Member
Joined 2009
Paid Member
Some more musings. The zener diode is to protect things. It could be used as regulation.
The time constant 1CR of 4.7uf x 750R = 3.5mS. That is faster than the mains can charge the circuit so looks safe if a conventional PSU. I very much doubt a longer time period would be risky.
BC546 input seems an ideal choice.
All my ZTX753 values are pure guesses as some say the Vbe in this circuit is 550mV. Graphs say it could be.
I have pushed the LTP even further so as to look like a single transistor. This also gives slightly nicer drive. Whilst the VAS is a current to voltage converter it seems nice to give it an easier time. My brother measured IM distortion if this is ignored.
RF should be a common mode problem. An input choke will deal with it better than a shunt capacitor ( 330pF ). A simple aircored handmade one should work. No ferrite beads please. No output choke either. If a choke fitted the 330 pF should be reduced.
With RC 40V filtering to VAS and LTP I doubt if RF will take that route. It's only possible route is via the output emitters whilst switching to the LTP -ve input. If ever a better reason to have fast devices.
Non polar cap for better sound. Higher gain assists stablity and is useful.
Zin is a guess. It could be as low as 250R ( minimum gain spec ) if Naim clone. Realistically 500R reading ZTX graphs. Getting above 1K seems ideal. If the 68R to TR6 was changed you may get a better sound. My guess is 100R with modern devices in the output might be OK. Worth a thought. That would make the standard Naim Zin about 800R if in the higher gain region of the ZTX753.
I put VAS current at 11 mA as I have no idea where it could be in the 10 mA range. The transistors look safe at 440 mW. That's 50% of what claimed for them. Put a big collector pad on the PCB if you can. Put the pad at the edge of the heatsink area to assist soldering. That will make sure it cools. Usually SMD versions of transistors are better spec as the ideal heat transfer is possible and data is given in PDF's. Some double the dissipation!
Last edited:
Could be several causes. First, without signal, check all the voltages and currents in your circuit are what you expect.
Member
Joined 2009
Paid Member
where did you get that schematic from nigel, is it a real amp or did you draw it for this thread ? it looks closer to my thinking than I thought.
the one thing that stuck in my head from yesterday was the sound. I was listing to mono, from a cell phone running hi-res Youtube in a small bright room with only hard surfaces. But the sound was, in hindsight, different from my other amplifiers and has grabbed my interest. I don't know if I can describe it. My memory this morning is of a sound that was lively, almost gritty when the music is complex. There was a slight 'ess' on female vocal (may have been in the recording of course) but the voice sounded very good, almost like the lady was standing close to me. It didn't sound 'neutral', it was more 'alive' from lows to highs than I expected.
I do suspect that Cdom may need to be higher than 39pF in my build, with fast drivers and fast modern output devices. Or perhaps my assumption about the damping resistance of the FETs is not quite right and I need to add a small series resistor. Both options might be interesting to hear.
With a load I am drawing much more current through the output device. The fT of transistors increases with current, they get 'faster'. This is one clear difference between no-load and a load.
I am pleased about the results so far on the thermal compensation. It's not proven yet, but my approach may solve the long-running concern that folk have had with this amplifier.
the one thing that stuck in my head from yesterday was the sound. I was listing to mono, from a cell phone running hi-res Youtube in a small bright room with only hard surfaces. But the sound was, in hindsight, different from my other amplifiers and has grabbed my interest. I don't know if I can describe it. My memory this morning is of a sound that was lively, almost gritty when the music is complex. There was a slight 'ess' on female vocal (may have been in the recording of course) but the voice sounded very good, almost like the lady was standing close to me. It didn't sound 'neutral', it was more 'alive' from lows to highs than I expected.
I do suspect that Cdom may need to be higher than 39pF in my build, with fast drivers and fast modern output devices. Or perhaps my assumption about the damping resistance of the FETs is not quite right and I need to add a small series resistor. Both options might be interesting to hear.
With a load I am drawing much more current through the output device. The fT of transistors increases with current, they get 'faster'. This is one clear difference between no-load and a load.
I am pleased about the results so far on the thermal compensation. It's not proven yet, but my approach may solve the long-running concern that folk have had with this amplifier.
Last edited:
Member
Joined 2009
Paid Member
I re-wired my bench set up. I can't remember exactly what I was doing but I suspect I was making inappropriate use of the signal ground connection and had the output wires and input wires poorly separated.
Well, whatever I did seems to have cleaned up the square wave. Attached shows the 1kHz square wave into the same resistive load as yesterday with no changes to the amplifier. The 2nd attachment is what it looks like when you pull out the x10 timebase control.
I connected up my modified Sansui T-60 tuner and soldered up a 10kohm Alps pot to make a better source. Sound is now different than yesterday through my Blackberry. The bass is quite a lot 'more'. I'm letting it soak on the bench again to see how it settles in. Bias at around 5mV across Re.
Well, whatever I did seems to have cleaned up the square wave. Attached shows the 1kHz square wave into the same resistive load as yesterday with no changes to the amplifier. The 2nd attachment is what it looks like when you pull out the x10 timebase control.
I connected up my modified Sansui T-60 tuner and soldered up a 10kohm Alps pot to make a better source. Sound is now different than yesterday through my Blackberry. The bass is quite a lot 'more'. I'm letting it soak on the bench again to see how it settles in. Bias at around 5mV across Re.
Attachments
Member
Joined 2009
Paid Member
Attached - a few sample voltages.
I'm not yet convinced about the treble but overall sound is very nice. I need to try something other than my DIY full range speaker on this amp.
Attachments
That's a nice looking 1kHz test wave. Nicely rounded corner indicating good HF response and passively rolled off somewhere.
No over shoot indicating good compensation that suits your specific test load.
There is no need to test at 50kHz. The 1kHz sqw tells you all you need to know for an audio amplifier.
What did you do to the input filters to pass this test sqw?
There's something odd with the voltages measured on the bases of the input pair.
The base current into the input transistors has to pass from the audio ground/signal return through the 22k1 R1 and 4k75 R1B
if base current of TR1 bc546a is 1uA then the voltage at the base lead will be ~ 0.02685V
Your bc546a would need an hFE of ~356 to get to your measured 0.04V of input offset.
What is the base current of TR1?
Could you have measured using the bottom of R2 as your DMM reference? The top of R2 should be your reference for measuring input voltages.
What is the voltage across R2?
Last edited:
Very useful data. I keep thinking the beauty of this Naim design is we have very few choices. As far as I can see one remains.
VAS transistor gain and Cobo seems favourite. Self rightly says better the capacitance is external as it is higher grade. What he means is you are getting something like the worse ceramics ( not COG/NPO ) inside the device.
Self remarked about the internal capacitance measuring as if a lower grade type. I think his reasoning being the place of reference being not exactly the same as a collector base external capacitance, i.e. not the base exactly. Extra avoidable non linearity. Self I think was looking more to the 2 to 6 pF devices as ideal. I think this was also talking out against the idea that single input transistor amps didn't need Cdom as they were more stable. His feeling was that was because early designs often had very high internal capacitance at the VAS and not really delay from LTP emitter to emitter. ZTX753 is almost one of them at 30 pF.
Member
Joined 2009
Paid Member
confession time
I didn't match the LTP devices, just didn't want to to wait to find out if the darn thing was going to work or not. I had several new transistors on a strip of cut-tape so I cut two off and installed them. I do expect that they are fairly close - in a previous project where I had borrowed an Hfe meter I found that parts from the same batch were very closely matched already - there was one 'rogue' part in the batch. I don't have access to that Hfe meter anymore.
What I could do next is to build up the 2nd channel and match the devices. Then see if I can discern a difference, measured and subjective listening-wise. But I do note that a mis-balance is something we believe JV was aiming for in order to achieve a desirable harmonic profile. Mis-balance will be primarily a 2nd order harmonic and this is believed to be one of JV's intended goals.
The input filter on the amplifier is per the schematic. The only real change was to re-wire the external-to-pcb wires on the bench top to something a lot more sensible. In particular, I had some poor grounding choices, including using the signal-gnd input as the scope probe ground despite having installed a ground lift resistor on the pcb.
these were quick and dirty measurements and I should review them again. It's quite possible of course that I have a rogue device in the LTP. The datasheet for the VAS device I have shows Hfe is expected to be around 200 which would imply 0.05mA of base current which means that leg of the LTP will be running at round 0.58mA and the other leg at 0.7mA, about 20% mis-balance whereas the difference in base resistors is only 3% (if the resistor values are correct). I'll check it again.
Thanks Nico - it's always nice when this happens
Last edited:
One thing that jumps out of the page which I had forgotten is the VAS is a constant voltage clamp ( that sets an almost constant current ). 1K sets this with Vbe. The time it shows an important difference is if the ZTX753 is being starved and starts to switch off. Vbe is very low as many say for ZTX753. If we have the tail current one side or the other ( high or low ) we could see the fake single transistor distortion curve. Or if we like the least second harmonic curve we dial a 50/50 balance. Thus we need to make 1K smaller if a quest to get something " better ". As with my musing 1.23 mA here. It probably is a better result although not pushing the VAS any harder. I wonder if 810R and 22K would be better? When brave perhaps 510R 11K and CCS Re 330R. DC offset should be 5 mV. I feel 39pF VAS cap will still be OK. It's 69pF total.
I am going to say what I think happened at Naim about 1978. When the design was laid down Julian was learning electronics. The early NAP160 had a different sound. I suspect the device used as VAS was different with higher Vbe. I suspect Julian was unaware that it mattered and prefered this version. This might mean the VAS is where the bigger difference can be had then or now, be it it good or bad. Quite interesting two points exist for the 2nd harmonic balance. One of them wastes current.
I rather like your 0R22 + 4700 uF. In fact you could make it 10 000 uF + 0R1 and reduce the primary caps a tad. 10 000 uF 0R1 10 000 uF could be interesting ( R21 C10 etcetera) . It will sound a bit dymanically softer as a result over a Naim design I suspect.
It's interesting to see your 1kHz square waves. To be honest a little better than I imagined. Then I remembered. It was the standard test in the 1970's. You bet the correction networks were tuned to get that done. If the amplifier rolls at 38 kHz one would expect to see trouble at 2 kHz if the old Open University course I followed was right. Thus 3 kHz is about where I would think it starts to really show. Real music worst case is a 1 kHz squarewave. A CD player just about can do 1.16 kHz to the Open University mathamatical teaching ( F +1/3f3 + 1/5f5+ 1/7f7..........+ 1/19f19 to a good approximation). Triangle waves are a squared number version F+1/9f3+1/25f5.... . Nice to see how a simple filter and a squarewave can do that.
A Square Deal
I am going to say what I think happened at Naim about 1978. When the design was laid down Julian was learning electronics. The early NAP160 had a different sound. I suspect the device used as VAS was different with higher Vbe. I suspect Julian was unaware that it mattered and prefered this version. This might mean the VAS is where the bigger difference can be had then or now, be it it good or bad. Quite interesting two points exist for the 2nd harmonic balance. One of them wastes current.
I rather like your 0R22 + 4700 uF. In fact you could make it 10 000 uF + 0R1 and reduce the primary caps a tad. 10 000 uF 0R1 10 000 uF could be interesting ( R21 C10 etcetera) . It will sound a bit dymanically softer as a result over a Naim design I suspect.
It's interesting to see your 1kHz square waves. To be honest a little better than I imagined. Then I remembered. It was the standard test in the 1970's. You bet the correction networks were tuned to get that done. If the amplifier rolls at 38 kHz one would expect to see trouble at 2 kHz if the old Open University course I followed was right. Thus 3 kHz is about where I would think it starts to really show. Real music worst case is a 1 kHz squarewave. A CD player just about can do 1.16 kHz to the Open University mathamatical teaching ( F +1/3f3 + 1/5f5+ 1/7f7..........+ 1/19f19 to a good approximation). Triangle waves are a squared number version F+1/9f3+1/25f5.... . Nice to see how a simple filter and a squarewave can do that.
A Square Deal
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.