mr Tibbs
 
I do think it's worth it, because it means each smps has a slightly less complex load and can therefore perform better (you could try a temporary arrangement without pulling everything asunder). Leave it for a while if you want - it's like money in the bank, you can take advantage of it when it suits you.

Mr Tibbs

You were right. I was able to fairly easily just re-route the individual 24v lines to use each SMPS circuit to power an entire channel. In terms of the sound, it's more of what I heard initially; more music, tighter performance, better pace. Not a huge leap, but certainly enough to notice, especially in bass parts.

Thanks again for your help.

M
 
I'm posting an update after spending a little more than 2 weeks evaluating Mr Tibbs' SMPS mod, and two minor variations with how it's wired.

My goal in this most recent test period was to determine if my original SMPS configuration was better, or if Mr Tibbs' suggestions about using an SMPS to power each channel was better. I tried to standardize as best I could, using the same sources, music, conditions/time of day, number of manhattans, etc.

The first test period consisted of about a week of listening to the pre with the initial mod configuration: where each SMPS powered 2 sets of cards inside the 32, both left and right channels for each set. The second week or so was spent listening to the same sources/music, and in the same settings/scenarios, but with each SMPS configured to power all cards on a single channel (Mr Tibbs suggested configuration).

First, my initial reaction to this update, (which was done along with moving the star earth point inside the pre), stands unchanged: this is a most worthwhile improvement, and very much beyond what I expected it to achieve. In either config, this update makes the preamp very transparent, imparting very little sonic character to whatever is fed into it, and at the same time highlighting the musical aspects of that same material. It is so revealing that I was able to hear pitch instabilities in my TT that I traced to a worn/stretched belt. A new belt brought back the stability and even more powerful dynamics and music. (It also widened the gap between the TT and any other source I have). It amazes me how easily I can now hear sonic differences between different sources. DVD player, old Proton 440 tuner, Rat Shack a/v switch box, they all clearly have a sonic signature, with strengths and weaknesses easy to spot. But somehow they all have a sense of flow about the music that eventually comes out.

After giving both an extended review, I still prefer the original configuration, but concede that the second resolves more detail, particularly at higher volumes. The original has a sense of musical control and focus that the second config seems to lack, at least to my ears. The second has a little more depth, air, and an appealing sparkle in the upper registers. Both are dynamic and lifelike, and ruthlessly revealing of source quality. I suspect my preference may be in masking of issues or problems either in the sources themselves, in the design and implementation flaws in the 32, in my construction and installation of the mod, or a combo of all of the above. Don't really care what it is, as I'm so pleased with the result that I don't see myself doing anything but buying more music and better source gear for quite some time. The provibial "whole new music collection" thing.

One other finding of Mr Tibbs that I can underscore: the quality of the incoming 24v to the SMPS is important. My stock SNAPS was easily bested by a DIY HiCap. More control and resolve of everything, particularly in the mid and lower octaves. Also both SMPS configs sound noticably better after 3-4 days of warm up.

This mod goes against the typical approach to PSU design for Naim preamps, but it has worked wonders for me. The 0v signal is clearly a very critical part of getting quality music out of these circuits, and this approach seems to pay excellent attention to that concept.

Again, thanks to Mr Tibbs for the initial suggestion, and help through the install.

M
 
Good feedback thanks. I'm surprised you prefer powering each set of cards with a single SMPS, rather than each channel. But as you say, if you prefer the sound then no problem. One thing to consider - are both SMPS's identical (both physical and electrical properties)? If they're not, then you're preferred config makes more sense. Or do you still have the relay or bulb powered? Again, having only one of these hanging off a single channel would upset the balance somewhat....I would expect....not at that stage with my 62 yet.

Richard
 
Both are very good, and if I hadn't heard both I would be plenty happy with the per-channel arrangement. I don't have the relay card installed, nor LED powered. SMPS and circuits are identical. Others may want to try Mr Tibbs' version first. (We need a 2-out-of-3 winner here to know for sure which is best. ;-) )

Also, can anyone confirm that a NAC 82 and on up (102 etc), is powered per channel, rather than per stage, with 1 HiCap (or 2 in the case of the 82)? On the older pres, one HiCap rail powers the gain circuit, and the other powers everything else. Wondering if Naim deviated from that with the newer models.

M
 
Don't really care what it is, as I'm so pleased with the result that I don't see myself doing anything but buying more music and better source gear for quite some time. The provibial "whole new music collection" thing.

That's the best outcome possible.

One other finding of Mr Tibbs that I can underscore: the quality of the incoming 24v to the SMPS is important. My stock SNAPS was easily bested by a DIY HiCap. More control and resolve of everything, particularly in the mid and lower octaves. Also both SMPS configs sound noticably better after 3-4 days of warm up.

Yes that's very true. It would have been a great bonus if the DC converters had made the quality of the primary supply a complete non issue, but we can live with that. At least they are not as fussy as any regulator I've come across.

Mr Tibbs
 
I haven't touched anything in the preamp for an age now, so figured I should open the bonnet and check the oil, so to speak. The outputs of all the 1086 reg's and the two Traco's were checked and found to be spot on, plus everything was running well within temp spec's. Reliability seems fine, and shouldn't present a problem in the longer term.

That done, I wanted to try a couple of minor changes. Mike noted that he preferred not having a Traco to each channel of the preamp. In other words, one Traco should feed the filter/buffer of both channels, and the other should feed the output stages of both channels. In a way, this is how Naim feed the power from a hicap to the stages of the standard 102.

I tried this before (way back when I first experimented with the DC converters) but thought the Traco's one-per-channel was best. I was wrong and so Mike is right! Yes there is a tiny loss of inner-detail, but the music as a whole seems to gel better and appears more real as a result. Nice one, Mike :)

Another thing I tried was to introduce a resistor between each 1086 and respective stage (until now there was simply a small hand-wound inductor at each 1086 output). I tried a couple of values but in the end, 27R came out on top. This change in sound is fairly subtle, but definitely an improvement (to my ears). Same sort of result as above, but this time with no (or almost no) loss of detail.

Lastly, I wanted to 'standardise' the hicap by converting the split 0V output back to one single 0V. The idea here is simply that if it doesn't affect the performance when feeding Traco's, then it is best to make it standard again.
I'm happy to say that there was no change at all that I could hear.

While the 102 was out of service, and with the benefit of a now standard Hicap, I bunged in my trusty 42.5 for a couple of hours. The whole lot was hooked up in completely standard Naim fashion. The result was eye-opening enough that I even converted the 102 back to bog standard (easily done with the Traco boards), for a try alongside the 42.5

I'll tell all later :)

Mr Tibbs
 
Another thing I tried was to introduce a resistor between each 1086 and respective stage (until now there was simply a small hand-wound inductor at each 1086 output). I tried a couple of values but in the end, 27R came out on top. This change in sound is fairly subtle, but definitely an improvement (to my ears). Same sort of result as above, but this time with no (or almost no) loss of detail.

Bang on with this one. I've tried numerous different ways of powering these circuits and have found that the 27R/47uF filter right on the circuits give the best 'balanced' sound.
Had a brief flirtation with VBEs but found there are negatives as well as positives to this approach.
My ears tell me that a linear reg followed by the RC filter is the solution.

Cheers,

Jim.
 
My ears tell me that a linear reg followed by the RC filter is the solution.

Agree. I think this has been mentioned before (by Martin Clark), but that 27R/47uF filter is just right for taking over where the reg is beginning to pass noise.

Mr Tibbs
 


advertisement


Back
Top