Did you use a passive LPF between the amp and your measurement box? I ask because Audio Precision say their kit is next to useless (in terms of getting realistic measurements) if you don't have one. The high level of HF clobbers the opamps inside.
http://www.ap.com/products/accessories/aux0100
Last edited:
When I transported my speakers (OB) the copper foil coil in the series filter had shorted.. Then one channel of my amplifiers was loaded with three 8" drivers in parallel (between 2 and 3 ohm) with 70uF in parallel again..
When I connected up I did not listen, but put the system on repeat playing at 90 - 95dB average and more than 105dB peak and let in "burn in" for several days..
The first time I sat dow to listen I observed that the left channel was sounding dull and that it was caused by the short.
Removed the short and it played OK again.
It is balanced class A JFET SIT amplifiers with a 20F PSU that have been powered on and have played for close to 10 years. There is no PCB and it is used only silver wires with air as insulation.
Earlier this year I "killed" one amplifier for the first time as I by accident connected 9VDC to one of the balanced inputs.
However the only problem was the JFET regulator transistor that got so hot that the soldering melted. Replaced that and everything was OK again.
The 1 watt class A JFET SIT amplifiers was "benchmarked" (3 - 4 years ago) against the McIntosh MC1.2KW mono amplifiers driving JBL K2 S9800 speakers and McIntosh tube preamplifier.
The only issue the McIntosh MC1.2KW was superior in was playing loud, but playing at normal listening levels even bass control was better with the JFET SIT amplifiers.
That peak meter was a Swiss product, a 1 height 19" rack pack, I don't remember the company's name, something like ACR or some such. It was swiped from a studio owned by a friend, who didn't much like it, and was with me for a few months. I did like it, it had a 40 dB dynamic range (which is very handy), a switch for 8/4 Ohms, another for VU/Peak, another for peak hold and yet another for dot/bar mode. Very complete, I'd say.
Take a wild guess what I did with it the first time it got to me? I grabbed my screwdriver and opened it up, naturally. And lo and behold, inside I saw mostly discrete components, two TI TLO op ams and a few three point regulators. The manufacturer obviously felt that the usual LM 3915/3916 setup, typically used when you want an extended dynamic range, wasn't good enough.
I have a project on stock doing the same. One of these days, I will complete it, depending on time.
Could this be a reason why so many commentators assume the the HF peak signals are very much lower than all the other frequencies in music samples?
I don't know but they are probably the same people who speak of 'low frequency transients' which makes no sense since all transients are HF by their very nature.
Btw German broadcasters (the 'D' in DIN) use -9dB as the highest level when recording digitally using Peak Program Meters to avoid clipping (reaching 0dBFS).
Btw German broadcasters (the 'D' in DIN) use -9dB as the highest level when recording digitally using Peak Program Meters to avoid clipping (reaching 0dBFS).
Looking for opinions and experience here.
To my ears, generally Japanese capacitors in the PSU tend to make amps sound bright to overbright. German caps tend to make them sound mightier in the bass, and tend to shift the tonal balance; sometimes, they can be bass heavy, other times they are simply better balanced than Japanese ones.
So, I suppose the ideal might be to use in parallel one Japanese and one German cap. This MIGHT allow for an ideal balance.
My question is: has anyone tried this, and if so, what were the results? Improved balance, or not? Given the prices I need to pay for quality caps, I wouldn't want to try it just for the hell of it.
To my ears, generally Japanese capacitors in the PSU tend to make amps sound bright to overbright. German caps tend to make them sound mightier in the bass, and tend to shift the tonal balance; sometimes, they can be bass heavy, other times they are simply better balanced than Japanese ones.
So, I suppose the ideal might be to use in parallel one Japanese and one German cap. This MIGHT allow for an ideal balance.
My question is: has anyone tried this, and if so, what were the results? Improved balance, or not? Given the prices I need to pay for quality caps, I wouldn't want to try it just for the hell of it.
I actually have a stock of Panasonic caps, although for decoupling, I prefer to use Nichicon. In part, no doubt, because I have a solid supplier for them.
For the PSU, which is what I referred to, I have not tried Mundorf, so that is an idea. But, aren't they sort of very expensive?
I have tried Aerovox BBH and didn't like them, I had a feeling they were sort of deadening the sound, as if they were slowing it down. Could be wrong, of course.
Bypassing might change sound quantity especially if a RF sensitive device . I borrowed a Techtronics 800 Mhz storage scope a few years back . I was horrified to see what was on the power rails of a budget priced Teac CD player . It was like seeing bacteria under a microscope of water you had been drinking . As Crystal said in the first 20 bit Delta Sigma chips . Solder 0.1 uF ceramic preferably across the chip 0 to Vcc . This will often be the only defense against power supply RF noise . A few mm away being dramatically less effective when a DAC . The 0.1 uF is no big deal . It was the drawing of how to do it that impressed me . I put the scope to the ones on the Teac and sure enough it was dramatically better local to the cap . I always solder a 10 nF across +/- of op amps . This does more than two separate ones ( 3 is ideal , Usually 2 x 10 uF Panasonic FC and 1 x 10 nF ceramic COG if possible ) .
An interesting musing, especially to me, as I have been musing on such topics myself.
Specifically, use an AC or servo assisted DC topology? While nominally the servo version should be better as it kicks out all decoupling caps out of the direct signal path (though some might remain in compensation and filter functions, output Zobel network). However, most DC servoed ampa do not sound any better to me than AC coupled amps, and some AC couled amps in fact sound way better than most DC Servo amps. My point is, simply using this or that topology guarantees nothing.
Can one actually hear capacitors? I can't speak for all of them (and anyway, I haven't actually tried all of them), but in case of power supply cašpacitors, I firmly believe that one can indeed hear them. As I said before, most Japanese large value filter capacitors (where "large" is say 4,700uF and upwards), sometimes the differences are very pronounced. Japanese caps do mid and high ranges very well, but in my view, fall pretty short on the bass range. German made capacitors are, in my view, much better balanced, in fact they don't so much lag in the treble range as they do bass much better, and obviously, this changes the timbre of the entire sound. Subjectively, with German made caps, one has a deeper and better controlled bass in better proportion to the mid and treble than the Japanese ones.
Initially, I thought this was a quirk of mine, but eventually, I tried comparing two amps made EXACTLY (fully electronically equal down to 0.5% - took some time, but I did it) the same way, with the only difference being that one used Elna for Audio caps and the other use Fischer & Tausche caps (actally, while both had 10,000 uF capacity, F&T was rated for 63V, Elna was rated for 71V). I very definitely preferred the F&T version, to me it sounded more even handed, while the Elna version was definitely more in the mid and treble.
Perhaps oddly enough, after much experimenting, I discovered that Nichicon caps work best fpr me for stage decoupling, although I might add that if I have any cap with a value of 1 uF and higher, I bypass it with another propylene or polycarbonate value of 20-30 times lower value, always a Wima or Siemens.
I have found that mounting in parallel 100 uF, 2.2-3.3 uF and 100-220 nF caps will usually clean up just about any mess in a power line, HOWEVER, soldering 100 nF caps across op amp + and - supplies to op amp ground is still a great idea.
Anyone see R10 ? Got me beat . Measuring the - ve op amp inputs confirms 1K2 . Too fine to solder there . Hypex Ucd 180 ST . 1.34 v sensitivity at 26 dB gain . I think 400 mV suits me better . Other end of R11 R12 seems my best choice . If only a Philips MRS 25 for this one resistor ( 7.5 mm body , 10 mm spaced holes if so ) .
Status is that I have started filling up the initial board. I have ordered some items and will have to wait for them.
As for its general fate, I repeat - this is a project I plan to put in the public domain, including the GERBER files someone making the boards will need to make them properly. Obviously, you included.
But YOU are on your way to cause me much grief. I am thinking about making additiona boards, to expand the output stage to double its initial potential, i.e. to 6 output pairs, using what Bryston calls "outrigger boards". Basically, instead of the last transistor pair, you use some wiring to take the action to the next board, which should contain another four pairs (remember, this was you lose the thord original pair, so you're down to two pairs on the original, and so need 4 additional pairs to make the 6 pair mark).
Must be your good nature which helps you suckle two mothers' milk, while telling the third mother that you're an orphan.
And I have not spared you a needling here and there, have I?But seriously, this can be done, I have seen it done on several DIY projects, and as I said, even the very professional Bryston does it.
But make no mistake Wayne, this will take time to organize. At this moment, I am working in parallel on two completely separate power amp projects, one being the Centurion, the other a 50 WRMS per side power amp for a friend who wants to make them for money (to be followed by another model rated at 100 WRMS per side later on, it's still just a schematic on the screen). And there's a preamp in the works, completely discrete, of course, but that one will also have to wait its turn.
Give me the weekend to reason this out, I think it can be done. See if the lower board can perhaps be redone to accept twice the number of output devices.
Last edited:
I was thinking about no wiring at all, power trannies directly soldered to the heat sink.
- Status
- Not open for further replies.