| audiorob |
Hi,
I'm a computer scientist who is lost in a EE's domain, so please forgive my ignorance.
I am trying to build my first DIY amp. I thought I would build the Zen 5 amp, as it can be built with point to point wiring. I thought I would build a test board and power it with my workbench power supply (before trying to build that big power supply). The max voltage on the workbench power supply is 30V, so I can get +/- 15V rails (at 3A). I _think_ (I don't _know_) this is enough voltage to at least get some noise from the amp (which is my goal before building the big power supply). Can someone confirm this guess?
I built the amp following Fig. 3 of Nelson's, "The Zen Variations - Part 5 The Complementary Zen" article. I'm as sure as I can be that I have built the circuit correctly. Because I'm using +/- 15V rails, I put two 47.5K resistors in parallel for R5 and for R6 (the value of R5 is ~24K, and the value of R6 is ~24K) to be able to use the pots to set the bias.
One of the first things I noticed was that by using only a Volt Meter, I really couldn't get the DC offset to vary much. I could not get the DC offset to vary across 0V. I tried a different meter, but with about the same results. I resorted to connecting 2 oscilloscope probes between the gate resistors and the gates (between R7/Q2 and the other between R8/Q8). With the scope, I could bias the transistors equally (but opposite polarity). The test bench power supply has an ammeter, so I adjusted the pots to fed about 1.5A equally though Q1 and Q2. The VM values across R9&10 and R11&12 matched Nelson's measurements (Ie: ~350mV). I don't remember the exact value, but the DC offset was very small (<.01mV). I've tested so many things that I'm getting fuzzy about my measurements, but I think I'm supplying either ~8V or ~12V to the gates.
When applying a 1V 1KHz sine wave signal to the inputs, I do see a sine wave at the gates. However, I don't really get much at the output.
I connected a speaker to the amplifier output and the test bench power supply ground. I connected a preamp output to the amp input and the test bench power supply ground. I applied a signal, but nothing. If I reconnect the signal generator to the input, and apply about a 3V sine wave, I can faintly hear a tone in the speaker. But 3V is a lot bigger signal that I thought the amp would need (I was thinking that ~1.5V would drive the amp to full output).
Does anyone know what input voltage is required to drive this amp?
Does anyone have any idea what I'm doing wrong, or what other tests I could do to debug this further?
thanks,
Robert
PS - Nelson also mentions, "... nor is there any reason not to bypass the 47 uF electrolytic caps with your favorite lower value film types." I _do_not_ understand this capacitor preference stuff. I do not know what a reasonable value for a film type cap would be. I do not understand what the value of a film type cap over an electrolytic is. Also, aren't most film type caps non polarized? Wouldn't putting a non polarized cap in this circuit in place of the electrolytics be a bad thing? Could some one help me with this topic? Pointing me to a book or a link is fine.
thanks again,
rf |
|
|
| Nelson Pass |
After a discussion on the phone, we worked out that
it was the test power supply. (It's always something
that makes sense afterword)
:cool: |
|
|
| audiorob |
I'm sure that no one is surprised, but for the next person to tries this... Nelson was of course right. I took his suggestion and got a second test power supply and connected it to the first, using the connection point as ground, and the amp works! Now to build that big power supply! I'm sure this is old hat for most of you, but for me, it is really a thrill to get this thing working. I can't say enough good things about Nelson...
Robert
PS - it's the little yellow board sitting on the flat side of a heat sink. |
|
|
| choky |
you have one big problem and you must solve it ASAP-your bench is not enough messy,and you'll hardly finish any decent amp this way
;) |
|
|
| Nelson Pass |
Typical computer guy - reserves the mess for his code
;) |
|
|
| Zen Mod |
| quote: | Originally posted by Nelson Pass
Typical computer guy - reserves the mess for his code
;) |
yup-look-only digits are bouncing on this picture...
no smoke,no beer,no leftovers of any kind....
poor fella'
:clown: |
|
|
| AndrewT |
Hi,
is that 8 readouts I see before me?
The eyes of a fly to take in all that info at once.
Can anyone instruct us how we can use a PC to collect all this info in just a few mS and then peruse the data at our leisure? |
|
|
| audiorob |
If anyone is interested, I have built one channel and one power supply, all per
Nelson's design (to the best of my ability). I have connected everything, and
done a little testing (which is about all I know how to do). :)
With a 1KHz triangle wave, I get 27W into 12 ohms (I don't have any 8 ohm
power resistors), with 3.8V input. I also looked at some square waves up to
20KHz. There is a little bit of corner rounding at 20Khz, but no overshoot or
ringing. This is a "test" setup, with long leads going to the transistors. Not
that things are bad or anything, but I think the 20KHz corners will get even
better when I mount things "for real" and have shorter leads.
Anyway, it sounded well enough for me to build the second channel, and
some mono boxes. More pictures soon. |
|
|
| audiorob |
| quote: | Originally posted by AndrewT
Hi,
is that 8 readouts I see before me?
The eyes of a fly to take in all that info at once.
Can anyone instruct us how we can use a PC to collect all this info in just a few mS and then peruse the data at our leisure? |
HI AndrewT,
Well, aside from the VM's and such, I use a PC/USB based O-scope. It is
made by EasySync Ltd, which is in the UK, I think. The model I use is a
DS1M12 Stingray. I think it can do at least 250KHz and 50VDC. I am a
CS guy, not a EE (, so I don't know much about scopes), but I think this is
on the slow (very slow?) side of conventional scopes. However, it is PC
based, and can capture data, as you asked...
Maybe I'll be able to post some pics of traces from the PC based scope
after I get this thing into a box. I'm not sure how big the pics of traces are,
but I know we have a posting limit of 1MB. So, no promises...
-- Robert |
|
|
| AndrewT |
Hi Audiorob,
your PC scope will get you by. But it will miss out a lot of important information, either due to too infrequent sampling or due to software smoothing over the data to generate the pics to be displayed.
Buy yourself a real, non digital, scope about 20MHz to 50MHz for audio testing.
Traditionally, a digital scope needed a sampling frequency of about 10 times the maximum display frequency to achieve accuracy and resolution. I am told that modern digital scopes can do a lot better than this. |
|
|
| audiorob |
Here is a pic of the main board and the small boards for the power supply circuit.
They are sitting on one of the heat sinks. The heat sinks for the main transistors
are 12"x19.5", but I think they need to be a little be bigger.
Although it is not shown, the main board is mounted in the center of the heatsink
with the transistors wired to the green terminal blocks. With about 2 inch leads,
the transistors are mounted on the heatsink with equally spacing. The front of the
board (with the POTS) then faces the top (or bottom) of the heatsink, so the
POTS are fairly easy to reach. Note that both POTS are on the same side of
the board so that I don't have to move very far to make adjustments.
You might be able to see that the caps on the main board are not an exotic brand.
Does anyone know if, say Black Gate caps would really make a difference? I
think the BG caps are about $2.50 each, but I haven't tried them because there
are minimum orders... Also, I have no idea what Nelson means when he mentions
bypass caps, so if they make a difference, and if anyone would like to enlighten
me, I need some schooling.
On the other hand, I have learned a lot from doing this project. And it has been
a lot of fun! Pics of the amps in boxes soon...
-- Robert |
|
|
| Mad_K |
| quote: | Originally posted by audiorob
H I
think the BG caps are about $2.50 each, but I haven't tried them because there
are minimum orders... Also, I have no idea what Nelson means when he mentions
bypass caps, so if they make a difference, and if anyone would like to enlighten
me, I need some schooling. |
The reason we are using plastic film (polyester, polypropylen) bypass caps across electrolytic caps is that the electrolytic caps becomes more inductive with higher frequencies. Placing a film cap (1uF is a good value) across the electrolytic lets the higher frequencies bypass the electrolytic via the film cap.
Conclusion: Much better sound in the mid/high frequencies. Some people actually likes the more romantic sound you get from not bypassing the el.cap. Using a tantalum electrolytic makes the sound even more "romantic". |
|
|
| Zen Mod |
| quote: | Originally posted by audiorob
Here is a pic of the main board and the small boards for the power supply circuit.
They are sitting on one of the heat sinks. The heat sinks for the main transistors
are 12"x19.5", but I think they need to be a little be bigger.
Although it is not shown, the main board is mounted in the center of the heatsink
with the transistors wired to the green terminal blocks. With about 2 inch leads,
the transistors are mounted on the heatsink with equally spacing. The front of the
board (with the POTS) then faces the top (or bottom) of the heatsink, so the
POTS are fairly easy to reach. Note that both POTS are on the same side of
the board so that I don't have to move very far to make adjustments.
You might be able to see that the caps on the main board are not an exotic brand.
Does anyone know if, say Black Gate caps would really make a difference? I
think the BG caps are about $2.50 each, but I haven't tried them because there
are minimum orders... Also, I have no idea what Nelson means when he mentions
bypass caps, so if they make a difference, and if anyone would like to enlighten
me, I need some schooling.
On the other hand, I have learned a lot from doing this project. And it has been
a lot of fun! Pics of the amps in boxes soon...
-- Robert |
first-AVOID terminal blocks for any semiconductor,especially output transistors;just solder them
second- bypass caps are small ones ,soldered in parallel to bigger ones;they are ,presumably, there to speed up things (ya know-mambo jumbo about parasitic inductivity in large caps etc) ;in some cases there is benefit in bypassing,in some other cases isn't ;
your ears will decide ;try with some 1uF bypass film caps,then decide what you like. |
|
|
| audiorob |
First, a big thanks to Mad_K and Zen Mod for the explanation and information. I
will add experimenting with the caps to my "to do" list. Thank you again!!!
Zen Mod - initially, I didn't understand the jibes about the mess (this is my first
home built amp). Now, I understand. :) In my case, I have a mess in about
5 different places where I work (not enough room at a single work place!). I
still have a huge mess to clear. But I'm going to listen to this thing first.
This is just a pic of me biasing one monoblock. The "bonnet" (heatsink) is up
so I can better reach the POTS and take temps.
-- Robert |
|
|
| audiorob |
This iis a trace of a 20KHz square wave through my Zen 5 amp. I
admit that I thought that the traces would be more square, but again,
this is not my field of expertise.
Does anyone know if this is good, bad, or normal? Is there a
formalized or generally accepted method of performing square wave
tests? I may not have performed this test correctly. I only have the
vaguest idea of what I'm doing. For example, this trace has no load
except for the O-scope. Traces into an 8 ohm load look far worse,
but that is true when applying the signal through the resistor alone
(no amp, just the resistor).
Assuming I have done this correctly, I wonder if this would improve with
the bypass caps. Does anyone know?
BTW, this trace is from the USB/PC based scope mentioned in previous
posts.
Thanks,
Robert |
|
|
| audiorob |
Here is another trace, this time using 1KHz triangle waves, just before the
amp clips. The yellow trace is the input. The magenta trace is the amplifier
output connected to an 8 ohm resistor load.
So, do we call this a 6V input or a 3V input?
The amp is swinging +/- 17V. If wattage is VA, then we need to know
A. From the equation V=IR, we know V and R. I'm guessing that we use
V=17 and not V=34. So:
17V = I * 8 ohms
17V/8 ohms = I
2.125 Amps = I
Then using W=VA we have:
W = 17V * 2.125 Amps
W = 36.125
Does this sound about right to everyone? (I have to ask because *I* would
have said that V should have been 34. But using that value in the equations
yeilds 4.25A and 144.5 Watts, which seems unreasonable.)
Does that look right?
thanks,
Robert |
|
|
| Zen Mod |
looks good enough to me,taking in account that you have pretty crappy 'scope ;)
find attached one pdf ,author can be known from file properties-it's bergerons ;
look for caps named C102 ,C104,C105
102 and 105 are there for stability reasons,and 104 is there to shape your rectangle waves.........
btw-I like that future case for your amp |
|
|
| Mad_K |
| quote: | Originally posted by audiorob
Assuming I have done this correctly, I wonder if this would improve with
the bypass caps. Does anyone know? |
Not likely.
| quote: | Originally posted by audiorob
The amp is swinging +/- 17V. If wattage is VA, then we need to know
A. From the equation V=IR, we know V and R. I'm guessing that we use
V=17 and not V=34. So:
17V = I * 8 ohms
17V/8 ohms = I
2.125 Amps = I
Then using W=VA we have:
W = 17V * 2.125 Amps
W = 36.125
Does this sound about right to everyone? (I have to ask because *I* would
have said that V should have been 34. But using that value in the equations
yeilds 4.25A and 144.5 Watts, which seems unreasonable.)
Does that look right?
thanks,
Robert |
To get W RMS you need to divide your answer (36W) by 2. So yes, it is correct (18W RMS) ;) |
|
|
| AndrewT |
Hi,
when defining your triangle wave you can use peak to peak (Vpp) or just the ref to peak voltage (Vpk). If it was asymetrical you would define the DC offset as well.
Peak power into a load using sinewave signal is P = Vpk^2 / R = Ipk^2 * R = Vpk * I. All three give the same answer. Use the version for which you know the operands.
The power is the peak power divided by two, P = Ppk / 2 = Vpk^2 / 2R.
Your DMM reads rms (AC average) when doing these measurements.
In this case the power using rms measurements is P = Vac^2 / R, omitting the divide by two. |
|
|
| audiorob |
Zen Mod said:
> btw-I like that future case for your amp
The cases are completed, and the amps are installed. I have been listening
to it for a couple of days.
-- Robert |
|
|
| Zen Mod |
what's temp of these barbecue bars?
:clown: |
|
|
| audiorob |
First - thanks AndrewT and Mad_K for looking at my wattage calculations.
Now for the square wave stuff. Mad_K was right that the bypass caps
made no difference for the square wave (or at least the square wave at
20KHz).
After looking at the Aleph 30 schematic (in the Zen Mod posting), I took a
guess at tried a 10pf, 12pf, and 15pf mica cap around R2. This also
seemed to make no difference in the 20KHz square wave. I did connect
and test a Harmon Kardon Citation 16 amplifier. I used the same
methodology as I used on the Zen 5, and the Citation produced a very
nice 20KHz square wave. So I think my method of testing is ok.
I guess I'm curious about _why_ the 20KHz square wave is not so
square. I would have expected this amp to be able to reproduce higher
frequencies than this. Are my expectations unrealistic? Should I care
about this at all? If I should care, what do I do about it?
thanks,
Robert |
|
|
| audiorob |
| quote: | Originally posted by Zen Mod
what's temp of these barbecue bars?
:clown: |
Well, I do not believe my meter measures aluminum correctly. However,
I believe that when idling, the steady temp is about 50-55C. When playing
music, the steady state temp is between 40 and 45C.
-- Robert |
|
|
| Zen Mod |
and what's exact schematic of yours aleph?
somewhere (probably in input LTP) you have wimpy current,or....what type of output mosfets you use and how many?
for test-try disconnecting pair of output mosfets (re-set overall current) and then look have you any changes in squares....
btw-is the same case exactly with both channels?
second btw-you can try to disconnect AC modulation of CCS (220UF cap) ,to help us in determining what part is responsible for this rounding- CCS or "active" half of output.
when I think a little ,you can try this first,disconnecting pairs of mosfets second |
|
|
| Babowana |
| quote: | Originally posted by audiorob
I guess I'm curious about _why_ the 20KHz square wave is not so
square. I would have expected this amp to be able to reproduce higher frequencies than this. Are my expectations unrealistic? Should I care about this at all? If I should care, what do I do about it? |
The rounding off looks somewhat too much at 20kHz. Do you have the square wave at 100kHz, too? I think that the rounding off is mainly due to internal capacitance on the signal path combined with resistance, which form a kind of integrator so that high frequecy roll off starts too early. Probably, you need to carefully look into whether there is any additional internal capacitance in the signal path . . . accidently added . . . |
|
|
| AndrewT |
Hi,
the input filter will round off the HF square waves.
It is normal to inject test waveforms after the input filter. However some amplifiers become unstable when the input is loaded this way.
I see your sample rate is 1Mb/S. This is far too slow to detect oscillation. You only have about 24 samples per half wave. Is this a clue to the rounding? Software taking a guess at what it thinks it should show you? |
|
|
| audiorob |
| quote: | Originally posted by Zen Mod
and what's exact schematic of yours aleph?
somewhere (probably in input LTP) you have wimpy current,or....what type of output mosfets you use and how many?
<snip>
btw-is the same case exactly with both channels?
<snip>
|
This is a Zen 5 (complementary Zen), that was built exactly as specified (to the
best of my ability) in Nelson's paper, with the following exceptions:
1) Almost all of my resistors are 1/2 watt, not 1/4 watt. R1 and R2 are 1/4 watt.
All resistors are 1%. R7 and R8 are matched. R5 and R6 are matched.
2) R9-R10 were replaced by a single 0.27 ohm 15W resistor, as well as R11-R12.
These are also matched.
3) The power supply caps C4-C9 are 40,000uF.
For completeness sake:
The transformers are 1KVA, 30V secondaries. I am regulating the power
supply rails at +/- 30V. The power supply uses aluminum bus bar, and 12
gauge wiring, except for the voltage regulation section. I used 16ga. wire to
the voltage regulation transistors drain and source, and 24ga. wire to the gates.
I used 16ga wire from the power supply to the amplifier board. I think I have
posted pictures of the power supply before putting them into boxes, but let
me know if we need more.
The amplifier circuit uses point to point wiring, and is half the size of Nelson's
schematic of the circuit (Fig. 3). I do use terminal blocks to connect the
transistors to the circuit, for power to the circuit, and for the input. I used
24ga wire for the inputs, and 16ga wire for the output. All wiring is stranded
copper, and are reasonably short lengths.
The transistors are IRFP240 and 9240's (one of each for amplification, one
of each for voltage regulation).
Yes, the square waves traces look the same between the left and right
channels (on the bright side, I'm consistant!). I do not know what an "input
LTP" is, so I don't know how to address that issue.
Here is a picture of the board, mounted on the heat sink, and connected
to the FETs. The wires to the FETs are about 2.5 inches long. The heat sink
is 12 inches tall by 19.5 inches wide. The board is about 3 inches by 5 inches.
thanks,
Robert |
|
|
| AndrewT |
Hi,
S P A C E D out, man!
would it help if the board was quarter of the area? 50mm by 60mm.
Should the FET gate resistors be attached to the gates?
NOT tens of cm away. The resistor body should be electrically within a few millimetres of the gate exit from the plastic package.
Is this the heatsink that lies flat instead of standing up?
How thick is the backplate?
A good guide for effective heat distribution is that ten times thickness is the effective limit for heat conduction to the extremities. Any further and the heatsink efficiency falls off. |
|
|
| Zen Mod |
| quote: | Originally posted by audiorob
This is a Zen 5 (complementary Zen), that was built exactly as specified (to the
best of my ability) in Nelson's paper, with the following exceptions:
1) Almost all of my resistors are 1/2 watt, not 1/4 watt. R1 and R2 are 1/4 watt.
All resistors are 1%. R7 and R8 are matched. R5 and R6 are matched.
2) R9-R10 were replaced by a single 0.27 ohm 15W resistor, as well as R11-R12.
These are also matched.
3) The power supply caps C4-C9 are 40,000uF.
For completeness sake:
The transformers are 1KVA, 30V secondaries. I am regulating the power
supply rails at +/- 30V. The power supply uses aluminum bus bar, and 12
gauge wiring, except for the voltage regulation section. I used 16ga. wire to
the voltage regulation transistors drain and source, and 24ga. wire to the gates.
I used 16ga wire from the power supply to the amplifier board. I think I have
posted pictures of the power supply before putting them into boxes, but let
me know if we need more.
The amplifier circuit uses point to point wiring, and is half the size of Nelson's
schematic of the circuit (Fig. 3). I do use terminal blocks to connect the
transistors to the circuit, for power to the circuit, and for the input. I used
24ga wire for the inputs, and 16ga wire for the output. All wiring is stranded
copper, and are reasonably short lengths.
The transistors are IRFP240 and 9240's (one of each for amplification, one
of each for voltage regulation).
Yes, the square waves traces look the same between the left and right
channels (on the bright side, I'm consistant!). I do not know what an "input
LTP" is, so I don't know how to address that issue.
Here is a picture of the board, mounted on the heat sink, and connected
to the FETs. The wires to the FETs are about 2.5 inches long. The heat sink
is 12 inches tall by 19.5 inches wide. The board is about 3 inches by 5 inches.
thanks,
Robert |
you know for proverb "short circuit between two earphones" ?
that's exactly what happened to me-I meant all the time on Aleph 5..........
I'll see Zen5 schmtic and rethink.....
btw- tip regarding gate stoppers is worthwhile |
|
|
| audiorob |
| quote: | Originally posted by Babowana
The rounding off looks somewhat too much at 20kHz. Do you have the square wave at 100kHz, too? I think that the rounding off is mainly due to internal capacitance on the signal path combined with resistance, which form a kind of integrator so that high frequecy roll off starts too early. Probably, you need to carefully look into whether there is any additional internal capacitance in the signal path . . . accidently added . . . |
I have not tried 100kHz. 20kHz is the highest that I tried.
Thanks Babowara, it was my thought too that some type of RC circuit was
causing the rounding. Here are some things that I know could cause more
resistance or capacitance than expected:
Cold or sloppy (generally bad) solder joints.
Long, or heavy guage wire.
Coils of wire.
Overheating wire.
I'm very much a novice at this, so I don't know if that list is exhaustive. Are
there other things I should try to find and avoid?
Now that I'm thinking about it, I do have a 16ga wire that runs from the drain
of 1 transistor to the drain of the other transistor (about a 3 inch length) with
a couple of solder connections to it (pretty much as it is depicted in the
Fig. 3 schematic in Nelson's article). I used 16ga wire there because I thought
that would be the section that has the most current. Might that be a possibility?
If it is, then surely the 12 inches (or so) of 16ga wire I use to connect the board
to the speaker bananna plugs would be a problem too. As would the test leads
on some of my probes. Maybe that is not the problem area. BTW, I tested a
40 foot length of Monster 11ga wire with 8 connectors will pass a 20kHz square
wave fairly accurately. Based on that, doesn't the 3 inch length of 16ga wire
sound unlikely to be the cause?
I would guess the maximum current the Zen 5 could deliver to the speakers
is about 3 amps. I'm guessing that I should use the smallest guage wire that
will carry 3 amps without getting too hot. Does that sound about right? Does
using 16 to 24ga stranded wire on this project sound reasonable?
And thinking about that, this project is so small, and the parts are so close
together that I only used about 3 wires on the board. The parts are pretty
much soldered together by their leads. And their leads were even cut to
shorter lengths. I guess there are 6 wires that go to the transistors, but I
wasn't counting those, since they connect to the board by a terminal block
(mentioned in the last post).
Has anyone else ever tried tracing a square wave through their Zen 5? If
so, what did you get?
thanks,
Robert |
|
|
| Babowana |
| quote: | Originally posted by audiorob
Yes, the square waves traces look the same between the left and right channels (on the bright side, I'm consistant!). |
The last picture looks like Mel Gibson on the cross on the wall. :D
If the music is ok . . .
I would forget the sq. waves . . . |
|
|
| Babowana |
The somewhat long leads might create a "stray-wiring capacitance", but it could degrade the very high-frequency response . . . e.g. > 100kHz or higher . . .
I was thinking more about possibility of unexpected additional capacitance between the gate and the drain. |
|
|
| audiorob |
| quote: | Originally posted by AndrewT
Hi,
the input filter will round off the HF square waves.
It is normal to inject test waveforms after the input filter. However some amplifiers become unstable when the input is loaded this way.
I see your sample rate is 1Mb/S. This is far too slow to detect oscillation. You only have about 24 samples per half wave. Is this a clue to the rounding? Software taking a guess at what it thinks it should show you? |
Thanks AndrewT.
I don't know that this circuit has any input filter (again, I'm a novice at this,
so I could be wrong).
WRT the sampling rate and precision of the scope (just to make sure we're on
the same page) the manufacturers state that the hardware is taking 1 million
samples per second, with a word size of 12 bits (12 bits per sample). Yes, I
agree that is about 20 samples per division, and about 24 samples per half
wave.
I don't know if the software interpolates data points, or exactly how the program
uses the collected data (beyond the most obvious things). But I did the same
test on a Harmon Kardon Citation 16 amp, and the scope displayed a good
square wave trace. I didn't save the trace, but I could do it again if you're
interested.
I think your point is that the scope is not sampling fast enough to show what
is really happening at the rounded corner, and you're probably right. However,
I think that the scope is showing us what it sees. I also tried some traces of
1kHz and 10kHz. Their patterns look indicative of the 20kHz trace.
Here is a 1kHz trace. The yellow trace is the input and the magenta trace is
the output. The 10kHz trace looks similar, with more rounding at the corner.
thanks,
Robert |
|
|
| audiorob |
| quote: | Originally posted by AndrewT
Hi,
S P A C E D out, man!
would it help if the board was quarter of the area? 50mm by 60mm.
Should the FET gate resistors be attached to the gates?
NOT tens of cm away. The resistor body should be electrically within a few millimetres of the gate exit from the plastic package.
Is this the heatsink that lies flat instead of standing up?
How thick is the backplate?
A good guide for effective heat distribution is that ten times thickness is the effective limit for heat conduction to the extremities. Any further and the heatsink efficiency falls off. |
HI AndrewT,
Thanks again for the replys - and everyone else too!
Yes, the transistors are spaced to better distribute the heat load.
The gate resistors are about 6 or 7 cm (if memory serves me, it is
about 2.4 cm/inch, right?) from the gate. The wire to the gate is 24ga
stranded cu. Is that too far? Even if it were too far, the effect would
not be to round a square wave, would it?
The backplate is 0.3 inches thick. That basicly fits your 10x rule.
And thanks for that. I did not know the 10x rule. Good info.
Yes, it is the heatsink that is the top piece in the pictures of the box.
Each voltage regulator has its own heatsink, which are black, and
mounted on the inside of the box.
thanks,
Robert |
|
|
| AndrewT |
Hi,
the fins within a 3inch radius of each device will work very well. The fins beyond the 3inch radius will work progressively less well as the distance increases.
Unfortunately with your sink lying on it's back the effective fins are surrounded by warm air from the less effective fins and so cooling of ALL parts of the sinks are less effective than what the manufacturer intended.
Any chance you could re-jig your layout to get the fins vertical?
The FET gate resistors MUST be attached very close to the FET package on the shortest possible leg lengths.
The Datasheet gives some clue by specifying the inductance of the leads in the package.
I have seen and used resistors bent right next to their body soldered just 3mm from the package. That is short, much shorter than on a PCB. |
|
|
| Babowana |
| quote: | Originally posted by AndrewT
The FET gate resistors MUST be attached very close to the FET package on the shortest possible leg lengths. |
In the class room, already been many discussions among students with Papa . . .
Shorter distance is better, but somewhat long is still ok . . .
Still ok with respect to the lead inductance . . .
Still ok with respect to the stray capacitance . . .
Still ok with respect to the parasitic oscillation . . .
No unexperienced policeman . . . let's be flexible . . . |
|
|
| AndrewT |
Hi Bob,
are you suggesting that I should not tell him?
or
that my use of MUST is going too far?
or
some other? |
|
|
| Babowana |
| quote: | Originally posted by AndrewT
use of MUST is going too far |
:smash: |
|
|
| AndrewT |
Hi Bob,
thanks.
How far is acceptable? |
|
|
| Babowana |
I have no absolute answer . . .
I have used up to about 200mm-long wire between the gate pin and the 220ohm resistor, without detecting any trouble. |
|
|
| AndrewT |
Hi,
thanks for your experience.
I could never recommend that sort of lead length between the gate resistor and the FET gate.
Even reducing it by a factor of ten to just 20mm would have me worried about oscillation |
|
|
| Babowana |
| quote: | Originally posted by AndrewT
I could never recommend that sort of lead length between the gate resistor and the FET gate.
Even reducing it by a factor of ten to just 20mm would have me worried about oscillation |
You could recommend a short lead . . . as everyone does . . . |
|
|
| Blues |
Looking at internal pictures of an Aleph J, the Gate resistors are nowhere near the mosfets. The Source resistors are right there with the fets though.
Noticed how NP can also make electrons go on tight 90 deg. corners in his pcb tracks? ;)
http://6moons.com/audioreviews/firstwatt3/alephj.html |
|
|
| Babowana |
| quote: | Originally posted by Blues
tight 90 deg. corners |
Papa never recommends it . . .
He keeps his secrets and heads on . . .
err . . . selfish . . . no time to learn new things from others. . . |
|
|
| RoboMan |
Robert,
I was unable to see your photos until I disable my firewall and anti-sypware.
I made two sets of ZenV3. The 20KHz output is very similar to what you've got. So it is absolutely normal to have the round-corners.
Nevertheless, I'm pretty happy with the result considering the simplicity of the circuit and the superb sound quality.
I think it is a common problem with MOS FET that caused by the Gate-Source and Gate-Drain capacitances. The 221 ohm resistor at the Gate together with the Gate-Source capacitance will form a RC circuit. Worse of all is that the Gate-Drain capacitance will provide negative feedback path feeding some of the output back to the gate.
That is why although the Gate-Source resistance is very high, we still need a low impedance source to drive a MOS FET so that we can charge the G-S cap quickly and drain away the feedback signal from the G-D cap.
If you can provide a very low impedance signal source, the output may look better. I think you can't do much to improve Zen5.
Zen V9 may provide a better square wave output as the input is isolated from the output.
Could somebody having a ZenV9 publish a 20KHz output photo for comparison?
You made very nice cases.
One suggestion though:
It will increase the efficiency of your heat sinks if the fins are installed vertically. It is too late to say that now, but may be useful when you are making another one (ZenV9 may be?). |
|
|
| AndrewT |
Hi,
my discussion with Babowana has brought to my attention the possibility that instead of a single pole roll off due to the gate resistor Rgate*Ciss/Cog the long leads from gate resistor to FET gate may have sufficient inductance to form a two pole roll off at a much lower frequency. The new time constants being based on Rgate*Ciss/Cog + Lleads*Ciss/Cog.
Could this be a factor for the rolled off extreme treble?
Maybe oscillation is not a concern with NP's topology. |
|
|
| Babowana |
| quote: | Originally posted by AndrewT
Could this be a factor for the rolled off extreme treble?
|
Once I have made my own capacitor of about 5-10pF with twisted wires of about 60mm length (thanks to Papa’s tip . . .). Based on this experience, I presume that, even if the long leads of the gate and drain (separated each other) create a capacitance, the value would be ignorable low compared with the input capacitance of the MOSFET.
By the way . . . I am with RoboMan (post#45). |
|
|
| audiorob |
| quote: | Originally posted by RoboMan
Robert,
I was unable to see your photos until I disable my firewall and anti-sypware.
I made two sets of ZenV3. The 20KHz output is very similar to what you've got. So it is absolutely normal to have the round-corners.
Nevertheless, I'm pretty happy with the result considering the simplicity of the circuit and the superb sound quality.
I think it is a common problem with MOS FET that caused by the Gate-Source and Gate-Drain capacitances. The 221 ohm resistor at the Gate together with the Gate-Source capacitance will form a RC circuit. Worse of all is that the Gate-Drain capacitance will provide negative feedback path feeding some of the output back to the gate.
That is why although the Gate-Source resistance is very high, we still need a low impedance source to drive a MOS FET so that we can charge the G-S cap quickly and drain away the feedback signal from the G-D cap.
If you can provide a very low impedance signal source, the output may look better. I think you can't do much to improve Zen5.
Zen V9 may provide a better square wave output as the input is isolated from the output.
Could somebody having a ZenV9 publish a 20KHz output photo for comparison?
You made very nice cases.
One suggestion though:
It will increase the efficiency of your heat sinks if the fins are installed vertically. It is too late to say that now, but may be useful when you are making another one (ZenV9 may be?). |
Humm... Lots of topics...
Re: pictures/AV/firewall stuff:
The pictures should not have any viruses, etc. in them. While I admit that the
PC based O-Scope is running MS, I do not connect it to the Internet. I did
transfer (via sneaker net) the images to a Linux box (that is connected to the
Internet - Linux boxes being more secure than MS boxes), to convert the
images to .jpg's and upload them. I guess that is a long way of saying that
I don't think the pics could have viruses, etc. in them. But if I'm wrong, please
let me know.
Re: Zen 3, square waves with rounded corners:
Thanks for the info. I noticed that in some of the projects that
Babowana has on his website show similar traces. It was also interesting
that some of the traces that Babowana captured of lower freq. square waves
(say 50Hz or so) have a trapazoidal shape, where the leading edge overshoots
the target. I checked and mine does the same thing...
I have an X150 that uses the same MOSFETs. I haven't tested it yet, but
I would bet that Nelson has it making perfect 20kHz square waves. I'll verify
that, but if I'm right, I wonder how Nelson gets around this problem?!?!
And thinking of more traces and testing, I built another Zen 5 board for
workbench testing (instead of goofing with the amps that are built and
working). I cannot tell any difference in the traces between the test
board and the "production" boards.
I tried connecting the gate resistors directly to the gates. There was no
noticable improvement. I think this supports the MOSFET capacitance
issues you mention.
I measured the capacitance of the Dale resistors that I used. I was
surprised to find the the 220 ohm 1/2W gate resistors measured 70pf,
while the 47.5K ohm 1/2W feedback resistors measured .33pf.
I tried measuring the capacitance of misc. lengths of misc. gauge
wire that I used, but I was unable to measure any capacitance.
Re: signal source:
I'll need to look for the output impedance of the signal generator I'm
using. As far as music goes, I'm driving the Zen 5's with a Pass
Labs X-1 preamp, which has an output impedance of 200 ohms.
Re: cases:
Thanks about the cases. They match our furniture, and I really like them
too.
Re: heatsinks:
I got tired of fighting the heat, so I went and got a couple of "wall wart"
type of power supplies that have an adjustable DC voltage switch and
supply an amp of current. Then I got some of the quitest 12VDC, 120mm
fans I could find, and wired them to the supply. I turned the supply down to
7.5V and set 2 fans on the big heat sinks (the "top" of the box). You cannot
hear the fans when you ears are 6" from them. I wired the supply to the amp's
switch so the fans turn on with the amp.
BTW, this has made a HUGE difference in the sound quality of the amp. I
think the voltage would fluctuate too much (due to changes in temps) without
the fans. With the fans, the DC offset and the bias current vary about 25%
of the amount they varied without the fans. I was also able to increase the
bias current to 2.5 amps (and nothing has blown yet).
Re: new projects:
I probably will do more amp projects, but not yet as I'm still learning from
this one!!!
Thanks a bunch (and to everyone),
Robert |
|
|
| Nelson Pass |
| quote: | Originally posted by Blues
Noticed how NP can also make electrons go on tight 90 deg. corners in his pcb tracks? |
Quantum electrodynamics teaches that an electron arrives
at it destination through the "sum of all paths" approach,
with some of the electron cutting across the corner and
some of it by way of Andromeda.
:cool: |
|
|
| audiorob |
| quote: | Originally posted by audiorob
<snip>
I measured the capacitance of the Dale resistors that I used. I was
surprised to find the the 220 ohm 1/2W gate resistors measured 70pf,
while the 47.5K ohm 1/2W feedback resistors measured .33pf.
<snip>
|
Uhh - I'm a dork. Please make those units nf (10^-9) not pf (10^-12).
Robert |
|
|
| Babowana |
| quote: | Originally posted by Nelson Pass
Quantum electrodynamics teaches that an electron arrives
at it destination through the "sum of all paths" approach,
with some of the electron cutting across the corner and
some of it by way of Andromeda. |
Ayo . . . some of poor eletrons might arrive at the destination somewhat late, noseblooding after :headbash: against the 90 deg corner wall . . . hehe
What is Andromeda? |
|
|
| RoboMan |
| quote: | Originally posted by audiorob
Re: signal source:
I'll need to look for the output impedance of the signal generator I'm
using. As far as music goes, I'm driving the Zen 5's with a Pass
Labs X-1 preamp, which has an output impedance of 200 ohms.
Robert |
Re: pictures/AV/firewall stuff:
Oh, that is the problems of how this forum works and my settings. Nothing to do with your pictures.
Re: signal source:
I suppose you drove your Zen5 directly with the Sig Gen. Since you have the X-1 preamp ready, why not drive Zen5 through X-1 and see.
The Zen needs a really low impedance source. |
|
|
| Babowana |
| quote: | Originally posted by audiorob
Babowana has on his website show similar traces. It was also interesting that some of the traces that Babowana captured of lower freq. square waves (say 50Hz or so) have a trapazoidal shape, where the leading edge overshoots the target. I checked and mine does the same thing... |
My that amp has an input coupling cap forming a RC network with a R referring to the ground. When I give the input a square wave signal (e.g.50Hz), the network sees the initial input voltage 100%, and then, the cap and the R share the input voltage time after time. That is the reason of the slope on the top part of the responding square wave. If I reduce the size of either the cap or R, the slope increases, indicating the low frequency rolloff is getting faster. And, the shape of the slope (not exactly linear to the microscope eyes) depends on the character of RC time constant.
. . . just my poorly-polished interpretation . . . |
|
|
| AndrewT |
Hi,
is the slope on the square wave a phase indicator?
Which tends to be much more sensitive than response fall off.
A few (tens) degrees of phase advance (error) will be more easily detectable than -0.01db of response error.
Is it phase advance or delay at the bass end? |
|
|
| Babowana |
| quote: | Originally posted by AndrewT
Hi,
is the slope on the square wave a phase indicator?
Which tends to be much more sensitive than response fall off.
A few (tens) degrees of phase advance (error) will be more easily detectable than -0.01db of response error.
Is it phase advance or delay at the bass end? |
Hi,
I do not know further details than in post #53.
I just like to simplify and apply for a simple engineering -- my way.
The input RC network is a leading network.
At fCL (critical low frequency, -3dB), the phase angle is about +45 deg.
At frequency of 0.1fCL, about +84 deg.
At frequency of 0.01fCL, almost +90 deg.
These are with one pole . . . |
|
|
| RoboMan |
| quote: | Originally posted by AndrewT
Hi,
is the slope on the square wave a phase indicator?
Which tends to be much more sensitive than response fall off.
A few (tens) degrees of phase advance (error) will be more easily detectable than -0.01db of response error.
Is it phase advance or delay at the bass end? |
I will interpret the falling horizontal slope on the square wave as a response fall off indicator instead of phase indicator. It does not show phase shift. A rising horizontal slope will indicator increase in low frequency response (or relatively a decrease in high frequency). You can try it with a preamp that having tone controls.
If you want to measure phase shift you have to include the input signal and compare side by side with the output signal. That means you'll need a dual traces oscilloscope, synchronized to one of the signal.
The horizontal line of a square wave respresents the low frequency response and the vertical line represents the high frequency response. You are right that the slope is more sensitive than measuring the real responses (e.g. by a sinewave). |
|
|
| Blues |
| quote: | Originally posted by Nelson Pass
Quantum electrodynamics teaches that an electron arrives
at it destination through the "sum of all paths" approach,
with some of the electron cutting across the corner and
some of it by way of Andromeda.
:cool: |
Yes...also remember these aren't ordinary electrons but Pass electrons in the Aleph Dimension and will always make 90 deg corners. Maverick Aleph electrons that find their way to Andromeda must go through dark matter, of low electron density, where they enslave ordinary electrons to get across lightyears of space and time. They'll miss the party unless these maverick electrons modulate light as carrier and will travel at lightspeed.
Just like gate resistors placed right at the threshold, >90 deg copper traces at corners are IMO over rated.:cannotbe: |
|
|
| audiorob |
| quote: | Originally posted by RoboMan
<snip>
Re: signal source:
I suppose you drove your Zen5 directly with the Sig Gen. Since you have the X-1 preamp ready, why not drive Zen5 through X-1 and see.
The Zen needs a really low impedance source. |
Well, I checked my signal generator's manual, and its output impedance is
50 ohms. As that is less than the output impedance of the X1, I guess there
is no reson to test through the X1. So the search continues...
Robert |
|
|
| Babowana |
.
Robert,
Why you worry about your wave shape of 20kHz?
From your picture, I read the risetime (10%V to 90%V) of about 3.2 uS.
So, the upper critical frequency (-3dB) is:
fc = 0.35/3.2 * 1000 = 109kHz
The amp has bandwidth upto 100kHz.
Isn't it enuf? |
|
|
| AndrewT |
Hi Bob,| quote: | | I read the risetime (10%V to 90%V) of about 3.2 uS |
how much of that rise time is invented by the software or can it be relied on to be a reasonable guide? |
|
|
| Babowana |
| quote: | Originally posted by AndrewT
Hi Bob, how much of that rise time is invented by the software or can it be relied on to be a reasonable guide? |
All tools (incl. my eyes and brain) have tolerances . . .
At the rate of % error . . .
I approximate the risetime myself from his picture, assuming his tool is the one good enough and well calibrated . . . |
|
|
| Babowana |
| O, I calibrate my eyes and brain every morning at 7 o'clock. In general, it takes half an hour. It actually depends . . . |
|
|
| AndrewT |
Hi Bab,
I have seen your F-3 formula or something similar before. F-3db=350/uSrisetime
What limitations are there on it's application?
BTW. I was not referring to your eyes nor your technical ability.
I think (an opinion) his computer is telling him lies at the frequencies and sampling rates he is working at.
Another opinion, he should get a real oscilloscope preferably analogue before he comes to rely on computer generated data too often. |
|
|
| Babowana |
| quote: | Originally posted by AndrewT
What limitations are there on it's application? |
O, I just added kidding on me myself, trying to blow off my own hard time lasting these days. . . :D
It is according to the book. The amplifier typically has one dominant bypass (RC) circuit that roles off the voltage gain at a rate of 20dB per decade until f(unity) is reached. The critical frequency (fc) of this bypass circuit is given by
fc=1/(2x3.14x RC)
where RC = time constant of the internal RC circuit.
In general, calculation of the internal RC value is uneasy . . . Without knowing the hidden RC value, however, we could find out the fc . . . using oscilloscope measurement.
We apply a square wave at "any" certain high frequency to the amp's input and see output response through the oscilloscope. Do we see that the vertical leading edge of the square wave appears rounding-off . . . ? In this condition, we could measure the risetime (from 10% to 90% amplitude), which is always 2.2 times of the time constant, i.e. the measured Tr = 2.2RC. Then, RC=Tr/2.2 . . . We bring this into the above fc calculation formula.
fc= 2.2/(2x3.14xTr) = 0.35/Tr
I think that there is no other condition we need to consider.
We could do the same with the lower critical frequency, measuring the fallingtime (the top line of the square wave). |
|
|
| AndrewT |
Hi Bab,
thanks for that and where it comes from.
What if the input RF filter and the amplifier HF response nearly co-incide?
The falling time, is that the slope on the top of the square wave? |
|
|
| Babowana |
| quote: | Originally posted by AndrewT
The falling time, is that the slope on the top of the square wave? |
Yes, it is also the RC curve . . . |
|
|
| AndrewT |
Hi, Bab,
looking at a 100Hz square wave with a peak to peak height of 10Vpp.
The slope on the top is about 10% of the peak height i.e. 500mV
The half cycle takes 5mS.
The slope has fallen 10% in 5mS so would fall 80% in 40mS.
RC=40mS/2.2 or 40mS*2.2? which?
High pass frequency (-3db) = 1/2/Pi/RC and for 40mS, F-3=8.8Hz or 1.8Hz. |
|
|
| Babowana |
| Pls, refer to this for the low fc. |
|
|
| audiorob |
| quote: | Originally posted by Babowana
.
Robert,
Why you worry about your wave shape of 20kHz?
From your picture, I read the risetime (10%V to 90%V) of about 3.2 uS.
So, the upper critical frequency (-3dB) is:
fc = 0.35/3.2 * 1000 = 109kHz
The amp has bandwidth upto 100kHz.
Isn't it enuf? |
Hi Babowana,
It is not that I am so worried about the 20kHz square wave, or the
bandwidth of the amp. The amps sound surprisingly good as they
are. It is more that this project is a learning tool for me (ok, and lots
of fun too), and the square wave traces do not match my expectation.
I would have expected a 20kHz square wave to look perfect through
such a simple (as in very few components) amp that uses modern,
switching type of FETs. Also, 1970 vintage amps had better square
wave performance than what my Zen does. This is not my field of
study, so maybe my expectations are unrealistic. I do not understand
why the traces are not perfectly square at 20kHz, but I would like to
understand.
Also, I do not have any distortion testing equipment, so I was using the
equipment I have to kind of "eyeball" the amp's performance. I have a
signal generator and a very slow O-scope. I looked at traces of sine,
triangle, and square waves. Square waves showed the biggest change
(do we call this distortion?) from the original at the lowest frequency of
the different wave shapes. I was surprised (not disappointed, just
surprised) that the amp did not reproduce a better square wave trace
(but what do I know).
So next I wondered if I had done something wrong when building the
amp. I still don't _know_ the answer to that, but I'm leaning towards
what I see is the normal behavior of the amp, and that I built the amp
correctly. But if I did not build the amp correctly, I want to discover
what I did wrong and correct it.
If I built the amp correctly, I would like to research how the amp could
be modified to produce a better square wave. I would like to listen
to the amp to hear if that makes a difference. Similarly, I would like to
experiment with stabilizing the current source (and voltage) maybe
by cascoding transistors. How would this sound? Would it
reduce the amount the bias and DC offset drift? I expect that this might
lead to another architecture, but I don't know. I think I would like to
play with some of these things, and gain a better understanding of what
they really do.
So is a 100kHz bw enough? Well, yes (again, I am surprised how
good this amp sounds) and no (I hope to learn more from this project). :)
Thank you for your bw calculation and all your comments.
Robert |
|
|
| Babowana |
| quote: | Originally posted by audiorob
. . . Thank you for your bw calculation and all your comments . . .
. . . The amps sound surprisingly good as they are . . .
. . . the amp did not reproduce a better square wave trace . . . |
Not at all, I just enjoying in sharing our experiences . . . :D
Good sound? Good motivation for me . . .
If we reduce the gain of 10 to 5, by increasing R1 from 470 to 1K, the rounded leading edge would be better squared off . . . at high frequencies. |
|
|
| audiorob |
| quote: | Originally posted by Babowana
Not at all, I just enjoying in sharing our experiences . . . :D
Good sound? Good motivation for me . . .
If we reduce the gain of 10 to 5, by increasing R1 from 470 to 1K, the rounded leading edge would be better squared off . . . at high frequencies. |
I tried it, and you are quite right. BTW, I think the square
waves look better across the low freq. too (better across
the 20-20kHz range).
But does this also mean that even more voltage will be
required to drive this thing? With R1 at 470 ohms, my amps
require almost 4V to get full output. Would the impedance
matching input buffer of Zen 4 work for us too?
Robert |
|
|
| Babowana |
| quote: | Originally posted by audiorob
Would the impedance
matching input buffer of Zen 4 work for us too? |
I think so . . . |
|
|
| audiorob |
| quote: | Originally posted by Babowana
I think so . . . |
I've been trying ot build an input buffer (or a 1st stage of amplification),
but I'm confused by the voltage I'm getting from the output stage.
Nelson mentions that the gain of the Zen 5 amp is slightly less than
R2/R1. With the specified values of R1 and R2, that was a gain of
slightly less than 10. It was taking a little more than 3V (6V peak to
peak) input to drive my Zen 5 to full output. But full voltage output was
at about 17V (34V peak to peak). This is a gain of about 6. I really
don't expect to get 30V (60V peak to peak) output, but I expected to get
higher than 17V.
Also, the spice simulations that I've done show an output > 17V (but
maybe I'm doing something wrong in spice).
I've also dropped the gain of the output stage by increasing the value of
R1 (to increase damping factor, and to raise the impedance of the input
to the output stage) and added a first stage voltage amplification stage.
I'm not overdriving the first stage, yet I still hit the 17V wall for the output.
And the spice models for this show that I can go to 28V.
Can anyone give me a clue about why the output is limited to 17V ???
thanks,
Robert |
|
|
| Babowana |
| quote: | Originally posted by audiorob
Nelson mentions that the gain of the Zen 5 amp is slightly less than R2/R1. With the specified values of R1 and R2, that was a gain of slightly less than 10. It was taking a little more than 3V (6V peak to peak) input to drive my Zen 5 to full output. But full voltage output was at about 17V (34V peak to peak). This is a gain of about 6.
Can anyone give me a clue about why the output is limited to 17V ???
|
Since the gates of the gain MOSFETs are assumed as somewhat relaxed virtual grounds due to the voltage-dividing resistors, I have estimated that the amp gain would be less than 10, but must be greater than 8, i.e. in-between 8 and 10 (20V rails).
I will try to measure the gain with my ZV5 as soon as I fly back to Shanghai this weekend . . . and will post the result in my thread . . . |
|
|
| Zen Mod |
| quote: | Originally posted by Babowana
........
I will try to measure the gain with my ZV5 as soon as I fly back to Shanghai this weekend . . . and will post the result in my thread . . . |
and then you can try to respond on thousand mails from Babothanchoky to Babothanyou ;) |
|
|
| Babowana |
| quote: | Originally posted by Zen Mod
thousand mails
|
Err . . . happy hours soon over . . .
Sure, i was 3-weeks away from mail boxes . . .
I expect thousands mails . . .
And, two-double spam mails . . .
Will respond to all! :smash: |
|
|
| audiorob |
Do you guys use spice to model your circuits? If so, is spice accurate?
The reason I ask (in addition to what I mentioned in my previous post)
is that Nelson said that the output impedance of the Zen 5 was (high at)
about 2 ohms. The spice model I have shows the output impedance to
be 6.774720e-01 (call it 0.7) ohms. I believe that Nelson is VERY
careful about what he writes, so I would bet on Nelson's answer. And
if spice is wrong, how much can you really use spice to model these
kind of circuits???
As I understand it, damping factor is the output impedance / the
nominal impedance of the speaker (or close enough...). In the paper,
Nelson mentions that the damping factor is about 4 for the Zen 5. I
assume that is the output impedance of about 2 ohms divided by a
"normal" speaker's impedance of about 8 ohms. If spice is correct, that
brings the damping factor value to about 11.
Nelson also said that you could increase the value of R1 to increase
the damping factor (at the expense of gain). But the demo version
of pspice shows that distortion also increases with increases in R1,
while spice show that it decreases. However, the demo version of
pspice does not allow you to use IRF240 and IRF9240 MOSFETS. The
closest part it allows you to use is IRF9140 and IRF150 MOSFETS,
which is what I used in my pspice model. In spice, I used the IRF240
and IRF9240 parts.
Oh, and no, I am not going to spend the $6000+US on an unlocked version
of pspice that would allow using the IRF240/9240 parts. So, other than
that, does anyone have any comments?
thanks,
Robert |
|
|
| steenoe |
| quote: | | And, two-double spam mails . . . |
The spammers are to be ......killed, sorry!!
What can we do to stop the spammers??? If anyone has a good idea, let us know:)
Steen:) |
|
|
| Zen Mod |
| quote: | Originally posted by steenoe
The spammers are to be ......killed, sorry!!
What can we do to stop the spammers??? If anyone has a good idea, let us know:)
Steen:) |
nice to see ya back;
probably missed some announce,but -I hope that this silence from Steen Master was just another funny vacation ;)
anyway-with POP accounts e-mail remover is hell of a proggy - I use it everyday.........
but-if you use (as you use ;) ) web mail account ........then I dunno what to recommend,except one nice '45 and frrrrrtload of booooolitzzzzz
;) |
|
|
| Babowana |
| quote: | Originally posted by audiorob
Nelson mentions that the damping factor is about 4 for the Zen 5. I assume that is the output impedance of about 2 ohms divided by a "normal" speaker's impedance of about 8 ohms. If spice is correct, that brings the damping factor value to about 11.
|
Pondering on the coffee mug . . .
The output impedance must be decided by the source resistor, transimpedance of the FET and the voltage gain figure. I believe that the original ZV5 has an actual output impedance of about 2 ohms.
I wonder if Papa would say something different . . . |
|
|
| audiorob |
| quote: | Originally posted by audiorob
As I understand it, damping factor is the output impedance / the
nominal impedance of the speaker (or close enough...). In the paper,
Nelson mentions that the damping factor is about 4 for the Zen 5. I
assume that is the output impedance of about 2 ohms divided by a
"normal" speaker's impedance of about 8 ohms. If spice is correct, that
brings the damping factor value to about 11.
|
Uhh... Oops!
I'm sure it is obvious to everyone that I have the equation inverted. It
should say that damping factor is the speaker's nominal impedance
divided by the amplifier's output impedance.
I hate when I do that...
Robert |
|
|
| audiorob |
As mentioned in Babowana's, "Papa, I want a zen 5 amp" thread, I did some
experiments with rail voltage. I posted the results there, but they really should
have been posted here, as this is the "problems wth zen 5" thread. That is to
say, sorry for the cross post...
As weird as this sounds (or, *I* do not understand this), this is what I have
discovered. With the "standard" (ie: per Nelson's paper) Zen 5, and re-biasing
for the different rail voltages:
+/- 20V Rails -> +/- 17V (34Vpp) output before clipping
+/- 25V Rails -> +/- 18V (36Vpp) output before clipping
+/- 30V Rails -> +/- 17.8V (35.6Vpp) output before clipping
I guess +/-20V rails is the sweet spot.
Does anyone have an idea why the output voltage does not rise above +/- 18V,
regardless of rail voltage? Could this be due to a max Vgs of 20V for the
IRFP240 and IRFP9240? I guess I need to trace voltages on the gate to see
if that might be it. Other than that, I'm out of ideas...
thanks,
Robert |
|
|
| audiorob |
| quote: | Originally posted by audiorob
As weird as this sounds (or, *I* do not understand this), this is what I have
discovered. With the "standard" (ie: per Nelson's paper) Zen 5, and re-biasing
for the different rail voltages:
+/- 20V Rails -> +/- 17V (34Vpp) output before clipping
+/- 25V Rails -> +/- 18V (36Vpp) output before clipping
+/- 30V Rails -> +/- 17.8V (35.6Vpp) output before clipping
I guess +/-20V rails is the sweet spot.
Does anyone have an idea why the output voltage does not rise above +/- 18V,
regardless of rail voltage? Could this be due to a max Vgs of 20V for the
IRFP240 and IRFP9240? I guess I need to trace voltages on the gate to see
if that might be it. Other than that, I'm out of ideas...
|
Uhh... never mind. My scope seems to have a 25V, apparenty pp scale, that
I thought was +/- 25V. I discovered this while measuring gate and source voltage
with the scope and got strange results. As the source pins are (almost directly)
tied to the rails, I measured rail voltage with the scope and got bogus numbers. I
found the different scale, and... well... there you have it.
Another example of a gibbon at the controls. :)
I haven't measured rail voltage vs output voltage yet. I wanted to post this
first, to hopefully stop any waisted effort caused by my stupidity.
My apologies to any following my bad info.
Robert |
|
|
| audiorob |
Hi all,
As I have the "stock" Zen 5 running and stable, I thought I would summarize some of the things I discovered as I tried this project. Hopefully these notes will help the next person who tries the Zen 5. No, I am not finished experimenting with the Zen 5, but those experiments probably should be in a different thread.
The very first thing I tried to do is to build the zen5 amplifier section only, and use a bench type of power supply to power it. My bench power supply is 30V, 3A. I tried to use what I will call a floating ground (I don't know if that is the right term or not) to get something like this:
-15V-----GND-----+15V
To make a long story short, that did not work. I needed to get another power supply (another thanks to Nelson - what an incredible guy). With the second bench power supply, I was able to get plus and minus voltages, and get the amplifier board working.
While getting that first board biased, it became apparent that I should have put both POTs on the same side of the board, so that I did not have to jump to different sides of the board to set the bias/offset.
Most of my remaining problems (arguably) are related to heat. I discovered very late in the project (uhh, I'm still discovering?) that as the different parts heat, and possibly heat at different rates, the parts vary the voltage and/or current going through them. This variation does not usually do good things for performance (sound quality). So one lesson that I learned is that having certain parts (like the output transistors) synced in temperature will help with bias and DC offset drift.
With regard to the output transistors, I was mostly concerned with dissipating their heat on heatsinks. I thought I would have one big (area wise) heatsink, and put the transistors on opposite ends of it. I made several mistakes in this regard. The output transistors should be mounted close to one another on the heatsink so that their temperature is basically the same. Next, as a rule of thumb, you get the best efficiency from the part of your heatsink that is within 10 times the base thinkness of the heat source (the transistors). My heatsink was very wide and long (like 12"x19" or so), but only .3" think. I should have had a heatsink with a much thicker base.
I made the same mistake with the transistors for voltage regulation. As your power supply rails vary in a similar way due to heat, they should be mounted close to one another on the same heatsink. That heatsink should be sized to handle the heat load of both transistors.
I think it could be argued that the same concept (ie: temperature syncronization) holds true for the source resistors too. This probably does not make nearly the same difference as syncing the output or voltage regulation transistors, but stablizing voltage is incredibly important.
I don't know if others think there is a problem with the amp reproducing square waves or not. In my efforts to reproduce better looking square waves, the length of wires (within reason) made no discernable difference. I cannot say that I fully understand Miller effect, but my guess about issue of the amp rounding 20KHz square waves is due to Miller effect and the high Ciss of the IRFP240 and IRFP9240 transistors. I also think that the "ramped" low frequency (say 100Hz) square waves is due to low input impedance. Increasing the input impedance does improve the look of lower frequency square waves, but it changes a lot of other parameters of the amp too. I am a novice at this. I could be very wrong about both.
As to the issue of sound quality, this is so very subjective that I am hesitant to address the issue. I will try, but please recognize that these are my opinions only. My amp required almost +/-4V (8Vpp) to drive it to full output. I think this makes the music reproduced by the amp to be somewhat uncompressed. To me, this was done so much as to make some quieter passages inaudible. I have added a first stage to my amps, which I think corrects the situation (+/-1.5V in -> full output). However, I think that discussing modifications to the zen 5 is beyond the scope of this thread and I hope to start a new thread on that topic soon. With the changes mentioned above, I think this amp sounds very good - but I'm sure I'm biased. It has very good bass, but I think for my speakers, it could use more damping factor. Simply for comparison, I think the Zen 5 has better bass than my X150. The midrange is also very good. However, *I* have not heard anything with better midrange than an X150, and I still have that opinion. I also think the Zen 5 has very good high end - almost as good as the X150.
So, there it is. I hope this information helpfull to others. In this regard, I would like to thank all of you that helped me with this project. It has been (and still is) a LOT of FUN. I hope to see you in a "zen 5 mods" thread.
Build carefully and be safe,
Robert |
|
|
| Babowana |
| quote: | Originally posted by audiorob
I think the Zen 5 has better bass than my X150. The midrange is also very good. However, *I* have not heard anything with better midrange than an X150, and I still have that opinion. I also think the Zen 5 has very good high end - almost as good as the X150.
|
My experience on J-Low
Bass: very good +
Mid: very good +
Treble: very good +
Warm and sweet
Even at high volume, never bawl at me
Lovely
For your info, today I use input coupling caps of 6.8uF film,
replacing the original 47uF electrolytic
:eguitar: :eguitar: :eguitar: |
|
|
|
