Hi Toni
Nice to read you are OK, Steve's post reminded me I hadn't seen any trace of you and I was a little worried.
I have continued to work on the Class H we discussed but now with switch mode rail trackers.
Have you considered or sim'ed this?
I have a very simple circuit that works better than I expected.
Still not quite satisfactory however, this switch mode stuff is complicated - non linearity in the FET capacitance, and parasitics matter.
Power FETs may have almost infinite input impedance at DC but need a lot of drive to switch fast for efficiency.
And I can't analyse the loop with my favourite Tian tools.
So I'd like to have another pair of eyes look it over.
You interested?
Otherwise, stay safe and healthy.
I hope the European Covid situation improves for everyone.
Best wishes
David
Dear Dave,
nice to hear from you! Has been a long time ago we have discussed your ideas regarding class H.
And yes, I have simulated my (non switching) class-H variant with ngspice as detailed as possible but not breadboarded yet. This is the reason I haven't posted any schematics at all.
Of course I am interested to get an update to your current state of development
!
Winter is coming and there is more spare time for "fighting" with the solder iron ...
BR, Toni
P.S.: stay healthy too!
nice to hear from you! Has been a long time ago we have discussed your ideas regarding class H.
And yes, I have simulated my (non switching) class-H variant with ngspice as detailed as possible but not breadboarded yet. This is the reason I haven't posted any schematics at all.
Of course I am interested to get an update to your current state of development
!Winter is coming and there is more spare time for "fighting" with the solder iron ...
BR, Toni
P.S.: stay healthy too!
Last edited:
Of course I am interested...
I will email you the full ASC and details.
In the meantime here is the schematic of the outputs and switch mode class H.
Super simple, just the switch FET and one transistor to combine the functions of comparator and FET driver.
I use Bruno Putzeys' trick - the filter also contributes the phase shift that makes it oscillate.
*I missed the FFT frequency scale in the snip.
So, at 20 kHz and 400 W it's 60 dB down for 2nd and 3rd harmonics, open loop - no feedback at all!
With the front end and feedback it should be <0.001% distortion @ 20 kHz, if anyone needs a >400 W tweeter driver
Main issue now is to make the FET switch transitions cleaner to reduce losses.
If this is too hard then I can always have a classic, linear Class H supply for the low volt baseline before the switchers kick in.
The only downside is a small increase in power supply circuitry.
I am pretty safe because Australia has had excellent success in the Covid battle.
Almost complete elimination of local transmission (we still have cases in quarantine from travellers who return to safety)
In the next week or two it is likely local transmission may be finally stamped out.
Attachments
Last edited:
I should add that this is just a proof-of-concept schematic.
No attempt has been made to optimise the transistor selection, I just used the models I already had from my Class B+ amplifier.
The outputs can be lower breakdown Vceo and lower power so should be faster and more Hfe.
Similarly the comparator/driver.
The switch FETs have low total power dissipation but substantial peaks as they switch so SOA will be the limit, I haven't checked this yet.
The schematic uses linear rated FETs but faster, lower loss FETs may be possible, especially if a more sophisticated driver circuit is added.
Also the OPS emitter resistors can be reduced to further improve efficiency.
Best wishes
David
No attempt has been made to optimise the transistor selection, I just used the models I already had from my Class B+ amplifier.
The outputs can be lower breakdown Vceo and lower power so should be faster and more Hfe.
Similarly the comparator/driver.
The switch FETs have low total power dissipation but substantial peaks as they switch so SOA will be the limit, I haven't checked this yet.
The schematic uses linear rated FETs but faster, lower loss FETs may be possible, especially if a more sophisticated driver circuit is added.
Also the OPS emitter resistors can be reduced to further improve efficiency.
Best wishes
David
Last edited:
Instead of 4 rails I have rails that track the output, the terms are used inconsistently but I call this Class H.
The rails are labelled on the schematic and their behaviour is shown on the transient sim attachment.
The trick is that the rails are driven by a simple switch mode down converter so they are quite efficient.
Thanks, there is a trade off between better filtered rails and slew rate.
So it's not optimised for square waves.
It is possible to have a separate fast, linear but inefficient path for transients.
Yamaha did this in some of their pro amplifiers.
In practice I will probably use this for the subwoofer and woofer amplifiers that are band limited anyway so not a problem for my application.
I don't need more slew rate than 600 W power at 20 kHz I think
Best wishes
David
Last edited:
That alters the current available to turn on/off the FETs as well as the FET gate potential, so there are some constraints.
Probably needs a more sophisticated FET driver circuit but I was surprised how well it works with only one transistor!
Best wishes
David
Maybe using a totem pole driver so we get more switch off current?
And yes - it looks too simply to be true but I think it could work if the pcb layout is done correctly - we are talking about a ~ 350 - 400kHz switching frequency. Maybe we need current limiting for the FETs to avoid destroy during overload conditions. I think we should invite Chocoholic to have a look at the design. He is one of the Class D experts here in the forum - his LiteAmp Class D series is working now for years like a charm.
BR, Toni
The attachment shows the Tian loop probe (the V sources labelled "X") in the positive side feedback loop, and the Return Ratio.
So you can see it's a nice oscillator with plenty of feedback
Basically Q3 and Q4 drive the FET and are driven by the output of the FET.
What is neat is that the Q3 (& 4) emitter current acts as a supply to the output transistor in parallel with the FET so there is no wasted current.
It is even possible for Q3 to supply any excess demand that the FET can't keep up with in a transient (needs a bit more complexity however).
Yes, that's the obvious next step.
With this simple circuit it's near 1 MHz and seems OK.
Need to find more suitable transistors before serious optimisation of the frequency to best balance rail ripple/noise and switch losses.
I would like to try GaN FETs, their characteristics look well suited.
My idea is to use the FETs to switch off the power if there is DC offset, over temperature, a short at the output, and so on.
Protect both the amp and speakers, combines the function of power supply fuses and speaker relay and saves on relays and fuses.
The properly developed circuit may need a resistor in the FET gate, zener protection etc. but the proof of concept circuit actually keeps Vgs and current reasonable just with the circuit resistances.
I would love him to.
I opened a few threads on some ideas, one in the Class D forum, but there was not much interest.
Most people would rather build yet another Blameless clone, it seems.
To be fair - it is certainly easier than to develop a new circuit
Best wishes
David
Attachments
Last edited:
Just two updates.
Do you have any ideas for totem pole driver?
I can't think of a nice, simple way to do this and still keep the power supply simple too.
Ideally I would like just +plus and -minus supply rails, avoid the complexity of sub rails to run the totem pole drivers.
But drivers powered off the main rails use up headroom and decrease the efficiency.
Feels like there should be some way to do it, maybe bootstrapped, but I can't see it yet, can you?
Second is that "Choco" hasn't been active on DIYaudio at all this year.
If you have his email address perhaps you could ask him to have a look?
Best wishes
David.
Dear Dave,
totem pole: maybe something like this - see picture. Unfortunately the switching frequency highly depends on input frequency and voltage swing too...
I have simulated 2k input and the switching frequency changes between > 2MHz and 500kHz. IMHO > 500kHz switching frequency may lead to very high switching losses in the IXYS FETs. The upper plane shows the gate voltages of the IXYS FETs.
The example is dangerous as the IXYS could be destroyed because the totem pole drivers deliver to high gate voltages depending on the rail voltages. We would need 2 extra supplies for the FET drivers.
BR, Toni
P.S.: I will try to contact Chocoholic using his private address ...
totem pole: maybe something like this - see picture. Unfortunately the switching frequency highly depends on input frequency and voltage swing too...
I have simulated 2k input and the switching frequency changes between > 2MHz and 500kHz. IMHO > 500kHz switching frequency may lead to very high switching losses in the IXYS FETs. The upper plane shows the gate voltages of the IXYS FETs.
The example is dangerous as the IXYS could be destroyed because the totem pole drivers deliver to high gate voltages depending on the rail voltages. We would need 2 extra supplies for the FET drivers.
BR, Toni
P.S.: I will try to contact Chocoholic using his private address ...
Attachments
Last edited: