Hi Michael. Thanks for posting your thoughts on the amps so far and for letting them get some sun in a slightly warmer climate 
Technically speaking, what impresses me the most about the MP7720 chip is how little heat is generated. Tripath chips like that TA2024 and TA2021B get pretty warm even when they're just idling but the 7720 only gets hot when you run it near and above it's clipping point. Plus it has the same power output capability (when using 12V rails) and the chip is a lot smaller and easier to solder.
As you pointed out, they sound smoother and more laid back compared to the more "in your face" sound some of the Tripath chips have. I feel the smoother, laid back sound gives a more honest reproduction of the recording without being fatiguing. And as I've said many times, dual mono amps can give you some of the best imaging you've ever heard. That can really bring some recordings alive because good channel separation not only helps the width of the soundstange but also the depth, even though speaker positioning is even more important. I would encourage everyone to experiment with speaker placement because you don't know what you're missing until you find that sweet spot and the soundstage just comes into focus and the speakers disappear. It's comparable to listening to a good pair of full range speakers for the first time
Technically speaking, what impresses me the most about the MP7720 chip is how little heat is generated. Tripath chips like that TA2024 and TA2021B get pretty warm even when they're just idling but the 7720 only gets hot when you run it near and above it's clipping point. Plus it has the same power output capability (when using 12V rails) and the chip is a lot smaller and easier to solder.
As you pointed out, they sound smoother and more laid back compared to the more "in your face" sound some of the Tripath chips have. I feel the smoother, laid back sound gives a more honest reproduction of the recording without being fatiguing. And as I've said many times, dual mono amps can give you some of the best imaging you've ever heard. That can really bring some recordings alive because good channel separation not only helps the width of the soundstange but also the depth, even though speaker positioning is even more important. I would encourage everyone to experiment with speaker placement because you don't know what you're missing until you find that sweet spot and the soundstage just comes into focus and the speakers disappear. It's comparable to listening to a good pair of full range speakers for the first time
I have only heard from a few people, 2 of which have posted their opinions so far - Corloc and panomaniac. Everyone else... ??? I've sent out emails asking how they're doing, if they need any help, have any questions, etc but haven't heard back from anyone. Maybe they're just tired of me emailing them 
I hope to do another group buy sometime in January with updated PCBs (no major changes) and possibly a regulator board, since the 7720 chip needs stable +/-12V rails and I really like the sound with my current setup that uses regulators. I wanted to design a small SMPS for it but haven't had any time between other projects and work to even think about where I'd want to go with that.
You should be able to get the 1360 directly from Coilcraft in small quantities. http://www.coilcraft.com/general/order_us.cfm
I hope to do another group buy sometime in January with updated PCBs (no major changes) and possibly a regulator board, since the 7720 chip needs stable +/-12V rails and I really like the sound with my current setup that uses regulators. I wanted to design a small SMPS for it but haven't had any time between other projects and work to even think about where I'd want to go with that.
You should be able to get the 1360 directly from Coilcraft in small quantities. http://www.coilcraft.com/general/order_us.cfm
I've been in sporadic contact with Mike who is building mine (12 days with no power puts a crimp in internet access)... hopefully mine will make tunes early in the new year -- with this current experience, i might also rig it to run off a couple car batteries too (iPod driving stock T-Amp and A126 sounds quite sweet if a bit lacking in the bottom).
dave
dave
planet10 said:
I know he is proficient with surface mount soldering and the kit is quite simple, so I would have expected him to have both boards built in an hour or two at most. I know I'm capable of faster than an hour but I know where the parts are supposed to go.
planet10 said:
You like the sound with just a stock T-amp eh? When (if?) you get these modules your "bottom" shall be faithfully restored.
Banned
Joined 2002
planet10 said:
Yes, the lack of bass is really due to two things: low value input coupling caps and a weak power supply. If you increase the value of the coupling caps to 2.2uF or higher the high pass cutoff will be reduced below 10Hz. Going from 8AA batteries to a well regulated supply (or larger capacity SLA battery) will also raise the frequency response in the bass region.
The D10.1s have no coupling cap, and therefore have response down to DC. They also benefit from well regulated supplies, as do basically all amplifiers.
MPS 7720... wired as a UCD?
Has anyone tried using the MPS chips as a UCD?
Its a power comparator as I see it.
The UCD topology solves some of the problems hysteretic class D's have when they clip, and the distortion is lower.
They do have about 20mV of hysteresis, which needs to be overcome, i.e. a comparator in front of it...
Thanks
PortlandMike
Has anyone tried using the MPS chips as a UCD?
Its a power comparator as I see it.
The UCD topology solves some of the problems hysteretic class D's have when they clip, and the distortion is lower.
They do have about 20mV of hysteresis, which needs to be overcome, i.e. a comparator in front of it...
Thanks
PortlandMike
Re: MPS 7720... wired as a UCD?
No but I can easily modify a board so that feedback is taken solely after the output filter. The recommended configuration of resistive pre-filter feedback and capacitive post-filter feedback is a good combination sonically.
You are correct that functionally it is a power comparator. The distortion can be reduced with post filter feedback (as well as a few other nice benefits) but that doesn't necessarily mean it will sound better.
Where did you see the 20mV hysteresis figure?
Portlandmike said:
No but I can easily modify a board so that feedback is taken solely after the output filter. The recommended configuration of resistive pre-filter feedback and capacitive post-filter feedback is a good combination sonically.
Portlandmike said:
You are correct that functionally it is a power comparator. The distortion can be reduced with post filter feedback (as well as a few other nice benefits) but that doesn't necessarily mean it will sound better.
Portlandmike said:
Where did you see the 20mV hysteresis figure?
MPS as UCD?
Brian.
I tried to get this type of configuration to work but managed to fry the chip. Somehow I don't think the inputs are going directly to a comparator. I couldn't figure out just what they have done. I do think the idea has merit as eliminating the input filter cap along with post filter feedback would seem a good route to better sonics.
Please keep us up on what you are finding out.
Roger
Brian.
I tried to get this type of configuration to work but managed to fry the chip. Somehow I don't think the inputs are going directly to a comparator. I couldn't figure out just what they have done. I do think the idea has merit as eliminating the input filter cap along with post filter feedback would seem a good route to better sonics.
Please keep us up on what you are finding out.
Roger
Hi Roger. Good to see you back here!
That's right, I forgot that you tried to do this. Would you like me to send a couple replacement chips?
As far as I could tell, and from the datasheet specs, the two inputs are direct comparator inputs. They have an absolute max rating of Vss-1V to Vdd+1V and a common mode voltage range of Vss to Vdd-1.5V; common fare for a comparator.
I haven't tried to close the feedback loop around the output filter because I don't want to nuke any more chips and I would be guessing at the resistor and cap values for the phase lead network. I understand that the combined phase response of the output filter, phase lead network, and propagation delay of the power comparator sets the idle switching frequency where the total response crosses approximately -180 degrees, but I don't know the propagation delay of the power comparator and I don't have the equipment to measure it.
That's right, I forgot that you tried to do this. Would you like me to send a couple replacement chips?
As far as I could tell, and from the datasheet specs, the two inputs are direct comparator inputs. They have an absolute max rating of Vss-1V to Vdd+1V and a common mode voltage range of Vss to Vdd-1.5V; common fare for a comparator.
I haven't tried to close the feedback loop around the output filter because I don't want to nuke any more chips and I would be guessing at the resistor and cap values for the phase lead network. I understand that the combined phase response of the output filter, phase lead network, and propagation delay of the power comparator sets the idle switching frequency where the total response crosses approximately -180 degrees, but I don't know the propagation delay of the power comparator and I don't have the equipment to measure it.
BWRX said:
I thought about this a bit and remembered a simulation analogspiceman came up with some time ago. I dug it up, changed some things around, and can simulate the predicted switching frequency based on the filter/feedback component values, overall propagation delay, and estimated capacitance of the comparator input.
It should be pretty easy to determine the approximate propagation delay of the MP7720 by using the simulated values in the actual circuit, measuring the idle switching frequency, comparing it to the simulated value, and tweaking the model until the same fsw is achieved.
Time permitting, I'd like to give that a try this week and see what happens.
BWRX said:
Rise and fall times are about 5nS.
Dead time is about 50nS.
Prop delay is about 50nS, but it will depend on overdrive.
I think ideally, you'll want to put a comparator or even a 10x op amp in front of it to elemenate or at least reduce the hysteresis.
I'd use the following values, and btw this is for a bridge design, but I think they will be about the same. These will get you in the ball park and not blow up.
Feedback: 10k || (1k+470p)
input: 1k
Inductor 10uH
output caps: 1uF accross the bridge, 100n to ground from each output.
Have Fun!
Mike
Hi Mike.
Where did you see or how did you come up with your rise/fall time, dead time, and propagation delay figures?
I wouldn't add another comparator because it would add more propagation delay and complicate things. I do use an op amp volume control/buffer between my source and amps because the current components I use yield an input impedance of approximately 10k.
The values I'm using in my simulation are Ci=33p, Ri=2k, Rf=16k, Rflead=1k, Cflead=47p.
The output filter values are L=10u and C=0.47u which is what I currently use. I can tweak the cap value but want to keep the inductor value fixed. This gives a cutoff frequency of roughly 75kHz with a slightly underdamped response (Q>0.707) for loads above 4ohms.
These values result in a switching frequency of about 700kHz assuming a propagation delay of 150ns according to the simulation (single ended, non-inverting topology). My modules are single ended, inverting so I will have to do some rework to get them wired properly.
Where did you see or how did you come up with your rise/fall time, dead time, and propagation delay figures?
Portlandmike said:
I wouldn't add another comparator because it would add more propagation delay and complicate things. I do use an op amp volume control/buffer between my source and amps because the current components I use yield an input impedance of approximately 10k.
Portlandmike said:
The values I'm using in my simulation are Ci=33p, Ri=2k, Rf=16k, Rflead=1k, Cflead=47p.
The output filter values are L=10u and C=0.47u which is what I currently use. I can tweak the cap value but want to keep the inductor value fixed. This gives a cutoff frequency of roughly 75kHz with a slightly underdamped response (Q>0.707) for loads above 4ohms.
These values result in a switching frequency of about 700kHz assuming a propagation delay of 150ns according to the simulation (single ended, non-inverting topology). My modules are single ended, inverting so I will have to do some rework to get them wired properly.
BWRX said:
B,
I can't confirm your values right now, but mine are very usable.
Yours are way different. In fact I think they are just wrong in as much as I can tell, but there your chips. I didn't re-run my simulation, and values were a bit off for things, but I'm confident they will not smoke for what that's worth. I believe it runs steady state at ~600kHz.
I see your lead network not kicking in until about 200k, which is just too high, you will have at best very little negative feedback compared to what I gave you. It also does depend on supply voltage and I was using 24V.
You might try simulating distortion. small step times. and a 2^n number of cycles. That will fill you in on the amount of distortion, but unless your simulating dead time its going to be very optimistic. An AC analysis is the best way to determine feedback values. Just keep in mind these are not ideal comparators and if you have to much NF the signal at the comparator input will be so small that the prop delay will increase.
If you understand the UCD and distortion you will find that propagation delay is of only minor issue. It will not have a linear affect on the switching frequency, so its well worth it otherwise you'll have far less negative feedback and in fact, the hysteresis may even cause you to have a lower switching frequency.
Its dead time that causes distortion at the root, and NF that reduces this. the UCD rocks at this if implemented correctly, the NF that is, it does nothing for dead time.
An op amp in the volume control will do nothing for hysteresis. Also, if your single ended, you really need to rid yourself of coupling caps. If you've got coupling caps in there and you think it sounds good, take them out and it will be better to much better depending on how much your spending. That means split supplies, and often bridging is just a better way to deal with it. If you don't want to use a comparator, that's fine, but you'll need to use a good sounding fast op amp like the LM4562. At a gain of 10 that's got 5.6Mhz bandwidth approximately, and will be dead stable. But don't kid yourself, op amps have prop delay. (5.6Mhz*6.28)^-1 is an estimate. ~30nS or so. Put it all(op amp, comparator...) in the simulation, its trivial with switchercad.
Also, bridge or no, use a diff amp type network around the UCD. Go to Hypex's web site and you'll see what I'm talking about.
In fact, adding the dual op amp they use is a great idea, but that will be in addition to the comparator or 10x gain amp used to reduce hysteresis. If your not going to reduce or eleminate the hysteresis, the hysteresis will likely reduce the NF such that much of the UCD benifit will be lost, and also I think you'll get all the nasty disadvantages of the prefilter feedback in clipping, that is in band switching noise right at clipping that the UCD is far more tolerant of.
As for your first question,
I always look for the DS to provide my data, and if what I want isn't there, I talk to the manufacture.
What is Ci?
Regards,
Mike
Hi Mike.
Attached is the circuit I was using in switchercad to simulate the switching frequency. If it is incorrect would you mind telling me how to fix it? As I said, it is a simulation that analogspiceman came up with a year or two ago. I just entered it into switchercad and changed the part values.
analogspiceman also came up with something to simulate distortion. I may try to get something similar working if you think it will help analysis.
I thought UCD is a phase shift topology? Doesn't that make propagation delay a very large issue, especially around the frequencies I'd like the amp to be switching at? The propagation delay is the major factor in determining the upper limit on the possible switching frequency, is it not?
The op amp buffer/volume control isn't meant to alter the hysteresis. Simply to buffer the source from the pot and to provide a low impedance drive for the circuit. Getting rid of the coupling cap was the whole reason I chose the single ended design with +/- rails.
Since these amps are working off of 12V rails they obviously don't need a lot of gain. A 10x gain stage in front of the amp would mean you'd want the class d portion around unity gain...
I know op amps have propagation delay, but that shouldn't matter since they are not inside the feedback loop. How would you use them inside the feedback loop?
I'm working towards the instrumentation configuration using the class d stage as the differential amp.
I thought hysteresis could just be lumped into the propagation delay figure. Guess I need to do some more reading!
I couldn't find any of that info in the datasheet and had emailed MPS inquiring about the dead time a year ago but got no response. I was just wondering how you managed to get that information.
So all in all, it sounds like you have a working prototype built using the MP7720 in a UCD configuration. How does it sound?
Thanks for resparking my interest in this!
Brian
Attached is the circuit I was using in switchercad to simulate the switching frequency. If it is incorrect would you mind telling me how to fix it? As I said, it is a simulation that analogspiceman came up with a year or two ago. I just entered it into switchercad and changed the part values.
Portlandmike said:
analogspiceman also came up with something to simulate distortion. I may try to get something similar working if you think it will help analysis.
I thought UCD is a phase shift topology? Doesn't that make propagation delay a very large issue, especially around the frequencies I'd like the amp to be switching at? The propagation delay is the major factor in determining the upper limit on the possible switching frequency, is it not?
The op amp buffer/volume control isn't meant to alter the hysteresis. Simply to buffer the source from the pot and to provide a low impedance drive for the circuit. Getting rid of the coupling cap was the whole reason I chose the single ended design with +/- rails.
Since these amps are working off of 12V rails they obviously don't need a lot of gain. A 10x gain stage in front of the amp would mean you'd want the class d portion around unity gain...
I know op amps have propagation delay, but that shouldn't matter since they are not inside the feedback loop. How would you use them inside the feedback loop?
I'm working towards the instrumentation configuration using the class d stage as the differential amp.
I thought hysteresis could just be lumped into the propagation delay figure. Guess I need to do some more reading!
I couldn't find any of that info in the datasheet and had emailed MPS inquiring about the dead time a year ago but got no response. I was just wondering how you managed to get that information.
So all in all, it sounds like you have a working prototype built using the MP7720 in a UCD configuration. How does it sound?
Thanks for resparking my interest in this!
Brian
Attachments
BWRX said:Hi Mike.
Attached is the circuit I was using in switchercad to simulate the switching frequency. If it is incorrect would you mind telling me how to fix it? As I said, it is a simulation that analogspiceman came up with a year or two ago. I just entered it into switchercad and changed the part values./
Can't spend the time now, maybe later, sorry.
It is insightful.
Maximizing switching frequency is not the goal, but maximizing negative feedback IMO. 20mV of hysteresis is going to limit the amount of NF you can get seriously.
This stage of 10x gain doesn't increase the system gain, just reduces the affect of the hysteresis. It is IN the FB loop. We need a picture I'm afraid. L8R.
Yes, in calculating the swithing frequency, but it really causes a hit in potential NF.
email is how I got it. Phone calls work better though.
So all in all, it sounds like you have a working prototype built using the MP7720 in a UCD configuration. How does it sound?
Thanks for resparking my interest in this!
Brian
Sorry to say I don't. But plan to. Problem is its best not done as a circuit hack, since I really think you want a comparator in the end. btw. Check out a LMH7720. If your running on 12 volts, it might be an easy hack. Also, put your part at over 20V. It will be much better. Class D's always like more volts, and there is minimal penalty for this.
Mike
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.