As long as the moderators aren't after youBruno Putzeys said:
it's OK but you can also take a place in the Audio Vendor's Bazaar and talk as much as you want about your products... which I think are very interesting......This can be experimented. Use an oscilloscope to look at the power supply voltage and look for ringing in the 4MHz range. This ringing is very obvious when no damper cap is present.ghemink said:
See whether the BG or the original cap does better.
If you're uncomfortable probing around a "live" circuit, it's best to keep the caps that are on the board.
Bruno Putzeys said:
Hi Bruno,
Thanks again. I have been probing "live" circuits (need to bring a scope from work) but probing a class D amp is quite tricky because of EMC that disturbs the measurement (GND cable of scope probe picks up EMC). However, a 4Mhz signal is much higher than the switching frequency so should be able to see that. The scope I used has a spectrum analyzer built in so resonances maybe traceable even at a relatively low level.
For now, I`ll take your advice and remove those BGs (the sound of the totally unmodded UcD modules is already very good). When I have the time for it, I`ll get in a scope and check. I guess decoupling the OPAMP and use other coupling caps has probably the most effect (if any).
Best regards
Gertjan
ghemink said:
Just my $0.02 - if you already had them in there why not listen to the amp? The only tweaks worth while are those you can hear in your system. Saves you all the probing and validates the tweak where it matters.
Another thing - I read somewhere that Blackgate N and Nx type caps should always be grounded with a wire to system ground, not just relying on the PCB traces. Supposedly this increases their effectiveness. Would this be something that makes placement of the caps easier? It may look like a nasty hack, but according to the data sheet it's a good idea to run a decent gauge wire to ground directly from the cap.
Peter (saving his $$$ for some UCD400s)
Jan-Peter said:
To get rid of all op-amp and elcap choices
I would like to choose the option above and bypass the NE5532/AD8620 stage. My sources (tuner & CD) deliver normally approx. 1V r.m.s., my tube pré has sufficient gain and can deliver high voltage swings, and the bare UCD has a gain of 4.5, so driving the UCD180 to earbleeding levels should be possible.
BUT... I do have some questions about impedance matching.
The tube pré output is unbalanced and inverting. So I would connect signal + to the inverting bare UCD input, and the signal ground to the noninverting bare UCD input. So, what impedance does my pré see at his output. Is it 1.8kOhm or 11.8kOhm?
And I haven't even taken the UCD ground connection into account. Help!
Re-BUT... The output impedance of the tube pré is 1.5kOhm.
If the input impedance is 11.8kOhm, it's maybe possible using short wiring, although not optimal. Or am I completely wrong?
Still, it would be nice to avoid an additional input/buffer stage.
Any comments/advises, please ????
Yves
pburke said:
Hi Peter,
I agree that listening is the most important part. However, I made many changes at the same time so if one change has benefit and the other one cancels that benefit by causing new problems, then I don`t know if anything I did had a positive effect. The only way to find that out is to make one change at a time. Which is a lot of work since I have to do all the mods one by one to verify their effect and I need to have two stereo amps all the time and modify one of them each time to be able to verify the improvement with respect to the previous version. I guess this is the most scientific way to do (assuming that subjective listening is a scientific way
). Actually, I may end up doing it that way. That will also give more valuable input for others.Best regards
Gertjan
Hi,
I wonder a bit about the impedance discussion. It should be quite easy as in the RF area, where one connects outputs and inputs with equal(?) impedances together. The only "loss" is the given by the voltage divider ratio of both impedances.
The same is valid in the audio field. You should observe the input (and cable, parasitic) capacitance to avoid a simple RC-lowpass which may limit the upper frequency. Easy to calculate, isn't it?
The higher the impedances, the higher the susceptibility to EM noise, therefore: shielding and symmetric wiring - see posts before.
I would suggest to open two new threads:
- classD amp wiring (audio connections)
- classD amp power supplies
and to push all related posts into these threads.
The questions and answers are very interesting and the information would be focussed.
Regards, Timo
I wonder a bit about the impedance discussion. It should be quite easy as in the RF area, where one connects outputs and inputs with equal(?) impedances together. The only "loss" is the given by the voltage divider ratio of both impedances.
The same is valid in the audio field. You should observe the input (and cable, parasitic) capacitance to avoid a simple RC-lowpass which may limit the upper frequency. Easy to calculate, isn't it?
The higher the impedances, the higher the susceptibility to EM noise, therefore: shielding and symmetric wiring - see posts before.
I would suggest to open two new threads:
- classD amp wiring (audio connections)
- classD amp power supplies
and to push all related posts into these threads.
The questions and answers are very interesting and the information would be focussed.
Regards, Timo
The unbuffered UcD should be seen as a one-opamp differential amplifier with 1.8k input resistors and 8.2k feedback resistors. Together with the on-board buffer it forms an instrumentation amplifier circuit. The gain of the UcD alone is roughly 8.2k/1.8k=4.5.Tangui said:The tube pré output is unbalanced and inverting. So I would connect signal + to the inverting bare UCD input, and the signal ground to the noninverting bare UCD input. So, what impedance does my pré see at his output. Is it 1.8kOhm or 11.8kOhm?
And I haven't even taken the UCD ground connection into account. Help!
Re-BUT... The output impedance of the tube pré is 1.5kOhm.
If the input impedance is 11.8kOhm, it's maybe possible using short wiring, although not optimal. Or am I completely wrong?
Still, it would be nice to avoid an additional input/buffer stage.
If you remove the input op amp, you get to see different input impedances. The impedance on in+ is 10k (=1.8k+8.2k), the impedance on in- is 1.8k to a "virtual source" V=0.82*V(in+).
The dirty method is to drive the noninverting input of the UcD and use the inverting input as reference. This gives the easiest load for your preamp.
CMRR will be severely degraded though. To retain cmrr, both inputs must be driven from sources with an identical impedance. If your preamp has a known output impedance (certain resistance in series with a certain capacitance), make a "dummy" output that is the same impedance tied to the preamp's local ground. This provides you with a balanced output, where the "hot" output is that from the preamp, the "cold" output is the dummy output.
As noted somewhere else in the thread, a balanced output doesn't have to mean symmetrical signals. One signal may be ground but the impedances on both lines have to be the same.
Then you can proceed to the UcD through a balanced cable as described elsewhere.
Actually this measure is a good idea even if you use the UcD with its onboard buffer present, but it becomes indispensible if you try to live without it.
You quote the preamp as having an output impedance of 1.5k. The gain of the power amp becomes 8.2k/(1.8k+1.5k)=2.5 times.
This means a lot of voltage from your preamp. Unfortunately the preamp's output impedance is most probably non-linear (cathode follower), so under these conditions distortion may become unexpectedly high.
On the subject of tube vs solid state preamps I'd like to note that in my experience, where solid state preamps "go wrong" is at the potentiometer. Owing to their very linear grid capacitance, tubes interface exceedingly well with high impedances such as pot meters. This is their strong point. Elsewhere in the signal chain, op amps hold their own well against tubes. Because of that, you may find that using the op amp on the UcD board is most likely the optimum choice.
Thank you, Bruno, for all those precisions.
A gain of 2.5 would become too less, a "dummy" output and XLR-out is not an option, so I'm afraid I'll have to stick to the existing design .
Not a problem, though. It will make life easier .
In this case, I'ld go for a bi-fet low Ibias opamp with 100k input resistors. Cables: RCA cinch, no XLR.
Yves
A gain of 2.5 would become too less, a "dummy" output and XLR-out is not an option, so I'm afraid I'll have to stick to the existing design
.Not a problem, though. It will make life easier
. In this case, I'ld go for a bi-fet low Ibias opamp with 100k input resistors. Cables: RCA cinch, no XLR.
Yves
This is one of the better known differences between half-bridge and full-bridge operation. This pumping is not a big problem though.Konrad said:
Most people get scared of "pumping" because they saw it happening when the amp is connected to a lab supply. Because a lab supply has no storage caps at the output, the effect is very strong (small charge into no cap is large voltage).
If you have a normal supply with normal supply caps (min. 4700uF per supply per channel), pumping is very minimal even at 10Hz (unlikely to exceed 5V).
If people simply tried this experiment it'd save them a lot of time panicking about it.
Bruno Putzeys said:
Hi Bruno,
Yes I understand your point. Still I can`t resist the idea of bridging two UcD180 modules using a symmetrical signal and connecting V+ of the source to the + input of one amp and to the - input of the other amp and the V- signal to the - input of the first amp and the + input of the other amp. Of course bypassing the opamps and the coupling caps. Power supply pumping issues are then a non-issue. The gain per UcD is about 4.5, but it doubles by bridging them. With 3V RMS out of the preamp, I can get 27V RMS out of the bridged setup. With a 4Ohm load this should be just OK to handle with the UcD modules. A total of 180W RMS would be delivered this way and the current that each UcD delivers would be the same as with one UcD in a 4Ohm load driven to 27V RMS out. Of course the input impedance would now be about 900Ohm for both the + and - input. I hope the OPA2134 in my preamp can handle that load, I`m afraid the distortion goes up quite a bit for the OPA2134 with a 900Ohm load since the datasheet of the OPA2134 shows a clearly increased distortion with a 600Ohm load in comparison with a 2k load.
On the other hand, the NE5532 is also seeing almost the same load (1kOhm) when driven symmetrically. And others have reported good results with the OPA2134 and AD8620 that all would see a 1k load when driven symmetrically.
So the above idea may work.
Best regards
Gertjan
I'm back.Bruno Putzeys said:
On the face of it, "getting an SMPS to work" may be simpler than UcD (as the large amount of working power supplies of various quality shows), making a really good SMPS is not to be sneezed at.
Off the shelf (OTS) SMPS's tend to have the following failings:
1) Maximum peak power = average power
A 100W power amp draws 200W peaks. An OTS supply will therefore have to be selected for 200W. Of course it'll be overrated then, too large and too expensive. I'm unaware of SMPS's where thermal design and regulation (sagging instead of cutting out) are optimized for audio use. As a result, using a standard SMPS will make your product lighter but much bigger than a 50 Hz transformer!
2) Terrible EMI. The layout quality standard of most SMPS's is years behind on that used by seasoned class D designers.
3) Ridiculously large Y caps between primary and secondary (made necessary by incorrect choice of transformer form factor and pinout).
4) PFC, if present, overshoots at around 17 Hz (current feedforward never heard of, although many of these controller chips have dedicated input pins for this)
All methods necessary to make good SMPS's are known and understood, but only by a small number of experts (I have good reasons to believe analogspiceman is one of them).
Unfortunately, these true experts number hardly more than the number of class D experts. Until one of these few takes time out to do something special for audio, we're still better off using linear supplies.
I am personally trying to find time to make such a "good smps", more specifically a single stage PFC+galvanic isolation, the aim being to put the storage on the secondary side, get high efficiency, good pfc but only basic regulation.
Unfortunately my experience with high-voltage (primary side) design is not very great so a product won't be ready anytime soon. In the meantime - all eyes on analogspiceman
Been busy with work and taking care of a very ill parent. 
So today I've been playing with an LTspice simulation of a 180 W UcD style class d amplifier with a power factor corrected power supply (with and without voltage balancing on the outputs).
The pfc front end has been set up with a voltage loop crossover frequency of 17 Hz (with a second pole at 16 Hz), a 16 Hz double pole low pass filter on the rectified ac feedforward squared divisor signal, and a 10 Hz double pole low pass filter on the output current feedforward multiplier signal. Override operating limits are 4 amps on the input current and 400 volts on the output voltage. AC input range for proper operation is 80 to 264 volts and nominal dc output is 375 volts. A 330uF capacitor is used for primary energy storage.The pfc front end is followed by a simple dc to dc step-down supply for the isolated rail voltages. Its step-down ratio is 0.12, yielding a nominal 45 volts or so for the class d rails. A 10,000uF capacitor is hung on each rail so that total output energy storage is just slightly less than that on the 330uF boost capacitor. The rail voltages are balanced by an auxiliary circuit that can easily be disconnected by setting its series resistance to a high value.
The output load is my guess at an equivalent RLC network for a nominal 4 ohm woofer with a 34 Hz box resonance. For the test run shown, the amp is overdriven with a 15 Hz clipped sinewave burst. Without active balancing the positive rail voltage gets pumped up to a whopping 59 volts at the end of the burst.
With the balancing circuit in place it remains safely well under 50 volts at all times. An interesting note about the voltage balancing circuit: by designing it with a certain controlled level of impedance, it can avoid having to handle the large reactive currents required to keep the two output capacitors in lock step, yet still nicely squelch the pumping effect. (I intend to cover my ideas on balancing circuits in another post.)
I could post the simulation files here as a zip upon request, or they will probably show up shortly in the files sections of the LTspice or audioexperiment Yahoo Groups.
Regards -- analog(spiceman)
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.